THOMPSON BOWCHER FONTAINE SCHONTHAL 2019 The Cambridge Handbook of Systemic Functional Linguistics PDF

The Cambridge Handbook of Systemic Functional Linguistics
Introduction 1
Wendy L. Bowcher, Lise Fontaine and David Schoenthal
Part I. SFL: The Model:

1. Firth and the origins of systemic functional linguistics: process, pragma, and
polysystem 11
David G. Butt
2. Key concepts and the architecture of language in the SFL model 35
Jonathan J. Webster
3. Semantics 55
Miriam Taverniers
4. The clause: an overview of the lexicogrammar 92
Margaret Berry
5. The rooms of the house: grammar at group rank 118
Lise Fontaine and David Schoenthal
6. Context and register 142
Wendy L. Bowcher
7. Intonation 171
Wendy L. Bowcher and Meena Debashish
8. Continuing issues in SFL 204
Mick O'Donnell
9. The Cardiff model of functional syntax 230
Anke Schulz and Lise Fontaine
10. SFL in context 259
Christopher S. Butler
Part II. Discourse Analysis within SFL:

11. Models of discourse in systemic functional linguistics 285
Tom Bartlett
12. Cohesion and conjunction 311
Maite Taboada
13. Semantic networks 333
Andy Fung and Francis Robert Low
14. Discourse semantics 358
J. R. Martin
15. Appraisal 382
Susan Hood
16. SFL and diachronic studies 410
David Banks
17. SFL and multimodal discourse analysis 433
Kay L. O'Halloran, Sabine Tan and Peter Wignell
18. SFL and critical discourse analysis 462
Gerard O'Grady
Part III. SFL in Application:

19. Language development 487
Geoff Williams
20. Applying SFL for understanding and fostering instructed second language
development 512
Heidi Byrnes
21. Language and education: learning to mean 537
Peter Mickan
22. Systemic functional linguistics and computation: new directions, new challenges
John Bateman, Daniel McDonald, Tuomo Hiippala, Daniel Couto-Vale and Eugeniu
Costetchi 561
23. Clinical linguistics 587
Elissa Asp and Jessica de Villiers
24. Language and science, language in science, and linguistics as science 620
M. A. K. Halliday and David G. Butt
25. Language and medicine 651
Alison Rotha Moore
26. Language and literature 690
Donna R. Miller
27. Language and social media: enacting identity through ambient affiliation 715
Michele Zappavigna
28. Theorizing and modeling translation 739
Erich Steiner
29. Language typology 767
Abhishek Kumar Kashyap
Index 793
Preface
This volume has been several years in the making. It was first conceived of
in 2013, when Cambridge University Press approached Lise Fontaine with
the possibility of including a Handbook on Systemic Functional Linguistics as
part of its series of Handbooks on Language and Linguistics. Recognizing
this as a wonderful opportunity, Lise, Geoff Thompson, and Wendy
Bowcher discussed the possibility of co-editing the volume. It was decided
that Geoff would take the lead, and in consultation with various scholars,
including Michael Halliday, he developed the conceptual framework for the
book – a volume with a comprehensive, somewhat historical but also
forward-looking overview of Systemic Functional Linguistics. Later, after
Geoff’s untimely death, David Schönthal was invited to join the editorial
team. As editors, we encouraged contributors to include both theoretical
and practical details where possible – the latter noted by Halliday in
personal correspondence as being an important part of the character of a
‘handbook’. Finding contributors to this volume was difficult, but in a
positive way, as there are so many scholars around the world with expertise
in the various areas covered who could have been approached. The final
line-up, we feel, offers a wide scope of perspectives from a range of estab-
lished and emerging scholars, some expert in more than the field of
research which they have written about in this volume. We would like to
take a moment here to thank all our contributors for the effort and
expertise they have brought to this collection. Readers will notice that at
the beginning of some of the chapters there is a note of tribute to several
scholars who have passed away since the volume’s inception: Chapter 4
pays tribute to Geoff Thompson (see also the tribute to Geoff at the
beginning of this volume), Chapter 7 to Bill Greaves, Chapter 23 to Johna-
than Fine, and Chapter 26 to Ruqaiya Hasan. We felt it was important to
include these tributes – to Geoff himself as the person who really got this
project off the ground, and to Geoff and all the other scholars who have
been such an important influence not only in the development of the
Downloaded from https://www.cambridge.org/core. The Librarian-Seeley Historical Library, on 20 Aug 2019 at 08:30:31, subject to the Cambridge Core terms of use,
available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316337936.001
xxx PREFACE
theory and practice of Systemic Functional Linguistics evident throughout

the book, but in their commitment to furthering the field through their
encouragement, generosity, and dedication to mentoring SFL scholars
around the world. In the final stages of preparing the manuscript, we
became aware that Emeritus Professor Michael Halliday, the founder of
Systemic Functional Linguistics did not have long to live. It was not long
after we had submitted the manuscript that he passed away, on 15 April
2018. It is a great privilege and honour to have included in the volume
Chapter 24 ‘Language and Science, Language in Science, and Linguistics as
Science’, which he co-authored with David Butt. After enquiring as to
whether this was the last work that Michael Halliday penned, David Butt
kindly offered to write a brief note on the nature of the co-authorship of
this chapter. This note is presented at the end of Chapter 24.
Introduction
Wendy L. Bowcher, Lise Fontaine, and
David Schönthal
I.1 The General Scope of This Volume
The Cambridge Handbook of Systemic Functional Linguistics presents a clear and

comprehensive overview of one of the most appliable and socially progressive
linguistic theories available today. As a social semiotic theory of language,
Systemic Functional Linguistics (SFL) prioritizes language from the perspec-
tive of systems of meaning – how these systems are shaped, and, at the same
time, how they play a role in shaping human social systems and how they
relate to other systems of meanings within society. With its origins firmly
grounded in a functional approach to language, SFL brings into theoretical
rigour the concept of function in relation to language in a unique and robust
way. That is, in SFL theory, function is ‘meaning in context’. It is not simply
equated with ‘use’, but is considered a property of language at every level of
description (cf. Halliday 1985: 17). Moreover, SFL places importance on its
‘appliability’. M. A. K. Halliday, the founder of SFL theory, has described a
linguistic theory which is appliable as one which
tackles problems and tries to answer questions – but questions that are
asked, and problems that are raised, not by professional linguists so much
as by other people who are in some way concerned with language, whether
professionally or otherwise. There are large numbers of such people: educa-
tors, translators, legal and medical specialists, computer scientists, students
of literature and drama, . . .; and it is their ‘take’ on language that is being
addressed, at least to the point of clarifying what sorts of questions can
usefully expect to be asked, and whether or not there is any hope of coming
up with an answer.
(Halliday 2013:128)
Unlike theories of language that separate ‘langue’ from ‘parole’ and which
consider parole as a somehow ‘flawed’ version of language, SFL recognizes
the ‘symbiotic relation’ between langue and parole:
Downloaded from https://www.cambridge.org/core. Stockholm University Library, on 30 Oct 2019 at 11:18:10, subject to the Cambridge Core terms of use, available
at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316337936.002
2 WENDY L. BOWCHER, LISE FONTAINE, AND DAVID SCHÖNTHAL
language as an already produced and an evolving system is predicated on the

symbiotic relation between parole and langue – between process and its
product – so we may claim that the system enables the efficacy of process
while process fashions the very rules by which it attains this efficacy.
(Hasan 2009:310)
This symbiotic relation is also seen in the relation between language and
society, and is explained through the social semiotic perspective on lan-
guage that SFL holds:
The universal characteristics of parole – its orderly variation, its flexible regu-
larities – are functional (Halliday 1970): they have their origin in the relations of
parole to the community’s living of life, while at the same time, the various
dimensions of a community’s social contexts of living depend on parole for
their creation, maintenance, and evolution. Language as a social semiotic is
predicated on this mutual relation between parole and social contexts.
(Hasan 2009:309–10)
These principle perspectives of SFL theory underpin the contributions in

the present volume, with these contributions covering the theory’s origins,
architecture, key concepts, levels of analysis, and areas of application. Key
terms are defined within the chapters, and key concepts are cross-
referenced where relevant. Such cross-references to chapters within this
handbook are given by reference to the author of the relevant chapter – for
example, ‘see Butt, this volume’ to guide the reader (in this case) to
Chapter 1. While the volume could be read from beginning to end, it is
not necessary for readers to do so. Rather, readers wishing to understand a
specific area of the theory or its application can refer to those chapters most
relevant to their area of interest.
I.2 The Structure and Contents of This Volume
The chapters in this volume cover a comprehensive range of theoretical

perspectives and applications of Systemic Functional Linguistics written by
some of the world’s foremost SFL scholars, including M. A. K. Halliday, the
founder of SFL theory. As editors, we have endeavoured to compile a
volume that can be used primarily as a reference tool with descriptions
and explanations of theoretical concepts and discourse analytical tools
along with some exemplar analyses. There are also ample citations pointing
readers to further literature wherein more detailed information and ana-
lyses can be obtained. Many of the chapters include a brief discussion of
possible future directions in which research might be conducted or issues
be further investigated and resolved. Ultimately, we hope that readers will
not only become better informed about the various features of SFL theory
and the value it can bring to solving societal problems, questions, and
ambiguities in which language features, but that they may be inspired to
Introduction 3
pursue some of the challenges and issues raised within the volume – be
they theoretical or practical in nature.
The volume is divided into three parts: Part I, ‘SFL: The Model’; Part II,
‘Discourse Analysis within SFL’; and Part III, ‘SFL in Application’. The
following paragraphs summarize the contents of the chapters in each of
these sections.
I.3 Part I – SFL: The Model
As the introductory part of the volume, Part I, ‘SFL: The Model’, has been
designed in such a way as to cover the core features and terminology of the
SFL framework. The organization of this part is designed in part to reflect
the key perspectives on the theory. Opening the part, in Chapter 1, David
G. Butt lays out in considerable detail the origins and history of how the SFL
approach evolved. Halliday’s interest in a language-based linguistics is
shown to derive directly from J. R. Firth, and, by better understanding
Firth’s concerns, the reader gains valuable insight into Halliday’s develop-
ment of SFL. A description of key terms in the SFL model is then given in
Chapter 2, by Jonathan J. Webster. These two chapters form a necessary
background for the more specific chapters that follow.
In Chapter 3, Miriam Taverniers takes up the central concept of seman-
tics and explores how it is conceptualized and modelled in SFL theory. In
particular, she teases out the different conceptions of semantics within SFL.
Importantly, she relates the key concepts of abstraction, patterning, and
actualization to stratification and metaredundancy. Chapters 4 through 7,
then, combine to provide a detailed discussion of four key approaches or
perspectives on language in the SFL framework: the clause, units of the
clause, context, and sound patterns. The multifunctional view of the clause
is detailed by Margaret Berry in Chapter 4 on the lexicogrammar. Berry
presents a concise analytical overview of the clause from the experiential,
interpersonal, and textual metafunction. In Chapter 5, Lise Fontaine and
David Schönthal present a critically engaged description of the units of
‘group’ and ‘phrase’. After reviewing the different units below the clause,
they go on to challenge the distinction between the units of ‘group’ and
‘clause’. Context and its relation to text type is examined by Wendy
L. Bowcher in Chapter 6, as she details the concepts of context and register
within the model. Specifically, Bowcher discusses the history of these
two concepts and their relation between one another, and reviews
seminal SFL research on context and register. Chapter 7, by Wendy
L. Bowcher and Meena Debashish, details how intonation and English tone
groups are situated within the SFL framework. Not only do the authors
offer a usable description of English intonation, they also raise important
issues related to topics currently under debate and possible areas of future
research.
The final three chapters of Part I each step outside the standard SFL
model in different ways. In Chapter 8, Mick O’Donnell explores aspects of
the model which are under debate within the field. He outlines the main
points of interest, showing where there is scope for significant contribu-
tions from researchers. Chapter 9, written by Anke Schulz and Lise Fon-
taine, presents the model of functional syntax as developed within the
Cardiff Grammar. As a model with its roots in SFL theory, the Cardiff
Grammar shares many of the same principles as outlined in the other
chapters in this part. However, there are some important differences illus-
trated in this chapter. In the final chapter of Part I, Chapter 10, Christopher
S. Butler situates SFL in its theoretical context in relation to other func-
tional approaches, or what he refers to as ‘functional-cognitive space’. This
chapter, based on a detailed comparison of sixteen different models, shows
that while some differences are highlighted, there are also some interesting
points of shared concerns.
I.4 Part II – Discourse Analysis within SFL
The second part of the volume contains eight chapters which present
various discourse analytical tools developed within the framework of SFL
theory. This part begins with Chapter 11, by Tom Bartlett, who first
describes SFL from a discourse analytical perspective and then discusses
some of the main approaches to discourse analysis within SFL. This chapter
includes critical comments on some of the analytical approaches and effect-
ively sets the background for the chapters which follow.
Chapter 12, by Maite Taboada, focuses on cohesion and conjunction. This
chapter describes Halliday and Hasan’s (1976) early work on cohesion, and
includes a brief description of cohesive harmony (Hasan 1985; Khoo 2016), as
well as work on rhetorical structure theory (RST) (Mann et al. 1992). In her
discussion of conjunction, Taboada briefly points out some of the differences
between Martin’s (1992) and Hasan’s (1985) descriptions. The chapter also
includes a brief description of some areas in which Halliday and Hasan’s
model of cohesion and coherence has been applied, such as in computational
studies and foreign language teaching and learning.
Chapter 13, by Andy Fung and Francis Robert Low, focuses on semantic
networks as developed by Hasan (1996), but the chapter first situates
this discourse analytic framework in relation to other semantic-level
approaches within SFL. The chapter describes the basic unit of analysis,
the ‘message’, and demonstrates the utility of this framework through an
analysis of a mother–child interaction. The latter part of the chapter illus-
trates how semantic networks have been applied, with a focus on pedagogic
and journalistic discourse.
Chapters 14 (J. R. Martin) and 15 (Susan Hood) describe two related analyt-
ical frameworks: discourse semantics and the system of appraisal. Martin
Introduction 5
briefly explains how his ‘discourse semantics’ differs from Halliday and
Hasan’s (1976) concept of coherence and cohesion before elaborating on
various systems within his framework. The second part of the chapter
presents a text analysis demonstrating the application of the different dis-
course semantic systems and the value of this analytical approach in high-
lighting threads of ideational, interpersonal, and textual meanings
throughout a text, and how these threads relate to each other. Towards the
end of the chapter, Martin discusses the interpersonal system of appraisal,
which provides a natural segue to Hood’s chapter on Appraisal. Hood first
situates the appraisal system within SFL theory and then introduces the
three main appraisal sub-systems: attitude, graduation, and engage-
ment. The latter part of her chapter discusses some of the ways that the
system of appraisal has been applied, ending with a discussion of the
current trends in Appraisal research, including multimodal studies,
research into legal language, and studies of identity and affiliation.
Chapter 16, Diachronic Studies by David Banks, while not technically
about a discourse analytic approach, is about a discourse analytic perspec-
tive – diachrony – and Banks illustrates some of the discoursal features that
are focused on by SFL researchers whose data and analytical perspectives
would fall within the domain of diachronic research.
Chapter 17, by Kay L. O’Halloran, Sabine Tan, and Peter Wignell, covers
multimodal discourse analysis, a particularly fruitful and growing area
within the SFL theoretical framework, abbreviated as SF-MDA (Systemic
Functional Multimodal Discourse Analysis). The chapter highlights the
features of SFL theory which inform SF-MDA before presenting an exemplar
analysis of a multimodal text, an internet webpage, using the analytical
tools of SF-MDA.
The last chapter in this part, Chapter 18, by Gerard O’Grady, outlines the
relationship between SFL theory and Critical Discourse Analysis (CDA),
particularly that developed by Norman Fairclough (1989, 2015). After
tracing the development of CDA, O’Grady describes Fairclough’s method-
ology, pointing out its compatibility with SFL theoretical perspectives and
analytical concepts. Towards the end of the chapter, O’Grady presents a
critique of the criticisms of SFL-inspired CDA, and in the last part of the
chapter he outlines some of the particularly productive areas of CDA
research which make use of SFL theoretical tools.
I.5 Part III – SFL in Application
This third part of the volume presents several fields of research in which
the theory of SFL has been applied, with the first three chapters
(Chapters 19, 20, and 21) on different areas related to language develop-
ment and learning. Geoff Williams’ chapter describes Halliday’s ground-
breaking work on child language development (Chapter 19), and is
organized around two main thrusts of the SFL approach: how research into
language development informs the SFL theory of language, and how learn-
ing language goes hand in hand with learning culture and in developing
one’s position and identity within society. Williams relates how this latter
perspective, in particular, is demonstrated through research into
mother–child interaction conducted by Ruqaiya Hasan and analyzed using
her semantic networks (e.g. Hasan 1996).
Chapter 20, by Heidi Byrnes, moves on to the application of SFL in the
field of second language development. She first locates SFL within the
general domain of second language acquisition research, and then discusses
the application of several key features of SFL theory and L2 teaching and
learning, noting, in particular, the value brought to the L2 teaching/learn-
ing environment of the SFL approach to the description of language and
language in use, as well as its description of a lexicogrammar rather than
the typical and ‘unsustainable’ separation of lexicon and grammar. She also
raises the significance of SFL research into and description of grammatical
metaphor in second language learning and teaching.
In Chapter 21, Peter Mickan outlines how SFL theory has been applied
within the field of general education. Mickan’s chapter covers work focus-
ing on early childhood and primary school education, secondary school,
tertiary education, and finally teacher-training and educational research.
Underlying all these chapters is the principle that learning language and
learning through language is a process of ‘learning how to mean’.
In Chapter 22, John Bateman, Daniel McDonald, Tuomo Hiippala, Daniel
Couto-Vale, and Eugeniu Costetchi note the long history of connection
between SFL and computational linguistics, mentioning Halliday’s involve-
ment in ‘some of the earliest attempts to achieve automatic translation
systems in the 1950s’ and his key role in some of the ‘most well-known
language-oriented systems to emerge in computational linguistics and Arti-
ficial Intelligence in the 1970s and 1980s’. The chapter discusses recent SFL-
related research, noting some of the challenges that a meaning-based
theory of language poses for computational models, but also the distinct
and far-reaching possibilities that SFL can offer the field.
The next three chapters (Chapters 23, 24, and 25) are connected in terms
of their focus on SFL in relation to science and medical research. Chapter 23,
by Elissa Asp and Jessica de Villiers, concerns clinical linguistics. After
briefly describing the field of clinical linguistics, Asp and de Villiers present
an overview of research that falls within the SFL theoretical approach,
including work on schizophrenia, neurodevelopmental and neurocognitive
disorders, Alzheimer’s disease, and aphasia. They also discuss the signifi-
cance of SFL-informed research and some future directions.
The focus in Chapter 24, by Michael Halliday and David G. Butt, is
‘science’ and scientific language. They describe the part that language has
played in the development of science and how scientific language, as a
register, has come to construe knowledge and experience in a specific and
Introduction 7
‘uncommonsense’ way, which they argue is in line with the social purposes
of science. Another argument that the authors make is that science lan-
guage, or ‘verbal science’, operates in a similar way to verbal art (see Miller,
this volume) in that there is a ‘symbolic articulation’ – that is, conventional
language choices realize metaphorical constructions of meanings to create
novel understandings.
Alison Rotha Moore, in Chapter 25, comprehensively reviews SFL-
informed research into language and medicine. She describes the kind of
health problems and medical contexts in which research has been con-
ducted, such as HIV, emergency services, surgery contexts, and health
curricula. She discusses the analytical tools used and then outlines the
achievements SFL researchers have made in this field. The chapter also
suggests some of the directions this kind of research can take and the
possible knock-on improvements that could emerge in healthcare and in
the healthcare system.
The next chapter (Chapter 26), by Donna R. Miller, entitled Language and
Literature, presents some of the most innovative work on the analysis of
‘verbal art’ available. Miller presents a historical recount of the field of
stylistics and the place of Halliday’s and Hasan’s work in relation to this,
noting the possible reasons why Halliday’s work has been acknowledged
outside the circle of SFL scholars, whereas, surprisingly, Hasan’s has
received little recognition. Several key influential figures emerge in her
chapter, such as Jakobson and Mukařovský. The chapter describes Hasan’s
systemic socio-semiotic stylistics (SSS) model and its value and insights for
understanding the ‘art’ in verbal art, demonstrating that literature is not
like other varieties of language, but is a special variety.
In Chapter 27, Michele Zappavigna describes current work in the appli-
cation of SFL tools to analyzing social media platforms such as Twitter,
Facebook, Instagram, Weibo, etc. Specifically, she focuses on the way SFL
tools can unlock the construction of ‘identity’ and ‘affiliation’ in such
platforms, pointing out that this kind of investigation derives from Firth’s
(1950) work on the language of persons and personalities and his concepts
of ‘communion of feeling’ and ‘the user in uses’ (see also Martin 2009). The
chapter also discusses issues involved in collecting social media data.
Chapter 28, by Erich Steiner, focuses on translation studies and the
usefulness of SFL ideas and concepts for theorizing and modelling the
process of translation. His chapter compares translation with multilingual
text production and interpreting, and discusses the relationship between
translation, text variation, and paraphrase. It also includes a discussion on
the SFL perspectives on equivalence, the translation of registers or text
types, and the role of the translator in the process of translation. Steiner
describes some of the SFL tools for text analysis that are relevant to transla-
tion and possible future directions of SFL-informed translation studies.
The last chapter in the volume is Abhishek Kumar Kashyap’s chapter on
language typology (Chapter 29). Kashyap briefly traces the development of
the field of language typology, the kinds of questions that typologists ask,
and some of the different theoretical approaches taken. The chapter covers
the major contributions to language typology from an SFL perspective and
applications of language typology to other fields such as translation, inter-
cultural communication, and language teaching and learning.
References
Fairclough, N. 1989. Language and Power. Harlow: Longman.

Fairclough, N. 2015. Language and Power. 3rd ed. Abingdon: Routledge.
Firth, J. R. 1950. Personality and Language in Society. The Sociological Review
42(1): 37–52.
Halliday, M. A. K. 1985. Part A. In M. A. K. Halliday and R. Hasan, Language,
Context, and Text: Aspects of Language in a Social-semiotic Perspective. Geelong,
Vic.: Deakin University Press. 1–49.
Halliday, M. A. K. 2013. Putting Linguistic Theory to Work. In J. J. Webster,
ed., The Collected Works of M. A. K. Halliday, Volume 11: Halliday in the
21st Century. London: Bloomsbury. 127–42.
Halliday, M. A. K. and R. Hasan. 1976. Cohesion in English. London: Longman.
Hasan, R. 1985. Part B. In M. A. K. Halliday and R. Hasan, Language, Context,
and Text: Aspects of Language in a Social-semiotic Perspective. Geelong, Vic.:
Deakin University Press. 52–118.
Hasan, R. 1996. Semantic Networks: A Tool for the Analysis of Meaning. In
C. Cloran, D. Butt, and G. Williams, eds., Ways of Saying, Ways of Meaning:
Selected Papers of Ruqaiya Hasan. London: Cassell. 104–30.
Hasan, R. 2009. Rationality in Everyday Talk: From Process to System. In J. J.
Webster, ed., The Collected Works of Ruqaiya Hasan, Volume 2: Semantic
Variation: Meaning in Society and in Sociolinguistics. Sheffield: Equinox.
309–52.
Khoo, K. M. 2016. ‘Threads of Continuity’ and Interaction: Coherence,
Texture and Cohesive Harmony. In W. L. Bowcher and J. Y. Liang, eds.,
Society in Language, Language in Society: Essays in Honour of Ruqaiya Hasan.
Basingstoke: Palgrave Macmillan. 300–30.
Mann, W. C., C. M. I. M. Matthiessen, and S. A. Thompson. 1992. Rhetorical
Structure Theory and Text Analysis. In W. C. Mann and S. A. Thompson,
eds., Discourse Description: Diverse Linguistic Analyses of a Fund-Raising Text.
Amsterdam: John Benjamins. 39–78.
Martin, J. R. 1992. English Text: System and Structure. Amsterdam: John
Benjamins.
Martin, J. R. 2009. Realisation, Instantiation and Individuation: Some
Thoughts on Identity in Youth Justice Conferencing. DELTA: Documentação
de Estudos em Lingüística Teórica e Aplicada 25(SPE): 549–83.
1
Firth and the Origins
of Systemic Functional
Linguistics
Process, Pragma, and Polysystem
David G. Butt
1.1 J. R. Firth (1890–1960): First Impressions and

Paradoxes of Plain Statement
‘[M]eaning’ is a property of the mutually relevant people, things,

events in the situation. Some of the events are the noises made
by the speakers. (Firth 1964:111)
The ideas and principles of Professor John Rupert Firth are an essential
source of what is important and distinctive about the development, after
Firth’s death, of Systemic Functional Linguistics (SFL). Firth was the first
Professor of General Linguistics in England (1944) at the School of Oriental
and African Studies at London University (SOAS). He also headed the
Department of Phonetics in London. Though first trained in history,
between the outset of WWI and 1928, Firth had worked as a teacher of
English and as a professor in British Imperial India (at Lahore, now Paki-
stan) as well as in Afghanistan and East Africa. He later returned to these
communities to conduct further descriptions of languages. His students
and colleagues are notable for the extent and depth to which they
developed the study of languages of these regions, as well as languages
of East and South East Asia. Firth emphasized the importance of
de-Anglicization, and of looking back at one’s own language from the
perspective of another culture. This is a form of ‘de-familiarization’ quite
remarkable in a person who seemed a conservative Yorkshireman. Yet, as
emphasized by Roman Jakobson after their two meetings, Firth shared with
the pioneer of British linguistics, Henry Sweet (1845–1912), an ‘unusual
courage to see the world’ with his own eyes ‘irrespective of the environ-
mental usage, habit and predilection’ of conventional thinking (Jakobson
1966:242). For others, especially to phonemicists and morphologists in
America, this ‘unusual courage’ appeared to be eccentricity, and a lack of
Downloaded from https://www.cambridge.org/core. Columbia University Libraries, on 11 Aug 2019 at 08:12:20, subject to the Cambridge Core terms of use, available
12 DAVID G. BUTT
commitment to being methodical or complete in his explanations. Thomas

(2011:179–80) offers a recent synopsis of such reactions; but note the
related misgivings expressed by one of Firth’s colleagues at SOAS (see
Palmer 1968:4).
The thesis of this chapter is that those emphasizing the eccentricity and
‘obscurity’ of the linguistic approach urged by J. R. Firth have misconstrued
the rational basis of Firth’s motivations, and produced a ‘conventional
wisdom’ that continues to obstruct our understanding of the potential of
linguistics amongst the human sciences. Firth, in essence, sought a
language-based linguistics. This is to say Firth would not follow other
theorists who deferred to psychological, philosophical, or logical units as
if such units offered some extrinsic grounding and escape from relativity –
namely, the relativity inherent in one group of human speakers examining
patterns of speech in another community. Furthermore, by contrast with
Bloomfield, and then with Chomsky, Firth’s emphasis was on the linguist
making statements of meaning (certainly not making statements against
the possibility of such statements!). To these issues – the relativity of the
language we use for describing languages and of the primacy of meaning –
Firth was, in my view, punctiliously consistent. Both these issues are of
increasing importance in human sciences: for instance, on the one hand,
through a growing scepticism concerning universals in language and the
commensurability of frameworks by which languages are described; and,
on the other, through the growing emphasis on interpersonal meaning,
that is, meaning over and beyond reference and truth conditions (see
Section 1.3). Firth (1968:97) put his views succinctly:
My own approach in general linguistics and especially in the study of mean-

ing in purely linguistic terms dates back to about 1930 when the linguistic
movement in philosophy was also arousing interest. My main concern is to
make statements of meaning in purely linguistic terms, that is to say, such
statements are made in terms of structures and systems at a number of
levels of analysis: for example, in phonology, grammar, stylistics, situation,
attested and established texts. I do not attempt statements about a speaker’s
or writer’s thoughts and intentions, ideas and concepts – these are for other
disciplines.
We might note at this point that there is little difficulty in citing what Firth
recommended for linguistics, and also in establishing what he believed a
linguist should abjure. Firth’s view is at the polar extreme from Chomsky’s
claim that linguistics is a ‘branch of cognitive psychology’ (Chomsky
1972:1). Similarly, the idea that linguists might concede the domain of
semantics to a philosopher of ‘intentions’, like J. R. Searle (1969), or that
we pass pragmatics over to the maxims of a logician like H. P. Grice (1989),
would have seemed a total abrogation of the roles and tool power of
linguistics. For a start, such an approach assumes a diminution of linguis-
tics, namely, that each of the strata of language could be contracted out and
Firth and the Origins of SFL 13
managed separately and autonomously, with linguists left to ‘mop up’

the syntactic and lexicographical curiosities beside a few eccentric
phoneticians.
To explain my thesis concerning Firth and the role of linguistics in the
human sciences, I will emphasize the Firthian notions of meaning and
context of situation, with some inclusion of phonetics, phonology, and
the issues surrounding ‘prosody’ – the latter supplying a metaphor for
the pattern and process on all the levels of the linguistic order in the
Firthian approach: viz. ‘semantic prosody’. I recognize that my views may
be at odds with certain of the expressed views of very great scholars, some
of whom had direct contact over years with Firth. Perhaps, being ‘in medias
res’ can create its own difficulties in the sorting out of any history of ideas.
On the other hand, I may simply be wrong on various matters due to my
distance from the era, and due to my own limitations in linguistics. But one
can only attend to the relevant texts, and strive to reconcile sources –
working for a linguistic, scientific, and sociohistorical coherence. My read-
ing is that such coherence has not been achieved, despite there being
certain advantages for such reconstrual of Firth’s work in the current
intellectual milieu.
The argument here, through meaning and context, has not been an
obvious pathway for positive evaluations of Firth: meaning and context of
situation have even been a source of some embarrassment for those evalu-
ating Firth’s ideas, for some of those inside the orbit of Firth’s influence, as
well as for outside commentary (see Langendoen 1968; Lyons 1968).
A recent overview by Kachru (2015:72–93) is a valuable addition to evalu-
ations of Firth in that Kachru’s career spanned decades of Firth’s influence,
including an exploration of the ways in which context might be invoked
in linguistic description.
A principled approach to context appears to be a primary step for Firth in
the general problem of ‘turning language back on itself’, not merely an
addendum or codicil. Firth emphasizes social process (the living of life) and
the engaged, human body anticipating a structured future, the ‘conser-
vation of the pattern of life’ (Firth 1957:143). The expression ‘Turning
language back on itself’ was Firth’s characteristically concrete way of
drawing attention to the ‘timeless and ineffable’ character of linguistic
categories (Firth 1968:39), and of avoiding the veneer of technicality in
the logician’s term ‘metalanguage’ (Halliday 2002 [1985]). Metalanguage
too easily takes on a false status as if representing prelinguistic, or even
presemiotic, categories. Firth’s formulation reminds us that meanings can
only be rendered by another version of themselves: hence the ineffability.
For example, to suggest that [+Animate] or [+Male] take us deeper into a
semantic or logical atomism would be simply to overlook the ideologically
loaded and relativistic value these terms have already in English (and
certainly between speakers of different Englishes). The category terms that
represent the complex of relations of a language should not be mental,
14 DAVID G. BUTT
acoustic, or even social primes that are extrinsic to languages. Linguistic

categories are tools that bring out a pattern or regularity in relation to an
immediate analytic purpose. They are therefore ‘ad hoc’ and not to be
considered as existents in themselves.
In fact, the positive and negative in Firth’s views are like points one can
plot across a theoretical phase space (Cohen and Stewart 1995:198–210).
Were one to miss the ‘co-ordinates’ of theory announced or implied in one
paper, their repetition in an adjacent paper is awaiting any reader willing
to track the ‘thematic motifs’ of this approach to the study of meaning. On
the one hand, Firth can be alerting us to the idealization of the system in
the work of Saussure and of Hjelmslev (the latter’s semiotic ‘calculus’; see
Firth 1957:140); yet, on the other hand, Firth can be seen applying the
notions of valeur and paradigm without compromise, in a strictly relational
characterization of all linguistic units (see Palmer 1968:7).
There is no contradiction in this; though, at first, paradox there may
seem to be. On the Saussurean side, Firth rejects the superordinate struc-
turalism implied by Saussure – the Russian critics have it right, we are told:
Saussure presents the system as if ‘in rebus’ (Firth 1957:181, emphasis in
original). In relation to Hjelmslev, Firth understands the motivation for a
‘semiotic calculus’ to serve semiotic endeavours in the way mathematics
has served the physical sciences; but he rejects the practical necessity of
what must be a step into greater abstraction or Platonism.
Fresh formulations and succinct exempla (as in the systematic interpret-
ation of Sitwell’s expression Emily coloured primulas; see Firth 1968:15–18)
reward the careful, reflective reader with the step by step demonstration of
Firth’s approach, and of the reasons for his rejection of psychological units
in semantics. The economy and clarity of the demonstration affords an
interesting comparison with Shklovsky’s attempt to explain the impact of
Mayakovsky’s lexis in Russian poetry (see Shklovsky 1972:127–33).
Firth’s ‘A Synopsis of Linguistic Theory 1930–1955’ offers a compressed
summation, usefully examined in the context of the various papers of his
junior colleagues (Firth 1962). In the Synopsis, Firth assists us in establish-
ing the central tenets of his writings by quoting his own works with a
careful ‘weighing up of every sentence’ (an assurance cited by his editor; see
Palmer 1968:4). His editor was, in fact, Frank Palmer, a leading proponent
of the ‘London School’. But Palmer (1968:4, emphasis in original), in a burst
of frustration, characterizes the Synopsis as ‘even less coherent and consist-
ent than de Saussure’s Cours de linguistique générale’ (one reason offered by
Palmer is that the sub-divisions do not seem to align with any unified
subject which each might purport to explain).
Evaluations, even by those who worked close to Firth, fall across a
spectrum of such bursts of frustration and bafflement: why not a fully
worked out theory? Why no pages of the much anticipated book? How
could one claim there is meaning made at all levels? How can linguistics
operate without recourse to distinctions like thought in relation to
expression? Those at greater distance, in time and space, raise other issues –
for example, the claim that it is inconsistent of Firth to state the anomalies
of phonological practice using a phonemic alphabet.
As an initial response to such reactions and interpretations of Firth’s
ideas, it is advisable to consider two points: first of all, the problem of
anachronism in seeing the past according to the fashion of speaking in
succeeding decades; and secondly, to ask how a genuinely different way of
seeing problems can achieve a rational evaluation in any science with a
strong prevailing set of assumptions.
Certain ways of seeing scientific and linguistic issues in particular have
made it difficult for many linguists to construe what Firth was delivering
with the brevity of ‘gnomic utterances’. Strevens emphasized, on the other
hand, in his editorial comments (Firth 1964), that Firth wrote his papers for
expert audiences of linguistic specialists, to whom the assumption of
expertise was naturally extended. The problems arise with the beliefs a
reader brings. If reading or hearing Firth in the late 1940s, the prevailing
views were positivist and Bloomfieldean. ‘Language’ (Bloomfield 1933)
espoused a behaviourist psychology and eschewed any ‘science’ in seman-
tics – any science of meaning. Perhaps more problematic was the ideal of
what could be a science at all. The discovery procedures of the era meant
treating levels like phonology and grammar as if they needed to be man-
aged autonomously: quite the inverse of Firth’s ‘meaning is made at all
levels’. Bloomfield’s statement may not do justice even to Bloomfield’s own
linguistic methods (e.g. his work on Menomini, including categories
like ‘verbs of being’ (Bloomfield 1962:274); and ‘verbs of undergoing’
(Bloomfield 1962:298)). Firth’s reactions were patently clear, but bewilder-
ing to the adherents of Bloomfield: the phoneme is dead, and Nida’s work
on morphology is ‘“nonsense” . . . added to “nonsense”’ (Firth 1957:170).
A decade and more later, with Chomsky’s work, syntax is presented as
formal and autonomous; meaning is deferred and passed over to philoso-
phy; the individual is the domain of study; the assumption of a genetically
based universal grammar is used against any evidence of the ‘typical actual’
of language behaviour; a language is the collection of sentences generated
by the formal rules of the language; and intuitions of grammaticality
become the least assailable form of linguistic testimony.
As the adequacy of Chomsky’s assumptions in linguistics was variously
contested, semantics returned but via formalisms from logic (entailment;
presupposition), on the one hand, and from ‘speech act’ theory in philoso-
phy, on the other. The study of context or pragmatics was also re-invented,
but through a kind of ‘Alice in Wonderland’ idiom of maxims of conversa-
tional logic. These supposed reinvigorations of semantics and cultural
analysis were two doses of North Atlantic ‘conventional wisdom’ which
were, from their dominant spokespersons at least (i.e. Searle and Grice,
respectively), without ethnographic evidence or any historical basis or
cross-cultural complexity. Grice himself, in his idiom of gentle whimsy,
16 DAVID G. BUTT
is purported to have claimed that his maxims were just special versions of
what a ‘decent chap ought to do’ (cited by Thomas 2011:219). Had not
Malinowski and Firth already died, their reactions to these developments
might well have ended them. But my point is that by the time Firth and
Malinowski are relinquished to the status of ‘proto-pragmatists’ (Levinson
1983:xii), they have been occluded each decade for quite different reasons –
firstly, due to assumptions that most have relinquished in linguistics; and
secondly, because we now draw our semantic inspiration from speech act
theory, Gricean maxims, the psychological load of relevance, or other
sources extrinsic to the recording of actual linguistic exchanges. If we
reflect on the decades of the dominant American linguistic theories after
WWII, the idea of paradox can be invoked most strongly against each of the
prevailing assumptions of each decade. Consider, for instance: the rejection
by both Bloomfield and Chomsky of the study of meaning; the promotion of
intuitive competence over observable behaviour; the dramatic genetic
speculations about how recent language might have evolved; the idea that
the basis of language might not have been for communication between
people; and the final reduction of UG to recursion (and Merge) only
(Chomsky and McGilvray 2012:16–20, 245).
There were, then, at least on the face of things, both paradoxical state-
ments and what might be called serious revisions. Yet, all these statements
have been influential principles for a period in modern linguistics. One
might be reminded then that dogma and rhetorical forcefulness character-
ize much of what achieves a high degree of visibility in science (see Brooks
2011: Chapter 6). Firth’s approach, with its emphasis on actual languages
and variation, deserves fresh evaluation in the light of Halliday’s develop-
ment of Firth’s polysystemic approach to linguistic description. Much as
with the work of Sapir, important work can be brought back to a more
rational evaluation.
1.2 Restricted Language
The efficacy and consistency of Firth’s linguistics can be seen clearly in his
emphasis on ‘restricted languages’: the variation in relation to social pur-
poses and varying social contexts. By setting the scale of analysis close to
the ‘typical actual’ of social events, Firth’s approach helps resolve (or
‘dissolve’) many of the practical and theoretical conundrums of ‘doing
linguistics’. These include, for example:
(a) whether linguists are accounting for a collective phenomenon or the

ability of the speaking individual;
(b) whether linguistics studies a unified relational system in an idealized
social or psychological space, or a local, specific group of people with
their own semantics and characteristic sociality; and
(c) whether linguistics should attend to the dominant motif of a unity

in which everything holds together (after Meillet’s dictum) or
whether linguists should emphasize the evidence of change and vari-
ation (e.g. Halliday 1974; 1978; Harris 1981; 1990; Le Page and
Tabouret-Keller 1985; Mufwene 2001).
Firth’s approach assists us in reconciling these dichotomies in what Halli-

day likes to describe as ‘eating our cake, and having it too’. By addressing
register variation, the researcher takes up the participants as a recipro-
cating and dynamic system: ‘There can be no reciprocal comprehension if
there is no situation’ (Firth 1962:30). Language is not about individual
minds but the adhesion of significant actions. Being close to the anthropo-
logical stream of actions, there has to be an account of the social milieu,
and of the way, for example, speakers construe the appropriateness of one
channel over another. This is the germ of what Halliday’s context theory
refers to as ‘mode’. In the first place, this is the issue of channel – how the
text is carried and how the text is organized in relation to the demands of
the channel. For instance, the choice of giving or demanding information
may be a semantic generalization with direct relevance to the grammatical
stratum (viz. the selection of mood in Western grammatical tradition). But
this generalization needs to be related to whether one is acting face to face
or in a more mediated mode of meaning. Similarly relevant to social
meaning is whether the giving or demanding is processed and experienced
in ‘real time’ or whether ‘delayed’ (i.e. offering the respondent time to
weigh up a reply). Particular instances of giving or demanding information
can only be accorded a linguistic value when contextualized, when embed-
ded in the experiential matrix of specific social exchange.
1.3 ‘Mutual Expectancies’: Pattern, Process, and Pragma
Firth’s principles appear to be all directed to bringing out the way mean-
ingful practices in a culture ‘hang together’ or cohere through an ensemble
of organized behaviours. From the access we have to the ‘aggregate of
experience’ (Firth 1962:1), we can follow the social relations and their role
in events down to the posture of body and voice in the reciprocation of a
social situation. It is the expectancies that hold the social show together.
These mutual expectancies can be regarded in probabilistic terms: when we
get ‘A’, to what degree does that predict (or prehend) the presence of ‘B’,
and of ‘C’, or of ‘B: C’, etc.? Clearly, this suggests we will be investigating
aspects of text beyond the domains of normative grammar and lexicology,
and beyond phonemic ideas of sound. Such an open sense of co-occurrence
encompasses more too than the ideas that Firth passed down to us
in collocation (the co-presence of word-like units) and colligation (the clus-
tering of grammatical categories). Mutual expectancy opens up for
18 DAVID G. BUTT
consideration any regularity that carries a value in the text; and this
breadth of conception leads Firth to step away from assuming the efficacy
of conventional ‘units’, preferring to use the term ‘piece’. With this neutral
‘place-holder’ for sequence and order, Firth can propose statements that
are not limited by the procrustean assumptions of classical, Western
‘syntagma’. One might ‘net in’ co-occurrences that are significant in char-
acterizing the text, or the persona of an author, or the typical personae of a
group with their restricted language, their trade language, their signs of
solidarity or ‘insider’ talk, their total sociality, their semantic variety.
Yet, it still needs to be asked, how then do we proceed under the guidance
of these principles? To produce ‘statements of meaning’, the language
needs the separation of its various polyphonic strands, of its ensemble of
congruent levels. For Firth, the serial contextualization of a language piece
is like the diffraction of white light through another medium: one sees that
the totality of the white beam is actually a spectrum of waves all contrib-
uting at different scales, but with each essential to the patterning across all
scales. Hence, ‘meaning is made at all levels’. And that cross-level relation
must be handled according to ‘the differences that make a difference’
(Bateson 1982).
The alliterative terms – patterns, pragma, and polysystem – supply three
of the motifs that guide Firth’s approach to structure and function.
Structure is that syntagmatic order of ‘mutual expectancy’ within a given
social event – not the simple sequence of the structuralist’s fixed units.
Function is the profile of relations that pertain to paradigmatic aspects of
that order – essentially, where a ‘piece’ fits in all up and down the relations
on one level and then, where crucial to the issue under investigation, across
those relations from other levels that determine the value of the ‘piece’. For
example, the expression He kept popping in and out all afternoon challenges
segmental descriptions at the ranks in the grammar of verb/verbal group;
but statements of meaning relating to this wording would also need to
include that it only fits into specific social situations, between certain
participants, face to face, and somewhere contextually between personal,
confidential exchange and gossip. It is part of a network of relations that
narrow down the potential of wording to what might be thought of as its
place in a ‘restricted language’. Put quite simply, a polysystemic profile
suggests that the piece could not appear just anywhere. It carries with it a
penumbra of collocational and colligational patterns that reflect the habits
of a social membership and of a personality: an instance of the ‘typical
actual’ for a personality in a social moment.
The linguistic importance and cross-disciplinary resonance of this obser-
vation can be brought out by again reflecting on ‘mutual expectancy’, and
the more unusual, technical word from Whitehead’s ‘process’ interpret-
ation of nature: namely, prehension (Whitehead 1979:379). Considered
from a phonetician’s point of view, mutual expectancy encompasses much
that linguistics up to today has overlooked about patterning in language.
The spectrum of phonetic expectancies allows for ‘a closer approach to the

social texture of language’ (Laver 1980:5). Laver, in quoting Firth and
discussing variation, emphasizes all ‘recurrent, patterned, phonetic activ-
ity’ (Laver 1980:5) that characterizes a speech community, not just the
distinctive contrasts which purportedly define its relation to a natural
language. Laver’s approach, grounded in the phonetic leadership and teach-
ing of David Abercrombie, illustrates the systemicization of voice quality
and what Firth referred to as a total vocal ‘posture’ – a ‘vox’ – which carries
the ‘habitus’, the patterns of living expressed by the posture: ‘Surely it is
part of the meaning of being American to sound like one’ (Firth 1957:192).
This connection to forms of life is explicit in Firth’s numerous footnotes to
Wittgenstein’s later work (Glock 1996:329).
Then there are the mutual expectancies between the complex of variables
that produce the semantic clusters of different registers. Halliday (following
Jean Ure and others) uses register as the key term for a language variety
according to the immediate use. Firth used ‘restricted language’. By this he
emphasized that nothing modifies the meaning of a wording more than a
change of context. Language working to a particular purpose in a conven-
tional social event takes on its valeur from the habitual appearance in that
restricted environment. He even suggested that dictionaries could be organ-
ized around such restricted languages. This would mean that lexicographers
would not need to think of every definition of a word as isolates, but that
the range of a word or piece could be displayed by ‘serial contextualization’
(a term that could be applied to all levels of analysis, thereby doing the
theoretical work of Saussure’s valeur). While register and restricted lan-
guage are ways of describing the variation in meaning-making practices,
in semantics, such variation is realized in choices at different levels of
language. Such choices include probabilities of co-occurrence: e.g. hypotaxis
vs. parataxis; transitive ‘relations by place and order, by particles and by
case’ (Firth 1962:18); cohesion devices, regular or bunched in their disper-
sion; and the various ‘filters’ and ‘lenses’ of the ‘time camera’ of tense – that
is of ‘grammatical time’ based on what linguistic systems actually ‘do’:
Confining ourselves to English as the language of description, let us face the

facts and admit that such words as time, past, present, future and all the rest
of the ‘temporal’ nomenclature, have been employed with gross carelessness
to describe notions supposed to characterize the verb . . . Each language has
its own means of handling ‘experiential’ time, has its own ‘time-camera’ so
to speak, with its own special view-finders, perspectives, filters, and lenses.
It is childish to draw excessively over-simplified linear diagrams to deal with
such linguistic structures and systems. The point is they are not time-
systems but linguistic systems.
(Firth 1962:19)
So the same complex layering of continua in phonetics was to be seen at

other levels mutatis mutandis. Firth asks, for instance, where are the
20 DAVID G. BUTT
morphemic exponents of the grammatical roles of the Latin word pedibus

(feet, ablative: on foot)? Namely, how should the word be divided to display
the meanings it contributes?
A single word isolate such as Latin pedibus might have to be considered in a

sentence structure in which the categories in colligation would include gender,
number and person, and the noun-substantive itself. But where are all these if
the grammarian looks at the word pedibus itself? The exponents of the cat-
egories are ‘cumulative’ in the word and also discontinuous in the sentence. . . .
The mutually expectant relations of the grammatical categories in colligation,
however, cannot be regarded as necessarily having phonological ‘shape’.
(Firth 1962:14)
The extension of meaningful relations through mutual expectancy also

motivates the concept of semantic prosody, a concept that addresses many
difficulties in text analysis and linguistic corpora, particularly in relation to
any account of interpersonal meanings. In this concept we also see the
lexical and grammatical discussed in terms of collocation and colligation.
In computational linguistics, collocation becomes powerful because it can
be easily given mathematical specification. Firth’s insight: you can ‘know a
word by the company it keeps’ is well illustrated by corpus work that can be
directed to diachronic and/or synchronic goals. The grammatical counter-
part – colligation – is more elusive: I take it to net in the proximities of
grammatical categories over and beyond the order that they require by
agreement. This may be a new domain of research addressing the implicit
patterns of choice across stylistics and discourse analysis.
We might claim, as a result of restricted language and register studies, that
we do not speak a grand system which we all hold mentally in a common
collective: rather, we work through a patchwork of differently habituated
social exchanges, all variously local or strange, with variable boundaries on to
other exchanges. We are marked by our memberships and habits in relation
to these exchanges. Hence, our personalities are on display in the ‘roles’ that
we inhabit and by the verbal lines that we know and contribute. Firth
reminds us, then, not only of our vocal posture but also of the need to know
our rhetorical roles. In and out of various group memberships, we each
constitute, metaphorically, our own ‘figure’ of speech. Thus, the collective
and the personal are reconciled through semantic enquiry (Firth 1957:177).
As we move from phonetic perspectives to contexts, semantics, and
grammatical expressions, we have the tool power to consider problems that
appear extrinsic to linguistics, but which, in fact, have a direct bearing on
the forms that language takes around us, and within us. Class and ideology
take on an extended significance, a significance brought out into plain sight
by the statistical treatment of messages in the semantic variation work of
Hasan from the 1980s (see Hasan 2009). The evidence is drawn from
principal component analysis of consistency of choice across up to fifty
semantic parameters. These parameters, in their combinations and
different configurations, represent strategies within interaction that clearly

aligned with social class differences in the Australian urban context. In fact,
the semantic parameters that distinguished class are illustrations of what
Firth called a semantic prosody: a motivated consistency across diverse
resources in the construction of meaning.
Here the use of ‘mutual expectancy’ can be further appreciated as a
radical theoretical motif. It forces us to consider not only ‘the company
that a word keeps’, nor only the overt conventions of grammatical agree-
ment and government (the theoretical compass points for navigating in
synthetic languages like Latin and Greek), but all the ensemble effects that
enact the social positioning of speakers – the fabric of discourse over and
beyond the grammatical ‘rules’. Hasan’s central concern was how speech
realized the social roles of class and gender. Her data were indicative of
Bernstein’s social theory, with social roles being part of a lived experience
coded by implicit expectations of how the world ‘hangs together’, and of
how one plays a role within the conditions offered by class membership.
The semiotic foundations of class, of codes, and of control were in mothers’
talk, particularly in relation to a continuum from individuation to the
assumption of shared ‘local’ values. This is to say, following Firth, that
there was a deep semantic prosody in the ways that mothers projected the
child’s place in the world back to the child. One prosodic consistency was of
a world of individuated access to evidence and the ‘appearance’ of negoti-
ated authority; the other was of a strong acceptance of a settled order in
which knowledge and authority came unquestioned with natural relation-
ships: viz. in deference to age and intrinsically with the status of parents.
This continuum was based on a configuration of each mother’s pattern of
choices – that is, the degree to which their messages encoded the sources and
significance of differing points of view (including the point of view of the
child involved). As the explicit framing of ‘point of view’ diminished in
mothers’ talk, the assumption of a world with a fixed order became more
insistent in the thousands of messages that flowed around and through the
child. It is hard to think of a more Firthian demonstration of the diversity of
our modes of meanings: the ‘collective consciousness’ of Saussure’s social
fact is not a unity in which all the parts hold together (after Meillet’s
dictum). Rather, language becomes a means for engineering and maintain-
ing covert difference around positional authority in ‘the living of life’ (Hasan,
personal communication). By being continuous between human external
and internal experience, we can see in the mothers’ talk a potent ideological
medium. On this last point Firth (1968:199) quotes Whitehead (1938):
The human body is that region of the world which is the primary field of
human experience but it is continuous with the rest of the world. We are in
the world and the world is in us.
The concepts of ‘mutual expectancies’ and Hasan’s semantic variation are

congruent with the co-presence (or absence) of forms suggested by Whorf
22 DAVID G. BUTT
(1956:83) in the notion: ‘configurational rapport’. Essentially, this configur-

ation means that certain elements of language co-pattern because they are
part of a wider, insistent ‘fashion of speaking’ in the culture (Whorf
1956:158). It is by such ways of meaning that a culture manages experience.
Whorf emphasized covert categories – those distinctions that are not overtly
coded in the grammar, but which display a reactance – a consequence – in
discourse texture. One needs experience in the culture to interpret these
categories: simple examples might be what constitutes a female name in
varieties of English, or what is a feminine name in forms of Chinese. Whorf
also drew attention to cryptotypes – those aspects of the language that are
restricted or influential for no obvious motivation. Consider such issues for
English: for example, certain verbs that cannot be used transitively; the lack
of a passivized form of the verb be in English, even though other verbs can
substitute for be with full passive potential (see the ‘represents / is repre-
sented by’ test in identifying clauses in Halliday and Matthiessen 2014:
Chapter 5). These phenomena may be thought of as analogous to the ‘hidden
variables’ theory in physics, as discussed by Bohm (1952a;b): essentially an
‘implicate order’ behind the explicate order of phenomena accounted for by
normative science (in this case, normative grammar).
1.4 So What Would Be a Firthian Approach in Situation,

and for Sound?
1.4.1 Context of Situation

In bringing attention to work developed before 1960 – work that merits
attention and further development – I will draw on the contextual study by
Mitchell (1957). One needs to work with ‘[p]rocesses and patterns of life in the
environment’ that can be ‘generalised in contexts of situation . . . Order and
structures are seen in these, and after examining distribution in collocations,
“pieces”, words and morphemes, [these] may be arranged in ordered series,
resulting in systems and sets of systems’ (Firth 1968:24). Consequently, one is
not seeking a mere enumeration of environmental details (what Hasan
distinguished as the ‘material situational setting’). There have to be some
criteria of relevance, what in the environment leaps out to have a consequen-
tial role in the direction that meaning takes. Mitchell himself takes Firth’s
use of ‘technical’ (as in a ‘technical language’) and turns it into a further,
useful distinction: technical terms in Mitchell refer to those aspects of
patterning which give rise to the particular character of a situation, while
non-technical terms are relevant but generalizable across a number of con-
texts. Firth (1968:200) rounds off the issue of abstraction with:
The abstraction here called context of situation does not deal with mere
‘sense’ or with thoughts. It is not a description of the environment. It is a set
of categories in ordered relations abstracted from the life of man in the flux
of events, from personality in society.
Mitchell uses three ‘attested’ texts of sales in markets and shops in Cyrena-
ica. From these, he exemplifies the most significant contextual distinctions
made by Firth. In particular, the final discussion of negotiations over a fast
horse – a prestige purchase, involving long-standing traditions and commu-
nity expertise – illustrates the practical necessity of beginning at context
rather than treating context as a final form of pragmatic ‘tweaking’. Mitch-
ell also raises a number of problems, issues to which I would also attest
from my own collaborations in surgical care and psychiatric contexts.
These problems include the following: the absence of talk on the activity;
the presence of oblique talk; and the difference between talk on the job and
talk explaining the job; and (I would add) talk that runs a bulletin and guide
which, as a commentary, brings all participants to a shared understanding
of how the process is progressing. Mitchell certainly emphasizes ‘mutual
expectancy’ and even cites prehension (Mitchell 1957:39, 49, 54, 55), noting
the cumulative effects of unfolding connections between relevant words
and actions. In SFL, following Halliday, the process of choices unfolding has
been referred to as logogenesis. The term encompasses the changing values
in the text as new choices accord with what has gone before, and as these
choices direct the changing expectancies as to what is to come in the light
of the most recent choices in the potential. Consider how generic elements
or stages are related to both order and succession (Mitchell 1957:43, 47).
The link between ‘habitual’ patterns, collocation, and expectation is
expressed by Mitchell (1957:55). An interesting distinction is also made
plain between personalities and persons (Mitchell 1957:36–59): the differ-
ence is between the role you are playing and the physical being involved.
Firth emphatically rejected the idea of individuals in language; but he
invoked ‘personality’ (as did Sapir). These distinctions lead to Mitchell’s
four-column tabulation of Text, Translation, Personality, and Stage. At
other points there are signs of some emergence of the more recent style
of Pragmatics: ‘essential conditions’ (Mitchell 1957:36) suggests J. L.
Austin’s influence (although Austin’s 1955 lectures were not in print until
1962); and mutual expectancy appears to encompass overt connections like
those that were later referred to by ‘adjacency pairing’ (Mitchell 1957:59).
Mitchell (1957:34) is working out the ‘complex pattern of activity’ per-
taining to a central pattern (the way ‘a line needs to be distinguished as to
its role in either a rectangle, or as the radius of a circle’). In relation to
the sale of the horse in the market, the exchange takes semantic directions
that the outsider is unlikely to predict or construe. The quoting of a
quatrain of traditional poetry is not unique to the Bedouin stages and styles
of negotiation – many negotiations across cultures invoke gnomic sayings
or apothegms. But the complex relationship between the horse’s speed and
the rider’s good fortune, on the one hand, and the whorls on the horse’s
flanks, on the other, are opaque for anyone not enculturated to Bedouin
horsemanship. These issues are detailed (Mitchell 1957:70–1) and are strik-
ingly parallel to Malinowski’s example of canoe racing, relying as they do
24 DAVID G. BUTT
on traditional formulas, the specific locale (mountain and water for

Malinowski’s example), and the activity (competitive paddling).
Some contact I have had with East Timorese cultures (see accounts of
Tetum from Ainaro district, especially in de Corte-Real (1998); but also see
Hull et al. (1998)) have produced significant parallels. For example, what we
might categorize as texts of greeting or official welcoming have a complex
multivalence: yes, they are performed as a ceremonial greeting to officials
from another community; they typically involve a senior male ‘keeper of
the traditional knowledge’ – a ‘lianai’. But they also urge the power of the
local community in an intimidating manner. In short, they threaten or
warn against any false action in relation to the local community. Footage of
an Indonesian government representative ‘greeted’ in this way (before
Timor-Leste achieved independence from Indonesian authority) has a very
unsettling effect. The strikingly elaborate war canoes of the Trobriand
Island communities trading in their precious shells or ‘kula’, as discussed
by Malinowski and others, appear to have been similarly ambivalent:
a situational meaning of respect of ritual and of ‘Don’t mess with us’.
In relation to East Timor, de Corte-Real and I attended a meeting with an
Indonesian military commander in Dili who directed de Corte-Real most
civilly to stick to the literary aspects of the performance genre, not to the
political and nationalist implications. The multifunctional meaning of
action and wording was clear to the officer.
What Mitchell achieves is a ‘context first’ outline of linguistically motiv-
ated categories. Mitchell does not here systemicize the sound or other
(intervening) levels of analysis, although various issues of grammar and
numerous observations on rendering the speech phonetically are noted
throughout. But, unlike the strong accounts we have of prosodic analysis
in published anthologies,1 illustrations of context categories have not been
pursued to a similar degree. This is an indication of some bafflement
amongst those working with the neo-Firthian traditions. Interest in the
systemicization of context has produced significant work, especially that
led by Hasan’s consistent attention to the problems of turning ‘context of
situation’ into a concept with greater tool power. Now there are a number
of nodes of contextual modelling, each distinctive: in the UK (e.g. Cardiff );
in Hong Kong; and in Germany; as well as in Australia.
Firth is clear on the point of systemic representation, including in rela-
tion to context. Language needs to be dispersed, like white light, across its
spectrum of different ‘wave lengths’, and addressed as a polysystemic
phenomenon. Such an approach appears to yield the most practical descrip-
tions since there is something universal about language: its organization as
a system of systems:
1
See articles by Allen (1953); Halliday (1959; 1963); Henderson (1987); Mitchell (1975); Palmer (1970); Robins
(1957).
I have suggested that language is systemic. . . . We may assume that any

social person speaking in his own personality will behave systematically,
since experienced language is universally systemic. Therefore, we may study
his speech and ask the question, ‘What is systemic?’
(Firth 1957:187)
1.4.2 Sound
While Firth and Halliday are both cited for systemicization in phonology
and the semantics of intonation, the consistency of their work (along with
the work of Abercrombie at Edinburgh) is notable when viewed against the
current trends in phonological and phonetic debate. In an attempt to set
the record right, Palmer emphasizes that Firth’s distinctive approach can be
tracked in publications from 1934 to 1937 (see Palmer 1970:x–xi). The
polysystemic principle was argued in Firth’s (1935:51) discussion of Marathi,
where he points out that eight /n/ sounds needed to be recognized as
linguistically and functionally distinct, and therefore not the same unit even
if seemingly identical phonetically. This was part of the broader contextual
principle: namely, that one had to recognize not only the phonetic contexts,
but also ‘lexical and grammatical functions’ (Firth cited in Palmer 1970:xi).
The value of an item was dependent on the systemic character of ‘recurrent
contexts’. So too, there were the issues of y and w prosodies; the urging of
the importance of the syllable and of extended phenomena in phonological
analysis; and the notion of ‘articulation types’. Palmer recommends the
completeness of Henderson’s (1949) analysis of Siamese in that she shows
a full hierarchy of prosodies: in sentence; sentence parts; polysyllables and
sentence pieces; in syllables and syllable parts; and in consonant and vowel
units. Firth’s arguments against rigid phonemic methods concerned the
unreality of segmentation and of discovery procedures. These arguments
were bewildering to linguists of his day, despite the vigorous critique of
the phoneme concept by Twaddell in the USA (see Anderson 1985) and the
questions raised by other linguists in the USA (Palmer (1970:x) cites Harris
and Hockett in this regard). Today, views homogeneous with those of Firth
are part of the ‘natural’ background in the study of phonetics. Since Firth
emphasized that the /d/ that was word initial was not the same as /d/ word
final since the two acoustic elements operated in distinct systems (i.e. with
differing ‘valeurs’), then we might regard this as a rigorous application of
relational thinking (suggesting some affinity with the extreme relationalism
of Hjelmslev, a linguistic alter ego with whom Firth enjoyed frequent
exchanges and debates). We can see that Firth’s declarations against Saus-
sure’s work – namely, that it creates a system ‘in rebus’ – need to be seen
alongside what may be the strictest application of Saussure’s relational
universe (see the comment on valeur by Palmer 1968:xx).
The problems urged by Firth in relation to phonetics and phonology are
usefully explained and criticized in an evaluation of Linguistic Thought in
26 DAVID G. BUTT
England 1914–1945 (see Love 1988:151–62). After setting out Firth’s

approach, Love argues for the efficacy of (and necessity of ) our use of
abstractions and ‘scriptic conventions’, contrary to Firth’s attitudes. With-
out such conventions, Love claims, Firth is unable to separate the ‘preor-
dained structure’ from the ‘unique-ness’ of an utterance (Love 1988:163–4).
More significant is the claim that Firth’s principles would open ‘the flood-
gates’ to comparisons right across a language (Love 1988:162) without any
counter principle to limit analysis. I disagree with Love’s ultimate judge-
ments: first of all, Firth’s focus on ‘restricted languages’ suggests that the
‘floodgates’ to the whole language are an artificial or hypothetical, rather
than a practical, concern.
Second, one needs to keep in mind that the machinery of analysis is ad
hoc for Firth; yet, all discussion has to start somewhere. Consequently, yes,
you start your discourse where you meet the discourse of others – hence,
grammar in Europe will set out from terms like transitivity, mood, and
tense, although the end of discussion may be in changing the valeur of
these concepts, or with the development of other categories altogether.
In phonology, this may mean starting out from the ‘scriptic conventions
used for stating and citing utterances’ (Love 1988:164). I would suggest that
the problem for Firth is then how to improve upon the specific account, not
how to ensure an absolute final account, or to supply exhaustive paradigms,
as Love appears to assume. Like parenting (and probably all sciences), it is a
case of achieving a ‘good enough’ or improved result. Love’s final remarks
do not give sufficient weight to the pragmatist implications of Firthian
‘ad hoc-ery’ and to the role given to the operational status of linguistic
techniques. The role of the term ineffability also seems to me undervalued
for its significance (see by contrast the emphasis in Halliday 2002 [1985]). It
is surprising to see Firth being criticized by Love for underestimating the
difficulties involved in ‘turning language back on itself’ when so much of
Firth’s writing is directed to bringing out the implications of just those
difficulties.
Firth’s insights attempt a ‘renewal of connection’ with the facts and
their contiguities in an actual language – much as Whitehead originally
conceived of intellectual process as having to take flight up to the thin
air of abstraction and then to return to a grounding in the realia of the
experiential matrix that is our primary resource for demarcating sense
from non-sense.
1.5 Posture and the Communicating Body
Firth consistently refers to the body going out to the world – of the ears
being active in exploring sounds; of the brain as a guide to acting and
moving in ‘situations’; of the human as engaged in the pursuit of a ‘joy’
at structure. Firth sets out in linguistics with the explicit assumption of this
active engagement with the world, quite the opposite of the passive, tabula
rasa notion often foisted upon empiricists in human sciences. Firth’s citing
of sources, however, needs to be regarded for its own cultural context and
the ways in which Firth’s approach is oriented to current thinking.
Firth cites Charles Sherrington’s work for its emphasis on movements
and proprioception as ways into understanding emotions – the ‘felt ME’ –
and as the way of building relations with other humans. Firth responds
positively to Sherrington’s view that the brain is a ‘manager of muscle’, a
view in accord with the fact that current analysis suggests that the cerebel-
lum, although only 10 per cent of the brain’s volume, encompasses more
than fifty per cent of its neurons.
More centrally, Firth argues for the continuities between human inner
and outer worlds, with the brain creating a form of ‘mutual grip’ between
the world and us. This side of Firth’s thinking is congruent with the
position of neuropsychologists like Trevarthen (1998; see also Panksepp
and Trevarthen 2009). Trevarthen, having worked with Sperry, Bruner,
and Halliday, has argued for a related going out to the world in neonates,
a bridge-building through intersubjectivity to person-ness. Trevarthen
(in Stensæth and Trondalen 2012) summarizes his own thinking thus:
In 1974, emphasising the rhythmic properties of expressions, I said, ‘when a

newborn is alert and coordinated, its still very rudimentary movements
have, nevertheless, the pace as well as the form of activities such as looking,
listening, and reaching to touch, from the start. This can be perceived and re-
acted to unconsciously by an older person. As the person approaches the
infant, acting gently and carefully as people tend to do instinctively to a
baby, then all the emanations from this approach have rhythmical proper-
ties that are comparable with those inside the movement-generating mech-
anisms of the infant’s brain. From this correspondence I believe the infant
builds a bridge to persons.’
Firth’s concepts of ‘posture’ and ‘person’ suggest a general account of a

communicating body; and certainly this is an area in which the emphatic
statements of Firth go under-interpreted. Contemporary discourses on
‘embodiment’ are a reprise on ideas emerging from 1906 to the 1950s.
From 1960 to 1990, a narrower metaphor for brains became dominant in
the developments of artificial intelligence: the image of a disembodied
computer engaged in the perhaps 10 to the power of 120 moves of a game
of chess. The contemporary emphasis on the body is hardly an innovation,
but a return to an earlier engagement with ideas recently occluded by that
‘starkly narrow’ agenda in American linguistics during the period after the
1950s (Levinson 1983:xii). Narrow claims concerning the task of linguistics –
claims concerning universal grammar, mental organs, poverty of the stimu-
lus, autonomous syntax, genetically based competence, etc. – exerted a
hegemonic control over what could be declared ‘science’ in linguistics (see
e.g. Ellis 1993).
28 DAVID G. BUTT
1.6 Prehension in Science and the Arts: The Zeitgeist
Prehension is a term central to Whitehead’s theory of natural, cosmic

cohesion: the interconnectedness between all ‘occasions of experience’.
Aspects of this approach attracted numerous important scientists (e.g.
C. H. Waddington in biology; Joseph Needham in bio-chemistry, also a
specialist on science and civilization in China; Ludwig von Bertalanffy in
evolutionary systems thinking; Charles Hartshorne in philosophy). For
most scientists such a notion could have been judged as not more than a
promissory note on the emergence of influence across vast spaces, between
different scales of events, and in a continuum between outer and inner
experiences. For the claims across so many phenomena, Whitehead’s work
has even been scoffed at for obscurantism and ‘metaphysics’ (a term of abuse
between many who claim to be in natural sciences). Curiously, we all now live
with experimental confirmation of electron ‘entanglement’, that is, between
sub-atomic particles at ‘out of contact’ distances (Aspect et al. 1982; DeWitt
2004:288–305; Whitaker 2006). Consequently, the idea of ‘negative prehen-
sion’ – Whitehead’s idea that each region of the universe bears the influence
of the configuration of the rest of the universe – should be revisited as
prescient, not absurd. It is interesting to note that a leading theorist of
prosodic theory (Allen 1970 [1951]) actually used ‘action at a distance’ as an
argument against cases of assimilation and dissimilation in phonemic
theory. This illustrates, at least, that linguists did observe the epistemological
aspects of physical theory in reflecting on linguistics as a science.
The metaphor of prehension may seem a curious choice for Firth – a
philosopher’s term cited by a linguist who is vehemently against ‘arm
chair’ philosophizing in linguistics. Yet here too, there is significant con-
sistency. Whitehead’s project was nothing less than reversing the mislead-
ing consequences of atomisms and ‘fallacies of misplaced concreteness’. He
was Chancellor at London University in the 1930s. Did Whitehead and Firth
converse on this matter? – I do not know. But Firth took the title of
Whitehead’s 1938 book, ‘Modes of Thought’, into his own writing: as
‘Modes of Meaning’. Firth’s oft-used expression for applying the contextual
framework was to extract ‘pattern’ from ‘the general mush of goings
on’; again, both key idioms from Whitehead’s writings between the
World Wars.
Within the linguistics of text, the idea of prehension is well motivated
and practical. Firth’s own search for differences which make a difference
semantically only needed computers to track ‘latent patterns’: an approach
taken up indefatigably by John Sinclair, a supervisee and colleague of Halli-
day. Sinclair’s work (a real legacy in its own methods and style) was much
guided by Firth’s concept for seeking new patterns in lexis and new pieces
in grammar, and by ‘semantic prosody’ (Sinclair 1966; see also reference to
‘latent patterning’ in Sinclair and Coulthard 1975). Every choice in a text
sets up a new probability about what is impending, and at the same time
modifies the value of what has preceded, namely, the choices that have
brought the logogenetic unfolding to this point. The expansion of
computer-based corpus studies has not only activated Firth’s concepts of
collocation and colligation, but has also given a practical, probabilistic
character to register studies and for ‘restricted languages’ (Halliday and
James 1993; Teich 2003; Bartsch et al. 2005; Matthiessen 2015b).
1.7 Conclusion
Unlike most other influential theories of language and linguistics with

active proponents since WWII, Systemic Functional Linguistics (SFL) is
continuous with linguistic history: its proponents do not claim to have
made a dramatic or ‘revolutionary’ break, but rather, a rational integration
of linguistic experiences including an emphasis on languages beyond
European studies. Furthermore, unlike the formal theories of our era, and
quite contrary to the way it is referred to from the ‘outside’, SFL involves no
neologisms or terms without deep roots in linguistic or pedagogical or
rhetorical traditions. Halliday’s studies of Chinese have always provided a
lens through which meanings in European and world languages are
brought into a depth of field not characteristic of theories that set out from
the Western classical tradition. Consider, for example, the assumption
around Subject deletion or ‘Pro-drop’ in generative linguistics.
Those working in SFL need to be clear about the potential, and the limits,
of their tradition without the distraction of boundary disputes or apoc-
ryphal sketches of linguistic history. For limits there are. It might be
argued, for instance, that despite the efforts of Halliday and Hasan (1985)
with their work on context networks, SFL practitioners as a cohort have not
delivered on the explication of the context to meaning relation. These long-
cited notions may not have been brought up to what Matthiessen
(2015a:178) refers to as ‘an industrial-strength representational system’ –
the concepts have not been doing sufficient work in the discourse descrip-
tions of SFL. They have not added sufficient value where it might have made
an important contribution, for example, to narrowing down the referential
domain of computational linguistics.
Nevertheless, the notion of ‘situation’, along with the idea of language
variation according to use (i.e. register), have already had a significant
influence in enlivening linguistics and language education in England,
Australia, and elsewhere. Then there is the work on semantic variation
(Hasan and Cloran 1990), on Rhetorical Structure Theory (see especially
Mann and Matthiessen 1983), on computational modelling developed in a
number of centres (in Germany, at Macquarie in Australia, and by two
generations of the Cardiff group of SFL researchers), as well as the general
linguistic approach to the mind set out and illustrated in Construing Experi-
ence through Meaning (Halliday and Matthiessen 1999). All are examples of
30 DAVID G. BUTT
the robustness of the SFL model. Furthermore, the ideas around context
have insulated SFL from the autonomous formalisms that Duranti states
had taken over from the anthropological linguistics in America (see Duranti
2009:3).
Firth was first of all a naturalist of languages – the languages of Eastern
and African cultures. Firth was focused on instances of language rather
than on the essence of what language ‘is’. Like Henry Sweet before him,
Firth’s efforts were directed to shifting the axis of linguistic discussion
away from the focus on a Eurocentred, reconstructed classical tradition
towards the ‘typical actual’ of current speech. His interests were data
driven and problem oriented: from apprehending the polyphony of human
articulation to improving the system of ‘shorthand’ for rapid recording of
speech. In his writings he tended to reprise his speeches before learned
societies, a spoken context in which much could be taken for granted.
On the other hand, he wrote two clear and prescient books on language
matters for a wider public that had, between the Wars, a demonstrated
keenness to absorb knowledge from experts. Between linguists, it might be
said that he insisted on consistency of theory, part of which meant being
vitriolic at any reifying of theoretical abstractions. Firth was politically
conservative, yet found common ground with the Russian critique of Gene-
van structuralism as well as with Halliday’s Marxist perspective on linguis-
tics. Most remarkable, perhaps, is that he demanded of his students and
colleagues that they defamiliarize their worlds through the semantics of a
non-European language, adopting the lens of linguistic evidence and only
looking back at English with a renewed, extended grammatical
imagination.
References
Allen, W. S. 1953. Relationship in Comparative Linguistics. Transactions of

the Philological Society 52(1): 52–108.
Allen, W. S. 1970 [1951]. Some Prosodic Aspects of Retroflexion and Aspir-
ation in Sanskrit. In F. R. Palmer, ed., Prosodic Analysis. London: Oxford
University Press. 82–90.
Anderson, S. R. 1985. Phonology in the Twentieth Century: Theories of Rules and
Theories of Representations. Chicago: University of Chicago Press.
Aspect, A., J. Dalibard, and G. Roger. 1982. Experimental Test of Bell’s
Inequalities Using Time Varying Analyzers. Physical Review Letters 49(25):
1804–7.
Bartsch, S., R. Eckart, M. Holtz, and E. Teich. 2005. Corpus-based
Register Profiling of Texts from Mechanical Engineering. Proceedings
of the Corpus Linguistics Conference CL 2005 1(1). Available online at: www
.birmingham.ac.uk/research/activity/corpus/publications/conference-arch
ives/2005-birmingham.aspx. (Last accessed 25/07/2017.)
Bateson, G. 1982. Difference, Double Description and the Interactive Desig-

nation of Self. In F. A. Hanson, ed., Studies in Symbolism and Cultural
Communication. Kansas: University of Kansas Publications in Anthropol-
ogy. 3–8.
Bloomfield, L. 1933. Language. New York: Holt, Reinhart and Winston.
Bloomfield, L. 1962. The Menomini Language. New Haven: Yale University
Press.
Bohm, D. 1952a A Suggested Interpretation of the Quantum Theory in
Terms of ‘Hidden’ Variables, I. Physical Review Letters 82(2): 166–79.
Bohm, D. 1952b. A Suggested Interpretation of the Quantum Theory in
Terms of ‘Hidden’ Variables, II. Physical Review Letters 82(2): 180–93.
Brooks, M. 2011. The Secret Anarchy of Science. London: Profile Books.
Chomsky, N. 1972. Language and Mind. New York: Harcourt Brace
Jovanovich.
Chomsky, N. and J. McGilvray. 2012. The Science of Language: Interviews with
James McGilvray. Cambridge: Cambridge University Press.
Cohen, J. and I. Stewart. 1995. The Collapse of Chaos: Discovering Simplicity in a
Complex World. London: Penguin.
de Corte-Real, B. 1998. Mambae and Its Verbal Art Genres: A Cultural Reflection of
Suru-Ainaro, East Timor. PhD Thesis, Macquarie University.
DeWitt, R. 2004. Worldviews: An Introduction to the History and Philosophy of
Science. Oxford: Blackwell.
Duranti, A. 2009. Linguistic Anthropology: A Reader. Chichester: Wiley-
Blackwell.
Ellis, J. 1993. Language, Thought and Logic. Evanston: Northwestern
University Press.
Firth, J. R. 1935. The Technique of Semantics. Transactions of the Philological
Society 34(1): 36–72.
Firth, J. R. 1957. Papers in Linguistics 1934–1951. London: Oxford University
Press.
Firth, J. R. 1962. A Synopsis of Linguistic Theory 1930–55. In J. R. Firth et al.,
eds., Studies in Linguistic Analysis: Special Volume of the Philological Society.
Oxford: Blackwell. 1–32.
Firth, J. R. 1964. The Tongues of Men and Speech. Edited by P. D. Strevens.
London: Oxford University Press.
Firth, J. R. 1968. Selected Papers of J. R. Firth 1952–59. Edited by F. R. Palmer.
London: Longmans.
Glock, H. J. 1996. A Wittgenstein Dictionary. Oxford: Blackwell.
Grice, H. P. 1989. Studies in the Way of Words. Cambridge, MA: Harvard
University Press.
Halliday, M. A. K. 1959. The Language of the Chinese ‘Secret History of the
Mongols’. Oxford: Blackwell.
Halliday, M. A. K. 1963. Intonation in English Grammar. Transactions of the
Philological Society. 62(1): 143–69.
Halliday, M. A. K. 1974. Language and Social Man. London: Longman.
32 DAVID G. BUTT
Halliday, M. A. K. 1978. Language as Social Semiotic: The Social Interpretation of

Language and Meaning. London: Edward Arnold.
Halliday, M. A. K. 1984. Language as Code and Language as Behaviour:
A Systemic-Functional Interpretation of the Nature and Ontogenesis of
Dialogue. In R. Fawcett, M. A. K. Halliday, S. M. Lamb, and A. Makkai,
eds., The Semiotics of Language and Culture: Language as Social Semiotic, Vol. 1.
London: Pinter. 3–35.
Halliday, M. A. K. 2002 [1985]. On the Ineffability of Grammatical Categor-
ies. In J. J. Webster, ed., The Collected Works of M.A.K. Halliday Volume 1: On
Grammar. London: Continuum. 291–322.
Halliday, M. A. K. and Z. L. James. 1993. A Quantitative Study of Polarity and
Primary Tense in the English Finite Clause. In J. Sinclair, M. Hoey, and G.
Fox, eds., Techniques of Description: Spoken and Written Discourse. London:
Routledge. 32–66.
Halliday, M. A. K. and R. Hasan. 1985. Language, Context and Text: A Social
Semiotic Perspective. Geelong: Deakin University Press.
Halliday, M. A. K. and C. M. I. M. Matthiessen. 1999. Construing Experience
through Meaning: A Language-based Approach to Cognition. London: Cassell.
Halliday, M. A. K. and C. M. I. M. Matthiessen. 2014. Halliday’s Introduction to
Functional Grammar. London: Routledge.
Harris, R. 1981. The Language Myth. New York: St. Martin’s Press.
Harris, R. 1990. On Redefining Linguistics. In H. G. Davis and T. J. Taylor,
eds., Redefining Linguistics. New York: Routledge. 18–52.
Hasan, R. 2009. The Collected Works of Ruqaiya Hasan, Volume 2: Semantic
Hasan, R. and C. Cloran. 1990. A Sociolinguistic Interpretation of Everyday
Talk between Mothers and Children. In M. A. K. Halliday et al., eds., Learning,
Keeping and Using Language: Selected Papers from the 8th World Congress of
Applied Linguistics, Sydney, 16–21 August 1987. Amsterdam: John Benjamins.
Henderson, E. J. A. 1949. Prosodies in Siamese: A Study in Synthesis. Asia
Major New Series 1: 189–215.
Henderson, E. J. A. 1987. J. R. Firth in Retrospect: A View from the Eighties.
In R. Steele and T. Threadgold, eds., Language Topics: Essays in Honour of
Michael Halliday. Amsterdam: John Benjamins.
Hull, G. 1998. The Basic Lexical Affinities of Timor’s Austronesian Lan-
guages: A Preliminary Investigation. Studies in Languages and Cultures of
East Timor 1: 97–202.
Jakobson, R. 1966. Henry Sweet’s Path toward Phonetics. In C. E. Bazell
et al., eds., In Memory of J. R. Firth. London: Longmans. 242–54.
Kachru, B. B. 2015. Socially Realistic Linguistics: The Firthian Tradition. In
J. J. Webster, ed., The Bloomsbury Companion to M. A. K. Halliday. London:
Bloomsbury. 72–93.
Langendoen, D. T. 1968. The London School of Linguistics. Cambridge, MA: MIT
Press.
Laver, J. 1980. The Phonetic Description of Voice Quality. Cambridge: Cambridge

University Press.
Le Page, R. B. and A. Tabouret-Keller. 1985. Acts of Identity: Creole-based
Approaches to Language and Ethnicity. Cambridge: Cambridge University
Press.
Levinson, S. 1983. Pragmatics. London: Cambridge University Press.
Love, N. 1988. The Linguistic Thought of J. R. Firth. In R. Harris, ed.,
Linguistic Thought in England 1914–1945. London: Duckworth. 148–64.
Lyons, J. 1968. Theoretical Linguistics. Cambridge: Cambridge University
Press.
Mann, W. C. and C. M. I. M. Matthiessen. 1983. Nigel: A Systemic Grammar for
Text Generation. Marina del Rey: Information Sciences Institute, University
of Southern California.
Matthiessen, C. M. I. M. 2015a. Halliday on Language. In J. J. Webster, ed.,
The Bloomsbury Companion to M. A. K. Halliday. London: Bloomsbury.
137–202.
Matthiessen, C. M. I. M. 2015b. Register in the Round: Registerial Cartog-
raphy. Functional Linguistics 2(9): 1–48.
Mitchell, T. F. 1957. The Language of Buying and Selling in Cyrenaica:
A Situational Statement. Hesperis 44: 31–71.
Mitchell, T. F. 1975. Principles of Neo-Firthian Linguistics. London: Longman.
Mufwene, S. S. 2001. The Ecology of Language Evolution. Cambridge:
Cambridge University Press.
Palmer, F. R. 1968. Introduction. In F. R. Palmer, ed., Selected Papers of J. R.
Firth 1952–59. London: Longmans. 1–11.
Palmer, F. R. 1970. Prosodic Analysis. London: Oxford University Press.
Panksepp, J. and C. Trevarthen. 2009. The Neuroscience of Emotion in
Music. In S. Malloch and C. Trevarthen, eds., Communicative Musicality:
Exploring the Basis of Human Companionship. Oxford: Oxford University
Press. 105–46.
Robins, R. H. 1957. Malinowski, Firth, and the ‘Context of Situation’. In
E. Ardener, ed., Social Anthropology and Language. London: Tavistock.
33–46.
Searle, J. R. 1969. Speech Acts. Cambridge: Cambridge University Press.
Shklovsky, V. 1972. Mayakovsky and his Circle. New York: Dodd, Mead and Co.
Sinclair, J. 1966. Beginning the Study of Lexis. In C. Bazell, J. Catford,
M. A. K. Halliday, and R. Robins, eds., In Memory of J. R. Firth. London:
Longman. 148–62.
Sinclair, J. and R. M. Coulthard. 1975. Towards an Analysis of Discourse:
The English Used by Teachers and Pupils. Oxford: Oxford University Press.
Stensæth, K. and G. Trondalen. 2012. Dialogue on Intersubjectivity: An
Interview with Stein Braten and Colwyn Trevarthen. Voices: A World
Forum for Music Therapy 12(3). Available online at: https://voices.no/index
.php/voices/article/view/682/568#31. (Last accessed 29/05/2017.)
34 DAVID G. BUTT
Teich, E. 2003. Cross-linguistic Variation in System and Text. Berlin: Mouton de

Gruyter.
Thomas, M. A. 2011. Fifty Key Thinkers on Language and Linguistics. New York:
Routledge.
Trevarthen, C. 1998. The Concept and Foundations of Infant Intersubjectiv-
ity. In S. Bråten, ed., Intersubjective Communication and Emotion in Early
Ontogeny. Cambridge: Cambridge University Press. 15–46.
Whitehead, A. N. 1938. Modes of Thought. New York: The MacMillan
Company.
Whitehead, A. N. 1979. Process and Reality: An Essay in Cosmology. New York:
Macmillan.
Whitaker, A. 2006. Einstein, Bohr, and the Quantum Dilemma: From Quantum
Theory to Quantum Information. Cambridge: Cambridge University Press.
Whorf, B. L. 1956. Science and Linguistics. In J. B. Carroll, ed., Language,
Thought and Reality. Cambridge, MA: MIT Press. 207–19.
2
Key Terms in the
SFL Model
Jonathan J. Webster
2.1 Systemic Functional Linguistics
A Systemic Functional approach looks at how language functions to make

meaning in context of situation.1 Meaning is represented paradigmatically
in terms of systems of choice related to what is being talked about
(i.e. ideational); how those interacting are relating to one another through
what they say (i.e. interpersonal); and how ideational and interpersonal
meanings are turned into discourse (i.e. textual). Halliday and Matthiessen
(2014:33, emphasis in original) maintain that ‘functionality is intrinsic to
language: that is to say, the entire architecture of language is arranged
along functional lines. Language is as it is because of the functions in
which it has evolved in the human species.’ A functional approach to the
study of language offers insight into how language is learnt and how
language eventually evolves into the adult language system with these
three major components of meaning or metafunctions: ideational, interper-
sonal, and textual.2
2.2 Meaning in SFL Theory
Language is the instantiation of an indefinitely large meaning potential

through acts of meaning which simultaneously construe experience and
enact social relationships. Acts of meaning are also acts of identity, occur-
ring in contexts of situation. By means of my ‘act of meaning’, I participate
1
The selections for further reading included in the footnotes of this chapter are adapted from Bloomsbury’s
The Essential Halliday (Halliday 2009) edited by J. J. Webster, which includes selected extracts and additional readings
from the ten volumes of Halliday’s Collected Works for twenty key concepts in Systemic Functional Linguistics.
2
For further reading on ‘functions and use of language’, see Halliday 2003a:298–322; Halliday 2003b:47, 51–6,
68–74, 81–7, 270–80; Halliday 2007a:41–2, 50–3, 56–7; Halliday 2007b:88–92, 120–2.
36 J O N AT H A N J . W E B S T E R
in an act of interpersonal exchange, communicating my sense of my own

identity, my worldview, my interpretation of experience. An act of is an
instance of meaning formed out of an infinite meaning potential for reflect-
ing on the world and interacting with others in it.3
Language is a semiotic system.4 It is the means by which meanings are
created and exchanged. ‘Since language is the leading edge of meaning’,
writes Halliday (2013:211), ‘the leading edge of semiotics is linguistics.’
Halliday (2013:194) describes a language as being more than a semiotic
system – it is also ‘a system that makes meanings’: a semogenic system.
As he explains,
[t]he usual way we talk about language is by saying that language ‘expresses’
meaning, as if the meanings were already there – already existing, in some
formation or other, and waiting for language to transpose them into sound,
or into some kind of visual symbols. But meaning is brought about by
language; and the energy by which this is achieved, the source of its semo-
genic power, is grammar.
(Halliday 2013:194–5)
In his book Reinventing the Sacred, Kauffman (2008:193–4) writes that ‘while
the human mind, central to our human embodied agency, is sometimes
algorithmic and sometimes computes, it does some things we do not yet
understand; it makes meanings’.
Halliday (personal communication) describes language as ‘a basic human
resource with potentially immense power, which is hidden, partly because
people are genuinely not aware of how much they are, in fact, depending
on it’. We depend on language to construe the world around us and
describe our feelings within, and exchange this meaning with others.
Meaning, in SFL theory, is not limited to referential meaning, i.e. word
meanings. We use language not only to construe experience but also to
enact social relationships, and create the discourse.
A text is an instance of meaning, a construct of meaning that is formed
out of a continuous process of choice among the innumerable interrelated
sets of semantic options organized into three main functional components
or metafunctions: ideational, interpersonal, and textual.5 Operating in par-
allel, the three metafunctions comprise the total meaning potential of a
language.
3
For further reading on ‘act(s) of meaning’, see Halliday 2002a:201, 206, 354–6; Halliday 2002b:50, 52; Halliday
2003a:171, 174, 355–74, 375–89; Halliday 2003b:11–12, 14–15, 18–20, 113–43, 212–18, 239, 245–6,
249–50, 327–52; Halliday 2005a:198–202; Halliday 2013:253, 264.
4
For further reading on ‘semiotics’, see Halliday 2002a:196–218, 384–418; Halliday 2002b:23–84, 150–2; Halliday
2003a:2–7, 93, 113–15, 116–24, 131–7, 147–51, 171, 192–8, 199–212, 213–31, 275–7, 355–74, 375–89,
390–432; Halliday 2003b:6–27, 90–112, 140–3, 157–95, 212–26, 250–66, 281–307, 327–52; Halliday
2004:43–4, 53–5, 102–34, 216–25, 198–202; Halliday 2007a:81–96; Halliday 2007b:179–86, 193–6, 259–63.
5
For further reading on ‘metafunction’, see Halliday 2002a:21–36, 390–2; Halliday 2003a:15–18, 248–50, 277–8;
Halliday 2003b:209–25, 332–3, 335–6, 338–41, 343–4, 346, 348–9; Halliday 2005a:200–2, 215–22; Halliday
2007b:183–4.
Key Terms in the SFL Model 37
These sets of semantic options constitute a semantic system or system of

meaning potential which is distinguishable from other semiotic systems by
the fact that it is based on grammar. The semantic system is one of three
overlapping levels, or strata, which together comprise the whole linguistic
system. Between the semantic system above and the phonological and
morphological realization below is the lexicogrammar.6
What enables the meaning-making potential in language is the lexico-
grammar. While the three metafunctions are arguably universal to every
language – every language user needs to be able to use language to construe
experience, enact social relationships, and create discourse – the means by
which meaningful choices are lexicogrammtically realized is language-
specific. The complex realization of options from the three metafunctions
becomes evident in the clause as a lexicogrammatical unit.
2.3 Grammar in SFL Theory
SFL is first and foremost a theory about how language works at the level of
grammar.7 ‘[T]hinking about meaning means thinking grammatically’,
writes Halliday (2013:207). A grammar is a theory of experience of everyday
life. It is that abstract stratum of coding between meaning and expression; it
is a resource for making meaning. Just as linguistics is language about
language – or ‘language turned back on itself’ (Firth 1957:181) – grammatics
is a theory of grammar; it is a theory for explaining how the grammar
constructs a theory of experience. Grammatics is theorizing about a theory;
it is a theory of a second order, a part of a more general theory of meaning.8
In categorizing the grammar, the grammarian comes at the task from
three perspectives, each of which corresponds to a different stratum. First,
there is the higher stratum of semantics. Here, the grammarian’s perspec-
tive is from above. Second is the stratum of lexicogrammar, where the
perspective is from around. The third perspective is from below and looks
at the morphological and phonological realization of meaning. In a func-
tional grammar, priority is given to the perspective from above, as form
follows function, and the meaning of an expression will decide its phono-
logical and morphological realization.9
6
For further reading on ‘semantic systems’, see Halliday 2002a:196–218, 310–11; Halliday 2002b:23–8, 45–52;
Halliday 2003a:323–54; Halliday 2003b:90–8, 109–12, 115–25, 281–94; Halliday 2007a:345–6; Halliday
2007b:131–3, 143–4, 153, 158, 164–6, 183–4, 186–95, 256–7.
7
For further reading on ‘theory and description’, see Halliday 2002a:37–42, 58–61, 72–7, 86, 98–9, 106–17, 158–72,
396, 403–6, 414–15; Halliday 2003a:7–15, 37–47, 199–212, 327–30; Halliday 2004:53–8; Halliday
2005a:227–38; Halliday 2005b:156–63; Halliday 2006:294–322; Halliday 2007a:136–9, 149.
8
For further reading on ‘grammatics’, see Halliday 2002a:296–8, 365–6, 369–73, 384–6, 416–17; Halliday
2003a:264–5, 274–6, 286, 362, 373, 385; Halliday 2005a:213–38.
9
For further reading on ‘trinocular vision’, see Halliday 2002a:398, 402, 408–9; Halliday 2003a:202–5, 254–5, 266;
Halliday 2005a:231.
To the ‘architecture’ of grammar, Halliday and Matthiessen (2014:55)

attribute
the dimensions that define the overall semiotic space of lexicogrammar, the
relationships that inhere in these dimensions – and its relationship to other
sub-systems of language – to semantics and to phonology (or graphology).
Thus, according to systemic functional theory, lexicogrammar is diversified
into a metafunctional spectrum, extended in delicacy from grammar to
lexis, and ordered into a series of ranked units.
The fundamental categories for the theory of grammar are unit, structure,
class, and system. Rank is the scale on which the units are ranged.
Rankshift occurs when a given unit is transferred to a lower rank.
A structure is made up of ordered elements. Sequence is one formal expo-
nent of the more abstract notion of order. Delicacy has to do with the depth
of detail, ranging along a cline from least delicate at one end, i.e. primary,
to those small infinities at the opposite ‘where distinctions are so fine that
they cease to be distinctions at all’ (Halliday 2002a:48), i.e. secondary.
Structural types include configurational, prosodic, and periodic. Experi-
ential meaning can be accounted for in terms of the configuration of
process, participant, and circumstance. Interpersonal meaning tends to be
prosodic, involving intonation. Textual meaning is more periodic, with the
flow of discourse understood less in terms of discrete constituents than
wave-like movements.10
2.4 Realizing Ideational Meaning
Based on the findings from his study of the English language, Halliday
observed how each functional component or sub-component produces its
own distinct dimension of structure. For example, experiential meaning,
i.e. the ‘construing experience’ function (a sub-component within the Idea-
tional Metafunction), is realized as the structural configuration of process
as an integrated phenomenon involving participant(s) and circumstance(s).
In Systemic Functional Grammar (SFG), the structural configuration of
process, participant, and circumstance is referred to as a clause’s transitiv-
ity structure.
We talk about our experience of the world in terms of processes, plus the
participants and circumstances that enter into them. Processes are typically
realized as verbs which may describe an action, or a feeling, or a state of
being, or a way of behaving, either happening in the world around us or
within our own consciousness. Processes are often accompanied by
10
For further reading on ‘structure and rank’, see Halliday 2002a:40–9, 75–81, 95–105, 106–17, 118–26, 196–218;
Halliday 2002b:24–5, 27, 79–80; Halliday 2005a:29–36; Halliday 2005b:xii–xxix, 154–63, 249–51.
mention of who or what is participating, and possibly also of the circum-

stances – where/when/why/how.
The same experience may be described differently in different languages.
Even within the same language, the same experience may be expressed very
differently, suggesting some difference in how the same event is construed.
For example, if instead of saying This reminds us that . . . the speaker says This
is a reminder that . . ., the speaker has reconstrued the mental process to
remind – into a participant, a reminder, with all the qualities of noun-ness.
There is no one way to construe what is going on. The same happening
can be reconstrued differently depending on the speaker’s perspective on
the event being described. As Halliday (2003a:16, emphases in original) puts
it, ‘the grammar is not merely annotating experience; it is construing
experience – theorizing it, in the form that we call “understanding”.’
How one construes the world and what is happening will be evident in
how you talk about the world.
While processes may be grouped into different types on semantic
grounds – processes of ‘doing’, ‘being’, ‘sensing’, ‘saying’, etc. – there is
no clear reason for stopping at this point and not identifying further
distinctions until we have distinguished between individual words. Instead
Halliday distinguishes types of processes on both semantic and lexicogram-
matical grounds.
Processes also may be distinguished on the basis of who/what qualifies to
be a participant in that process. For example, because sensing verbs are
verbs of consciousness, the one sensing, the Senser, must be either human
or human-like, in other words, have a sense of consciousness. Unlike a
material process, the second participant in a mental or verbal process
may be a that-clause. A relational process may also take a that-clause, but
only if the noun as Subject is equivalent in meaning to what is expressed in
the that-clause, e.g. The concern is that rain might ruin the picnic.
Corresponding to the different process types – material, mental, rela-
tional, verbal, etc. – are different role designations. While it is reasonable to
talk about participating entities such as Actor and Goal in a material
process (or Agent and Affected if analyzed according to an ergative inter-
pretation), it would not be sensible to use the same role labels for partici-
pants in some other kind of process, such as a relational process, in which
one is describing the attributes of some entity.
Besides experiential meaning, another sub-component within the Idea-
tional metafunction deals with inter-clausal logico-semantic relations,
such as whether clauses of equal grammatical standing are combined
paratactically, or whether one main clause is hypotactically combined
with a dependent clause. Besides taxis relations, logico-semantic relations
also include whether the clauses are related through elaboration (e.g.
exemplification, clarification), extension (e.g. addition, alternation),
enhancement (e.g. embellishment, qualification), or projection (e.g. indirect/
direct speech).
2.5 Representing Interpersonal Meaning
Language functions to facilitate exchange, both of information and of

goods and services. The interpersonal metafunction deals with our use
of language to interact with our partners in the exchange. In any
exchange, there is give and take. When we ask a question – Is it raining? –
we anticipate an answer. If my communication is successfully understood
by my listener(s), then they will know what I expect of them, and,
depending on their ability and willingness, respond appropriately to what
I have requested. If we want someone to do something for us – Please turn
down the volume on the television – then such a request would typically
employ an imperative. A declarative typically expresses a statement, typ-
ically giving information; an interrogative expresses a question, typically
asking for information; an imperative expresses a directive; and an excla-
mative expresses an exclamation.
Such choices are realized structurally in English by what Halliday calls
the Mood element, consisting of the Subject and Finite. The sequence of
Subject before Finite realizes a declarative, e.g. The man fell; Finite before
Subject realizes a yes/no interrogative, e.g. Did the man fall? In fact, the
relationship between mood and illocutionary act is not one-to-one but
instead many-to-many. While a declarative typically expresses a statement
intended to give information, it may also function to make a request of the
listener. For example, the statement It sure is cold here could in fact be a
request to someone to turn off the air-conditioner. To understand the
speaker’s intended meaning, the situation context and the linguistic
co-text are essential.
Modality is another aspect of interpersonal meaning related to the
expression of the speaker’s attitude about what they are saying. In English,
modality is expressed by the use of modal adjuncts (possibly by certain
adverbs like surely, possibly, or by various thematic structures such as it is
possible that, or there is a possibility that), or through a small set of verbs
known as modal auxiliaries, e.g. can, may, might. Intonation also plays a
significant role in expressing ‘the particular tone of assertion, query, hesi-
tation, doubt, reservation, forcefulness, wonderment, or whatever it is,
with which the speaker tags the proposition’ (Halliday 2002a:205). Inton-
ation may be analyzed as a complex of three phonological systems, or as
Halliday puts it ‘systemic variables’: tonality, tonicity, and tone, which are
interdependent with rhythm.11
11
For further reading on ‘intonation’, see Halliday 2002a:55, 78, 90–1, 192–3, 205–7, 262–4, 269–70; Halliday
2002b:27–9, 32–6, 204–5, 232–3, 255; Halliday 2003b:50–1, 106–7, 162, 177, 184–9, 233, 317–19; Halliday
2004:69–71; Halliday 2005a:77–8; Halliday 2005b:57–70, 106–7, 139–40, 155–6, 161, 192–5, 213–15, 218,
237–86, 287–92; Halliday 2007a:71–3, 101, 158–9.
2.6 Representing Textual Meaning
The Textual Metafunction is concerned with textual meaning, or what

gives texture to a text. Texture is what makes a text into a coherent piece
of language, as opposed to simply being an unorganized string of sen-
tences. The more cohesive and coherent the text is, the greater is its
texture. On the one hand, cohesion deals with how successive sentences
are integrated to form a text, i.e. inter-sentence texture. Cohesion in a text
increases as the elements within a text become more mutually dependent
on one another for their interpretation. Cohesive ties cross clause and
sentence boundaries. Coherence, or intra-sentence texture, on the other
hand, has to do with fit to context. The organization of the message to fit
the context comprises two aspects: one aspect, which Halliday refers to as
the ‘hearer angle’, relates to the organization of the message so that it ties
up with the preceding text, with that which the hearer has already heard
about, i.e. the Given; the second aspect, the ‘speaker angle’, relates to how
the message is organized around what the speaker wants to say, or what
Halliday calls ‘Theme’.
Each clause is a proposition which contributes new information to
the text as a whole. Unlike Given information which is recoverable
from the preceding text or the immediate context of situation, New infor-
mation is something not previously mentioned; it is not recoverable.
Together, Given plus New information constitute information structure.
Intonation and rhythm, especially the pitch contour of speech, figure
prominently in the information system, foregrounding the new informa-
tion (New) from that which is otherwise recoverable from the discourse and
its context of situation (Given). New information typically culminates in an
element – i.e. the Focus – which is recognizable from the tonic prominence
which it receives. This connection between tone group and information
unit notwithstanding, information structure is as relevant to the analysis of
written texts as it is to spoken. In both spoken and written discourse, the
Given refers to something recoverable or locatable in the text or the context
of situation.
In terms of thematic structure in English, the Theme occurs in
clause-initial position. Because it is a matter of choice about what
comes first in the clause, it is therefore meaningful. Where the choice
is typical, we say it is ‘unmarked’. In a declarative clause, the grammat-
ical subject is the unmarked theme. In a yes/no question, the unmarked
theme is the request for the answer ‘yes’ or ‘no’, i.e. the finite element
of the verb: e.g. Did you eat? In a WH-question, the unmarked
theme is the request for information, i.e. the interrogative ‘WH-
element’: what, who, how, why, etc. In an imperative, the unmarked theme
is either the addressee(s), as in (You) leave quietly! (‘you’ being understood
even if not mentioned), or addressee(s) and speaker, as in Let’s leave
quietly!
Besides the grammatical Subject, other grammatical units which may

function as experiential/topical Theme include those that contribute to the
content of the utterance, from complements to circumstantial adjuncts,
and even the predicate in an imperative clause.
The Theme is underlined in each of the following examples (1) to (5):
(1) Bill left the party early.

(2) I don’t know.
(3) This afternoon we will go see a movie.
(4) The truth they could not accept.
(5) The one who baked the cake is my uncle.
The grammatical Subject as Theme of a declarative clause is typical

and unmarked, e.g. examples (1), (2), and (5), even when it contains an
embedded clause as postmodifier of the nominal-as-Subject. An adjunct
expressing circumstance is less typical and more marked, e.g. example (3).
The complement as Theme, e.g. example (4), is more marked because a
complement is least likely to be made thematic.
The Theme of a clause is first and foremost about the content of the
message, but may also express other meanings as well. For example, in the
sentence, But Mary, I do love your cooking. The initial conjunction But is
textual Theme, the vocative Mary is interpersonal Theme, and the gram-
matical Subject I is topical or ideational Theme. In English, this sequence of
thematic entities – textual, interpersonal, ideational – appears to be
the norm.
Thematic progression concerns the successive choices of what is rendered
thematic. Daneš (1974) identified three typical patterns of thematic pro-
gression: linear, continuous, and derived.
Linear thematic progression occurs when the new information found
in the Rheme of one clause occurs as Theme in the next clause. The flow
of information is New!Given. In the following example, from a journal
article appearing in Science (Bradshaw and Holzapfel 2006), notice
how the Theme of <c(lause) 2a> refers to the new information in <c 1a>
and <c 1b>; and the Theme of <c 3> refers back to the new information
from <c 2a>.
<p 1> <c 1a> Over the past 40 years, species have been extending their
ranges toward the poles <c 1b> and populations have been migrating,
developing, or reproducing earlier in the spring than previously (1–4).
<c 2a> These range expansions and changes in the timing of seasonal events
have generally been attributed to ‘phenotypic plasticity’ – that is, the ability
of individuals [<c 2b> to modify their behavior, morphology, or physiology
in response to altered environmental conditions (5, 6).] <c 3> Phenotypic
plasticity is not the whole story. <c 4a> However, recent studies show
<c 4b> that over the recent decades, climate change has led to heritable,
genetic changes in populations of animals as diverse as birds, squirrels,

and mosquitoes (see the first figure).
In continuous thematic progression, the same theme is repeated over subse-

quent clauses. This is illustrated in the following paragraph from a speech
given by Malala Yousafzai addressing the Youth Assembly at the United
Nations on her sixteenth birthday, 12 July 2013. Note the repeated thematic
references to those who shot her:
Dear Friends, on the 9th of October 2012, the Taliban shot me on the left
side of my forehead. They shot my friends too. They thought that the bullets
would silence us. But they failed. And then, out of that silence, came
thousands of voices. The terrorists thought that they would change our
aims and stop our ambitions but nothing changed in my life except this:
Weakness, fear and hopelessness died. Strength, power and courage was
born. I am the same Malala. My ambitions are the same. My hopes are the
same. My dreams are the same.
(Yousafzai 2013)
Derived themes are themes of successive clauses which may be regarded as

related to some superordinate topic or hypertheme. For example, in a
narrative account, we may notice repeated thematic references to signifi-
cant chronological points in the story, e.g. Early in the morning, By late
afternoon, or In the evening.
2.7 Formalizing Paradigmatic Relations
In the late 1950s Halliday worked in the pioneering Cambridge Language

Research Unit:
In this context it became necessary to represent grammatical features in

explicit, computable terms. I wanted to formalize paradigmatic relations,
those of the system; but I did not know how to do it – and I totally failed to
persuade anyone else of this!
(Halliday 2005a:138)
Subsequently, however, system networks were developed as a way to repre-

sent meaning potential not simply syntagmatically as an inventory of
sequentially ordered items, but instead primarily as paradigmatically
organized sets of options.
A system network is comprised of only AND and OR relations between
options. No distinction is made between linguistic information and
non-linguistic information; all information is stored and processed
the same way, as interconnected options in a vast network. Information
processing consists of the transmission of activation along path-
ways defined by the network, and changes in probabilities attached to
options.
Halliday (2003a:327) defines semantic networks as ‘a hypothesis about

patterns of meaning’. Semantic networks
• specify the range of alternatives at the semantic stratum;

• relate the hypothesized categories to some general social theory or
‘theory of behaviour’;
• relate those same categories to the categories of linguistic form.
(see Halliday 2003a:334)
The great problem with the system network as a form of representation, as

Halliday himself points out, is that it is a very static form of representation.
In an interview with Kress, Hasan, and Martin, Halliday commented:
It freezes the whole thing, and then you have to introduce the dynamic in
the form of paths through the system. Your problem then is to show how the
actual process of making paths through the system changes the system.
(Halliday in Martin 2013:110)
Each act of meaning, ‘each instance, no matter how minutely, perturbs

these probabilities and so changes the system’ (Matthiessen 2015:212).
Halliday uses the term ‘instantiation effect’ to describe how our acts of
meaning feed back into the system. With advances that have been made in
corpus building, quantitative profiling, or the assignment of probabilities to
choices within the system network, has become a realistic goal for linguis-
tic description.12
A related concept is the notion of ‘markedness’. In a skew where one
option is more frequent, this option is quantitatively unmarked, and the
less frequent option is marked. Where the options are equally probable,
neither is the ‘unmarked term’. Besides being quantitatively unmarked, an
option may be simpler and therefore identified as being formally
unmarked.13
In Saussurean terms, acts of meaning are ‘parole’, and the system of
meaning potential which these acts instantiate is ‘langue’. However, unlike
the Saussurean distinction between langue and parole, Halliday sees acts of
identity (parole) and meaning potential (langue) not as two distinct classes
of phenomena, but instead as only a difference in the stance taken by the
observer. Langue is parole seen from a distance, parole is langue up close
and in its context. What some perceive to be a dichotomy between langue
and parole, Halliday instead views as a cline of instantiation ranging
between system and instance. As Halliday (2005a:248) explains,
12
For further reading on ‘quantifying language’, see Halliday 2002a:70–2, 92–4, 166, 168–9, 352–68; Halliday
2003a:23–6, 122, 253, 404–13, 425–6, 430; Halliday 2005a:8–9, 13–19, 42–62, 63–75, 76–92, 93–129,
130–56, 157–90, 235–8; Halliday 2006:209–48; Halliday 2007a:310–16.
13
For further reading on ‘markedness’, see Halliday 2002a:305, 320–1, 326, 376–7; Halliday 2002b:28–38, 199–200,
205; Halliday 2003b:342–3; Halliday 2005a:22–3, 81, 88, 91, 96–7, 101–2, 131–2; Halliday 2005b: 5–54,
55–109, 110–53, 154–63, 193–5, 203, 220–31, 249–61, 264–86, 288–9; Halliday 2006:5–174, 209–48,
330–2.
Instantiation is a cline, modelling the shift in the standpoint of the observer:

what we call the ‘system’ is language seen from a distance, as semiotic
potential, while what we call ‘text’ is language seen from close up, as
instances derived from that potential. In other words, there is only one
phenomenon here, not two; langue and parole are simply different observa-
tional positions.
To illustrate his point, Halliday gives the analogy of climate and weather.
The weather is what we experience on a daily basis. Over time, we general-
ize this daily occurrence of the weather into our sense of what the climate
is. So the climate becomes a theory about the weather. Similarly, like the
notion of climate, the system of meaning – meaning potential – is a theory
of the text – the instantiation of meaning potential (see Halliday
2008:79–80). Just as the weather and climate are not two distinct and
separate phenomena, so too parole (instances of language) and langue
(system of language) are one and the same phenomenon.
2.8 Language in Flux
Language is always in flux. It is dynamic and changing. SFL views the

fuzziness and indeterminacy which one observes in actual usage to be both
necessary and positive, enabling language to achieve its richness. Halliday
describes indeterminacy in language as occurring (i) where distinctions in
meaning are more continuous than discrete; (ii) where meanings become
fused to the extent that one cannot select between them; and (iii) where
some domain of experience may be construed in contradictory or compet-
ing ways. This inherent fuzziness and indeterminacy cannot be ignored, but
instead deserves to be accounted for in the grammar.14
The direction of this change is towards growth. The meaning space is
constantly expanding. As society evolves, so too does language. New ways of
doing demand new ways of meaning, i.e. new registers.15
Phylogenetic evolutionary change in human language is one of three
histories identified by Halliday, the other two being the ontogenetic devel-
opment of each individual’s language, and the logogenetic unfolding of a
particular text.
14
For further reading on ‘indeterminacy in language’, see Halliday 2002a:399–402, 409–10; Halliday 2002b:33, 51,
139–40, 145–6; Halliday 2003a:54–5, 254–5, 266–7; Halliday 2005a:204–7, 211, 226–30; Halliday
2007b:193, 200.
15
Halliday defines a register as ‘a syndrome, or cluster of associated variants; and again only a small fraction of the
theoretically possible combinations will actually be found to occur’ (2002b(1990):168) rather than the obligatory
incidence of particular features. Dialects are identified by their users. Codes are patterns or speech habits of speakers
of the same language. For further reading on ‘varieties and variation in language: dialect, register, code’, see Halliday
2002b:17, 168–70, 231–4; Halliday 2003a:255–6, 268, 360, 362–3, 382–3, 416–17; Halliday 2005a:225–6,
248, 263–4; Halliday 2005b:214–16; Halliday 2007a:29–31, 240–3, 296–300; Halliday 2007b:5–40, 85–8,
93–7, 103–7, 115–16, 129–30, 138–9, 140–2, 147, 174–5, 181–3, 196–7, 205–9, 235, 242–3, 252–5,
259–61.
Phylogenetic evolutionary change is well illustrated by the language of

science, in which, as Halliday explains, there has been ‘a steady drift [in the
grammar of scientific discourses] towards things’ (Halliday 2004:129,
emphasis in original). Things have been foregrounded at the expense of
qualities, processes, and relations. While this nominalizing grammar ‘has
given us enormous powers over our physical and biological environment’
(Halliday 2004:47), so much so that ‘the scientist can make the world stand
still, or turn it into one consisting only of things, or even create new, virtual
realities’ (Halliday 2004:viii), still it has done so at the risk of alienating
learners and turning science into ‘the prerogative of an elite’ (Halliday
2004:179).
Over the course of history, as the need arose for new ways to theorize
the human experience, humankind has relied on the power of language
‘to reconstrue commonsense reality into one that imposed regularities
on experience and brought the environment more within our power to
control’ (Halliday 2004:xvii).
Halliday describes this metaphor-making potential, i.e. grammatical
metaphor, as ‘a concomitant of a higher-order, stratified semiotic – once
the brain splits content into semantics and grammar, it can match them up
in more than one way’ (Halliday 2004:123). It is as though there has been a
‘partial freeing of the lower-level systems from the control of the semantics
so that they become domains of choice in their own right’ (Halliday
2002b:131).
Halliday (2013:78, emphases in original) further explains this metaphor-
making potential as follows:
Metaphor, whether in its grammatical or its lexical sense, is a cross-coupling

between the semantics and the lexicogrammar. In lexical metaphor, which is
metaphor in its traditional sense, this is the replacement of one lexical item
(word or phrase) by another in the realization of a given meaning . . . In
grammatical metaphor, one grammatical category is replaced by another – a
word class, a structure unit, and often both; for example, in place of
‘she didn’t know the rules, so she died’ we have ‘her ignorance of the rules
led to her death’. Both grammatical and lexical metaphors are characterized
by semantic junction (this is the basis of the distinction between metaphor
and simile).
Historically speaking, this metaphor-making potential has been achieved

over ‘three successive waves of theoretical energy’ (Halliday 2004:46) –
generalization, abstractness, metaphor – each ‘tak[ing] us one step further
away from ordinary everyday experience’ (Halliday 2004:47), but at the
same time each step may be thought of as having ‘enlarged the meaning
potential by adding a new dimension to the total model’ (Halliday 2004:46).
Whether it is the ‘art’ in verbal art, or the ‘science’ in verbal science,
what is being crafted through the metaphor-making potential available in
language are hypotheses or models about the world experienced around us
and in us. Such is the semogenic power of grammar, that it enables us

to define ‘the basic experience of being human’. Hasan refers to the hypoth-
eses articulated in verbal art as themes about some aspect of social
life achieved through second-order semiosis – a kind of double articulation –
‘so that one order of meaning acts as metaphor for a second order
of meaning’ (Hasan 1985:100). Halliday adopts Mukařovský’s term
‘de-automatization’ to describe this ability ‘to interpret the grammar in
terms that go beyond its direct realizational function’ (Halliday 2002b:139),
allowing ‘the wholesale recasting of the relationship between the grammar
and the semantics’ (Halliday 2004:19).
Halliday’s notion of grammatical metaphor resembles what Kauffman
(2008) refers to as ‘Darwinian preadaptations’. Darwinian preadaptations in
biological evolution occur when pre-existing phenotypic features are detoured
to new uses, ‘explor[ing] . . . the “adjacent possible”, and thereby expand[ing]
the range of actuality in unforeseen and unforeseeable ways’ (Shaviro 2008).
Just as science does not have all the answers when it comes to predicting
the partially lawless evolution produced by Darwinian preadaptations, so
too the ceaseless creativity in language, while enabled by grammar, can
never be fully predicted by grammar – especially if that grammar is algo-
rithmic, computational, or connectionist in character.
Semantics and grammar are coupled together through what Halliday
(2004:94) calls ‘a relation of congruence’, in which things are realized as
nouns, processes as verbs, qualities as adjectives, relators as conjunctions,
circumstances as adverbs or prepositional phrases. However, we can
de-couple these congruent relations, and re-couple them in whatever way best
suits our purpose. Examples of grammatical metaphor include length, which
is ‘a junction of (the quality) “long” and the category meaning of a noun, which
is “entity” or “thing”, [and motion, which is] a junction of the (the process)
“move” and the category meaning, again, of a noun’ (Halliday 2004:xvi–xvii).16
The following three-step recipe illustrates how to turn a congruent state-
ment into its metaphorical alternative:
Congruent: If the item is exposed for long, it will deteriorate rapidly.

(i) Nominalize the Process (exposed ) exposure; deteriorate )
deterioration)
(ii) Make the Medium of that Process a ‘possessive’ modifier (the item / it
) of the item)
(iii) Express the relation between the two events (If . . . then ) will result in)
Metaphor: Prolonged exposure will result in rapid deterioration of the item.
16
For further reading on ‘grammatical metaphor’, see Halliday 2002a:219–60, 346–8, 358–60, 397; Halliday
2002b:23–84, 160, 164, 219–23, 226; Halliday 2003a:130–4, 139–76, 248–70, 282, 284–5, 384, 388, 415,
419–23; Halliday 2003b:339–40, 347–9, 367–9; Halliday 2004:7–23, 32–43, 49–101, 102–34, 143, 147–52,
156–7, 162, 171–9, 190–7, 214–16, 220–5; Halliday 2005a:42–62, 63–75, 196–212, 213–38; Halliday
2006:325–33, 339–44; Halliday 2007a:63–80, 105–10, 117, 123, 126–8, 301–3, 354–67, 379, 381; Halliday
2007b:239, 243–4, 278.
However, the choice between congruent and metaphorical is not binary,

but complementary. Rather, as Halliday (2004:84–5) illustrates, there may
exist a series of agnate forms at intermediate points along a cline of
metaphoricity:
(i) Osmolarity increases, so putrescine is rapidly excreted.

(clause nexus: paratactic)
(ii) Because osmolarity increases, putrescine is rapidly excreted.
(clause nexus: hypotactic)
(iii) That osmolarity increases has the effect that putrescine is rapidly
excreted.
(clause: two rankshifted clauses, finite)
(iv) Osmolarity increasing leads to putrescine being rapidly
excreted.
(clause: two rankshifted clauses, non-finite)
(v) Increasing of osmolarity causes rapid excreting of putrescine.
(clause: two nominal groups, verb as Head)
(vi) Increase of osmolarity causes rapid excretion of putrescine.
(clause: two nominal groups, mass noun as Head)
(vii) Increases of osmolarity cause rapid excretions of putrescine.
(nominal groups, count noun as Head)
Besides this complementarity between congruent and metaphorical, there

are other complementarities as well, such as that between speech and
writing, or the meta-complementarity between lexis and grammar. These
complementarities make it possible for our common-sense grammars of
daily life to accommodate multiple and possibly even contradictory per-
spectives on the same set of phenomena.
In SFL theory, beyond the language-internal strata of semantics, lexico-
grammar, and phonology, there exists the language-external contextual level.
The context of culture represents the systemic potential for choice
depending on how the context of situation is defined with respect to field,
tenor, and mode. What is going on, i.e. field, plays a role in determining
choices related to transitivity structure, or process, participant, and circum-
stance. Who is involved, i.e. tenor, deals with interpersonal choices from
the systems of mood and modality. How the exchange takes place will
influence choices involving cohesion, as well as thematic and information
structures.17
17
For further reading on ‘context of culture and context of situation’, see Halliday 2002a:29, 35, 201, 211, 217, 221,
225–31, 243, 246, 263, 283–5, 311, 357, 359, 405; Halliday 2002b:38, 44, 51–64, 150–2, 229–34, 243–4, 251,
254; Halliday 2003a:154–6, 185, 195–7, 210, 273, 279, 298–9, 358, 362, 382, 420; Halliday 2003b:81, 87, 95,
101, 111, 121, 134, 204, 207, 286–95, 302–4; Halliday 2005a:207–8, 217, 225, 238, 249, 256, 260, 266;
Halliday 2005b:199, 306–37; Halliday 2006:10, 13, 16, 20, 64, 355–7; Halliday 2007a:85–7, 94, 96, 271–90, 298,
300, 307, 311, 349, 354–67, 368–82; Halliday 2007b:59, 62, 77, 82, 90–7, 105, 110–20, 127, 130, 133–7, 140,
142, 172, 180–2, 184, 187, 192–9, 203, 209, 235, 242, 258–9, 262.
2.9 A Marxist Orientation
Listing what he identifies as the seven distinctive features of SFL,18 Halliday

includes ‘Marxism’. Marxism, he writes, has ‘always [been] part of my own
thinking: values of language and language varieties, language as political
process and political tool; linguistics as (critical component of ) a science of
meaning <more recent, of course>’ (Halliday, personal communication).
Subsequently, in an interview,19 Halliday suggests that in order to study
language in Marxist terms, one should begin by questioning the views of
language that were then current in a non-Marxist society. So, for example,
when he returned to England from China in the 1950s, language studies
and linguistics privileged or prioritized standard languages over dialects;
written language over spoken; classical languages over modern ones;
formal language over colloquial; dominant languages over emergent ones;
literary language over everyday language; majority languages over minority
languages; ideational meaning over interpersonal. While acknowledging
work being done in dialectology, and on minority languages, Halliday
(personal communication) nonetheless argues, ‘The view of language which
was predominant was essentially derived from all those highly-valued
forms of language rather than the others.’ Having noted this skewed
emphasis on highly valued forms as compared with ‘socially-induced low
visibility forms of language’, Halliday, along with other members of the
Communist Party Linguistics Group20 sought to unpack how these prior-
ities had influenced the representation of grammar in mainstream linguis-
tics. They focused their attention on the languages of ex-colonial societies
which were struggling to gain status as national languages. Their aim was
to give value to these forms of language. As Halliday explains in the same
18
In a letter, Professor M. A. K. Halliday identified the following seven distinctive features of SFL: ‘(1) Quantitative studies
and probability: This has been a constant thread since “Linguistics and machine translation” (it was even raised in my
PhD thesis on the Secret History), with references also to Shannon Weaver’s theory of information (now at last
recognized for its importance, e.g. in recent work by Terrence Deacon). And as far as I know nobody has followed up
the work I did with collaboration from Zoe James on the probabilistic nature of grammatical systems. (2) Metafunction:
again a recurrent theme, with the point that metafunction determines the way that languages have evolved, and the
insistence that the interpersonal and textual metafunctions are equally fundamental, along with the ideational, to the
functioning of language as a semiotic system, the form taken by grammatical structures maximizing the possibilities for
different meanings to combine freely with one another. (3) Historical contexts of linguistics: Wang Li’s and Firth’s
departments were unusual (even unique?) in building the history of linguistics into their teaching of the subject; Wang
Li the Chinese tradition, Firth the European (and also the Indian, though I never learnt so much about that); more
recently, linguistics in the context of the history of ideas, relation of the human to the natural sciences (linguistics itself
being also physical, biological and social). (4) Marxism: early with Jeff Ellis, Dennis Berg, Jean Ure and others, mostly
existing only in typescript; not made explicit in the McCarthy era (I had lost too many openings because of it in those
years!) but always part of my own thinking: values of language and language varieties, language as political process
and political tool; linguistics as (critical component of ) a science of meaning <more recent of course>. (5) Unity of
lexis and grammar. (6) Continuity of protolanguage. (7) Thinking in terms of patterns of analogies, complementarities,
compatibility with other views and theories (where others only see contractions)’ (Halliday, personal communication).
19
This interview was conducted by Annabelle Lukin, David Butt, and myself with Professors Halliday and Hasan in their
home in 2012.
20
Jeffrey Ellis, Jean Ure, Dennis Berg, Trevor Hill, and Peter Wexler.
interview, ‘In my own background at that time, my own thinking at that

time, I would say “Yes, I was definitely trying to adopt what I would
consider a Marxist approach, and indeed, I still would”.’
Clarifying what he means by Marxist thinking, Halliday emphasizes that
he is ‘not at all saying Marxist means as it were there even in embryo in the
words of Marx himself’. Rather, ‘if Marxist thinking means anything, it
simply means clear and honest and objective thinking about, first of all,
whatever it is that you are focusing your attention on’.
One cannot avoid the influence of one’s personal ideology on one’s work
in science. To argue otherwise and say that ideology should not affect our
science is itself an ideological position. While acknowledging that Marxism
is part of the background to his own thinking about language and society,
Halliday is quick to add that this does not commit anyone who does
Systemic Functional Linguistics to working in left-wing politics.
Nevertheless, there is obvious compatibility between an SFL perspective
on language and the Marxist view that consciousness and language are
evolutionary and ‘are intertwined because of the social basis of the origins
of both’ (Holborow 1999:17). In the same interview with Lukin, Butt, and
Webster, Hasan described the view both she and Halliday have taken of
language as being ‘intensely social’:
Human beings are incapable of living, surviving alone. In this lies their
humanity. Everything in them is created through being part of a society.
Either a reaction to it, or a following of it, whichever form you take.
As he states in his interview with Kress, Hasan, and Martin in 1986 (Martin
2013:118), Halliday saw Firth’s approach to the study of language as being
‘perfectly compatible’ with a Marxist linguistics:
It seemed to me that, in fact, the ways in which Firth was looking at

language, putting it in its social context, were in no way in conflict with
what seemed to me to be a political approach. So that it seemed to me that in
taking what I did from Firth, I was not separating the linguistic from the
political. It seemed to me rather that most of his thinking was such that
I could see it perfectly compatible with, indeed a rather necessary step
towards, what I understand as a Marxist linguistics.
In a recent contribution to the Bloomsbury Companion to M. A. K. Halliday,

Halliday (2015:99) adds the following about how a Marxist orientation
impacted his own thinking:
It seems to me that this overall conceptualization of language, which had

been developing slowly in my own thinking across several decades, is essen-
tially – though not aggressively – marxist in its orientation. If I never
proclaimed this out loud, this was because it would be too much open to
misunderstanding: there are too many different ideas about what ‘marxist’
means, and most people nowadays wouldn’t think it was worth discussing.
One attribute of my ideas that is at least compatible with a marxist ideology
is the way they have always developed in conversation with other people,
often as a by-product of my activities as a teacher; I have tried to acknow-
ledge those who have been part of this enterprise, though I am conscious of
having done so only very inadequately. It is impossible to track the proven-
ance of scientific ideas, and with our modernist ways of thinking we attach
too much importance to the individual anyway. It was my privilege to
encounter so many congenial and thoughtful colleagues.
Halliday (2015:98) also credits a Marxist orientation with having motivated

the ‘appliable’ emphasis of Systemic Functional Theory:
I hoped that what I was trying to achieve as a linguist might make some
contribution to improving the human condition, however minuscule and
oblique. This is what I meant by calling the theory ‘appliable’. The term is
less specific than ‘applicable’, which denotes applicable to some specific task,
and therefore less immediate, and more indirect; its relevance is less obvi-
ous, but more long term. But other than this feature of being appliable, what
other aspect of the theory might be considered as marxist?
2.10 Appliable Linguistics
Practising appliable linguistics has indeed been a driving force behind the
development of Systemic Functional Linguistic Theory. Describing what he
means by ‘an appliable linguistics’, Halliday (2013:128) writes:
An appliable linguistics, as I understand it, is a theory which tackles prob-

lems and tries to answer questions – but questions that are asked, and
problems that are raised, not by professional linguists so much as by other
people who are in some way concerned with language, whether profession-
ally or otherwise. There are large numbers of such people: educators, trans-
lators, legal and medical specialists, computer scientists, students of
literature and drama, . . .; and it is their ‘take’ on language that is being
addressed, at least to the point of clarifying what sorts of questions can
usefully expect to be asked, and whether or not there is any hope of coming
up with an answer.
Early in Halliday’s academic career, he taught at Edinburgh. The fact that

many of Halliday’s students were likely to become teachers in the Scottish
school system prompted Halliday’s interest in discovering what linguistics
had to offer to these teachers as well as what could be learned from those
teaching at secondary and primary levels of education. When Halliday
moved back to London in 1963 to serve as Director of the Communication
Research Centre at University College, his experience working on several
projects related to educational materials development (Breakthrough to Liter-
acy, Language in Use, and, indirectly, Language and Communication) and his
involvement in the Linguistics Properties of Scientific English project contrib-
uted to his groundbreaking description of English grammar, entitled ‘Notes
on transitivity and theme in English’, which was published over three

successive issues of the Journal of Linguistics in 1967–1968 (see Halliday
1967a, 1967b, 1968). Halliday saw this development of a grammar for
educational purposes as addressing the need for social accountability.
Teachers would often ask Halliday about the language experience of
children before they came to school. Their questions raised Halliday’s
own interest in how children develop language. Halliday saw language
development as involving learning language or leaning how to mean, learn-
ing which takes place through language, and learning about language.
The better informed the language teacher is about language, the more
successful their pupils’ learning is likely to be.
Halliday’s interest in studying the language of science, for example, grew
out of his concern ‘to find the source of the difficulties faced by learners of
science’ (Halliday 2004:xx). One can only help the learner, insisted Halliday,
to the extent that one understands how the discourse works. Halliday’s
search led him back into the history of scientific discourses, out of which he
discovered ‘new strategies evolving: new ways of organizing the grammar
as a resource for making meaning’ (Halliday 2004:xv).21
More recently Systemic Functional Linguistics Theory has proved its
usefulness in attempts at using natural language as the means for achiev-
ing intelligent computing. A Systemic Functional form of semantic repre-
sentation has been employed in projects related to data fusion, fuzzy
reasoning, and the ability to construe the context of situation by inference
from the text.22
References
Bradshaw, W. E. and C. M. Holzapfel. 2006. Evolutionary Response to Rapid

Climate Change. Science 312(5779): 1477–8.
Daneš, F. 1974. Functional Sentence Perspective and the Organization of the
Text. In F. Daneš, ed., Papers on Functional Sentence Perspective. Prague:
Academia. 106–28.
Press.
Halliday, M. A. K. 1967a. Notes on Transitivity and Theme in English: Part 1.
Journal of Linguistics 3(1): 37–81.
Halliday, M. A. K. 1967b. Notes on Transitivity and Theme in English: Part 2.
21
For further reading on ‘language teaching, learning and development’, see Halliday 2002a:323–4, 349–51; Halliday
2003a:228–30, 273–4, 378–9, 384, 397–404, 429–30; Halliday 2003b; Halliday 2005b:297–305; Halliday
2007a; Halliday 2007b:63–4, 75–81, 118, 128–9, 175–6, 193–5, 212–13, 223–30.
22
For further reading on ‘linguistic computing’, see Halliday 2003b.
Halliday, M. A. K. 1968. Notes on Transitivity and Theme in English: Part 3.

Halliday, M. A. K. 2002a. Collected Works of M. A. K. Halliday, Vol. 1: On
Grammar. Edited by J. J. Webster. London: Bloomsbury.
Halliday, M. A. K. 2002b. Collected Works of M. A. K. Halliday, Vol. 2: Linguistic
Studies of Text and Discourse. Edited by J. J. Webster. London: Bloomsbury.
Halliday, M. A. K. 2003a. Collected Works of M. A. K. Halliday, Vol. 3: On
Language and Linguistics. Edited by J. J. Webster. London: Bloomsbury.
Halliday, M. A. K. 2003b. Collected Works of M. A. K. Halliday, Vol. 4:
The Language of Early Childhood. Edited by J. J. Webster. London:
Bloomsbury.
Halliday, M. A. K. 2004. Collected Works of M. A. K. Halliday, Vol. 5: The
Language of Science. Edited by J. J. Webster. London: Bloomsbury.
Halliday, M. A. K. 2005a. Collected Works of M. A. K. Halliday, Vol. 6: Com-
putational and Quantitative Studies. Edited by J. J. Webster. London:
Bloomsbury.
Halliday, M. A. K. 2005b. Collected Works of M. A. K. Halliday, Vol. 7: Studies in
English Language. Edited by J. J. Webster. London: Bloomsbury.
Halliday, M. A. K. 2006. Collected Works of M. A. K. Halliday, Vol. 8: Studies in
Chinese Language. Edited by J. J. Webster. London: Bloomsbury.
Halliday, M. A. K. 2007a. Collected Works of M. A. K. Halliday, Vol. 9: Language
and Education. Edited by J. J. Webster. London: Bloomsbury.
Halliday, M. A. K. 2007b. Collected Works of M. A. K. Halliday, Vol. 10: Language
and Society. Edited by J. J. Webster. London: Bloomsbury.
Halliday, M. A. K. 2008. Complementarities in Language. Edited by J. J.
Webster. Beijing: Commercial Press.
Halliday, M. A. K. 2009. The Essential Halliday. Edited by J. J. Webster.
London: Bloomsbury.
Halliday, M. A. K. 2013. Collected Works of M. A. K. Halliday, Vol. 11: Halliday in
the 21st Century. Edited by J. J. Webster. London: Bloomsbury.
Halliday, M. A. K. 2015. The Influence of Marxism. In J. J. Webster, ed.,
94–100.
Functional Grammar. 4th ed. London: Routledge.
Hasan, R. 1985. Linguistics, Language and Verbal Art. Oxford: Oxford
University Press.
Holborow, M. 1999. The Politics of English. London: SAGE.
Kauffman, S. 2008. Reinventing the Sacred: A New View of Science, Reason, and
Religion. New York: Basic Books.
Martin, J. R. 2013. Interviews with M. A. K. Halliday: Language Turned Back on
Himself. London: Bloomsbury.
Matthiessen, C. M. I. M. 2015. Halliday’s Conception of Language as a
Probabilistic System. In J. J. Webster, ed., The Bloomsbury Companion to
M. A. K. Halliday. London: Bloomsbury. 203–41.
Shaviro, S. 2008. Reinventing the Sacred (Stuart Kauffman). Available online at:
www.shaviro.com/Blog/?p=636. (Last accessed 16/05/17.)
Yousafzai, M. 2013. The Full Text: Malala Yousafzai Delivers Defiant Riposte
to Taliban Militants with Speech to the UN General Assembly. Independent,
12 July 2013. Available online at: www.independent.co.uk/news/world/
asia/the-full-text-malala-yousafzai-delivers-defiant-riposte-to-taliban-mili
tants-with-speech-to-the-un-8706606.html. (Last accessed 05/06/17.)
3
Semantics
Miriam Taverniers
3.1 Introduction
Semantics can generally be described as ‘the study of linguistic meaning’,

and this goes for Systemic Functional Linguistics (SFL) too. Beyond this safe
but rather uninformative definition, there are many different types of
interpretations of what (a) semantics can be, how big it is, what place it
has in a more general model of language, and how it is related to other
dimensions of the model. And this, too, goes for SFL. Whereas in general,
linguistic models can be distinguished in terms of how they view ‘linguistic
meaning’, within SFL different conceptions of ‘semantics’ exist side by side,
and different views of semantics are seen as complementarities that make
the theory rich, flexible, and adaptable (to different purposes). In other
words, it is not possible to give an overview of what ‘semantics’ means
(no pun intended!) in SFL. What is possible, and much more interesting, is
to explore how semantics can potentially be viewed in SFL, how those
different views of semantics can be understood, and how semantics has
been modelled in SFL.
The aim of this chapter is thus twofold: to survey different possible
conceptions of semantics in SFL by elucidating how these conceptions are
related to one another (i.e. by fleshing out what each conception high-
lights with respect to a certain architectural dimension of the theory),
and to explore how specific semantic analyses and models which have
been proposed in SFL can be understood against the background of these
possible conceptions. Section 3.2 explores different views of semantics in
SFL. These conceptions are put into a broader theoretical perspective
in Sections 3.3 and 3.4. Section 3.5 then looks at some recent specific
semantic models in SFL and places them against the background built
up in Sections 3.2 to 3.4.
56 MIRIAM TAVERNIERS
3.2 The Basics of Semantics in SFL: What, Where,

How, and Why?
In order to explain what ‘semantics’ can mean in SFL, we will consider the
following questions:
(1) What can ‘linguistic meaning’ be in a systemic functional perspective? and

Where is ‘semantics’ located in the overall architecture of the model?
(2) How can semantics be modelled?
(3) Why is semantics to be recognized as something in its own right? Why is it
theorized and designed as it is?
3.2.1 The Basis of Valeur, Stratification, and Meaning-Making:

WHAT and WHERE is Semantics?
The basic equation of ‘semantics’ with ‘meaning’ can be extended in two
directions. ‘Meaning’ does not occur as such: (i) it is expressed in a certain
way, i.e. linguistic meaning is encoded or ‘realized’ in linguistic ‘forms’ which
are further expressed in sound or in writing, and (ii) it is always ‘meaning-in-a-
context’, i.e. something (and this is not restricted to language) is meaningful in
a specific ‘context’. The conception of meaning between form and context is
the view that is highlighted in the familiar image of stratification in SFL, with
semantics as a stratum between context and lexicogrammar.
The two dimensions of ‘meaning’ pointed out above are also present in the
tool of the system network. The concept of ‘meaning’ as ‘that which is
realized in form’ underlies the relationship between a valeur in a network,
i.e. a systemic option (or rather, a path of systemic choices) and its realization
in a structure (specified in a realization statement, see e.g. Schulz and Fon-
taine, this volume). In the tool of the system network, the other, contextual
side of ‘meaning’ is present too, since systemic options that are grouped in one
network are those that are available in a particular setting, i.e. in the context
of a particular rank or unit, for example, the options of process type which
are available at clause level. In this view ‘meaning’ is the value – or valeur – of
an option in relation to other options that are available in a context, a value
that is realized in a structure, which thus is a token of that value.
Hence two basic conceptions of ‘meaning’ in SFL can be distinguished by
looking at where the relation of ‘realization’ or encoding occurs in the
model, and this is between strata, on the one hand, and between systemic
options and realization statements, on the other. In Figure 3.1 these two
conceptions of meaning/semantics are given in [1] and [2].1
1
Note that the duality of the initial two conceptions of meaning that are pointed out here is widespread in linguistics:
the former, as the type of meaning that is interfacing with context, has also been called ‘contextual meaning’, or
‘extra-linguistic meaning’ (and more specific sub-types of this are ‘reference’, ‘ontological meaning’, ‘speech act
meaning’, etc.), whereas the latter has been called ‘intra-linguistic’ or ‘formal meaning’ (with ‘sense’ as a sub-type), or
has been defined in relation to grammar as the ‘semantics of grammar’.
xt
te
n
co
ics
nt
a
sem

ram
-g
lex
ph
gra
realization statement
phon/
doings meanings wordings systemic option realization statement content expression language as a whole
context meaning form value token plane plane as ‘meaning-making’
as a ‘meaning potential’
context semantics lexicogrammar paradigmatic valeur syntagmatic structure
[1] [2] [3] [4]
Figure 3.1 Four types of conceptions of ‘meaning’ in SFL
A third conception of meaning (Figure 3.1:[3]) appears when a differenti-

ation between an expression side and a content side of language is priori-
tized: in that case ‘meaning’ is the content that is expressed in sounds/
writings. This idea underlies the basic structuralist split between ‘content’
and ‘expression’ in language, i.e. the two sides of a linguistic sign in Saus-
sure’s model, which Hjelmslev (1963) called content and expression ‘planes’.
In addition to those three views of ‘meaning’, there is a broader view
(Figure 3.1:[4]) which is highlighted in typical characterizations of language
from an SFL perspective: language as a whole is defined as a ‘meaning
potential’, and learning a language is learning ‘to mean’ (see the title of
Halliday 1975). Halliday’s adaptation of the familiar verb to mean to refer to
this overall meaning of language highlights this fourth, more general
conception of meaning: it is not just the semantic stratum that ‘means’,
nor the content plane, but language as a whole makes it possible for us to
‘mean’ in the various contexts of our human lives.
Hence we arrive at four basic conceptions of meaning/semantics: [1] mean-
ing as valeur; [2] meaning as one stratum in relation to other strata in a
stratified view of language; [3] meaning as the content side of language; and
[4] meaning as what characterizes language in general as a meaning-making
resource. These conceptions have been highlighted, focused on, and com-
bined in different stages in SFL, but they also lie at the basis of a primary
distinction between the Cardiff model of grammar and Halliday’s model of
grammar. In the Cardiff model the distinction between system and structure
is seen as primary, and falling together with the dividing line between
semantics and lexicogrammar. Semantics is seen as the stratum of the system
networks, whose options are then realized one stratum below, in lexicogram-
matical structures. In other words, in this model semantics is paradigmatic
and lexicogrammar is syntagmatic (see Schulz and Fontaine, this volume).
In the traditional SFL model, which is focused on in this chapter, the ideas
of meaning as valeur, meaning as an intermediate stratum, meaning as the
content plane of language, and meaning as what characterizes language as a
whole are not incompatible at all, but are intrinsically intertwined, and are all
useful and necessary, as we will see below. In this view the different concep-
tions of meaning pointed out above are all valid side by side. At the same time,
it has been explored what a semantics as a stratum in its own right can be and
how this can then be modelled, a topic we turn to in Section 3.2.2.
3.2.2 Three Interrelated Conceptions of HOW Semantics

Can Be Modelled
In this subsection we focus on the conception of meaning as a stratum in its
own right, i.e. that conception of meaning for which the nominal term
‘semantics’ is used in SFL. Afterwards (Section 3.3) we will return to how
this semantics is related to other dimensions that are called ‘meaning’ or
‘semantic’ in the broader perspective sketched.
Semantics 59
Asking the question of what a (separate) semantic stratum can be, more
specifically, correlates with asking what this stratum can contain, or how this
stratum can be modelled in relation to other – especially surrounding –
strata. In the visual metaphor of stratification, with the strata represented
as cotangential circles (since Martin and Matthiessen 1991; see Figure 3.1), the
higher levels of the language system are the more abstract ones, and the lower
ones the more concrete realizations. In keeping with this orientation, seman-
tics in general can be conceived of as ‘higher order valeur’: a ‘valeur’ which can
be recognized above the lexicogrammar, as a more abstract meaning.
Three complementary ways of modelling this ‘higher order valeur’ can be
distinguished. I will initially characterize them here, and we will then
return to the different types of modelling in more detail when looking at
specific semantic analyses (Section 3.3) after we have addressed the ‘why’
question (Section 3.2.3).
(a) Semantics as topological meaning: In one view, semantics is the stratum at

which areas of the lexicogrammar are regrouped into semantic
domains. This is the case when distinct lexicogrammatical phenomena
(in different networks and/or at different ranks or units) realize a
similar motif at a higher, semantic level. In a topological model
(see Martin and Matthiessen 1991) meanings are not organized system-
ically (or ‘typologically’), in networks of distinctive valeurs, but in terms
of their likeness along one or more dimensions: phenomena that
are similar in one or more respects are conceived of as areas that are
closer to one another and as belonging to the same domain in a larger
multidimensional space.2 The method of (re)grouping lexicogramma-
tically distinct phenomena into more abstract domains or components
has often been used in SFL in addition to modelling meanings in
networks.3 Two topological semantic concepts that are familiar in
SFL are the metafunctions and the motif of logico-semantic relations.
The metafunctions are first and foremost ‘semantic components’ of
language (e.g. Halliday 1977), and they are groupings of phenomena
which are dispersed over different networks at the lexicogrammatical
stratum, for instance, networks for different ranks (with, e.g., idea-
tional meanings realized in clauses, in clause complexes, and in
groups). Logico-semantic relations, i.e. expansions and projections
2
In this respect the topological modelling of (higher-order) meanings is similar to the method of semantic maps which is
used in functional typology (e.g. Haspelmath 2003).
3
Note that the ‘regrouping’ of distinct lexicogrammatical phenomena into topological ‘domains’ does not necessarily
imply that those domains are conceived as pertaining to a different (i.e. usually ‘higher’) stratum. Martin and
Matthiessen (1991) talked about regroupings within a stratum, with a focus on the lexicogrammar. This has been
interpreted by Halliday (1996:15) as suggesting a view of lexicogrammar as typologically organized and semantics as
topologically organized strata. This is also the view highlighted here, although it should be borne in mind that the
topological/typological distinction does not correlate with semantics/lexicogrammar per se (indeed, it perturbates
throughout the system as a fractal motif – see Section 3.4.2 below on fractality and 3.4.3 on the related concept of an
extravagant theory of language).
and their subtypes, can also be realized by different lexicogrammatical

means, in different units (see Halliday’s (1985) synoptic table of
‘expansion’, and Halliday and Matthiessen’s (1999) more extended
treatment of the various manifestations of ‘expansion’ and
‘projection’ across lexicogrammatical means).
(b) Semantics as discourse-structural meaning: In a second type of view, seman-
tics is the stratum at which patternings at the level of discourse are
modelled as patterns of unfolding text which are larger than the struc-
tures recognized at the level of lexicogrammar, where the maximal
unit, in terms of size, is the clause complex. Thus in this sense, the SFL
concept of cohesion is an intrinsically semantic concept (see Halliday
and Hasan 1976:4), and by extension, since it is cohesive relations that
create texture and hence make a text, a text is by definition a semantic
unit. Note that ‘discourse-structural’ meaning is here intended as a type
of structure which is different from lexicogrammatical structure, the
latter being structure in the traditional, narrower sense. In this vein,
too, cohesion is often characterized as a non-structural phenomenon.
(c) Semantics as higher-level systemic meaning: In a third conception of seman-
tics, the semantic stratum itself is organized in terms of system networks
that are superimposed upon lower lexicogrammatical networks. In this
view, options in semantic networks are realized by options in lexico-
grammatical networks. This conception underlies the model of speech
functions in SFL (see Halliday 1984). speech function is a semantic
system with options such as ‘command’ (asking for goods and services)
or ‘question’ (asking for or about information). Those semantic options
can be realized by (various) lexicogrammatical options from the system
of mood: e.g. a ‘command’ can be realized as an imperative (Open the
window), or as an interrogative (Could you open the window for me?).
Figure 3.2 summarizes those three design alternatives of a semantic

stratum in the overall architecture of SFL. These alternatives are not mutu-
ally exclusive but complementary. In much recent work on semantics in
SFL, the organization of the semantics in terms of system networks has been
an important objective, but in many cases this systemic modelling went
hand in hand with and/or was inspired by insights from topological and
discourse-structural views of semantics, as we will see below (Section 3.4).
3.2.3 A Closer Look at Stratification: The WHY of Semantics

The answer to the first why question – Why is it necessary to recognize a
semantic stratum? – is that a semantic stratum is needed in order to account
for ‘variability’ between expression and content functioning in different
contexts. Recognizing a semantic stratum is necessary when there is no
one-to-one relationship between expression and content, i.e. when a
coupling between a form/structure, on the one hand, and a meaning, on the
Semantics 61
context
semantics
[1] [2] [3]
lexicogrammar
Figure 3.2 Three design alternatives in modelling a semantic stratum: topological [1],
systemic/typological [2], and discourse-structural [3]
other hand, is not enough. For some systemicists or from one perspective, this
is always the case, for others it is the case for specific contexts (as we will see
below). In order to come to an understanding of how the recognition of a
stratum of semantics is motivated in relation to this concept of variability, it
is necessary to first clarify what exactly this variability is, and how different
research purposes can put this variability into different types of perspectives.
3.2.3.1 Variability and the Internal Stratification of the Content Plane

Theoretically, if there is no one-to-one relation between content and expres-
sion, there are two types of situations which are possible, and these two can
also occur in combination. This is visualized in Figure 3.3. At the left
(Situation [1]) is the type of situation that is familiar from the systemic
model of the lexicogrammar, with systems in which each valeur (or end-
point in a systemic path) is tied to a specific expression form, i.e. there is a
one-to-one relation between meaning and form. In the centre are two
complementary possibilities of variation: one-to-many, where one meaning
is dispersed over different expression forms (Situation [2]), or many-to-one,
where one form can be the expression of different types of meanings
(Situation [3]). At the right is the combination of Situations [2] and [3].4
4
Lamb (1962) (whose stratificational theory has been a source of inspiration in SFL) refers to those different situations
of variability between strata as ‘composite realization’ (=[2] here), ‘portmanteau realization’ (=[3]), and ‘interlocking
diversification’ (=[4]).
meanings
[1] [2] [3] [4]
f
forms
Figure 3.3 One-to-many [2], many-to-one [3], and many-to-many [4] conceptions of
variability between meanings and forms in relation to the default view of one-to-one
meaning-to-form couplings within lexicogrammar [1]
In research contexts where a semantic stratum comes into play (to be

explored in Section 3.5), the analysis and description of the meaning of a
linguistic sign as a content that is tied to an expression in a (lexicogramma-
tical) system (Situation [1] in Figure 3.3) is not sufficient or adequate. In
those cases, it is useful to ‘un-couple’ content and expression, and to
recognize an additional, higher level of content which is not covered in
the content in the lexicogrammatical system.5 In more general terms, one
specification of a content linked to its expression is not enough: the content
plane of language is split into two strata, i.e. lexicogrammar and semantics.
This is referred to by Halliday as an ‘internal stratification of the content
plane’ (Halliday 1976, 1998a; also see Taverniers 2011 for a further explor-
ation). The additional interface in the content plane, i.e. between the strata
of semantics and lexicogrammar, is set up in order to account for variability
between contents and expressions.
3.2.3.2 Semantics as an Interface and Trinocular Perspectives on

Variability
The description of the internal stratification of the content plane above
focuses on the relation between lexicogrammar and semantics as two
content strata of language. However, semantics is not only inserted above
lexicogrammar, but also ‘below context’: it is a stratum that is wedged
between lexicogrammar with its systemic model of linguistic forms and
what lies outside language, i.e. the non-linguistic context. In this sense, the
two content strata are interface strata, and semantics is an interface
stratum through which language interacts with extra-linguistic context,
i.e. with eco-social environment (see Halliday 2013). We saw above that the
stratification of the content plane into semantics and lexicogrammar
makes it possible to account for variability between contents and expres-
sions. This variability is an inherent feature of language functioning in
different contexts: it is in the interaction with different contexts, and in
the actualization of language in specific instances, that variation arises.
5
In Hjelmslev’s (1963) terms, a ‘connotative semiotic’ is then recognized (i.e. a semiotic that has a higher-order content
plane) (see Taverniers 2008).
Semantics 63
More specifically, stratification makes it possible to create ‘meanings’ that

are adapted to specific contexts and that are beyond what is (fixed or
‘codified’) in the (lexicogrammatical) system of a language. Stratification
makes it possible to use ‘forms’ in ways that go beyond their valeurs in
lexicogrammatical systems, for instance, to mean several things at the
same time, i.e. to be creative in a myriad of ways with the finite means
that are available in the formal units of the language.
Thus the concept of stratification is how SFL theorizes the way in which
language interacts with non-linguistic context. Language appears in its
material expression (i.e. in sounds and writing), but as a ‘semiotic’ system
it does not interact with the eco-social environment ‘directly’ through
sounds and writings: there are intermediate layers of content between
linguistic expressions and the contexts in which language functions. First,
there has to be a level of ‘meaning’ (there is a separation between ‘content’
and ‘expression’, which is the principle of double articulation); and second,
in order to account for variability, there have to be two content strata (this is
the separation between ‘semantics’ and ‘lexicogrammar’ within the content
plane) – and this is a hypothesis about how language functions in ‘context’.
In keeping with the visual metaphor of strata as organized with context at
the top and the expression in sound or writing at the bottom, any type of
phenomenon can be viewed from three perspectives. Focusing on semantics,
this stratum can thus be seen ‘from above’ (from context), ‘from below’ (from
lexicogrammar), and ‘from roundabout’ (from its own position in the strati-
fied model). The possibility of those complementary perspectives is referred
to as ‘trinocular’ vision or perspective (Halliday 1977, 1996), or ‘trinocularity’
(Halliday 2009:79–80). Different types of questions arise about what a seman-
tics can be, depending on the type of stratal perspective that is taken on
semantics as an interface. This type of perspective, together with the type
of variability between lexicogrammar and semantics that is focused on,
determines the more specific reasons for setting up a semantics within the
architecture of SFL, and also how this semantics can be conceived of.
3.2.3.3 A Specific Motivation for Setting Up a Semantics:

Grammatical Metaphor
The need for a separate semantic stratum came into sight in various
research perspectives in SFL in which (a type of ) variability between content
and expression came to be highlighted. One of those, which has been highly
influential in the theory as a whole, is the area of grammatical metaphor.6
6
Other areas in relatively early SFL studies which point to a semantic stratum in addition to a lexicogrammatical one
include the following: stylistics and socio-semantic variation (esp. with the notion of a de-automatization of a grammar à
la Mukařovský, e.g. Halliday 1982); language development (ontogenesis) (with the view that the adult language system
contains more strata – an extra content stratum? – than the proto-language of the child); and mood (with variability
between mood choices and socio-semantic roles, on the one hand, and between mood types and lexicogrammatical
realizations, on the other hand).
The concept of grammatical metaphor instrinsically has to do with the

lack of a one-to-one relationship between meaning and form, and it is
already present, in a pre-theoretical way, in earlier concepts such as mark-
edness and foregrounding, as well as congruent (typical) and incongruent
(atypical) links between semantics and lexicogrammar.
Grammatical metaphor becomes the more general label for what had
already been partly recognized in the interpersonal component, on the one
hand (where one speech function can be realized in different ways), and
similar types of variation in the ideational component, on the other hand.
Halliday (1984) calls the typical, default links between speech functions and
their realizations in lexicogrammar (through the system of mood) ‘congru-
ent’, and the less typical ‘alternative’ realizations ‘metaphorical’. In this
sense a command can be realized congruently by an imperative (Please, open
the window for me), or metaphorically by a modalized interrogative (Could you
open the window?). In Halliday’s (1984) first presentation of ‘grammatical
metaphor’, such incongruent realizations of speech functions are called
‘metaphors of mood’. Later, in Halliday (1985), within the interpersonal
component, a second type of metaphor is recognized, viz. ‘metaphors of
modality’, in which a modal meaning is realized, not by means of a modal
element in the clause structure (i.e. a modal verb or a modal adjunct: It will
(probably) be a good season) – which is regarded as the congruent realization
of modal meanings, but rather by other means which are beyond the clause
structure and which give a more ‘explicit’ wording of the modality
intended (as in I think it will be a good season; I suppose it will be good season;
It is obvious that it will be a good season; Everyone says it will be a good season).
Figure 3.4 presents a visual presentation of interpersonal metaphors in
terms of the variability between content and expression as sketched above.
In presenting the concept of grammatical metaphor, Halliday (1985) expli-
citly combines two complementary views of variability between forms and
meanings. He describes the traditional concept of what is ‘metaphorical’ as
highlighting a ‘many-to-one’ type of view of the relation between meaning
‘command’ ‘probability’
projecting
modalized relational clause
interrogative It is obvious that …
Could you …
modalized projecting
imperative* declarative modal elements mental clause
You should … will, probably* I think, I suppose …
*congruent realization *congruent realization
Figure 3.4 Interpersonal grammatical metaphors of mood (left) and modality (right)
theorized within a view of variation of one-to-many
Semantics 65
‘part of a
piece of furniture’ ‘event’
‘part of an
animal’*
leg expressed
expressed as nominal
as clause*
*literal meaning *congruent realization
Figure 3.5 The traditional view of metaphor (left), and an alternative view which is
equally inspiring in the study of grammatical metaphor (right)
and expression: there is one form, and this form has multiple meanings,
among which are at least one literal and at least one metaphorical interpret-
ation. This is the conception that prevails in thinking about metaphor in
general, and especially in traditional views of lexical metaphor (e.g. legs can
be ‘parts of animals’, i.e. have a literal meaning, or ‘parts of furniture’, as in
the lexical metaphor table legs). However, Halliday argues, the complemen-
tary view, i.e. of ‘one-to-many’, is equally important in analyzing grammat-
ical metaphor (see Figure 3.5, which shows the relation between those two
views). Thus in introducing ‘ideational grammatical metaphor’, Halliday
(1985) gives an example of a clause which realizes a process with its partici-
pants (i.e. a processual or ‘event’ meaning), and then shows that this pro-
cessual meaning could be realized by a nominal group, which functions as a
participant in another process configuration. The first example offered by
Halliday concerns the congruent wording Mary saw something wonderful,
compared to the incongruent A wonderful sight met Mary’s eyes or Mary came
upon a wonderful sight, where the process of ‘Mary seeing something wonder-
ful’ is construed as a nominal group, a wonderful sight.
In explaining grammatical metaphor as ‘one-to-many’ in addition to
‘many-to-one’, Halliday both connects grammatical metaphor with the
traditional concept of metaphoricity (‘many-to-one’) and brings the new
concept of grammatical metaphor more in line with what had earlier been
recognized in relation to typicality in the interpersonal metafunction
(where the ‘one-to-many’ view was the initial source of inspiration:
one speech function such as ‘command’ can have several realizations such
as ‘imperative’, ‘interrogative’, etc.). In terms of the stratal type of perspec-
tive, the approach to grammatical metaphor that is initially taken is that
from below: the starting point is a form, and the question is what this
form ‘means’, and then this ‘meaning’ is analyzed as having other
(alternative, metaphorical) ways of being realized. Section 3.5.1 explains

how flipping the traditional conception of metaphor, starting from above
and asking how one meaning can be realized, has led to a more comprehen-
sive investigation of grammatical metaphor.
3.3 Wider Perspective I, Flexibility: Enter Metaredundancy

and Probability
The overview of initial conceptions of ‘semantics’ in SFL has shown that

there is no single explanation of what (a) semantics is, where it is to be
found, how it can be modelled, if it is regarded as a separate stratum, and
why it should be recognized as such. Even in those cases where the focus
was on a (separate) semantic stratum, the description was couched in terms
of ‘certain contexts’, or ‘certain research purposes’ for which such a seman-
tics comes into view. Indeed, even a cursory reading of the systemic func-
tional literature will reveal that the familiar systems of the clause, viz.
process type, mood and modality, and theme and information, are
sometimes called ‘lexicogrammatical systems’, and sometimes ‘semantic
systems’. This (seeming) ‘indeterminacy’ in the description has to do with
the very conception of the interface between semantics and lexicogrammar
as by definition inherently flexible. In order to understand this flexibility, it
is necessary to put the notion of stratification in a wider perspective, and to
take into account another relation that has hitherto not been mentioned,
viz. the relation between the system and actual instantiations, or what is
called ‘instantiation’ in SFL. This wider perspective is that of viewing
language as a dynamic open system, and bringing in the concepts of
‘metaredundancy’ and ‘probability’.
3.3.1 Metastability and Metaredundancy

Since the mid-1980s, and under the influence of work by Lemke (1984),
Halliday has come to view language as a ‘dynamic open system’, or a
‘metastable system’. Such a system persists in interacting with different
contexts by being open, i.e. being maximally adaptable, and thus by con-
stantly changing, while at the same time avoiding change (i.e. it remains
stable at a meta-level). We saw above that the concept of stratification is
how SFL interprets the way in which language interacts with context or its
eco-social environment. The model of stratification also provides the ideal
basis to talk about how language can be metastable, but the concept of
dynamic openness puts the idea of stratification into a wider perspective.
As a semiotic system, language is able to continue to function because of
the combination of two features: the availability of ‘critical contrasts at
every level’, and ‘complex arrangements of articulation of these contrasts’
Semantics 67
(Lemke 1992), i.e. complex arrangements of the relations between those

options in the different strata. Thus, language achieves metastability
through the particular type of relationship between its coding layers or
strata. Those layers consist of things going together in a predictable way
(to limit change and thus ensure stability), while at the same time they
contain some ‘gaps and contradictions’ which provide room for innov-
ations, which originate as adaptations to changes (i.e. new requirements)
in the context (Lemke 1995:175).
Focusing on the relation between semantics and lexicogrammar, the
‘things going together in a predictable way’ are stable meaning–form rela-
tions (such as one-to-one relations), while the ‘gaps’ can be meanings that
do not have one default realization (meanings that are not (yet) codified)
and the ‘contradictions’ meanings that (come to) have multiple different
realizations, or the other way around. Thus, the notion of complex arrange-
ments between options in the different strata corresponds to the presence
of various types of variability that we distinguished above. This type of
relation between layers, based on such complex arrangements of variability,
is what characterizes ‘semiotic’ dynamic open systems, of which language is
the paramount example, and is called ‘metaredundancy’. In a pair of layers
with only one-to-one meaning–expression couplings, the layers are ‘redun-
dant’ vis-à-vis each other (there is no need to differentiate them, since the
one is always completely predicted by the other). When gaps and contradic-
tions are present, the relationship is said to be ‘metaredundant’: in the
system as a whole, there is redundancy to a certain/minimal extent, and
there is room for new connections between layers, i.e. there is room for
creativity or adaptations to new contexts. Innovations arise on top of and by
virtue of the existing system, and it is also in this sense that the extra layer
needed to model them is metaredundant with the existing system.
Within the perspective of language as a dynamic open system, the
concept of metaredundancy provides a new interpretation of the relation-
ship between strata, which is traditionally called ‘realization’ in SFL (see
Halliday 1991; Lemke 1995). In this perspective, the verb realize, too, is
reinterpreted as ‘redound’. What metaredundancy adds to realization,
besides the notion that it is an explicitly dynamic concept, is the fact that
it is a relationship of accumulative nesting, i.e. it is not sequential but,
rather, bidirectional. Thus, lexicogrammar redounds with phonology/
graphology; semantics redounds with the relation between lexicogrammar
and phonology/graphology; and context redounds with the relation
between semantics which redounds with the relation between lexicogram-
mar and phonology/graphology. This is the view that is most relevant when
focusing on the content side of language, i.e. ‘meaning’. The visual image of
the strata with the cotangential circles (which appeared around the time
that the theory of dynamic open systems became influential in SFL) ranging
from smaller (for the lower systems) to larger (for the higher systems)
represents exactly this view. However, while realization imposes a direction
xt
te
on
c
s
tic
an
m
xt
se
te
on
c
s
tic
an
m
se
s
tic
an
m
se
s
tic ar
an m
m am
gr
se
mar
ico
am
lex
gr
ico
lex
m ar
am ar
gr m
am
ico
gr
lex
ico
ph
ra
/g
lex
phon
ph
ra
/g
phon
Figure 3.6 Metaredundancy relationships between strata, starting from the top focusing
on the content side of language (left), and starting from the bottom focusing on the
expression side of language (right)
on the relation (i.e. it forces us to choose one thing that realizes the other,
as with the terms ‘signifier’ and ‘signified’), metaredundancy is neutral in
this respect. Accordingly, the opposite view with metaredundancy cycles
starting from the lowest stratum becomes a complementary frame of
interpretation. Here, phonology redounds with the metaredundancy
between lexicogrammar and semantics, and if context is also taken into
account, phonology redounds with the metaredundancy between lexico-
grammar and the metaredundancy between semantics and context. This
view, as Halliday (1992) argues, is the most relevant in research focusing on
phonology. Figure 3.6 is an attempt to capture accumulative metaredun-
dancy cycles in a three-dimensional way.
The conception of language as a dynamic open system puts the motiv-
ation for distinguishing between lexicogrammar and semantics into a
wider perspective. It is because of the requirement of adaptability of the
system to different contexts that the distinction between lexicogrammar
and semantics is needed, because it is exactly through the flexible relation
between lexicogrammar and semantics that language can be open, and that
gaps and contradictions, as room for innovation, can be allowed. In other
words, it is through variability between meanings and forms that language
can continue to function. It is through the specific type of relationship of
semantics, not just to lexicogrammar, but to the rest of the lower layers,
that language has semogenic power, that it is a ‘meaning potential’.
Semantics 69
In addition to the new perspective on the distinction between semantics

and lexicogrammar, the conception of language as a dynamic open system
also provides a wider frame of interpretation for the observation (see Section
3.2.1) that different conceptions of ‘meaning’ are possible and relevant at the
same time in SFL. The four different conceptions of ‘meaning’ distinguished
above have a place in the model of language as a dynamic open system:
• ‘Meaning’ as the stratum of semantics itself: this can now be viewed in

various ways, at least two of which are relevant here. (i) Within the
content plane of language, it is the stratum that redounds with lexico-
grammar, and focusing on content, the interface with lexicogrammar is
the place where variability between meanings and forms is made possible.
(ii) In language (content + expression), semantics is the highest stratum,
which metaredounds with the rest of the language system. This view of
semantics is proposed in SFL in connection with Hjelmslev’s notion of a
connotative semiotic (e.g. Halliday 1991). Here, semantics is seen as a
connotative layer which interacts with an existing semiotic (in Hjelmslev’s
terminology, the latter is a denotative system and functions as the expres-
sion of the connotative content). This is the idea of a new stratum coming
into existence on top of and by virtue of the existing system.
• ‘Meaning’ as the content plane of language: this is what metaredounds
with the expression side, with phonology.
• ‘Meaning’ as language as a whole: in this view language is interpreted as
a meaning potential (a potential that allows us to function in different
contexts): this is language metaredounding with context.
• ‘Meaning’ as the valeurs in system networks: this view is the least clear in
the visualizations with layers. However, if the concept of language as a
dynamic open system is carried through, this, too, is to be seen as a level
that redounds with the structures. For the valeurs are only an objectifi-
cation of language by the researcher. The relationship between the
valeurs and the structures is a relationship with much stability (one-to-
one relations), but here too, there is room for innovation in order for
language to persist as a metastable system.
These different conceptions of ‘meaning’ can be highlighted in the frame-

work of language as a dynamic open system. The validity of all those views
within SFL creates what can be seen as ‘indeterminacy’, as in, for example,
alternative characterizations of the familiar networks of mood, process
type, and theme as ‘semantic’ in some works, and as ‘lexicogrammatical’
in others. What counts is that both of these ‘content’ strata are meaning
(as a verb, i.e. meaning-making). This is explicitly captured in the idea of
metaredundancy: semantics metaredounds with the meaning that is already
in the relationship between lexicogrammar and phonology. It is in this sense,
too, that grammatical metaphor is theorized as ‘stratal tension’ (e.g. Martin
1992) (i.e. tension between a literal and a metaphorical meaning), because
both the lexicogrammar and the semantics convey meaning.
3.3.2 Time, Instantiation, and Probabilities

In discussing meaning and semantics in the framework of language as a
dynamic open system, so far in Section 3.3 the focus has been on the status
of strata and the relations between them, prioritizing the relation of real-
ization and its reinterpretation as metaredundancy. We have seen how the
concept of metaredundancy places the views of ‘meaning’ distinguished
above in a different theoretical perspective: both the types of ‘meaning’ that
are distinguished and the idea of variability between meanings and forms
are elucidated through metaredundancy. As a next step in exploiting this
theoretical frame of interpretation, a further dimension needs to be taken
into account, viz. time and the concomitant notion that is called ‘instanti-
ation’ in SFL. We will see that the same framework also provides a deeper
understanding of the kinds of ‘semantics’ that first appeared at the horizon
of SFL’s ventures into semantic theory (see Section 3.2.3.3).
One dimension that plays a crucial role in the picture of language as a
dynamic open system is that of ‘time’, and it will be noted that the descrip-
tion above is inevitably couched in terms of temporal meanings (witness
expressions such as ‘persist’, ‘constant’, ‘change’, ‘new’, ‘always’, ‘at the
same time’). Metastability is an inherently dynamic concept: a system that
is metastable is stable by ‘constantly changing’. Hence metastability is in
fact ‘dynamic stability’, although this term may seem a contradiction.
This apparent contradiction is due to the role of time, or rather, the way
in which, as researchers, we observe and objectify language, which only
occurs to us in its usage in unfolding time. What appears as a difference
between two states of language that are objectified for the sake of research
(i.e. two time slices across which language does not appear as ‘stable’) may
appear as actualizations of the same meaning potential if observed from a
further perspective, from a greater time depth. Vice versa, what is objecti-
fied as a stable system (in research, and also for the purposes of learning
a language as a technique) is just that: an objectification (abstraction) of a
system that is constantly in flux in a myriad of actualizations in different
situations. This is the paradox of dynamic stability.
As a meaning potential (which is virtually stable and thus learnable),
language predicts what forms and meanings can be actualized in a certain
context, and at the same time it is responsive to new requirements in those
contexts, because it contains gaps and contradictions. Hence language is
constantly open to allow and incorporate what appear as ‘innovations’
within its system, but what are in fact nothing more than actualizations
of the system in interaction with different environments. In relation to the
dimension of ‘time’, those innovations – which may or may not become
part of the stable system over time – can be conceived of in different ways,
depending on whether or not they are incorporated in the system and
depending on the research perspective that is taken. The research perspec-
tive refers to the time depth that is taken, i.e. what time frame of language
is made the object of study: language as a codified system or language in
Semantics 71
vivo in one particular instantiation (a text) or a particular setting. Innov-

ations can be conceived of as transformations of the system, as changes, as
deviations, as metaphors, as ‘complexifications’ of the system (adding a
further choice in the existing system, i.e. a more fine-grained distinction).
It will be noted that the different types of motivations for a separate
semantics that we have considered exemplify some of those different con-
ceptions of ‘innovations’.
In relation to the notion of ‘variability’ (between meanings and forms),
innovations can also be conceived of in different ways, which are complex
to describe because the dimension of time (and the research focus in terms
of time depth) also plays an intrinsic role. ‘Change’ can be based on diversi-
fication (from one-to-one to one-to-many or many-to-one), or it can be a
change in the existing proportionalities between options, e.g. options
may become more or less ‘at risk’ in certain environments, while other
options die out and new options appear. In essence, all change has to
do with changing probabilities in a system (see Halliday 1991). Any innov-
ation in language can be purely ‘accidental’, it can become part of a
situation-specific system (e.g. a register), or it can become part of the overall
system. In this sense, each instance potentially has an influence on
the system. Halliday (e.g. 1985) often uses a comparison with the weather.
Each weather situation is an instance of a climate, but also influences that
climate: when there are more wet winter days, the winter climate will
become wetter too. A change that becomes part of a system needs to be
incorporated into existing networks. If the change is based on diversifica-
tion, this leads to a complexification of the system wherein more fine-
grained distinctions will have to be set up.
When language changes (i.e. evolves), it is the couplings between layers
that change, or more precisely the relations of variability between strata:
relations are opened up and new relations appear, or, in other words, strata
are uncoupled and recoupled again. Grammatical metaphor, which has
been a key reason in SFL for distinguishing between the strata of semantics
and lexicogrammar, is conceived of in exactly those terms: it is seen as
‘a cross-coupling (decoupling, and recoupling in a different alignment)
between the semantics and the lexicogrammar)’ (Halliday 2008:16). Note
also how ‘metaphor’ in general is seen as the creation of new form-meaning
couplings which come into being on top of and by virtue of existing
couplings.
3.4 Wider Perspective II, a Design Rationale Based on

Multiperspectivism and Fractality: Enter Extravagance
Throughout the step-by-step characterization of ‘meaning’ and ‘semantics’

in SFL in this chapter, there has been a prevailing emphasis on ‘different
views’, ‘different possible conceptions’, and ‘perspectives’. This feature has
been explicitly thematized in SFL, at a meta-level, i.e. at the level of how we

think about how we can model language, or the level of our research
practice as linguists. Because it plays such an important role in understand-
ing how semantics is conceived of in SFL, it is useful to consider how this
idea of different views is given shape at the meta-level. There are two
interacting aspects to this feature, viz. ‘multiperspectivism’ (Section 3.4.1)
and ‘fractality’ (Section 3.4.2). The combination of these leads to a specific
type of theory that is called an ‘extravagant’ theory (Section 3.4.3).
3.4.1 Multiperspectivism
An inherent feature of SFL is that it is ‘multiperspectival’. Its theory of
language is based on a number of dimensions along which differentiations
can be made (see Taverniers 2002 for a further exploration of ‘differentiat-
ing dimensions’), the most important of which are the distinction between
metafunctions, between different strata (stratification, as well as the rela-
tionship of realization or metaredundancy), between ranks, between
system and instance (instantiation), and between syntagmatic and paradig-
matic modelling (see Halliday and Webster 2009). Within each of those
dimensions, a linguistic phenomenon can be looked upon in different ways,
and different perspectives are possible, depending on the viewpoint and the
focal depth:
• Perspectives can differ in terms of directions, when views from different

vantage points are possible. The stratal types of perspectives mentioned
above are a case in point: any phenomenon viewed in terms of
stratification can be looked upon from below, from above, or from
roundabout. Hence trinocularity is just one instance of this feature of
multiperspectivism.
• Perspectives can also differ in terms of focus. The investigator can single
out one focus point in a set of complementary views (e.g. look at a
phenomenon from one specific metafunction). Another example of dif-
ferences in focus is that where focal length differs in terms of time
depth, which is captured in the notion of ‘instantiation’ in SFL.
Any linguistic phenomenon can thus be looked at from different perspec-

tives, within one dimension, and also in terms of different dimensions, and
this leads to a theory that is explicitly based on complementarities. The
combination of complementarities leads to a richer theory about a complex
phenomenon, and it is through ‘shunting’ perspectives (see Halliday
1961:254) that the analyst can come to a better understanding and a more
detailed description of a phenomenon.
Grammatical metaphor can again be mentioned as a case in point. We
saw above how Halliday called for a shift in the perspective that is taken
on the variability between lexicogrammar and semantics that is inherent
in metaphor: a shift from the traditional view of ‘one expression has
Semantics 73
context
semantics
‘sequence
‘event’ ‘entity’
of events’
* * *
clause clause (nominal) group word

complex functioning functioning
in a clause in a group
*congruent realization
lexicogrammar
y/graphology
phonolog
Figure 3.7 Shunting perspectives in theorizing grammatical metaphor as a tension
between lexicogrammar and semantics
multiple meanings’ (a metaphorical and a literal one) to a view in which

one meaning can be expressed in various ways. By shunting the perspec-
tive in this way (see Figure 3.7 and compare Figure 3.5 above), and by
looking at metaphor from the semantics, SFL has been able to develop a
more comprehensive theory of ideational metaphor. This development
went in various steps, each further step adding more conceptual depth
to the theory of grammatical metaphor – but all steps hinge on the flipped
perspective compared to the more traditional view of ‘metaphoricity’.
First different types of realizations were conceived of in terms of degrees
of congruence (i.e. one meaning, such as an event meaning, can be
expressed in various ways, from more to less metaphorical). Concomitant
with this, the view developed that an ideational metaphor can be
‘unpacked’ in different steps, from relatively incongruent to more congru-
ent ways of wording the same meaning. Then the different realizations
were put in a type of implicational hierarchy, from most ‘relational’ or
context
semantics
ideational interpersonal textual
meaning meaning meaning
combined syntagm
lexicogrammar
y/graphology
phonolog
Figure 3.8 Shunting perspectives in theorizing the metafunctions as semantic
components of the system of language
‘processual’ meanings to condensed meanings with the noun as the end

point (see Halliday 1998a:211). This in turn laid bare different possible
points in a general ‘drift towards thinginess’ that characterizes those
registers which often make use of packaging meanings in a condensed
form, such as scientific discourse.
Another example of taking different perspectives is related to the concept
of the metafunctions. The metafunctions can be viewed as semantic zones,
which are realized in different lexicogrammatical areas. Alternatively, each
syntagm, i.e. each structure at the level of lexicogrammar, has all three
metafunctional meanings. This is a familiar SFL view of the multi-tiered
syntagm as a concerted meaning, i.e. a fusion of layers of metafunctional
meanings. Figure 3.8 shows how these two views on the metafunctions are
complementary, and it is clear that both are needed in order to come to a
deeper, richer understanding of the notion of the metafunctions as
Semantics 75
‘semantic components of the system’ (while at the same time (i.e. meta-
redundancy at play!) depending critically on the bundling of systems at the
lexicogrammatical stratum).
3.4.2 Fractality
By combining complementarities in order to come to a richer theory of the
complex phenomenon that language is, sometimes recurrent patterns are
discovered, and this feature is referred to in SFL as ‘fractality’ or ‘fractal
resonance’ (Martin 1995). Thus, fractality is an additional design feature
which is exploited at the meta-level, the level of the linguistic practice itself.
Put simply, a pattern that appears in one view, in one focus, at one level –
i.e. a type of modelling that is useful in accounting for a phenomenon
within that view – may also be recognized elsewhere, in another view that
appears by shunting the perspective. The newly recognized pattern is then
said to ‘resonate’ with the first one. Focusing on the meta-level, resonance
means that a model, a type of distinction, which works well for one area,
one view, or one dimension, is used as an inspiration to explore a different
area, view, or dimension of language.
The resonance may be between different ranks or units. This is the case
with the semantic motifs of expansion and projection (see Section 3.2.2).
A meaning such as causality (as a subtype of expansion: enhancement) can
be realized between clauses within a clause complex (This happened because
that happened.), within clauses through relational processes (This caused
that.), and within groups (The cause of that was . . .) (see Halliday and Mat-
thiessen 1999:222–6 on this and related types of fractal resonance). Such
resonating lexicogrammatical phenomena can be grouped together into a
topological domain, or a ‘(transcategorical) semantic domain’ (see the
notion of semantic domain in Halliday and Matthiessen 2004:593–4).
The resonance may also be inter-stratal. This kind of parallelism is
explored when the model of lexicogrammar, which is relatively familiar,
is used as an inspiration to tackle the stratum of semantics. This type of
inter-stratal fractality, which ties in well with the notion of metaredun-
dancy between levels of the system, will become clearer in Section 3.5,
which focuses on recent semantic models in SFL.
3.4.3 Extravagance
We have seen how different conceptions of semantics are not just possible
alternative views (see Sections 3.2.1 and 3.2.2), but exist side by side and
moreover, all appear, naturally, in a view of language as a dynamic
open system based on metaredundancy relations between its layers (see
Section 3.3.1). What has now been added, in Section 3.4, is the idea that this
varied view of semantics is ‘fostered’ in SFL. A theory which is based on
complementarities and fractal patterns resonating across its components, is
called an ‘extravagant’ theory by Halliday (e.g. 1998b), and it is argued that

linguistic theory as a meta-semiotic can only be extravagant, because lan-
guage itself is extravagant (see Halliday 2008). Language itself is extrava-
gant by having order while at the same time constantly allowing disorder in
its adaptation to different contexts. The linguistic theory that attempts to
understand language in all its facets therefore cannot but be extravagant as
well. Thus, the meta-semiotic is itself a dynamic open system, with some
things that have become conventionalized, but with gaps (evidently, and
luckily!) and, importantly, contradictions (alternatives, complementarities).
And note that drawing the parallel between language and linguistics (see
Halliday 1998b) is pointing to a fractal resonance between the two.
3.5 Recent Semantic Endeavours in SFL
3.5.1 Mapping Semantic Models in SFL

The aim of this section is to give a brief but systematic overview of semantic
models that have developed in SFL, most of them since the 1990s.
The individual models will not be presented in detail since the focus will
be on elucidating the nature of those models against the explanation of
‘meaning’ and ‘semantics’ that is sketched above, and in doing so, on
highlighting the specific contribution and hence ‘place’ of each semantic
model in relation to the theory as a whole. More specifically, this section
will show how different semantic models in SFL each flesh out a specific
dimension of a semantic stratum, i.e. a facet of semantics which comes into
view by taking a specific perspective. The perspectives that are taken on
semantics can be explained as originating from choices that are made in
terms of the complementarities/options that have been disentangled in the
preceding sections, viz.
• the way in which semantics is modelled in addition to the aim of

networking (i.e. topologically or discourse-structurally, or both) (see
Section 3.2.2);
• the role and the type of variability between semantics and lexicogram-
mar (see Section 3.2.3.1);
• the type of stratal perspective that is taken (see Section 3.2.3.2).
We will draw upon the various types of complementarities distinguished

above in exploring how recent semantic models are designed in SFL. In
doing so, we will come to a theoretically grounded understanding of recent
conceptions of semantics in SFL, of what a semantics can be, and of how it
can be modelled. In doing so, further on, we will once more re-visit the
crucial concept of stratification, and further flesh it out on the basis of the
specific models of semantics that will be under focus.
In turning to recent approaches to semantics in SFL, there is one ten-
dency that becomes clear: semantics has become ever more important in
Semantics 77
the theory as a whole (see Butler and Taverniers 2008). This is in vein with a
key aim in SFL since its inception, viz. to make a theory that prioritizes
meaning, to set up a system that is as semantic as possible (a ‘semanticky’
grammar, see Halliday and Matthiessen 2014:31), or a grammar that is
‘pushed’ as far as possible towards the semantics (see Halliday 1985:xix),
but the ‘semanticization’ of the model lies more specifically in the fact that
the modelling of the ‘semantic stratum’ has become an important research
aim across different areas and metafunctions.
In the later, more fully-fledged semantic endeavours in SFL, there is
continuity and also a major new research focus. There is continuity from
the initial motivations for separating semantics and lexicogrammar in that
central themes have remained important to the present day: grammatical
metaphor, and the semantics of speech function (as seen in Hasan’s model
of message semantics, see Section 3.5.3). Socio-semantic variation and a
focus on text also remain important. At the same time, it is possible to
recognize an oscillation in a specific approach to semantics, which is inher-
ently in accordance with the overall architecture of SFL, viz. the modelling
of (separate) semantic ‘networks’ which then interact with the lower net-
works in the lexicogrammar. This had already started in earlier work on the
interpersonal metafunction focusing on the relation between speech func-
tion and mood. In more recent models this approach is generalized to other
interpersonal domains (e.g. stance and sourcing of stances), to the idea-
tional metafunction and to the analysis of discourse. Hence, what we see in
the design of the stratum of semantics is fractality, with the method of
system networks being fractally extended from the lexicogrammar to
the semantics (see Matthiessen 2009:14–15).
Beyond this strong ‘systemic’ motif, the other two types of modelling are
also important sources of inspiration, viz. ‘topological’ and ‘discourse-
structural’ designs. As indicated above, those three types of designs are
complementary, and although the ultimate aim may be a systemic model
(as in recent semantic models in SFL), the topological and discourse-
structural views of semantics may be used as a primary source of inspir-
ation to disentangle ‘meanings’ which can then be ‘networked’ in a more
full-blown, systemic semantic model. It turns out that these different types
of designs, topological and discourse-structural, can be seen as very primary
lines of thought for conceptualizing ‘semantics’ (in addition to a third one
that will be introduced below). In other words, they are not only alternative
types of design, but they are also two complementary pre-systemic
approaches to ‘semantics’, and each of them is based in a specific vision
of what a semantics can be.
In all approaches to semantics, the question is, What ‘higher’ meaning
can be recognized, above the lexicogrammatical networks, that forms an
interface between lexicogrammar and context? In a ‘topological’ approach,
this question is explored in terms of different levels of ‘abstraction’. The
‘higher meaning’ is of a different level of coding in two ways: it is not exactly
what is already encoded in the lexicogrammar (otherwise no further

stratum of semantics is necessary), rather it goes beyond the lexicogramma-
tical meaning; and it is closer to non-linguistic context, with which it
interacts (see semantics as an interface stratum). For example, in the inter-
personal metafunction, ‘command’ is a higher-level, more abstract meaning
compared to a lexicogrammatical category such as ‘imperative’. The link
with the topological design is this: as soon as a higher-order meaning such as
‘command’ is recognized, the further question can be asked how this can be
linked back to the lexicogrammar. It then becomes clear that there are
different facets to realizing ‘commands’, i.e. ‘command’ at the stratum of
semantics is not just a semantic rendering of ‘imperative’ at the stratum of
lexicogrammar. More specifically, commands can also be realized by other
lexicogrammatical means than the imperative (such as different types of
modal verbs and/or adjuncts, see Figure 3.4). And hence in drawing this
connection between topological conception and differentiating semantics on
the basis of level of abstraction, we come full circle with the idea of variabil-
ity between two content strata (or between meanings and forms).
In a ‘discourse-structural’ approach, the question of what higher meaning
can be recognized is explored in terms of different levels of ‘patterning’. The
‘higher meaning’ is of a different level of patterning in two ways: it does not
just ‘consist of’ the structures and building blocks that are available in the
lexicogrammar, rather it goes beyond this lexicogrammatical patterning; and
it interacts with context. In this perspective, the text or discourse becomes
crucial at the level of semantics, since it is through texts that humans interact
with context (not through clauses). In modelling semantics as how we inter-
act with context through texts, one important cluster of questions therefore
will be how discourse is organized, how it is built, and what the basic
discourse units are. In relation to the method of the system network, this
question translates more specifically into a quest for the unit that is the entry
condition for semantic systems. Seen from below, i.e. in terms of how seman-
tics is beyond lexicogrammar, the crucial question becomes what type of
patterning takes place above and beyond the highest lexicogrammatical unit,
i.e. beyond the clause or clause complex. Importantly, again, the discourse
units do not ‘consist of’ lexicogrammatical units such as clause and clause
complex, but rather, they are realized by different lexicogrammatical means
(see Halliday and Matthiessen 2004:587). In discourse, these lexicogramma-
tical means occur sequentially, in unfolding text. Hence what will be import-
ant in the discourse-structural conception of semantics is how a discourse
pattern emerges from a sequence of lexicogrammatical patterns unfolding in
text. And this is how the link with variability can be made here: a discourse
pattern does not have a one-to-one relation with one lexicogrammatical unit
or pattern, but rather is realized through a sequence of (recurring) lexico-
grammatical forms in unfolding text.
The abstraction-based and pattern-based approaches to semantics which
are taken in recent semantic models are summarized and presented
Semantics 79
antics
antics
grammar
grammar
s em
s em
co
co
i i
lex lex
abstraction-approach pattern-approach
to semantics to semantics
design topological discourse-structural
higher-order higher-order abstraction: larger-order patterning:

meaning closer to context structures bigger than
lexicogrammatical structures
semantics what we can mean larger structures

as interface in between not composed of lexgram
what we can do building blocks,
& what we can say but realized by lexgram resources
Figure 3.9 Abstraction-based and pattern-based approaches to semantics, highlighting

a topological and discourse-structural design, respectively
visually in Figure 3.9. This figure shows how the approaches can be under-
stood as interpretations of the variability relation between a higher mean-
ing and a dispersed realization in the lexicogrammar.
As indicated above, the abstraction-based and pattern-based approaches
have been complementary sources of inspiration in modelling a semantic
stratum, and in setting up networks for this stratum. This has been the case
across the metafunctions, but there are differences in the roles and relations
between these two approaches for the different metafunctions. We will now
turn to specific semantic models in SFL. We will first look at each metafunc-
tion, paying attention to what exactly the abstraction and patterning
approaches have resulted in for those metafunctions, i.e. in setting
up semantic models that flesh out the specific nature of each metafunction.
After that we will turn to a third approach to theorizing a semantics, comple-
mentary to the abstraction-based and pattern-based conceptions, viz. a
register/probability-approach, which has been applied more globally across
the metafunctions.
3.5.2 Semantics from a Textual Perspective

At the level of lexicogrammar the textual metafunction comprises the systems
of theme and information, which are realized in the clause, and
cohesion, which is a label for a number of sub-systems (such as refer-

ence, conjunction) which are realized non-structurally. The textual
metafunction can be characterized by the following distinctive features. It
(i) is concerned with the creation of texture;

(ii) is regarded as a second-order metafunction, because it allows mean-
ings from the other two metafunctions, interpersonal and ideational,
to be brought together in syntagms and in larger, coherent wholes;
(iii) does not only comprise structural resources (theme and informa-
tion) like the other metafunctions do, but also non-structural ones.
Now, these defining features of the textual metafunction will help us

understand how this metafunction is conceived of at the semantic level.
The textual metafunction is intrinsically concerned with texture, and so
the ‘patterning approach’ is a natural route for approaching textual seman-
tics. In fact, the system of cohesion, with its ‘non-structural’ realization
across clauses (e.g. lexical cohesion realized by lexemes spread across
the text) is already a recognition of a higher-order pattern (beyond the
‘structural’ resources), and this conception ties in with the idea of the
textual metafunction as a second-order metafunction. In his model of
discourse semantics, Martin (1992) reinterprets three sub-systems of cohe-
sion at the semantic level (which for him is the level of discourse, of
higher-order patterning). In this framework, cohesion is not seen as part
of a second-order (textual) metafunction, but as occurring at a higher
stratum, that of discourse semantics. In discourse semantics, cohesive
resources are seen as belonging to different metafunctions: there are idea-
tional, interpersonal, and textual discourse-semantic systems. Each of those
higher-order semantic systems is a pattern-based reinterpretation of a
component of the earlier system of cohesion, which was already regarded
as ‘different’ from other lexicogrammatical systems, because cohesive
resources are non-structural. In addition to thus recognizing the ‘pattern’-
dimension of cohesion, the overall reconception of the notion of ‘cohesion’
as a higher stratum (rather than a ‘second-order’ metafunction) can be seen
as a remodelling of cohesion in terms of degree of ‘abstraction’. In this
sense, the pattern-based and abstraction-based conceptions of semantics are
both integral aspects of the model of discourse semantics.
3.5.3 Semantics from an Interpersonal Perspective

With regard to the interpersonal metafunction, a system of speech func-
tion is set up above the lexicogrammatical ones of mood and modality, in
order to deal with how speech functions are realized by dispersed lexico-
grammatical resources, and above (see Section 3.2.2) this was given as a
good example of the abstraction-based approach to semantics. An elaborate
model of speech functions which is context-specific (the context being
interactions between mothers and their children) is Hasan’s (1996) message
Semantics 81
semantics. However, a general speech functional model has been part of the
theory for a long time (see an early overview in Halliday 1984) and has been
incorporated into textbooks and presentations of the theory.7 In analyzing
the mapping between speech functions and their realizations in lexico-
grammar, some meaning–form couplings can be seen as the ‘default’ ones
and this leads to the concept of interpersonal grammatical metaphors of
mood (see Section 3.2.3.3). However, in another view, such ‘typicality’ is not
relevant or is at least problematic (in relation to mood as well as modality)
(see Hasan 2010:287).
Speech functions also feature in a pattern-approach to semantics, which
highlights relations between them in sequences, such as pairs consisting of an
initiating and a responding move (as in the traditional concept of adjacency
pairs) and in larger stretches of discourse, especially dialogue. This interper-
sonal dimension which focuses on the scaffolding of interaction is usually
referred to as negotiation or exchange, as in the framework of discourse
semantics (Martin 1992; Eggins and Slade 2005; Martin and Rose 2003).
In addition to speech function and exchange structure, a further inter-
personal area for which a conception of a semantic stratum is essential is
the expression of evaluation, which is modelled in the sub-theory of
appraisal. Appraisal deals with how attitudes and values are conveyed,
how those values are sourced, and how interactants are aligned in relation
to those values (White 2015). One source of inspiration (see White 1999) for
setting up appraisal theory was the concept of interpersonal grammatical
metaphors of modality, which revealed that one interpersonal meaning
such as a modal value of ‘probability’ (see examples in Section 3.2.3.3) can
be realized in various ways in the lexicogrammar (see Figure 3.4). In
appraisal theory, this conception is extended to a range of other types of
interpersonal values in addition to modality, for instance, affective mean-
ings and value judgements, which are realized in dispersed ways, through
interpersonal resources such as forms of modality, through experiential
resources – explicitly or through connotation, and/or through logical
resources (e.g. a conjunction but signalling a concessive meaning of
counter-expectation). Hence appraisal theory is a good example of an
abstraction-based approach to semantics which (re-)organizes various types
of lexicogrammatical means at a higher level in topological areas.
Appraisal meanings can also be looked at from a pattern-approach to
semantics, which then focuses on how attitudes are negotiated in a text and
how different alignments are set up across the text (e.g. Martin and Rose
2003; Martin and White 2005). A specific patterning, i.e. sequential, aspect of
appraisal that is useful to mention is that of ‘semantic prosody’. The combin-
ation of various appraisal resources in a text can form a pattern which sets a
tone or an attitudinal mood which is spread across a stretch of discourse, and
7
For overviews in textbooks, see, for instance, Thompson 1996: Chapter 4; Martin et al. 1997: Chapter 3; Halliday and
Matthiessen 2014: Section 4.1; Matthiessen 1993: Section 5.1.2.
which unfolds with varying degrees of strength, just like musical prosody
(see Martin and White 2005:59). For instance, certain words can trigger
positive or negative connotations in other words that occur in their neigh-
bourhood (see Louw 1993). Note, too, that the featuring of ‘connotation’,
both as a dimension of appraisal (the evocation, rather than explicit con-
strual, of evaluative meanings) and in the concept of semantic prosodies, as
an intrinsic aspect of interpersonal semantics, resonates with a general
conception of the stratum of semantics as a ‘connotative’ layer added to
the ‘denotative’ system of language (i.e. semantics interpreted in terms of
Hjelmslev’s model of a connotative semiotic, see Section 3.3.1 above).
3.5.4 Semantics from an Ideational Perspective

At the level of lexicogrammar, the ideational metafunction deals with the
building blocks of representational content through which experience is
construed, especially process types at the level of the clause, and with
relations between experiences, especially clause combining dealt with in
the systems of taxis and logico-semantic relations. Within the idea-
tional component, the former is called the experiential metafunction, the
latter the logical metafunction.
At the level of semantics, the ideational metafunction with both sub-
components is reinterpreted in Halliday and Matthiessen’s (1999) model of
the ideation base of language, which contains basic building blocks of
experience (‘experiential’), and combinations of them (‘logical’). In this
sense, the model of the ideation base is in effect a semantic model of the
rank scale, the primary options ranging from ‘elements’ (which can be
processes, participants, or circumstances) through ‘figures’ (by default real-
ized by clauses in the lexicogrammar) to ‘sequences’ (by default realized by
clause complexes). In linking the semantic system to the lexicogrammar,
the notion of typicality plays a crucial role, as indicated in the expression
‘by default’ in the previous sentence. Hence the ideation base incorporates a
model of ideational grammatical metaphor, which crucially hinges on the
notion of a rank scale (nominal groups are more condensed than clausal
realizations, and a clause is more condensed than a clause complex). Thus,
the ideation base is an ontological or phenomenological semantics, i.e. a
semantics of what types of entities, qualities, and relations between entities
there ‘are’; or what types of ‘phenomena’ humans conceptualize (signifi-
cantly, the entry point for Halliday and Matthiessen’s semantic network of
the ideation base is called ‘phenomenon’). It is clear that this is an
abstraction-based approach to semantics, the semantics here being an
interface between language and cognition or conceptualization, and again
what is modelled in the semantics are topological zones (types of ‘phenom-
ena’) which can be realized in dispersed ways in lexicogrammatical
resources. One specific component of their model, viz. the relations of
expansion and projection, which are seen as ‘transcategorical semantic
Semantics 83
domains’, has been referred to above (Section 3.2.2) as an example of a

topological approach to semantics, and the dispersion of realizations across
the lexicogrammar as an example of fractality (see Section 3.4.2). Expansion
was already regarded as dispersed through the lexicogrammar in IFG1 (see
especially Halliday 1985:306–7).
The components of expansion and projection (grouped as ‘sequence’ in
the ideation base) lend themselves to a pattern-approach to semantics,
precisely because it includes a concept of sequencing (projecting and pro-
jected element; expanded and expanding element), and thus are also useful
in a discourse perspective. There are at least two discourse-patterning
dimensions that are relevant: the unfolding of sequential occurrences of
projections and expansions through a text; and the fractal ‘replication’ of
such relations as also holding between larger units of text, such as between
clause complexes and between paragraphs (in the same sense as hyper- and
macro-Themes). This sequencing through relations of expansion and pro-
jection at text level is incorporated as conjunction in Martin’s discourse
semantics (see Martin 1992; Martin and Rose 2003). It is a focus of earlier
studies in the framework of Rhetorical Structure Theory (see Matthiessen
and Thompson 1988; Mann et al. 1992; Matthiessen 2002).
3.5.5 Semantics and Register

This section focuses on the relation between semantics and context. Here
the question is what semantics realizes from a higher stratum, and how the
relation between semantics and context can be modelled. Semantics, in this
view, is the stratum of registers as ‘semantic strategies’ (Matthiessen
2009:219), strategies being selections of options that are available in a
particular (institutional) setting and that, together, form a ‘procedure’ or
‘technique’ for functioning in that setting. The higher-level meaning that is
modelled at the semantic stratum in this case is thus a technique as a
combination of strategies, i.e. a grouping of meanings which can be made
in that context, and which are realized through different lexicogrammatical
resources that are ‘at risk’ in a particular setting. Resources that are ‘at risk’
in a context are those that are more likely to be selected because they
redound with certain aspects of the context. Because registers are defined
through the combination of options that are ‘at risk’ in a setting, they are
ways of setting the probabilities in the lexicogrammar stratum (see Halliday
and Matthiessen 2014:29), or, put differently, it is the specific combination
of those options at risk which are grouped at the semantic stratum as sets of
strategies, forming a ‘technique’ that works in a particular setting.8
8
It will be noted that in this way, semantics as register bears a fundamental similarity to that component of
interpersonal semantics which models ‘speech functional’ meanings through which social semiotic ‘roles’ are
enacted. This similarity is not surprising, since a register is a procedure for ‘functioning’ in a specific context, i.e. taking
a specific ‘role’ in an institutional setting. The similarity has been noted before in SFL, but it has not been studied
systematically (but rather has been seen as an ‘inconsistency’ or an unresolved issue). See Butler (2003), who also
antics
grammar
sem
co
i
lex
actualization-approach
to semantics
design probabilistic
higher-order higher-order grouping

meaning of strategies that
‘work’ in a context
semantics registers as gatekeepers:

as interface pre-selection of lexgram
resources ‘at risk’ in a context
Figure 3.10 An actualization approach of semantics, highlighting registers as

gatekeepers between context and lexicogrammar
This view of semantics, which I will refer to as an ‘actualization

approach’, can be seen as a third conception of semantics, complementary
to the pattern and abstraction approaches. It is visualized in Figure 3.10.
This approach is shown as a probabilistic interpretation, where at the
semantics there is a conglomerate of strategies, which together form a
procedure for functioning in certain contexts, and which activate certain,
but not other, lexicogrammatical options which are thus ‘at risk’ in this
context (shown by the check marks). Options that are not activated are
indicated in a lighter shade of grey and are thus backgrounded.
Here semantics can be seen as a gateway between context and lexicogram-
mar, and each register as a gatekeeper. More specifically, semantics is a
gateway between the language as potential, as a general code, and a
specific context, and a register is what is relevant, what is ‘activated’ in
a context. As complementary to abstraction and patterning-approaches to
semantics, this approach can be called an actualization view, because it
focuses on how language as a potential is actualized in specific techniques
that ‘work’ in specific settings. Note that register is also conceived of as a
connotative semiotic (see Butler 2003:383–90), a conception which again
refers to Gregory’s (1967) theory of register in which both dimensions pointed out are seen as two different aspects
of the interpersonal component, viz. one which relates to speech functions through which the relation between the
interactants are enacted (‘personal tenor’), and another which deals more broadly with the purpose of the text
(‘functional tenor’). In later work, the ‘purpose of the text’ is not tied to the interpersonal metafunction, but is realized
across the different metafunctions.
Semantics 85
resonates with the view of semantics as a connotative semiotic mentioned

previously (see Sections 3.3.1 and 3.4.3).
A final point to consider, in relation to semantics and register, is the
question of how exactly this gatekeeping function of registers between
lexicogrammar and context can be modelled. In SFL the relation of lan-
guage with context is interpreted in terms of the metafunctions. What
exactly, in the context, activates particular choices in a text is conceived
of as belonging to three metafunctional dimensions: (1) the content of what
is being talked about or the type of activity that is taking place (field –
ideational); (2) the relationship between the interactants (tenor – interper-
sonal); and (3) the role that language itself is playing as a medium (mode –
textual). In this way the complementarity between three metafunctions
resonates fractally through the different strata (see Hasan 2009:174), also
including context. This functional interpretation of language in relation to
context is referred to as the context-metafunctions hook-up in SFL, and
remains controversial to the present day (e.g. Hasan 2009, 2014).9
3.5.6 Mapping Semantic Models in SFL: Overview and Conclusion

The various specific conceptions of semantics that have come into view in
relatively recent semantic models in SFL are visually summarized in
Figure 3.11.
Three complementary approaches to semantics were distinguished which
each bear a particular relation to the design tool of the system network (and
the conception of language as a meaning potential, a network of options). The
abstraction and patterning approaches were both called ‘pre-systemic’: they
lead to topological areas of higher-order meanings or larger-sized patterning
in the semantics, which have dispersed realizations in the lexicogrammar.
Making internal distinctions within those zones and patterns forms a first
step into setting up networks at the level of the semantics, i.e. above the
lexicogrammar, and this is how the different semantic models pointed out
above developed. For each of the metafunctions, the abstraction and pattern-
ing approaches led to specific systems, which are summarized below the
stratal images in Figure 3.11. The actualization approach to semantics has a
different relation to the systemic dimension of language. Semantics is here
conceived of as a gateway between the overall code of language and specific
contexts. A register as a grouping of semantic strategies thus sets the prob-
abilities in the system – preselecting options across the different networks,
9
One problem, pointed out by Hasan (2009), is that when linguists determine what it is, in the context, that activates
particular choices in a text, they are already reasoning from the text (or with an imaginary text in mind), i.e. they always
reason from language. Hence context thus perceived is what Hasan calls ‘relevant context’, and this is different from the
more general non-linguistic ‘eco-social context’ which features in other interpretations of semantics. It will be noted that
this conception of ‘relevant context’ is itself a consequence and an inherent feature of language as a semiotic dynamic
open system with relations of metaredundancy between its strata, hence also between semantics and context-as-seen-
from-semantics.
approach to semantics
t
t
contex
contex
contex
a n t ic s
a n t ic s
a n t ic s
g rammar
g rammar
g rammar
se m
se m
se m
co
co
co
i i i
lex lex lex
abstraction-approach pattern-approach actualization-approach
interpersonal textual ideational
abstraction speech function ideation base

(message semantics)
specific metafunctional semantic models
te n o
identification e r
od
m
appraisal reference chains
tracking participants
interpersonal metaphor [earlier work on cohesion ideational metaphor
l
na
field
reinterpreted across
so
rper
l
ua
the metafunctions]
al
inte
text
tion
pattern negotiation conjunction
idea
sequencing of
t
exchange structure
contex
periodicity
a n t ic s
+ negotiation of projection & expansion
g rammar
attitudes (hyper-Theme, relations between
se m
macro-Theme) larger units of text
Rhetorical Structure Theory
co
i
lex
focus on enactment of creation of texture building blocks through

roles and stances which we construe experience
& &
how they are negotiated the relation between them
type of interactive semantics/ (discourse semantics: ontological/

semantics socio-semantics textual and across phenomenological/
metafunctions) conceptual semantics
Figure 3.11 Overview of semantic models in SFL in terms of differentiating dimensions in

the overall architecture (stratification, instantiation, metafunctional complementarity);
different design principles (systemic, topological, and discourse-structural); and the role of
variability between meanings and forms
and it links upward, to ideational, interpersonal, and textual aspects of the

context that are relevant and that redound with what is activated from
language. At the bottom right, the figure contains an attempt to visualize
the role of register as a semantic gatekeeper.
It is useful, as promised, to return once more to the crucial concepts of
stratification and metaredundancy, and briefly re-visit them in relation to
the types of ‘semantics’ that were explored in this section. It has become
clear above that the abstraction, patterning, and actualization approaches
to semantics are complementary. What I would like to highlight here is
that this complementarity also has a theoretical significance in relation
to how we understand stratification/metaredundancy. Metaredundancy
relations (see Figure 3.6 above) can be conceived of as ‘contextualization
relations’ (see Thibault 2004, who makes the same connection): each layer
in a semiotic system puts another (set of ) layer(s) into context. Stratifica-
tion, too, and in a more concrete sense, is a model for theorizing how
language (and its various coding layers) is ‘contextualized’. What the analy-
sis of ‘meaning’ and ‘semantics’ in this chapter shows is that the context-
ualization which semantics provides as a stratum that is of a higher-order
nature compared to lexicogrammar can mean at least three things: it can be
Semantics 87
higher-order abstraction (‘meanings’ that are closer to eco-social context,

analyzed in topological areas); it can be higher-order patterning (‘meanings’
that are larger in size, analyzed in sequential patterns); or it can be actual-
ization (‘meanings’ that are relevant, analyzed in registers). We have now
disentangled the dimensions of abstraction, patterning, and actualization,
for the sake of analysis, and for the mere sake of attempting to understand
‘language’, and, more specifically, in this chapter, what ‘semantics’ can be.
In language itself, all those contextualizations occur holistically and simul-
taneously in each instance of meaning-making.
As a final point, let us make the circle full and return to language as a
dynamic open system which is metastable, i.e. which continues to function
by changing. Each of the dimensions of semantics, viz. abstraction, pattern-
ing, and actualization, vastly enhance the openness of language and thus
the overall meaning potential: they are ways of organization (contextual-
izations!) that are not predicted by the lexicogrammar. At the same time,
they have a constraining function. This is clearest in the case of the actual-
ization approach, which highlights the role of registers as gatekeepers, but
it is also true for the abstraction and patterning dimensions, by definition,
because semantics interfaces between lexicogrammar and context. There is
bottom-up constraining of semantics towards context, because only those
‘meanings’ are available that can be realized in lexicogrammar with its
internal organization (such as the ‘amalgamation’ of different metafunc-
tions in each syntagm), and there is top-down restraining of semantics
towards lexicogrammar, because only those ‘meanings’ appear (topologic-
ally or as patterns) in semantics which the eco-social environment allows
(e.g. the types of speech functions are restrained by the types of social roles
in which we interact; the types of ‘phenomena’ that are distinguished are
restrained by the entities we find in our environment – this is why this
dimension of semantics is called ‘ontological’). In Figure 3.11, many-to-one
lines of variability between context and semantics have not just been added
to the register approach, but also to the other two approaches. And thus the
function of semantics as enhancing and constraining at the same time
finally brings us back to the concept of variability, with which we started
our exploration of the ‘why’ of semantics.
3.6 Summary
This chapter focused on ‘semantics’ and the related concept of ‘meaning’ in

SFL. It set out to explain what semantics is, how it is organized, where it is
found, and why it is regarded as something in its own right which is worth
looking at. Those questions were explored through a two-pronged approach
which determined the two-fold aim of this chapter: (1) on the one hand, to
distinguish different conceptions of ‘semantics’ and to disentangle the vari-
ous theoretical distinctions and perspectives which play a role in thinking
about semantics in SFL, and, (2) on the other hand, to look at various specific
proposals for recognizing a semantics in SFL and to place them against the
theoretical background of possible conceptions of semantics.
We started off in Section 3.2 by looking at where ‘meaning’ is situated in
the model of the system network and stratification in SFL, and by distin-
guishing three related ways in which semantics is designed in SFL. The ‘why’
question led us to the notion of variability relations between semantics and
lexicogrammar, and we saw how grammatical metaphor was an important
initial motivation for recognizing a (separate) stratum of semantics in SFL.
In Section 3.3, conceptions of meaning and semantics in SFL were placed
against a wider theoretical background, in two steps. First the concept of
variability was reconsidered against the background of viewing language as
a semiotic dynamic open system which is characterized by metastability
and by metaredundancy relations between its coding levels. In a second
step, the apparent indeterminate or multi-faceted view of meaning and
semantics in SFL was explained in relation to a design rationale that is
based on multiperspectivism and fractality.
In Section 3.4 we turned to specific semantic models in SFL. In connecting
how exactly semantics is fleshed out in those models to the different
conceptions of semantics which had been disentangled in the previous
sections, three basic approaches to semantics were distinguished, viz.
abstraction, patterning, and actualization. Those were linked to the ways
in which a semantics can be designed (topological, systemic/typological, and
discourse-structural), to the role of variability, and to stratification and
metaredundancy.
References
Butler, C. S. 2003. Structure and Function: A Guide to Three Major Structural-

functional Theories, Volume 2: From Clause to Discourse and Beyond. Amster-
dam: John Benjamins.
Butler, C. S. and M. Taverniers. 2008. Layering in Structural-functional
Grammars. Linguistics 46(4): 689–756.
Eggins, S. and D. Slade. 2005. Analysing Casual Conversation. 2nd ed. Sheffield:
Equinox.
Gregory, M. 1967. Aspects of Varieties Differentiation. Journal of Linguistics 3:
177–274.
Halliday, M. A. K. 1961. Categories of the Theory of Grammar. Word 17:
241–92.
Halliday, M. A. K. 1975. Learning How to Mean: Explorations in the Development
of Language. London: Arnold.
Halliday, M. A. K. 1976. Functions and Universals in Language. In G. Kress,
ed., Halliday: System and Function in Language. London: Oxford University
Press. 26–31.
Semantics 89
Halliday, M. A. K. 1977. Text as Semantic Choice in Social Contexts. In T.

van Dijk and J. S. Petöfi, eds., Grammars and Descriptions. Berlin: Mouton de
Gruyter. 176–225.
Halliday, M. A. K. 1982. The De-automatization of Grammar: From Priest-
ley’s ‘An Inspector Calls’. In J. M. Anderson, ed., Language Form and
Linguistic Variation: Papers Dedicated to Angus McIntosh. Amsterdam: John
Benjamins. 129–59.
A Systemic-functional Interpretation of the Nature and Ontogenesis of
Dialogue. In R. P. Fawcett, M. A. K. Halliday, S. Lamb, and A. Makkai, eds.,
The Semiotics of Culture and Language, Volume 1: Language as Social Semiotic.
Halliday, M. A. K. 1985. Introduction to Functional Grammar. London: Arnold.
Halliday, M. A. K. 1991. Towards Probabilistic Interpretations. In E. Ventola,
ed., Functional and Systemic Linguistics: Approaches and Uses. Berlin: Mouton
de Gruyter. 39–61.
Halliday, M. A. K. 1992. How Do You Mean? In M. Davies and L. Ravelli, eds.,
Advances in Systemic Linguistics: Recent Theory and Practice. London: Pinter. 20–35.
Halliday, M. A. K. 1996. On Grammar and Grammatics. In R. Hasan, C.
Cloran, and D. G. Butt, eds., Functional Descriptions: Theory in Practice.
Halliday, M. A. K. 1998a. Things and Relations: Regrammaticising Experi-
ence as Technical Knowledge. In J. R. Martin and R. Veel, eds., Reading
Science: Critical and Functional Perspectives on Discourses of Science. London:
Routledge. 185–235.
Halliday, M. A. K. 1998b. Linguistics as Metaphor. In A.-M. Simon-Vandenber-
gen, K. Davidse, and D. Noël, eds., Reconnecting Language: Morphology and Syntax
in Functional Perspectives. Amsterdam: John Benjamins. 3–27.
Halliday, M. A. K. 2008. Opening Address: Working with Meaning: Towards
an Appliable Linguistics. In J. J. Webster, ed., Meaning in Context: Imple-
menting Intelligent Applications of Language Studies. London: Continuum.
7–23.
Halliday, M. A. K. 2009. Methods – Techniques – Problems. In M. A. K.
Halliday, and J. J. Webster, eds., Continuum Companion to Systemic Func-
tional Linguistics. London: Continuum. 59–86.
Halliday, M. A. K. 2013. Meaning as Choice. In L. Fontaine, T. Bartlett, and
G. O’Grady, eds., Systemic Functional Linguistics: Exploring Choice. Cam-
bridge: Cambridge University Press. 15–36.
Halliday, M. A. K. and C. M. I. M. Matthiessen. 2004. Introduction to Functional
Grammar. 3rd ed. London: Edward Arnold.
Halliday, M. A. K. and J. J. Webster. 2009. Keywords. In M. A. K. Halliday and

J. J. Webster, eds., Continuum Companion to Systemic Functional Linguistics.
London: Continuum. 229–53.
Hasan, R. 1996. Semantic Networks: A Tool for the Analysis of Meaning.
In C. Cloran, D. G. Butt, and G. Williams, eds., Ways of Saying,
Ways of Meaning: Selected Papers of Ruqaiya Hasan. London: Cassell. 104–31.
Hasan, R. 2009. The Place of Context in a Systemic Functional Model. In M.
A. K. Halliday and J. J. Webster, eds., Continuum Companion to Systemic
Functional Linguistics. London: Continuum. 166–89.
Hasan, R. 2010. The Meaning of ‘Not’ is Not in ‘Not’. In A. Mahboob and N.
Knight, eds., Appliable Linguistics. London: Continuum. 267–306.
Hasan, R. 2014. Towards a Paradigmatic Description of Context: Systems,
Metafunctions, and Semantics. Functional Linguistics 1(9): 1–54.
Haspelmath, M. 2003. The Geometry of Grammatical Meaning: Semantic
Maps and Cross-linguistic Comparison. In M. Tomasello, ed., The New
Psychology of Language: Cognitive and Functional Approaches to Language Struc-
ture. Mahwah: Erlbaum. 213–42.
Hjelmslev, L. 1963. Prolegomena to a Theory of Language. Translated by F. J.
Whitfield. Madison: University of Wisconsin Press.
Lamb, S. M. 1962. Outline of Stratificational Grammar. Berkeley: University of
California.
Lemke, J. L. 1984. Semiotics and Education. Toronto: Victoria University.
23–62.
Lemke, J. L. 1992. New Challenges for Systemic-functional Linguistics:
Dialect Diversity and Language Change. Network 18: 61–8.
Lemke, J. L. 1995. Textual Politics: Discourse and Social Dynamics. London:
Taylor & Francis.
Louw, W. 1993. Irony in the Text or Sincerity in the Writer? The
Diagnostic Potential of Semantic Prosodies. In M. Baker, G. Francis, and
E. Tognini-Bonelli, eds., Text and Technology: In Honour of John Sinclair.
Structure Theory and Text Analysis. In: W. C. Mann and S. A. Thompson,
eds., Discourse Descriptions: Diverse Analyses of a Fund-raising Text. Amster-
dam: John Benjamins. 39–78.
Benjamins.
Martin, J. R. 1995. Text and Clause: Fractal Resonance. Text 15(1): 5–42.
Martin, J. R. and C. M. I. M. Matthiessen. 1991. Systemic Typology and
Topology. In F. Christie, ed., Literacy in Social Processes. Darwin: Northern
Territory University, Centre for Studies in Language and Education.
345–83.
Martin, J. R., C. M. I. M. Matthiessen, and C. Painter. 1997. Working with
Functional Grammar. London: Arnold.
Martin, J. R. and D. Rose. 2003. Working with Discourse. London: Continuum.
Semantics 91
Martin, J. R. and P. White. 2005. The Language of Evaluation. New York:

Palgrave Macmillan.
Matthiessen, C. M. I. M. 1993. Lexicogrammatical Cartography: English Systems.
Tokyo: International Sciences Publishers.
Matthiessen, C. M. I. M. 2002. Combining Clauses into Clause Complexes:
A Multi-faceted View. In J. Bybee and M. Noonan, eds., Complex Sentences in
Grammar and Discourse: Essays in Honor of Sandra A. Thompson. Amsterdam:
John Benjamins. 235–320.
Matthiessen, C. M. I. M. 2009. Meaning in the Making: Meaning Potential
Emerging from Acts of Meaning. Language Learning 59: 206–29.
Matthiessen, C. M. I. M. and S. A. Thompson. 1988. The Structure of
Discourse and ‘Subordination’. In J. Haiman and S. A. Thompson, eds.,
Clause Combining in Grammar and Discourse. Amsterdam: John Benjamins.
275–329.
Taverniers, M. 2002. Systemic-functional Linguistics and the Notion of Grammat-
ical Metaphor: A Theoretical Study and the Proposal for a Semiotic-functional
Integrative Model. PhD Thesis, Ghent University.
Taverniers, M. 2008. Hjelmslev’s Semiotic Model of Language: An Exegesis.
Semiotica 171: 367–94.
Taverniers, M. 2011. The Syntax–semantics Interface in Systemic Func-
tional Grammar: Halliday’s Interpretation of the Hjelmslevian Model of
Stratification. Journal of Pragmatics 43(4): 1100–26.
Thibault, P. J. 2004. Agency and Consciousness in Discourse: Self–Other Dynamics
as a Complex System. London: Continuum.
Thompson, G. 1996. Introducing Functional Grammar. London: Arnold.
White, P. R. R. 1999. Beyond Interpersonal Metaphors of Mood: Modelling
the Discourse Semantics of Evaluation and Subjectivity. Paper presented
at the 11th Euro-International Systemic Functional Workshop, July 1999,
Ghent University.
White, P. R. R. 2015. Appraisal Theory. In K. Tracy, C. Ilie, and T. Sandel,
eds., The International Encyclopedia of Language and Social Interaction.
Hoboken: Wiley. 1–7.
4
The Clause
An Overview of the Lexicogrammar
Margaret Berry
This chapter was to have been written by Geoff Thompson, but sadly he died before he
could write it. I would like what I have to say to be regarded as my tribute to Geoff.
Certainly I shall be drawing deeply on his work, particularly on the third edition of
his Introducing Functional Grammar (Thompson 2014).
4.1 Introduction
The chapter is intended to cover material which in most introductions to

SFL takes at least three chapters – long chapters. Inevitably, therefore,
I shall have to cut corners and oversimplify. I will give references to places
where more detailed discussions can be found. I shall assume knowledge of
the key concepts of SFL outlined in the preceding chapters (see in particular
Webster, this volume): the metafunctions, the notion of choice, systems,
system networks, rank and strata, and particularly the stratum of lexico-
grammar. I shall be concerned with choices from the three main metafunc-
tions – experiential, interpersonal, textual – and with the part these play in
the structure of the clause.
The chapter will be from the perspective of ‘mainstream’ SFL. Relevant
discussion from the perspective of another version of SFL – Cardiff
Grammar – can be found in Schulz and Fontaine, this volume. Following
Halliday and Matthiessen (2014), the language of exemplification will be
English, with only occasional footnotes referring to other languages.
SFL prides itself on being an ‘appliable’ linguistics. Applications are to be
discussed in later chapters of this volume. However, since for me the most
interesting application of the kind of analysis I am going to be discussing is
the application to discovering how children learn to write in different
I am grateful to Chris Butler, Jeff Wilkinson, and the editors of this volume for comments on the first draft of this chapter.
Of course I alone am responsible for any errors or misrepresentations.
Downloaded from https://www.cambridge.org/core. Nottingham Trent University, on 23 Aug 2019 at 07:38:48, subject to the Cambridge Core terms of use, available
The Clause 93
registers, I shall have this application always in mind. Most of my examples

will be taken from the work of an eleven-year-old girl. I will call her
Charlotte. Charlotte is in her first year at secondary school and is enjoying
the freedom she is being given to experiment with different styles of
writing for different purposes. She has provided me with a horror story,
an account of a day in her life, a history essay, work from her health and
nutrition course, and work from her science course. There will be no room
to discuss her differing registers fully, but I will make clear which of my
examples are taken from which of her pieces of work.
I am very grateful to ‘Charlotte’ for allowing me to use her work in this
way, and also to her mother for acting as go-between.
4.2 The Experiential Metafunction
The first metafunction to be discussed is the experiential metafunction. The

experiential metafunction is the function of language to represent our
experience of the world, to say what we want to say about the happenings
and states of affairs of the world and our responses to them. From the
perspective of the clause, SFL assumes that these happenings and states and
responses are represented as ‘processes’, with ‘participants’ in those pro-
cesses and ‘circumstances’ attendant on the processes.
Examples (1a) to (1c) are taken from Charlotte’s account of a day in
her life.
(1a) I pressed the snooze button twice more

participant process participant circumstance
(1b) I fed my hamster
participant process participant
(1c) I ’ll leave the house at eight
participant process participant circumstance
The experiential metafunction is responsible for the basic constituents of a

clause. (1b) is assumed to have three basic constituents, one process and two
participants; (1a) and (1c) each four, one process, two participants, and one
circumstance. We shall see later that the interpersonal and textual meta-
functions affect the structure of the clause in other ways.
SFL recognizes different types of process, each with its own types of
participant: material, mental, relational, verbal, behavioural, and existen-
tial processes. The first three of these are usually regarded as the
three main types, the other three as minor (e.g. Halliday and Matthiessen
2014:215). Halliday and Matthiessen report on a study of what they call
‘a registerially mixed sample of texts’ in which material processes were
the most frequent, relational processes the second most frequent, and
94 MARGARET BERRY
material process
behavioural process
mental process
clause
verbal process
relational process
existential process
Figure 4.1 Types of process represented as a system network
mental processes third. There will be room in this chapter only to discuss
the three main types.1
The types of process are presented as options in a system relevant to the
clause (see Figure 4.1). The options in the system are given the names of the
types of process. However, it is important to note that what is really being
chosen is a kind of package deal, a configuration of a process together with
relevant participants. Hopefully this will become clear as I discuss the main
types of process in detail.
4.2.1 Material Processes

Material processes are those which involve physical actions, such as pressed,
fed, and leave in examples (1a) to (1c) above. In (2a) and (2b), from Charlotte’s
history essay, the processes rushed and turned are again material processes,
as are grasp and emits in (2c) and (2d) from Charlotte’s horror story.
(2a) The English rushed boldly forward

(2b) The Normans turned their horses
(2c) Cold clammy hands grasp my own
(2d) A pearly aura emits from their body
A material process will have a participant representing the doer of the

action. SFL calls this participant the ‘Actor’. In (2a) to (2d), the Actors are
respectively The English, The Normans, Cold clammy hands, and A pearly aura. It
is possible however that, although a process has an Actor, the Actor will not
necessarily appear in the clause. In (3), a continuation of (1b), the Actor of
trekked is not mentioned in its own clause, but we know that the Actor is I as
this is recoverable from the previous clause.2
1
Halliday and Matthiessen (2014:215) comment that ‘[t]he minor process types appear to vary more across languages
than the major ones’.
2
|| indicates a clause boundary. I am using standard SFL notation, as set out in Halliday and Matthiessen (2014:ix–xi).
The Clause 95
(3) || I fed my hamster || and trekked downstairs ||
Often a material process will have a second participant representing the

person or thing that the action is done to. SFL calls this the ‘Goal’. In the
first half of (3) my hamster is the Goal, in (2b) their horses, in (2c) my own. Some
material processes on the other hand have only an Actor with no Goal. (2a),
(2d), and the second clause of (3) include circumstances, but have no second
participant; there is no person or thing that the action is done to.
We are here of course recognizing the traditional grammar distinction
between ‘transitive’ and ‘intransitive’, the difference between having an
object and not having an object, but SFL is focusing more on the semantics
than traditional grammar. SFL in fact uses the term ‘transitivity’ more
broadly than traditional grammar, to refer to the whole area of processes,
their participants, and their circumstances.
When analyzing texts, SFL uses ‘probes’ (e.g. questions) to distinguish the
different types of process and participant. The probes for material process are:
(a) Can the clause be rephrased in the form What X did was (to) . . .?
(b) If the process is taking place at the present moment, is the most
natural form in which to refer to it that of the continuous present
(i.e. the be + ing form)?
The processes in the two clauses of (3) pass both these tests. What I did was to
feed my hamster and What I did was to trek downstairs are both perfectly
acceptable. And if we switch the processes into the present, the most natural
way to refer to them would be I am feeding my hamster and I am trekking
downstairs. Other forms would have different implications. For instance,
I feed my hamster and I trek downstairs would seem to refer to habitual pro-
cesses rather than single processes taking place at the present moment.3
On the other hand, the process of ‘knowing’ in (4), from Charlotte’s
health and nutrition essay, fails both tests.
(4) I knew mango juice had a thick consistency
What I did was to know mango juice had a thick consistency sounds very strange;
‘knowing’ is not really a form of ‘doing’. And if we switch the process into
the present, I am knowing mango juice has a thick consistency sounds equally
strange. Much more likely would be I know mango juice has a thick consistency.
The two processes in (3) then are material processes. The ‘knowing’ process
in (4) is not. (Example (4) will be discussed later in Section 4.2.2 on mental
processes.)
3
The point about the continuous present being the most natural way of referring to a single present process is perhaps
truer of some registers than of others. If a register has something of the character of a commentary, the simple present
form can be used to refer to a present process. (2c) and (2d) are from a passage in Charlotte’s horror story that reads
like a commentary on what is going on. In this context, Cold clammy hands grasp my own and Cold clammy hands are
grasping my own would seem to be equally possible ways of referring to the present process, as are A pearly aura
emits from their body and A pearly aura is emitting from their body.
96 MARGARET BERRY
The first of the probes for material process can also be used as a probe for
the participant ‘Actor’. If the clause can be rephrased in the form What
X did was (to) . . ., where a constituent of the clause naturally replaces X, then
that constituent is representing the Actor. Applying this to (2a) and (2b),
repeated here for ease of reference,
(2a) The English rushed boldly forward

(2b) The Normans turned their horses
we get What the English did was to rush boldly forward and What the Normans did
was to turn their horses. The English and The Normans are the Actors of their
respective processes.
The probe for ‘Goal’ is an extension of the probe for material process and
Actor. If the clause can be rephrased in the form What X did to Y was (to) . . .,
where a constituent of the clause naturally replaces Y, then that constituent
is representing the Goal of the process. (2b) can be rephrased as What the
Normans did to their horses was turn them. Their horses is the Goal of the process
of ‘turning’. There is no equivalent to Y in (2a). (As we have already seen, in
traditional grammar terms, the process in (2b) is transitive, the process in
(2a) is intransitive.)
Applying this to examples (1a) to (1c), again repeated here for ease of
reference,
(1a) I pressed the snooze button . . .

(1b) I fed my hamster
(1c) I’ll leave the house . . .
we get What I did to the snooze button was press it and What I did to my hamster
was feed it. The snooze button and my hamster are the Goals of their respective
processes. However, there is a problem with (1c): i.e. What I’ll do to the house
is leave it. Here, I is not really doing anything to the house. The snooze button
is presumably in a different position as a result of the pressing. And a fed
hamster is presumably different from an unfed hamster. But there is no
change in the house as a result of the leaving. SFL regards the house as not a
Goal but an instance of another kind of participant which it calls ‘Scope’.
The processes in (1a) and (1b) have Goals and so are transitive. The process
in (1c) is regarded as intransitive, but having Scope. A Scope resembles
a Goal in that typically it occurs immediately after the process and in
that it does not include a preposition. But it represents the domain to
which the process relates, rather than something the process is done
to. Another example of a Scope (a made-up example this time) would be
the hill in (5a).
(5a) Jack climbed the hill

(5b) Jack climbed rapidly
The Clause 97
Figure 4.2 More delicate choices within material processes
Both (5a) and (5b) are regarded as intransitive. (5a) is intransitive with
Scope. (5b) is intransitive without Scope – it just has a circumstance.4
I have discussed the probes for material process, Actor, and Goal in detail,
in order to show how probes work. Unfortunately there will not be room in
this chapter to discuss in the same way the probes for the other categories
I am about to cover. For more on probes, see Fontaine (2013:85–91); Bartlett
(2014:48–82).
Moving on to discuss other distinctions within material process, I will quote
from Thompson (2014:96), who says ‘there are many different suggestions for
ways in which [material processes] can be subcategorized at more delicate
levels’. In fact Thompson himself (2014:95–7, 111–14), Berry (1975:151–2,
154–61), and Bartlett (2014:48–58) offer different selections from among the
possible distinctions. The most comprehensive account is that of Halliday and
Matthiessen (2014:224–44, see especially their Figure 5–10, on page 229), but
even that does not cover all the distinctions that have been suggested.
The distinctions I have chosen to present in this chapter are those which
in the past I have found most helpful in distinguishing different registers.
Figure 4.2 shows the distinctions I have discussed here so far and the
distinctions I am going on to discuss.
I have already discussed the transitive (having both Actor and Goal) and
intransitive (having Actor but no Goal) system and the further + scope and -
scope system. The transformative/creative system, the next system down in
Figure 4.2, distinguishes between processes that bring Goals into existence
(creative) and those which do something to existing Goals (transformative).
Thompson’s (2014:96) examples of this are as in (6a) and (6b).
(6a) I’ve just made the Christmas puddings

(6b) My Mum never eats Christmas pudding
(6a) is creative as the process of making actually brings the puddings into
existence. (6b) is transformative as the pudding exists before the eating of it
(or in this case the not eating of it!). Examples (7a) and (7b) are both from
Charlotte’s health and nutrition essay.
4
For discussion of different types of Scope, see Halliday and Matthiessen (2014:239–42).
98 MARGARET BERRY
(7a) Mango juice would add thickness to my smoothie

(7b) I used raspberries
(7a) is presumably creative as the process of adding creates the thickness.

But (7b) is transformative as the raspberries existed before Charlotte used
them in her smoothie.5
A set of distinctions that I have found particularly useful in distinguish-
ing different registers is that between concrete and abstract, with the
further distinction between animate and inanimate (again see Figure 4.2.)
All the examples of material processes I have given so far would count as
concrete material processes. But Halliday and Matthiessen (2014:243) note
that material processes may represent abstract doings and happenings.
They give examples from a passage of financial news, which include (8a)
and (8b).
(8a) AT&T’s stock slid Tuesday [sic!]

(8b) The disappointing forecast, which came as AT&T posted first-
quarter results . . ., dampened the enthusiasm . . .
It may seem a contradiction in terms to say that material processes can be

abstract, but processes such as those in (8a) and (8b) do pass the tests for
material processes discussed above: i.e. What AT&T’s stock did was to slide,
and What AT&T did was to post first-quarter results. And if we switch the
processes into the present, the most natural form for them would be AT&T’s
stock is sliding, and AT&T is posting first-quarter results.
Of course one may want to regard such examples as metaphorical, but
analyzing them as abstract versions of material processes shows the kind of
metaphor involved.
Examples of abstract material processes from Charlotte’s work, both
from her history essay, are (9a) and (9b).
(9a) My evidence comes from the Primary Source C

(9b) This point links to my next point6
Concrete material processes may be either animate or inanimate, this

distinction showing mainly in their Actors. As I explained in the introduc-
tion to Section 2, although the names of the options make it appear that we
5
As well as discussing the creating or transforming of Goals in transitive clauses, as I just have, Halliday and Matthiessen
(2014:231) discuss the creation of Actors in intransitive clauses. This is why I have followed them in presenting the
transformative/creative system as simultaneous with the transitive/intransitive system, instead of presenting the
transformative/creative system as dependent on the selection of transitive.
6
Halliday and Matthiessen may want to regard example (9b) as a relational clause – see what they say, for example,
(2014:265) about These plates went from the head to the tail. However, for me (9b) passes the tests for material
processes. We can say What this point does is link to my next point and This point is linking to my next point, though
these forms are perhaps unlikely in the register of Charlotte’s history essay.
The Clause 99
are simply classifying processes, in fact we are classifying package deals

consisting of configurations of process plus participant/s.
Examples (1a) to (1c) above, from Charlotte’s account of a day in her life,
have the animate Actor I. Examples (2c) and (2d) above, and (10a) to (10d)
below, all from Charlotte’s horror story, have inanimate Actors.7
(10a) Mist is swirling about me

(10b) A ghostly light penetrates the gloom
(10c) The wind will begin to howl like a wolf
(10d) The candles will flicker
The first paragraph of Charlotte’s horror story includes five material pro-
cesses. These all have inanimate Actors. And material process clauses with
inanimate Actors continue to feature in later paragraphs.
4.2.2 Mental Processes

The second type of process to be discussed here is that of mental processes.
Mental processes are not really processes of doing, but rather processes of
sensing. As we saw above, the sensing process of ‘knowing’, as in example
(4), fails both tests for material processes.
(4) I knew mango juice had a thick consistency
We cannot really say either What I did was to know mango juice had a thick
consistency or I am knowing mango juice has a thick consistency. ‘Knowing’ is not
a form of ‘doing’. And, if we switch the process into the present, the most
likely form for it would be the simple present – I know mango juice has a thick
consistency – not the continuous present which is most usual for material
processes.
The tests for material processes can thus be negatively used for mental
processes. Material processes pass the tests, mental processes fail them. (For
more on probes for mental processes, see Bartlett 2014:64–5; Fontaine
2013:87.)8
SFL recognizes four different types of mental process, as shown in
Figure 4.3.9
7
I am here disagreeing with Halliday and Matthiessen (2014:250), who say that in material clauses ‘the distinction
between conscious and non-conscious beings simply plays no part’. It is true that, in the class of material processes
viewed as a whole, Actors may be either conscious (animate) or non-conscious (inanimate), but this is not true of all
individual members of the class. Some material processes will normally have an animate Actor (e.g. trek in example
(3)), while others will normally have an inanimate Actor (e.g. flicker in example (10d)). There seems to be a cline of
material processes from this point of view. Examples with untypical animacy are usually viewed as metaphorical. For
discussion, see Berry (1975:151–2, 155).
8
Also see below. A mental process can take a clause as phenomenon. A material process cannot take a clause as goal.
9
For a more detailed system network showing more delicate choices within mental processes, see Halliday and
Matthiessen (2014:258).
100 MARGARET BERRY
Figure 4.3 A more delicate choice within mental processes
Perceptive processes include such processes as ‘seeing’, ‘hearing’, ‘feel-

ing’. Examples (11a) and (11b) are from Charlotte’s horror story, (11c) from
her account of the day in her life.
(11a) I can hear her rasping breath

(11b) I heard the cry of my best friend
(11c) You can see the parents spectating on the sidelines
Emotive processes include such processes as ‘loving’, ‘hating’, ‘enjoying’,

and ‘fearing’. (12a) and (12b) are from Charlotte’s day in her life, (12c) from
her health and nutrition work.
(12a) You hate sprinting

(12b) Since then I have loved sprinting
(12c) I love the smell and taste of lime
Cognitive processes include such processes as ‘thinking’, ‘believing’,

‘understanding’, and ‘conjecturing’, as well as the process of ‘knowing’
from example (4). (13a) and (13b) are from Charlotte’s health and nutrition
work, (13c) from her history essay.
(13a) I thought the mango juice would counteract the tart flavour of the
raspberries
(13b) I thought it would make an attractive red
(13c) I have concluded that the Norman tactics are the most important
reason for the Norman victory10
Desiderative processes include such processes as ‘wanting’, ‘wishing’, and

‘hoping’. (14a) is the only desiderative process I have found in Charlotte’s
work, from the day in her life. (14b) and (14c) are from Thompson
(2014:100).
10
It could be said that conclude passes the tests for material process. However, although there has not been room to
discuss it fully here, a further difference between mental and material processes is that mental processes can take a
clause as complement, while material processes cannot (see footnote 9). Conclude can take a clause, as in example
(13c).
The Clause 101
(14a) I hope my dream will come true

(14b) I don’t want any trouble
(14c) You may crave a cigarette
Mental processes will have a participant whose function is the sensing. SFL
calls this the ‘Senser’. Typically the Senser will be animate, indeed human.
All the examples I have so far given of mental processes have human
Sensers – the I’s and you’s. When a mental process occurs with an inani-
mate Senser, this is seen as an example of personification. Example (15) is
from Thompson (2014:98).
(15) We used to have a car that didn’t like cold weather
The thing that is sensed is called the ‘Phenomenon’. This may be a simple
Phenomenon, realized by a nominal group, such as any trouble in (14b), a
cigarette in (14c), cold weather in (15). Or it may be more complex, involving
another clause which itself contains another process. (11c), repeated here,
has the Phenomenon the parents spectating on the sidelines.
(11c) You can see the parents spectating on the sidelines
The Phenomenon of see includes another process; what is seen is the parents
spectating on the sidelines. Halliday and Matthiessen (2014:256) call this
type of complex phenomenon a ‘macrophenomenon’. In a macrophenome-
non the second process will be realized by a non-finite verb, usually the
‘-ing’ form.
Another type of complex phenomenon, called by Halliday and Matthiessen
a ‘metaphenomenon’, is found in (13a) to (13c) and (14a), again repeated here.
(13a) I thought the mango juice would counteract the tart flavour of the
raspberries
(13b) I thought it would make an attractive red
(13c) I have concluded that the Norman tactics are the most important
reason for the Norman victory
(14a) I hope my dream will come true
In a metaphenomenon, the second process will be realized by a finite

verb, and the metaphenomenon itself can be introduced by that. In (13c) the
metaphenomenon actually is introduced by that. A that could be inserted at
the beginnings of the metaphenomena in (13a), (13b), and (14a).
Halliday and Matthiessen (2014:256) distinguish two subtypes of meta-
phenomena: ‘idea’ type and ‘fact’ type. An ‘idea’ type introduces a new idea.
A ‘fact’ type represents something that has already happened. All Char-
lotte’s metaphenomena – in (13a) to (13c) and (14a) – are of the ‘idea’ type.
Halliday and Matthiessen’s (2014:256) examples of the ‘fact’ type are as in
(16a) and (16b).
102 MARGARET BERRY
(16a) He saw that they had left

(16b) He regretted (the fact) that they had left
All four main types of mental process – perceptive, emotive, cognitive, and
desiderative – can take simple Phenomena, but they vary in the types of
complex Phenomena with which they are associated. Macrophenomena
occur with perceptive and emotive processes, but not with cognitive or
desiderative processes. All four types can take metaphenomena, but per-
ceptive and emotive processes typically take ‘fact’ type, while cognitive and
desiderative processes typically take ‘idea’ type. For more on the possibil-
ities for the various types of mental process, see Halliday and Matthiessen
(2014:256); Bartlett (2014:64–5).
4.2.3 Relational Processes

The third type of process to be discussed is that of relational processes. (17a)
and (17b), from Charlotte’s horror story, and (17c) and (17d), from the day
in her life, are all examples of clauses with relational processes.
(17a) Everything is hazy and unclear

(17b) I am a ghost
(17c) The snooze button was a saviour
(17d) We are big breakfast people
In each clause a relationship is set up between an entity (Everything, I, The

snooze button, We), and a quality (hazy and unclear, a ghost, a saviour, big
breakfast people). Thompson (2014:101) comments:
Strictly speaking, neither of the basic experiential terms, ‘process’ and

‘participant’, is completely appropriate for this category. There is no process
in the normal sense of ‘something happening’; and although there are
always two concepts – one on each side of the relationship – there is only
one participant in the real world.
A description such as hazy and unclear is ‘hardly a prototypical participant’

(Thompson 2014:101). And in (17b) a ghost is not something separate from I,
but simply a description of I. But Thompson (2014:101) continues: ‘How-
ever, no grammatical term will cover equally well all the phenomena to
which we need to apply it, so we will continue to talk about process and
participants.’
SFL recognizes two main types of relational process: ‘attributive relational
processes’ and ‘identifying relational processes’. (17a) to (17d) above are all
examples of attributive relational processes, in which a description is attrib-
uted to an entity or group of entities.Attributive relational processes are each
assumed to have two participants: a ‘Carrier’, the entity or entities described,
and an ‘Attribute’, the description related. In (17a) to (17d) the Carriers are
The Clause 103
Everything, I, The snooze button, and We. The Attributes are hazy and unclear, a
ghost, a saviour, and big breakfast people. The Attribute may be adjectival, as in
(17a), or a nominal group, usually indefinite, as in (17b) to (17d).
Examples of identifying relational processes are (18a), from Charlotte’s
day in her life, (18b) and (18c), from her horror story, and (18d) and (18e),
from her history essay.
(18a) Breakfast is the most important meal of the day

(18b) The only reminder of my original features are my eyes
(18c) Dying was the easy part
(18d) My argument is that with power comes clumsy judgement
(18e) My third reason for why William won the Battle of Hastings is the
Norman tactics
In an identifying relational process clause, one entity is identified in rela-

tion to another. Breakfast, for instance, is identified as the most important meal
of the day. Usually something specific is identified by reference to a more
generalizable category. Breakfast is something specific, which is identified
by reference to the generalizable category of the most important meal of the
day. SFL calls the more specific participant the ‘Token’, while the more
generalizable category is called the ‘Value’. It is not always easy to decide
which is Token and which is Value. I would say that in (18a) to (18c) the
Tokens are Breakfast, my eyes, and Dying, with the most important meal of the
day, The only reminder of my original features, and the easy part as Values.
(Perhaps a clue to the fact that my eyes is Token is that Charlotte has made
the verb agree with my eyes rather than with what precedes it. My computer
does not approve of this and has underlined are in green!) Thompson
(2014:105) includes example (19).
(19) The explanation is that it is forbidden by the second law of

thermodynamics
He labels the that-clause as Token and The explanation as Value. By analogy

with this, Charlotte’s (18d) would have the that-clause as Token and My
argument as Value. Does that mean that in (18e) the Norman tactics would be
Token and My third reason . . . would be Value?
Identifying relational processes are similar to attributive relational pro-
cesses in that, although there appear to be two participants, there is only
one real-world entity. Breakfast and the most important meal of the day both
refer to the same real-world entity, though the relation between them
provides a new perspective. However, identifying relational processes differ
from attributive relational processes in that their participants are realized
by definite nominal groups or the equivalent, instead of the indefinite
nominal groups of attributive relational processes. Also, the identifying
relational process clauses are reversible in a way that attributive relational
104 MARGARET BERRY
Figure 4.4 A choice within relational processes
process clauses are not. It would be perfectly possible to say The most
important meal of the day is breakfast or My eyes are the only reminder of my
original features, instead of the versions in (18a) and (18b), and still sound
naturally idiomatic. While it is just about possible to reverse some attribu-
tive relational processes – e.g. Big breakfast people are we – the result would
sound marked and highly rhetorical.
There is a great deal more that could, and should, be said about relational
processes, but space does not permit. I will close this subsection simply by
saying that I think Thompson (2014:101–5, 122–7) is particularly helpful on
relational processes. For more detailed discussion, see Halliday and Mat-
thiessen (2014:259–300). Figure 4.4 shows the one choice within relational
processes that I have had room to discuss. There are many more. See Halli-
day and Matthiessen (2014:264).
4.2.4 Experiential Matters Not Covered in This Chapter

There are a number of experiential matters that I have not been able even to
touch on in this chapter. I have already indicated that I have not been able
to discuss the three minor types of process and their participants – behav-
ioural, verbal, and existential processes. See Halliday and Matthiessen
(2014:300–10). I have also not been able to discuss circumstances. For a
helpful table of possible types of circumstance, see Fontaine (2013:80);
Halliday and Matthiessen (2014:313–14). Most of my examples have been
of clauses with active voice; I have not been able to discuss their passive
voice counterparts. Thompson does discuss passive counterparts to active
clauses – see index to Thompson (2014) for detailed references. My
examples have mainly been prototypical examples – for discussion of more
problematic examples, see e.g. O’Donnell et al. (2009); Gwilliams and Fon-
taine (2015).
I have not said anything at all about an alternative way of looking at all
this – the ergative model as opposed to the transitive model. See Thompson
(2014:139–42); Halliday and Matthiessen (2014:332–55).
4.3 The Interpersonal Metafunction
SFL assumes that, simultaneously with making experiential choices of the

kind discussed in Section 4.2, we also make interpersonal choices – choices
The Clause 105
Figure 4.5 A simplified MOOD network for English
which have to do with the way in which we interact with other people – and
that these interpersonal choices also have an effect on the structure of the
clause.
Halliday (2002:203) says of the experiential options that they will tend to
be realized particulately – that is, a structure that represents experiential
meanings ‘will tend to have this form: it will be a configuration, or constel-
lation, of discrete elements’. We have already seen that the experiential
choices lead to the basic constituents of a clause, their number, and their
nature.
Halliday (2002:205) says of interpersonal meanings that they will tend to
be realized prosodically – that is, the interpersonal meaning is ‘strung
throughout the clause as a continuous motif or colouring’. This section will
attempt to show what Halliday means by this.
4.3.1 MOOD
SFL recognizes two main kinds of clause rank interpersonal choice, which
are usually discussed under the headings of MOOD and MODALITY. MOOD
choices will be discussed in this subsection, MODALITY choices in the
following subsection.
MOOD choices have to do with the forms we use when indicating the
kind of interaction in which we are engaged; e.g. whether we are making
statements or asking questions or giving commands. The main choices are
shown in Figure 4.5.11 Only independent clauses have access to these
choices, not subordinate clauses.
Concepts such as ‘declarative’, ‘interrogative’, and ‘imperative’ are of
course familiar in most approaches to the study of language. SFL’s particu-
lar take on them is to present them as options from which choices can be
made, and to relate them both to the semantics in the stratum above and to
the lexicogrammatical structure of the clause.
Relating them to the semantics is not a simple matter. While it is true
that in the unmarked cases questions are realized by interrogatives,
11
For a more detailed version of this network, see Thompson (2014:60). For an even more detailed version, see
Halliday and Matthiessen (2014:162).
106 MARGARET BERRY
statements by declaratives, and commands by imperatives, this is by no

means always so. Commands, for instance, can in certain circumstances be
realized by interrogatives or declaratives. Full discussion of this is beyond
the scope of the present chapter (but see Berry 2016). For a full discussion of
the relation between MOOD and SPEECH FUNCTION, see Martin (1992:
Chapter 2).
Also, of course questions etc. can be realized intonationally (see Debash-
ish, this volume). I am in this chapter concerned with the effect on the
lexicogrammar.
As far as the structure of the clause is concerned, realizations of the
MOOD options mostly affect the ‘Predicator’, the constituent which mainly
carries the transitivity process, and the ‘Subject’, a constituent which
carries one of the transitivity participants. (Which participant will depend,
for instance, on whether the clause is active or passive.)
The choice between ‘major’ and ‘minor’ is realized by the presence or
absence of a Predicator. In example (20), from Charlotte’s day in her life,
there are two major clauses, followed by a minor clause.
(20) I pressed the snooze button twice more, but eventually clambered
out of the warmth of my bed. A typical school morning.
The first clause has the Predicator pressed and the second clause the Predi-
cator clambered. But in the third clause, a minor clause, there is no explicit
process. Charlotte uses minor clauses in her horror story and in the day in
her life, but in her other writings minor clauses appear only as headings of
sections and subsections.
The choice between ‘indicative’ and ‘imperative’ is realized by the pres-
ence or absence of a Subject. The first clause in (20) has the Subject I and is
indicative. The second clause would also be regarded as indicative.
Although the Subject does not actually appear in that clause, it is recover-
able from the previous clause. On the other hand, the main clause in (21a),
again from Charlotte’s day in her life, is imperative. There is no Subject of
Imagine. Similarly (21b), from Thompson (2014:58), is imperative. There is
no Subject of Answer.12
(21a) Imagine you are standing on a race track

(21b) Answer no more than three of the following questions
For the realization of the choice between ‘declarative’ and ‘interrogative’,

we have to consider a possible split in the Predicator between the Finite part
of it and the main Lexical part. Examples (22a), (22b), and (22c) are all from
Charlotte’s horror story.
12
It is possible for imperatives to be accompanied by things that look like Subjects. Thompson (2014:59) gives the
example You listen to me, young man. However, Thompson argues that these are not ‘normal’ Subjects. The question
of what is meant by ‘Subject’ in SFL is a complex one. For varying views, see Halliday and Matthiessen
(2014:147–50); Fawcett (1999); Fontaine (2013:109–15).
The Clause 107
(22a) My eyes sparkle with tears

(22b) I shall now tell you my story
(22c) Will my mother cry at the sight of me
All three clauses are major clauses; they all have Predicators. And all three
are indicative; they all have Subjects. But in (22b) and (22c) the Predicators
are split between a Finite part and a main Lexical part: shall . . . tell and
will . . . cry. In (22a) the Predicator is not split – sparkle is itself a finite verb –
the Finite part and the main Lexical part have been conflated.
To return to the realization of the choice between ‘declarative’ and
‘interrogative’, in a declarative clause the Subject precedes the Finite part
of the Predicator, but in an interrogative clause the Subject follows
the Finite part. (22a) and (22b) are both declarative while (22c) is
interrogative.
This is in fact my working definition of the Subject, that it is the constitu-
ent in indicative clauses that inverts with the Finite verb to show whether
the clause is declarative or interrogative.13 Where there is already an
auxiliary verb to carry the finiteness, it is simply a matter of inversion.
We can turn the declarative (22b) into an interrogative simply by inverting
the I and the shall – Shall I now tell you my story. We can turn the interrogative
(22c) into a declarative simply by inverting the Will and my mother – My
mother will cry at the sight of me. Where there is not already an auxiliary
to turn a declarative into an interrogative, it is necessary to import one.
To turn (22a) into an interrogative, we would need to import do – Do my
eyes sparkle with tears. (For more on Subject and Finite, see Fontaine
2013:110–20.)
What I have just said is truer of yes/no interrogatives than of
wh-interrogatives. While it is the case that wh-interrogatives usually
involve inversion of Subject and Finite verb, this is not always so. What
really realizes a wh-interrogative is of course the presence of a wh-word –
who, what, where, when, why, how. When the wh-word is itself the Subject,
there is no inversion.
As well as the main MOOD options already discussed, there are other
possibilities. For instance, tag phrases may be used. A clause with declara-
tive form may be given an interrogative tag – e.g. My eyes sparkle with tears,
don’t they or I’ll now tell you my story, shall I. An imperative may be ‘softened’
by a tag – Imagine, will you, that you are standing on a race track or Imagine you
are standing on a race track, will you.
As usual there is much more that could be said. But hopefully I have said
enough to show what Halliday means by saying that the realizations of
13
I am here taking a more syntactic, less semantic view of the Subject than is probably usual in SFL these days. In my
view, we still need a syntactic perspective if we are to account for the ways in which the MOOD options are realized.
But see the references in footnote 12.
108 MARGARET BERRY
interpersonal meanings are ‘strung throughout the clause’. Experiential

options give rise to discrete constituents, those representing processes,
participants, and circumstances. The realizations of the interpersonal
MOOD options may involve the presence or absence of an experiential
constituent, the splitting of the constituent that carries the process, the
ordering of the Finite and the Subject, and/or the addition of a tag, and the
latter may be in the middle of the clause or at the end, as in the examples
I have just given.
4.3.2 MODALITY
‘Modality’ is the name which SFL gives to the whole area of clause rank
speaker/writer assessment of what is being said. If a clause is communi-
cating information, the assessment may be of the probability that the
information is true. In the made-up examples (23a) to (23c) different assess-
ments are made of the degree of probability. (The assumption is that all of
them are spoken on hearing someone arriving.)
(23a) That may be John now

(23b) That will be John now
(23c) That must be John now.
Or if a clause is negotiating an action, the assessment may be of the

desirability or practicability of the action. (24a) to (24c), again made-up,
offer different assessments of the desirability of going to the meeting.
(24a) You may go to the meeting (if you wish)

(24b) You ought to go to the meeting (it’s an important one)
(24c) You must go to the meeting (it’s crucially important)
Assessments of the truth of information are termed by SFL ‘modalization’.

Assessments of the desirability/practicability of an action are termed by SFL
‘modulation’. (In other approaches to linguistics, modalization is often
termed ‘epistemic modality’, while modulation is often termed ‘deontic
modality’.)
Each of these main categories can be subdivided. Assessments of infor-
mation may be about the probability of the truth of information, as in (23a)
to (23c). Or they may be about the usuality of the truth of the information,
as in (25).
(25) John is usually home by now
Assessments of actions may be about the desirability/permissibility/obliga-

toriness of the actions, as in (24a) to (24c). Or they may be about the
practicability of the action, in terms of the willingness/ability of the Actor,
as in (26a) and (24b).
The Clause 109
Figure 4.6 A simplified MODALITY network for English
(26a) I will go to the meeting (if you want me to)

(26b) I can go to the meeting (I’ve nothing else scheduled for that time)
Charlotte’s modalities are mostly of the willingness/ability subtype of the

modulation type of modality, as in (27a) to (27d), all from her horror story.
(27a) Ghosts can’t rest unless their death was fair

(27b) I will not do this
(27c) I could feel the blood trickling down my chest
(27d) I cannot stop myself
However, she does include a few assessments of probability and one assess-
ment of usuality. (28a) and (28b) are from the horror story, (28c) to (28e)
from the day in her life.
(28a) They might try to console themselves

(28b) Maybe their hearts will start to hammer in their chests
(28c) You may be wondering why I say this
(28d) You may share my opinion
(28e) Usually breakfast is toast and peanut butter
There is a further modality choice which cuts across the ones discussed so
far (see Figure 4.6).14
Cutting across the choice between modalization and modulation is a
choice among low, median, and high. This has to do with the strength of
the assessment. In examples (23a) to (23c), repeated below, (23a) represents
14
For a more detailed version of this network, see Thompson (2014:77). For an even more detailed version, see
Halliday and Matthiessen (2014:185).
110 MARGARET BERRY
an assessment of low probability, (23b) an assessment of median probabil-

ity, and (23c) an assessment of high probability.

(23b) That will be John now
(23c) That must be John now
In examples (24a) to (24c), again repeated below, (24a) represents an assess-

ment of low obligation, (24b) an assessment of median obligation, and (24c)
an assessment of high obligation.
(24a) You may go to the meeting (if you wish)

(24b) You ought to go to the meeting (it’s an important one)
(24c) You must go to the meeting (it’s crucially important)
The modality options may be realized in a number of ways. In most of the

examples I have given so far, the realization is by modal auxiliary verbs,
such as may, will, and must. Alternatively the realization may be by modal
Adjuncts. Or by a combination of modal auxiliary and modal Adjunct.
Alternative realizations of low probability modalization are shown in
(23a), repeated below, and (29a) to (29d).

(29a) That’s possibly John now
(29b) Perhaps that’s John now
(29c) That may possibly be John now
(29d) That may be John now perhaps
Where there is a modal Adjunct, such as possibly or perhaps, this may be at the
beginning of the clause, as in (29b), or in the middle as in (29a) and (29c), or at
the end as in (29d). Hopefully this can be regarded as another illustration of
what Halliday means by saying that the realizations of interpersonal mean-
ings are ‘strung throughout the clause’.15 Bartlett (2014:113) provides a
helpful table summarizing modality options and showing their realizations.
4.4 The Textual Metafunction
The third metafunction to be discussed is the textual metafunction. This

is the function of language to weave together the experiential and
15
Halliday and Matthiessen (2014:183–4) say ‘when we move around the languages of the world, we find a great deal
of variation in the grammaticalization of modality and other types of interpersonal judgement’. They go on to discuss
some of the differences.
The Clause 111
interpersonal meanings and present them as text. Bartlett (2014:131) says:

‘If experiential and interpersonal elements are the threads of a text, we can
say that the textual metafunction refers to the way in which these elements
are woven together to make a patterned cloth.’ This section will consider
the part that the structure of the clause plays in the weaving.
Halliday (2002:208) says that textual meanings will tend to be realized
‘culminatively’: ‘What the textual component does is to express the par-
ticular semantic status of elements in the discourse by assigning them to
the boundaries; this gives special significance to “coming first” and
“coming last”.’
As far as the clause is concerned, SFL regards two pairs of concepts as
being relevant to ‘coming first’ and ‘coming last’: Theme and Rheme, and
Given and New. Theme is relevant to ‘coming first’ in the clause, and New is
relevant to ‘coming last’, though the relevance of the latter is indirect, as
I shall explain. When considering the weaving together of experiential and
interpersonal meanings into the text, the questions to be asked are: Which
bits of meaning are regarded as Theme and so come first? Which bits of
meaning are regarded as New and so come last?
This section is mainly going to be about Theme and Rheme, but first a
brief word about Given and New.16
4.4.1 Given and New

Mainstream SFL assumes that Given and New are realized phonologically
through intonation (see Debashish, this volume, for a more detailed discus-
sion). However, it is necessary to say a little here in order to explain the
connection with the structure of the clause.
SFL recognizes an ‘information unit’ which is realized by a ‘tone group’ –
that is by a particular pattern of pitch movement. In the unmarked case a
tone group will be co-extensive with a clause. Again in the unmarked case, a
tone group will consist of a stretch of relatively level pitch followed by the
main pitch movement. The likelihood is that the relatively level stretch will
be presenting information as Given – that is, as information that is already
known – while the main pitch movement will be presenting information as
New, or particularly newsworthy. The most usual place for the main pitch
movement is on the last experiential constituent of the clause. Because of
this, the last experiential constituent of the clause gains an association with
newness or newsworthiness, and even in written language this is where one
tends to place what represents the main point of what one is saying.
In a recent article I wrote example (30a). The publisher’s copy editor
changed it to (30b). I changed it back to (30a). The methods of development
16
Textual meaning is sometimes regarded as ‘second-order’ meaning, since it is about the arrangement and relative
prominence of bits of experiential meaning and interpersonal meaning. For discussion, see Matthiessen (1992).
112 MARGARET BERRY
were the main point, and the new point in that paragraph, of what I was
saying, and I wanted them to have end focus.
(30a) This means that, like Child C, Writer 2 in both his passages has clear
methods of development
(30b) This means that, like Child C, Writer 2 has clear methods of devel-
opment in both his passages
It is probably for similar reasons that Charlotte preferred (18b), repeated

below, to (31a), even though this meant placing the Token after the Value
(see Section 4.2.3 above). Having introduced her eyes in New position in
(18b), she is then able to continue with (31b), the eyes now being repre-
sented as Given, with the tears now represented as New.
(18b) The only reminder of my original features are my eyes

(31a) My eyes are the only reminder of my original features
(31b) They sparkle with tears
‘Coming last’, then, is significant in the clause, as a result of the association,

in the unmarked case, between the last experiential constituent of the
clause and the main pitch movement of the tone group, this in turn being
associated with New information. This indirect association enables the
weaving of experiential meaning into the text to take account of the relative
givenness/newness of the experiential meanings on offer.17
4.4.2 Theme and Rheme

While New is relevant to coming last in the clause, Theme is relevant to
coming first. Theme is in fact the name that SFL gives to the beginning of a
clause. The rest of the clause is the Rheme. One might expect that what
comes first in the clause will always be Given. But things are not as simple
as that. In example (31b), They, which comes first, is indeed Given, in the
sense that its referent has been mentioned in the previous clause. But in
(32a) and (32b), again from Charlotte’s horror story, the most given items –
I and it – do not come first. Given and Theme are in principle distinct,
though they may co-occur.
(31b) They sparkle with tears

(32a) Tonight I shall visit their house
(32b) Like a glimmer of hope, it represents the tiny spark of humanity left
in the souls that wander the graveyard
17
For different definitions of ‘Given’ and ‘New’, see Berry (forthcoming).
The Clause 113
Halliday and Matthiessen (2014:89) say that the Theme is ‘that which
locates and orients the clause within its context’. The orienting may take
a number of different forms. For instance, the Theme, the first experiential
constituent in the clause, may be indicating that the text is staying with a
topic entity already mentioned, as in (31b). Or it may indicate a moving on
in time, as in (32a). Or it may simply be setting the scene, as in (32b).18
We have already seen that first position in the clause plays an important
part in English in realizing some of the interpersonal options. For a yes/no
interrogative, the finite verb will usually be in first position; for a wh-
interrogative, it will usually be the wh-word; and for an imperative, the
lexical verb may be in first place.
In a declarative clause, the most usual thing to happen will be that the
Subject will be at the beginning. When this happens, SFL calls it ‘Unmarked
Theme’. The first three clauses of the second paragraph of Charlotte’s
horror story all have Unmarked Themes.
(33a) I am a ghost
(33b) I wear a long white dress
(33c) My hair flows down my back in silvery strands
We then get the clauses about her eyes, quoted above, and then in the rest
of the paragraph the main clauses all have I Subjects, these being in first
experiential place and so being Unmarked Theme. This is the paragraph in
which Charlotte establishes the persona she has adopted for the story. The
repeated I’s as Unmarked Theme keep this persona firmly in view.19
Subject in first position then is the unmarked order for a declarative
clause. But other elements of the clause may precede the Subject for special
effects, in which case they are regarded as ‘Marked Themes’. Perhaps the
most common kind of Marked Theme is where an Adjunct, a constituent
representing a transitivity Circumstance, precedes the Subject. Examples
(34a) to (34d) are all from the day in Charlotte’s life.
(34a) Blearily I exposed one eye to the sunlight streaming through my

blinds
(34b) Before, I was driven to school
18
Opinions differ in SFL as to how much of a clause should be regarded as the beginning. Is the Theme just the very first
experiential constituent, or should it include any other experiential constituent near the beginning that has an orienting
function? I have said that They in (31b) has an orienting function in that it shows the text is staying with a topic entity
already mentioned. But I in (32a) and it in (32b) also have this function. Should they not also be included in the
Themes of their respective clauses? If one takes the notion of orienting function seriously, there are grounds for saying
that the Theme extends right up to the main lexical verb. Indeed in a language such as Spanish the Theme may even
include the lexical verb as orienting functions carried by early parts of clauses in English are carried by verbal inflections
in Spanish. For discussion, see Berry (1996).
19
Fries (1981) links Theme with what he calls the ‘method of development’ of a text. The pattern of the experiential
constituents selected to be Theme shows how the subject matter of the text is being organized and developed.
114 MARGARET BERRY
(34c) Suddenly it’s lunchtime

(34d) On such days I’ll spend the hours in peaceful tranquillity
Sometimes a whole subordinate clause may precede the Subject. Charlotte

makes a good deal of use of this option. Here are just a few examples, (35a)
from the horror story, (35b) from the day in the life, (35c) from the health
and nutrition work, and (35d) from the history essay.
(35a) As she drifts closer, I can hear her rasping breath

(35b) After breakfast has long passed, I’ll leave the house at eight
(35c) And as my smoothie was summery, I thought it would add a refresh-
ing aftertaste
(35d) Although this is a small detail, it shows that William cared about
his army
Rather less common are instances of a Complement preceding a Subject.

(The term ‘Complement’ is being used here to refer to an element which
represents a transitivity participant such as ‘Goal’ or ‘Attribute’, which in
English would more usually occur after the main lexical verb. In other
words it is the Object/Complement of traditional grammar.) I did not find
any instances of Complement Marked Theme in Charlotte’s writing. But in
(36), Lucy, a sixteen-year-old friend of Charlotte’s, uses both Adjunct
Marked Theme and Complement Marked Theme.
(36) In groups we worked out our things we needed to test to be able to

use as evidence to prove our hypothesis right or wrong. Later on after
having the results then we would work out the cross sectional area.
The width we measured 3 times and the depth 11 times to get it
accurate. The channel velocity we tested three times and used two
separate experiments which were with the hydro prop and recording
the time it took for the dog biscuit to travel across 10 metres of the
river at each site. We visited 4 different sites at each point of
the river.
The paragraph begins with an Adjunct Marked Theme In groups. Lucy is

clearly very aware of the group nature of the work in which she is engaged,
and this is picked up in the we Subjects. She also uses Adjunct Marked
Theme to indicate a shift in time reference – Later on after having the results
then. Complement Marked Themes are found in The width we measured 3 times
and The channel velocity we tested three times. And if we regard the depth 11 times
as an elliptical clause – the depth [we measured] 11 times – we would have
another Complement Marked Theme here. In the first sentence of the
paragraph, Lucy mentions our things we needed to test. Then later in the
paragraph, she uses Complement Marked Themes to highlight the things
tested.
The Clause 115
Earlier in this subsection, I drew attention to the part played by Theme

in realizing interpersonal options such as yes/no interrogative,
wh-interrogative, and imperative. So far in relation to declarative clauses,
I have been discussing the arrangement of experiential meanings. But
declarative clauses too can have interpersonal material in their Themes.
Modal Adjuncts, of the kind discussed in Section 3.2, can occur at the
beginning of a clause. (28b) and (28e), repeated below, provide examples
of this.
(28b) Maybe their hearts will start to hammer in their chests

(28e) Usually breakfast is toast and peanut butter
Halliday and Matthiessen do not regard Modal Adjuncts as Marked

Themes. That term is reserved for experiential material that precedes
the Subject. Instead they write of ‘multiple Theme’ (e.g. Halliday and
Matthiessen 2014:107). In (28b) Maybe their hearts would all be regarded
as Theme, with an interpersonal component Maybe and an experiential
component their hearts. Similarly in (28e) Usually breakfast would all be
the Theme, with an interpersonal component Usually and an experiential
component breakfast. In each case the Theme would be regarded as
Unmarked, since the Subject is the first experiential component, but
as multiple Theme, since it contains interpersonal material as well as
experiential.
Something else that can occur at the beginning of a clause, without
Halliday and Matthiessen regarding it as Marked Theme, is a word or
phrase which is specifically functioning to show how the text fits together.
Conjunctions would count here, as would conjunctive Adjuncts, such as for
instance, in addition, and therefore. In (37), from Charlotte’s horror story,
Besides would be regarded as a conjunctive Adjunct.
(37) Besides, I could never kill my parents
Where a multiple Theme includes material from all three metafunctions,

the likely order is specifically textual Theme followed by interpersonal
Theme followed by experiential Theme. Charlotte has a rather nice mul-
tiple Theme in the day in her life.
(38) I don’t know about you but, controversially in the evening when I’m going
to bed, I don’t feel tired
But would be textual Theme, controversially would be interpersonal Theme,

and in the evening when I’m going to bed would be experiential Theme. The
experiential Theme would be Marked Theme as it precedes the Subject I of
the main clause.
Figure 4.7 summarizes the declarative clause options I have been discuss-
ing in this subsection. As usual there is a great deal more that could be said.
Thompson (2014:147–81) is particularly helpful on Theme.
116 MARGARET BERRY
Figure 4.7 Thematic options for declarative clauses
For a more detailed network for Theme, see Thompson (2014:170). For a
much more detailed network, see Halliday and Matthiessen (2014:106).
4.5 Concluding Remarks
This chapter has discussed:
• choices from the experiential metafunction, showing how these lead to

the number and nature of the basic constituents of a clause;
• choices from the interpersonal metafunction, showing how the realiza-
tions of these are threaded in among the basic constituents;
• choices from the textual metafunction, showing how these relate to
what comes first and what comes last in the clause.
Many interesting questions have had to be left undiscussed, for reasons of

space. However, I have given references to where more detailed discussions
can be found.
References
Bartlett, T. 2014. Analysing Power in Language: A Practical Guide. London:

Routledge.
Berry, M. 1975. An Introduction to Systemic Linguistics, Volume 1: Structures and
Systems. London: Batsford.
Berry, M. 1996. What is Theme? A(nother) Personal View. In M. Berry, C. S.
Butler, R. P. Fawcett, and G. Huang, eds., Meaning and Form: Systemic
Functional Interpretations. Meaning and Choice in Language: Studies for Michael
Halliday. Norwood: Ablex Publishing Corporation. 1–64.
The Clause 117
Berry, M. 2016. Interpersonal Meanings, with Particular Reference to Getting

People to Do Things. Paper for seminar, Vigo, February 2016.
Berry, M. forthcoming. ‘Actually Given’ Versus ‘Presented as Given’ and ‘Actually
New’ Versus ‘Presented as New’: What Happens when the ‘Presented as’ Gets out of
Step with the ‘Actually’? Paper presented at the 2nd Round Table on Com-
municative Dynamism, Namur, Belgium, September 2016.
Fawcett, R. 1999. On the Subject of the Subject in English. Functions of
Language 6(2): 243–73.
Fontaine, L. 2013. Analysing English Grammar: A Systemic Functional Introduc-
tion. Cambridge: Cambridge University Press.
Fries, P. H. 1981. On the Status of Theme in English: Arguments from
Discourse. Forum Linguisticum 6: 1–38.
Gwilliams, L. and L. Fontaine. 2015. Indeterminacy in Process Type Classifi-
cation. Functional Linguistics 2(8): 1–19.
Halliday, M. A. K. 2002. Modes of Meaning and Modes of Expression: Types
of Grammatical Structure and Their Determination by Different Seman-
tic Functions. In J. J. Webster, ed., On Grammar: Collected Works of M. A.
K. Halliday. Vol. 1. London: Continuum. 196–218.
Benjamins.
Matthiessen, C. M. I. M. 1992. Interpreting the Textual Metafunction. In
M. Davies and L. Ravelli, eds., Advances in Systemic Linguistics: Recent Theory
and Practice. London: Pinter. 37–81.
O’Donnell, M., M. Zappavigna, and C. Whitelaw. 2009. A Survey of Process
Type Classification over Difficult Cases. In C. Jones and E. Ventola, eds.,
From Language to Multimodality: New Developments in the Study of Ideational
Meaning. London: Continuum. 47–64.
Thompson, G. 2014. Introducing Functional Grammar. 3rd ed. London:
Routledge.
5
The Rooms of the House
Grammar at Group Rank
Lise Fontaine and David Schönthal
5.1 Introduction
One morning I shot an elephant in my pyjamas

How he got into my pyjamas, I’ll never know! (Groucho Marx1)
Jokes like Groucho Marx’s famous ‘elephant in my pyjamas’ illustrate how we

process units at an intermediate level between words and clauses (see Fon-
taine 2013). This is also evidenced by other ambiguous structures such as I saw
the man from next door, where we get a different meaning depending on the
way in which the units relate to each other, i.e. the man from next door is the
person that I saw or I was next door when I saw the man. The distinction centres on
whether the man from next door functions as one unit or two. The main point
here is that there is an internal structure to the clause; it cannot be seen as a
string of words. As Halliday and Matthiessen (2014:362) explain, ‘Describing a
sentence as a construction of words is rather like describing a house as a
construction of bricks, without recognizing the walls and the rooms as
intermediate structural units.’ Understanding these intermediary units, i.e.
the rooms of the house, is essential to our understanding of meaning.
Within the framework of Systemic Functional Linguistics (SFL), these
units are viewed from the perspective of the function they serve in a higher
unit, which is, in this case, the clause. It is generally accepted that between
the level or rank of clause and word there are two different types of unit,
group and phrase, which are roughly equivalent. Halliday maintains that ‘a
phrase is different from a group in that, whereas a group is an expansion of
a word, a phrase is a contraction of a clause’ (Halliday and Matthiessen
2014:362–3). The differences between these two will be explained in detail
throughout Section 5.2 and 5.3. However, we can think of the difference in
We would like to thank Margaret Berry for her very useful comments on drafts of this chapter.
1
From the film Animal Crackers (1930).
The Rooms of the House 119
the following terms: groups form around a particular type of word, for
example, a noun forms a group called the nominal group, such as the
beautiful scenery, where the other words in the group somehow modify the
noun, in this case scenery. A phrase, in contrast, is less like a relationship
among words and rather combines two components through complemen-
tation, as in the prepositional phrase by the lake, where the lake is seen as a
complement to the preposition by.
In this chapter we explore these two types of intermediary unit in SFL. We
take as given that there are units larger than the word and smaller than the
clause. There is generally considerable agreement in terms of the descrip-
tions of these intermediary units (see McDonald 2017 for an excellent over-
view of the historical development of groups). However, there is also room
for debate. The aim of this chapter is to critically examine the theoretical
reasons for including two fundamentally different types of grammatical unit
between clause and word. The main question we ask is whether these
reasons hold in all cases, i.e. is it theoretically justified to maintain two types.
The way we will approach this is as follows. In the next section, we provide
an overview of the existing description of the units below the clause as
currently represented in the theory. Following this, Section 5.3 will present
and evaluate three main criteria for classifying a unit as either a group or a
phrase. These distinctions, drawn primarily from Matthiessen (1995) and
Halliday and Matthiessen (2014), include (i) the concept of (primary) class
and the relation between the functional potential of the unit and the ‘head
word’; (ii) univariate versus multivariate structures; and (iii) the role of
rankshifted units. In considering these perspectives on the units at the
intermediate rank between words and clauses, we conclude that there is no
significant theoretical or practical value in maintaining two different types of
unit at this level. We argue that it is important to ask questions such as those
we propose here in order to evaluate the strength of the position of the
theory and its usefulness in an appliable theory of language.
5.2 An Overview of Grammatical Units below the Clause
In this section, we provide an overview of the grammatical units which

function between the rank of clause and word. In Halliday’s important
1961 paper ‘Categories of the Theory of Grammar’, he describes the rela-
tionship between grammatical units and the concept of rank as follows:
The category set up to account for the stretches that carry grammatical
patterns is the ‘unit’. The units of grammar form a hierarchy that is a
taxonomy. . . . The relation among the units, then, is that, going from top
(largest) to bottom (smallest), each ‘consists of’ one, or of more than one, of
the unit next below (next smaller). The scale on which the units are in fact
ranged in the theory needs a name, and may be called ‘rank’.
(Halliday 1961:251)
120 LISE FONTAINE AND DAVID SCHÖNTHAL
Table 5.1 Example of the rank scale
Rank Scale Example
Clause unit The brown foxes were jumping over the laziest dog
Group unit [The brown foxes] [were jumping] [over the laziest dog]
Word unit [The / brown / foxes] [were / jumping] [over / the / laziest / dog]
The rank scale is described in Halliday and Matthiessen (2014:5) as ‘a

hierarchy of units, related by constituency’. The principle is then that any
unit at a given rank consists of units of a lower rank, e.g. a clause unit
consists of one or more group units and a group unit consists of one or
more word units.2
The terms, e.g. clause and word, reflect the type of unit at each rank and
each one then has different sub-classes, i.e. different types of clause, group,
and word. This is illustrated in Table 5.1 above. In this example, we have
included only the rank immediately ‘above’ the group unit, i.e. clause, and
immediately below, i.e. word, in order to simplify the illustration. It is also
possible for a unit to be embedded, or rankshifted, for example, when a
group unit contains a clause or another group unit. In addition to the
ranking of these units, every unit can form a complex which means that
it is possible for any unit to be multiplied recursively into a unit complex
such as a clause complex, group complex, or word complex.
While complexing is not central to the rank of unit between clause and
word, there is an important distinction to be made between a group, such as
the nominal group, and a word complex. This is a point that we will return
to in the discussion of the nominal group in Section 5.2.1 and in our
evaluation of the distinction between group and phrase in Section 5.3.
However, very briefly, it is worth explaining that while groups are often
discussed as groups of words, the group is a unit that consists of words
(units of a lower rank) and may include rankshifted units, as will be
explained below. It is a structure that has a relatively fixed order of con-
stituents. A word complex, on the other hand, is a linear arrangement of
two or more words (units of the same rank) involving some kind of depend-
ency, where the order is not fixed, i.e. a different order results in a different
meaning, as shown in the expressions in italics in examples (1) and (2).3
(1) the county plans to spread the money among a mixture of govern-
ment securities and bank investments (EnTenTen13, SketchEngine)
(2) I once worked in an investment bank (EnTenTen13, SketchEngine)
2
As introduced later in this chapter, we use the term ‘group unit’ to refer to the units at group rank, i.e. both groups and
phrases alike. Further, for a full account of how rank relates to the key SFL concepts of stratum, delicacy, and realization,
see Berry (2017).
3
For details of SketchEngine, see Kilgarriff et al. (2014) or online: www.sketchengine.co.uk.
The distinction is described by Halliday and Matthiessen (2014:437) as

follows: ‘[t]reating the group simply as a “word complex” does not account
for all these various aspects of its meaning’. They illustrate this point by
comparing ‘railway ticket office staff, which could be explained as a (univari-
ate) word complex, [with] that of these two old railway engines, which could
not’ (Halliday and Matthiessen 2014:437, emphases in original).
In addition, there is also a phonological rank scale (e.g. Halliday 1961;
Halliday and Matthiessen 2014; also see Debashish and Bowcher, this
volume). According to Halliday and Matthiessen (2014:15), the tone group
‘does a great deal of work in the construal of meaning: it organizes continu-
ous speech as a sequence of units of information’. In studying spoken
language, an understanding of the tone group is essential. See Halliday
and Matthiessen (2014) for a fuller account of the rank scale and Fawcett
(2010) for a critical review.
In Halliday’s first account of the rank scale, he explains the hierarchical
ranking of grammatical units as follows:
For English, for the two units between sentence and word the terms ‘clause’
and ‘phrase’ are generally used. It is at the rank of the phrase that there is most
confusion – because there are here the greatest difficulties – in the description
of English; one reason is that in English this unit carries a fundamental ‘class’
division, so fundamental that it is useful to have two names for this unit in
order to be able to talk about it: I propose to call it the ‘group’, but to make a
class distinction within it between ‘group’ and ‘phrase’.
(Halliday 1961:252–3)
Thus, Halliday proposes two types of unit, group and phrase, at the same
rank on the scale, effectively having two different types of unit with no
difference in rank. We will use the term ‘group unit’, as in Table 5.1, as an
umbrella term for units at this rank. This avoids having to say ‘groups and
phrases’ which is not only lengthy, but there is only one phrase and it
seems reasonable to have a single term for each rank along the scale. In this
paper, we have opted for the term ‘group’ rather than ‘phrase’ simply
because it is by far the more common term in Systemic Functional Linguis-
tics and as we will argue below can be applied to all intermediary units. The
difference between groups and phrases is described by Halliday and Mat-
thiessen (2014:262–3) as follows:
A phrase is different from a group in that, whereas a group is an expansion

of a word, a phrase is a contraction of a clause. Starting from opposite ends,
the two achieve roughly the same status on the rank scale, as units that lie
somewhere between the rank of a clause and that of a word.
This suggests that a group is in some sense more ‘word-like’ whereas a

phrase is more ‘clause-like’. This is a perspective that will be challenged in
Section 5.3. It is important to note here that work by Fawcett (1980, 2010)
and Tucker (1998, 2017) presents perhaps the most detailed account of
Figure 5.1 Word classes in SFL (adapted from Halliday and Matthiessen (2014:75))
these intermediary units, although this is from a slightly different SFL theor-
etical approach, the Cardiff Grammar (see Schulz and Fontaine, this volume).
There is another difference between group and phrase that must be covered
briefly before moving on to the description of each class of unit at the group
unit rank. This concerns the metafunctional nature of group units. In SFL, the
approach to lexicogrammatical description at the group unit rank parallels
the description of the clause in terms of the three main metafunctions. There
is an assumption that all group units express, at least to some extent, experi-
ential, interpersonal, and textual meaning. Halliday and Matthiessen
(2014:361) explain that ‘[a]lthough we can still recognize the same three
components, they are not represented in the form of separate whole struc-
tures, but rather as partial contributions to a single structural line’.4 When it
comes to logical meanings, however, we find that phrases, unlike groups, are
said to not express logical meanings. This metafunction explains one of the
key differences between groups and phrases. While groups have both experi-
ential and logical structure (Halliday and Matthiessen 2014:361–2), phrases
do not express logical structure (Halliday and Matthiessen 2014:425). This is a
distinction we will return to in Section 5.3 so that we can critically examine
some of the implicit assumptions that underlie this claim.
One final distinction must be discussed in relation to the description of units
in SFL. We mentioned above that within SFL, units are defined in terms of the
function they serve to express in the unit above, rather than by structural
similarities that groups might serve. This is what motivates Halliday’s classifi-
cation of units, and it explains why, for example, adverbials are classed differ-
ently than adjectives, but also why nominal groups and prepositional phrases
are seen as different classes in English, whereas in other languages they might
not be. The three primary classes of word are shown in Figure 5.1, where each
primary class corresponds to one of the ‘three main classes of group: nominal
group, verbal group and adverbial group’ (Halliday and Matthiessen 2014:362,
emphases in original). These classes are defined by the functions served in the
clause. For example, there is a tendency for nominal groups to function as
Subject (or Complement) and/or Actor (or Goal), whereas adverbial groups serve
4
A detailed account of the metafunctions at clause rank is given in Berry (this volume).
to express Adjunct or Circumstance roles. In the early development of SFL,

Halliday (1961:261, emphases in original) explains class and unit as follows:
A class is always defined with reference to the structure of the unit next
above, and structure with reference to classes of the unit next below. A class
is not a grouping of members of a given unit which are alike in their own
structure. In other words, by reference to the rank scale, classes are derived
‘from above’ (or ‘downwards’) and not ‘from below’ (or ‘upwards’).
As stated above, the principal distinction between units at this rank

concerns two main features: (i) the secondary type of class which forms
the basis for the unit, e.g. adverb vs. conjunction, and (ii) the nature of the
relationship between the elements of the group (e.g. the nominal group
consists of nominal elements). There are five groups generally recognized
within SFL (but see Fawcett 2010 and Schulz and Fontaine, this volume, for
some differences). Each has as its head element, a member of the relevant
class. For example, as suggested in Figure 5.1, the nominal group can have
as its head element a noun (Thing) or another nominal, e.g. an adjective
(Epithet), as will be explained below. Groups are formed by expansion of the
head through modification. The five groups are listed below in (3) to (7),
along with illustrative examples where the base (head) element is under-
scored. The example given in (8) is a prepositional phrase. The phrase in SFL
is not seen as having a word basis but rather as a type of reduced clause.
This is due to the relational function of the preposition and the comple-
mentation function of the nominal group. It is difficult to claim that the
complement nominal group, in this case the box, modifies the preposition in
any way. All examples in (3) to (8) have been attested in EnTenTen13 via
SketchEngine.
(3) a beautiful poem, nominal group (Ngp)

(4) was eating, verbal group (Vgp)
(5) very quickly, adverbial group (Advgp)
(6) right under, prepositional group (Pgp)
(7) just as, conjunction group (Cgp)
(8) in the box, prepositional phrase (PP)
In what follows, each group and phrase will be described in turn. Following
this, in Section 5.3, we will examine the criteria for the distinction made
here between the unit of group and phrase.
5.2.1 Nominal Group

The discussion of groups has so far presented the units in terms of logical
meanings, i.e. expansion through modification. However, this seems to
have been based on the idea of the group as word complex rather than on
Table 5.2 The nominal group (EnTenTen13, SketchEngine)5
Deictic Numerative Epithet Classifier Thing Qualifier
Those two popular lunch stops with an amazing view
the idea of constituency. As Matthiessen (1995:662) points out, when

viewed experientially, what is construed is not ‘an expansion of a Thing
(which is the logical perspective) but as a configuration of roles represent-
ing different aspects of a participant – a multivariate structure’. The nom-
inal configuration includes the following experiential elements: Deictic,
Numerative, Epithet, Classifier, Thing, and Qualifier. An example is given
in Table 5.2.
Each element of the Ngp contributes its own function. The Thing element
functions as the ‘semantic core’ (Halliday and Matthiessen 2014:383). It
classifies the entity being referred to (see ‘cultural classification’ in Fawcett
1980; Tucker 1998). In most cases, it also serves as Head element in the
logical structure, but there are cases where the two strands of meaning
diverge, as will be explained below. The Deictic has a determiner function
in the sense of indicating phoric relations (see Martin 1992) and indicates
whether the entity being referred to is specific or not and indeed whether
the addressee should be able to identify the referent. In this sense, it has the
potential to carry an implicature of definiteness or uniqueness in so far as
the Deictic element can specify the referent in some way (i.e. which thing it
is). The Numerative, as its name suggests, serves to specify the quantity or
amount of the thing being referred to. Epithets and Classifiers are modifiers
that either describe or depict the referent. Epithets are typically adjectives
that function as qualities of the referent, e.g. a lovely visit, whereas Classi-
fiers are typically nouns that sub-classify the referent along with the Thing
element, e.g. an office chair. This distinction is not strict, and it is possible for
there to be some overlap in this area. See Halliday and Matthiessen
(2014:376–8) for a detailed discussion.
The Qualifier element is defined generally as the element that occurs
after the Thing element (Halliday and Matthiessen 2014:381). By its nature,
the Qualifier is always expressed by an embedded (rankshifted) element,
but this is most frequently a PP. The relationship between the Thing
element and the Qualifier is different from that of the Classifier. With the
Classifier, it is reasonable to describe the relationship as an ‘is-a’ relation-
ship, e.g. ‘an office chair’ is a type of chair as compared to ‘a chair for the
office’. However, with the Qualifier, the relationship is relational. Halliday
and Matthiessen (2014:382) describe it as having a characterizing function
5
The original expression included only a as a Deictic: in order to include a Numerative, two, the Deictic was changed to a
plural one, those, and stop was made plural as well. Finding a naturally occurring example with all elements included
was very difficult.
where ‘the characterization here is in terms of some process within which

the Thing is, directly or indirectly, a participant. It may be a major process,
i.e. a relative clause; or a minor process – a prepositional phrase’.
There is a strong relation between Thing and the Head of the Ngp in
English. In SFL, we expect that the Head of the Ngp will be conflated with
the Thing element, as in Table 5.3, as this is the most usual instance
(Halliday and Matthiessen 2014:390). This combination makes sense some-
how in that the semantic core of the referring expression is what will occur
in the role of Head.
The main distinction between Thing and Head for Halliday and Matthies-
sen (2014:392) is that Thing is an element of experiential structure and
Head is an element of logical structure. The element that expresses the
Head determines the type of nominal expression, which could conflate with
the Thing element, e.g. common noun, proper noun, or personal pronoun,
but there could be other elements as Head, for example, Deictic or Numera-
tive. In other words, the Head and Thing are not necessarily conflated in the
Ngp. For example, consider the analysis of a cup of tea in Table 5.4. In this
case, the Head is a Deictic or Numerative and the expression is called a
‘measure nominal’ (Halliday and Matthiessen 2014:375).
Halliday’s Ngp structure also includes the potential for characterization
through the Qualifier in experiential terms. The analysis of the children in
blue hats in Table 5.5 shows the Qualifier element being expressed by a PP,
which will be discussed below.
One difficulty with the presentation in Halliday and Matthiessen (2014) is
that much of the fine detail concerning the experiential structure of the
Ngp is missing where it concerns the Qualifier. Halliday and Matthiessen
(2014:390) state that the distinction between Premodifier and Postmodifier
is not functional, but rather depends ‘on the rank of the modifying term’,
i.e. whether it is rankshifted (embedded) or not. The distinction for Halliday
Table 5.3 Experiential and

logical analysis of the white cup
the white cup
Deictic Epithet Thing
Modifier Head
Table 5.4 Analysis of a ‘measure nominal’ (adapted from Halliday and

Matthiessen 2014:392)
nominal group a cup of tea
experiential structure Numerative Thing
Premodifier Head Postmodifier
logical structure
is in the information structure, with the Postmodifier having ‘the greater

potential as news’ (Halliday and Matthiessen 2014:390). See Fontaine
(2017a) for a detailed discussion of the problems with the logical analysis
and also Fawcett (2010) for an alternative without these problems (see also
Davidse 2004; Ghesquière 2014).
5.2.2 Verbal Group

The structure of the Vgp in English is complex for a variety of reasons: the
way in which the Finite element works in English, the way verbs combine
to express complex (secondary) tenses, and the way ‘words’ combine to
form complex verbal lexemes (e.g. ‘to make up’). All Vgps have at least an
Event element, which corresponds to a lexical verb. Finite Vgps also have a
Finite element which serves to give the clause a point of reference; it
provides a bounded limit to the clause. The Finite may either be conflated
with the first auxiliary verb in the Vgp (see Table 5.6) or, in the absence of
any auxiliaries, directly with the main verb. The question of headedness
within the Vgp could be seen as depending on the perspective taken: for the
interpersonal metafunction, it would be the Finite element and for the
experiential metafunction, it would be the Event element, which is
expressed by the main lexical verb (see Table 5.6). Arguably, the Finite
element also carries textual meaning which is similar to that of the Deictic
Determiner in the Ngp (see Halliday and Matthiessen 2014:397). In addition
to these two elements, the Vgp could include one or more Auxiliary
Table 5.5 Analysis of the children in blue hats (adapted from Halliday and
Nominal group
the children in blue hats
Logical structure Premodifier Head Postmodifier
Experiential structure Deictic Thing Qualifier
Prepositional Phrase
Process Range
Nominal group
Logical structure Premodifier Head
Experiential structure Epithet Thing
Table 5.6 The verbal group (EnTenTen13, SketchEngine)

might have been being paid
Finite/AUXMOD AUXPERF AUXPROG AUXPASS Event
elements and/or a Polarity element. Although the Auxiliary element, AUX,

is optional, there may be more than one type expressed: e.g. modal, per-
fective, progressive, and/or passive auxiliaries, as illustrated in Table 5.6
(see also Table 5.13 for an additional example).
There has been some debate in the literature as to whether or not the
Vgp should be seen as a group unit. Notably, Fawcett (1980, 2000a, 2000b,
2010) has argued for treating all verbal items as elements of the clause, i.e.
at clause rank rather than below clause rank. Fawcett (2000a, 2000b)
outlines four main arguments which relate to the treatment of phrasal
verbs, the role of the Finite (or Operator) in mood structure, the problem
of discontinuous items such as adverbial insertion, and theoretical simpli-
city in the system descriptions. There is no space here to enter into this
debate, but see Morley (2000) for counter-arguments and Quiroz (2017) for
a useful discussion of the theoretical considerations of the issues. For a
discussion of the Vgp in other languages, see McDonald (2017); Caffarel-
Cayron (2017).
5.2.3 Adverbial Group

Halliday does not discuss the experiential structure of the Advgp in detail,
but only offers a very brief discussion of the logical structure of this unit
(see Halliday and Matthiessen 2014:419). For a very detailed account,
see Tucker (1998), although the theoretical underpinnings differ from
Halliday’s. As a group unit based on the lexical class of adverb, the Advgp
is described in logical terms by the Head element in relation to the
modifying elements as shown in Table 5.7. Experientially, however, this is
somewhat problematic given the potential for this group to have a rank-
shifted postmodifier, as shown below (see Halliday and Matthiessen
2014:422)
‘[T]he embedded clause serves to represent a standard of comparison’
(Halliday and Matthiessen 2014:492, referring to Fries 1977). The Head
element of the Advgp can be expressed by ‘an adverb denoting a circum-
stance as Head – for example, a circumstance of time (e.g. yesterday, today,
tomorrow) or of quality (e.g. well, badly, fast, quickly, slowly). Advgps serving as
modal Adjunct have an adverb denoting an assessment as Head – for
example, an assessment of time (e.g. still, yet, already) or of intensity (e.g.
really, just, only, actually)’ (Halliday and Matthiessen 2014:419).
Table 5.7 Description of the adverbial group (adapted

from Halliday and Matthiessen 2014:422)
much more quickly than I could
Logical Modifier Head Postmodifier
Experiential Extent Temperer Quality Standard
See Tucker (1998; 2017) for an alternative analysis, including a convin-

cing alternative account for adjective groups (for which see below).
5.2.4 Prepositional Group

The Head element of the Pgp is a preposition which can be modified. This is
in contrast to the prepositional phrase, which will be discussed below.
Examples from Halliday and Matthiessen (2014:423) include: right behind,
not without, all along, way off as in right behind the door, not without some
misgivings, all along the beach, way off the mark. In these cases, right, not, all
and way modify their respective preposition, and there is a kind of logical
relationship of modification such that the meaning of the preposition is
altered by the modifier. Some instances of prepositions appear complex in
form, especially in written language, where they are represented ortho-
graphically by more than one word. For example, in front of and for the sake of
(from Halliday and Matthiessen 2014:423) are considered to be complex
prepositions, which are effectively single lexical items similar to compound
words, i.e. they are not analyzed compositionally.
The Pgp is only discussed in terms of expansion of the preposition, i.e. in
terms of the logical structure of Head and modifier. Halliday and Matthies-
sen (2014) do not provide an account of the experiential structure. It is
possible that there is no real motivation for a Pgp since it might be best
accounted for as a word complex (see Matthiessen 1995:626). Also see
Fawcett (2010) for an approach that has unified the Pgp and phrase into
one unit.
5.2.5 Conjunction Group

The conjunction group is described by Halliday and Matthiessen (2014:423)
as consisting of a single conjunction such as and, or, or but, but also includ-
ing more complex6 ones such as as soon as, by the time, or in case. As with
other types of word classes, they ‘can form word groups by modification,
for example, even if, just as, not until, if only’ (Halliday and Matthiessen
2014:423, emphases in original). Furthermore, conjunctions serve to
express three main functions: there are ‘binders’ (i.e. subordinating con-
junctions) such as which, who, or where (Halliday and Matthiessen 2014:482);
‘linkers’ (i.e. coordinating conjunctions) such as and, or, or but (Halliday and
Matthiessen 2014:454); and ‘continuatives’, such as well, also commonly
known as discourse markers (Halliday and Matthiessen 2014:107).
However, although Halliday and Matthiessen (2014) consider the con-
junction group as an element below the clause, it works quite differently
6
Although this depends on the lexical representation assumed. If these are single lexemes, i.e. multi-word expressions
constituting effectively a single ‘word’, then their morphological composition may be more complex in terms of the
formation of the item, but nevertheless only a single item in the lexicon.
from the other groups. We do find conjunctions within groups, as in the

Ngp the cows and horses, but there is no potential in these cases for modifica-
tion and therefore it is not possible for a conjunction group to appear below
the clause. The conjunction group only occurs between clauses. Thus,
unlike other groups, it seems restricted to the clause rank, i.e. it cannot
be embedded or rankshifted. For these reasons, we will consider it as a unit
between clauses rather than below the clause, i.e. a unit at clause rank only.
Finally, as with the Pgp, the conjunction group may well be better
accounted for as a word complex rather than a group.
5.2.6 Prepositional Phrase

As mentioned above, next to the groups discussed so far, there is also one
phrase, which differs from groups in that it is considered to be a contrac-
tion of a clause rather than an expansion of a word (see Halliday and
Matthiessen 2014:362–3). This phrase is the prepositional phrase, such as
on the burning deck (from Halliday and Matthiessen 2014:424). It is analyzed
as a contraction of a clause because the preposition arguably is no different
from a verbal non-finite predicator, which Halliday and Matthiessen
(2014:425) illustrate with the examples near/adjoining (the house), without/not
wearing (a hat), or about/concerning (the trial). An analysis of a PP as a minor
Process is given in Table 5.8.
A PP, according to Halliday and Matthiessen (2014:311–12),
is an odd sort of hybrid construction. It has a nominal group inside it, as a

constituent, so it looks bigger than a group; and yet it is still not quite
a clause. In English, this nominal group inside a prepositional phrase is
no different from a nominal group functioning directly as a participant
in a clause, and in principle every nominal group can occur in either
context.
A further feature that, according to Halliday and Matthiessen (2014:425),

distinguishes the PP from groups, is that ‘they have no logical structure
as Head and Modifier, and cannot be reduced to a single element’. Thus,
PPs always have to consist of a prepositional element (the minor Process)
and a completive (Range), whereas Ngps such as apples, for example,
can consist of the Thing element only. We will critically engage
with these distinguishing features between phrases and groups in
Section 5.3 below.
Table 5.8 The prepositional phrase

Click on the image you’ve chosen
Process Location
Process Range
5.2.7 Adjective Group

The final group unit to be mentioned here briefly is the adjective group. The
status of adjectives and the adjective group is an area of debate within SFL
literature (see Tucker 1998, 2017; Fontaine 2013, 2017a). As mentioned
above, the adjective class is a subtype of nominal (see Figure 5.1). In an
example such as You’re very lucky in Table 5.9 (taken from Halliday and
Matthiessen 2014:391), the Complement, very lucky, is thus considered as a
Ngp where the Head element is expressed by an Epithet. Therefore, in IFG,
there is no need for the adjective group as another group unit, but see
Tucker (1998).
5.2.8 A Worked Example of the Clause in Terms of Units

below the Clause
To end this section, we will consider the group unit analysis of the clauses
in Table 5.10 and Table 5.11. Also see Berry (this volume) for a discussion of
the three metafunctions at clause rank.
Table 5.9 Epithet headed nominal group (adapted from Halliday and
You ‘re very lucky
Carrier Process: relational Attribute
Nominal Group Verbal Group Nominal Group
Premodifier Head
Epithet
Table 5.10 Sample clause analysis (EnTenTen13, SketchEngine)

the sharp gorgonzola was an unusual choice for a filling
Process:
Carrier Attribute
relational
Verbal
Nominal Group Group Nominal Group
Finite/
Deictic Epithet Thing Deictic Epithet Thing Qualifier
Event
Prepositional Phrase
Process Range
Nominal Group
Deictic Thing
Table 5.11 Clause analysis with Advgp (EnTenTen13, SketchEngine)

The backups can be performed very quickly
Goal Process: material Circumstance: Manner
Nominal Group Verbal Group Adverbial Group
Deictic Thing Finite/AUXMOD AUXPASS Event Temperer Quality
While these two analyses illustrate the group unit structure of two basic
examples, other more complex examples would give rise to further com-
plex issues in the description of these units. Unfortunately, it is beyond the
scope of this chapter to deal with these here, but see Fontaine (2013) for a
more detailed discussion of the structure of group units.
5.3 Criteria for Distinguishing between Group and Phrase
Having outlined the current description of the various intermediary units,

this section will take a much closer look at the distinction between groups
and phrases. There are three main distinctions between these two units.7
These are discussed in various sources, but the presentation in this section
will draw primarily on Matthiessen (1995, especially 627) and Halliday and
Matthiessen (2014). We will first consider the relationship between class
membership and how groups are formed. Then, we will examine the
distinction between groups and phrases in terms of multivariate vs. uni-
variate structures. In the final part of this section, we consider how each
type of group unit relates to the rank scale.
5.3.1 Notion of Class Membership

The principle of class of unit works on the assumption that the sub-classes are
similar in kind in terms of function in the rank above, e.g. while nominals,
verbals, and adverbials as group units are on the same rank, they are not the
same class because they serve different functions in the class above (partici-
pant, process, and circumstance respectively). One of the ways in which groups
are said to work differently from phrases is that ‘[groups] are groups of words of
the same primary class (nominals, verbals, or adverbials), so their functional
potential is related to the Head word’ (Matthiessen 1995:627). The description
of word class was given above in Figure 5.1. In this section, we consider
whether this description holds for each of the group units presented above.
While there is an argument to be made for identifying a three-part
distinction in the clause, i.e. participant, process, and circumstance (inter-
preted very broadly), it is more difficult to relate this to the structure of a
unit. Nevertheless, there is a clear ‘preference’, if we can call it so, for Ngps
to be the unit expressing participants in the clause and for processes to be
expressed by Vgps. As concerns circumstances, these are most frequently
expressed by Advgps or PPs, which suggests that PPs are more functionally
appropriate as members of the adverbial class. In addition, PPs can express
a participant role in certain clauses, i.e. as Attribute in relational attributive
clauses, or Recipient in verbal clauses.
7
These criteria do not address the criterion related to the function the unit serves in the unit above, but see Halliday and
Matthiessen (2014:362–4). We thank Margaret Berry for this point.
If we recall the word classification discussed above in Figure 5.1, the Pgp
(e.g. right behind, not without, all along) should, following the principle of class
membership, be made up of a group of words of the same primary class, but
if prepositions are in the verbal primary class and adverbs are in the
adverbial primary class, then the Pgp cannot be said to form a group.
Indeed, there is very little evidence for either a prepositional group or a
conjunction group, since they could be treated as word complexes at the
word rank.
There is a potential problem with these ‘compositional hierarchies’ (Hal-
liday and Matthiessen 2014:22). As McDonald (2017:252) explains, ‘Because
the conceptualisation of classes of word is dependent on their function in
classes of group, which are in turn dependent on their function in the
clause, only those word classes that commonly function as the Heads of
groups are “assigned” to a discrete group class.’ This is precisely why the
adjective class is a sub-class of nominal (see above). It is not entirely unrea-
sonable to class the adjective as a type of nominal, since its function is
always in relation to some noun, and its function is far more directly
related to expressing participant meaning in the clause than process mean-
ing or circumstance meaning. The apparent similarities between adjectives
and adverbs are considered as structural (i.e. morphological) rather than
functional, although see Tucker (1998). Recall that unit is not defined by
similarity of structure: see Berry (1975); Fawcett (2010) for a discussion of
the differences between class and type descriptions.
It is difficult to consider prepositions as a sub-class of verb other than in
the sense that they profile a relation (see Langacker 2016). The reason for
this is that in terms of class, its function in the unit above, i.e. in the PP, and
thereby in the clause, is not to express a process but rather to express a
circumstance. In fact the preposition rarely has a function in the clause (see
example (9) below and also Fontaine 2017b). In terms of class, the prepos-
itional units could belong to the sub-class of adverbials rather than verbals.
Berry (1975:76–7) maintains a useful distinction between class (defined by
function) and type (defined by structure), which allows, for example, clarity
when discussing the class of a unit (e.g. Advgp) vs. the type of formal item
(e.g. p, preposition, or c, completive). See also Fawcett (1980, 2010), who also
maintains such a distinction in terms of functional elements and structural
units and items (also see Schulz and Fontaine, this volume).
For the primary class of nominals, we do find within the Ngp word classes
identified as nominal (e.g. determiner, numeral, adjective, and noun as
shown in Figure 5.1), but also adverbial as in example (9), where the Ngp
is in bold.8
(9) The air outside feels weird and troubled.
8
Items such as ‘outside’ that have no complement are not treated as prepositions in SFL but rather as adverbials. This is
somewhat problematic, but see Fontaine (2017b) for a discussion of such prepositional items in SFL.
Here, since the qualifier outside is expressed by an adverbial class of word, in

such cases the Ngp is not formed of words of the same class as the primary
nominal class. Even if we argued that outside belonged to the verbal class, as
a preposition, the argument is unchanged.
While it is not clear that the Ngp or the Pgp is solely composed of
elements belonging to the same class, it does appear that the Vgp adheres
to this class principle, since the Vgp is effectively a group of verbal words
(e.g. might have been working). Even considering the case of phrasal verbs, the
group includes only verbal units since Halliday and Matthiessen (2014:413)
define phrasal verbs as ‘lexical verbs that consist of more than just the verb
word itself’. The treatment of multi-word expressions which may include
nominal units (e.g. kick the bucket) is significant since how they are classed
makes a difference to the theory as a whole, requiring an account of lexical
representation (see Fontaine (2017b) for a discussion of this point), however
this is beyond the scope of the current chapter.
The concept of class seems to be based on how the secondary class defines
what can occur as Head element in the logical structure of the group. There are
three main problems with this: (i) The relationship of secondary class to head
does not hold in all cases. In the verbal class, prepositions cannot function as
Head in a Vgp. (ii) There is an implication that all classes can be determined in
terms of logical meanings (i.e. having a Head element), which excludes the PP.
(iii) This view of class suggests that each class has constituency; by the rank
scale, however, constituency is best accounted for by experiential structure
rather than logical structure, i.e. the clause (at rank) is also a class, and while it
has an experiential structure, it has no logical structure.
To summarize this section, while the class principle for groups applies to
Vgps, it is not strictly the case for Ngps. Furthermore, we find that it could
be argued that the class principle suggests that prepositions and adverbials
belong in the same primary class due to their functional potential in the
clause.
5.3.2 Univariate vs. Multivariate Structures

The second distinguishing feature between groups and phrases is the type
of ‘variate’ structure they have, i.e. whether they are univariate or multi-
variate. The two types of structure describe the type of relation between
units (e.g. clauses, groups, words). Units on the rank scale are defined in
terms of constituency in the sense of being ‘composed of’ units from the
rank below, including the possibility of a rankshifted unit from the same or
from a higher rank. A univariate structure is defined as one where the
individual elements all have the same functional relationship (Halliday and
Matthiessen 2014:390). A multivariate structure, on the other hand, con-
sists of units with distinct functional relations (Matthiessen et al. 2010:148).
Roughly, groups are fundamentally seen as univariate structures, while
phrases are seen as multivariate structures which have no univariate
structure, although, as we shall see, this split is often blurred, which brings
into question the distinction between group and phrase units.
We can compare univariate and multivariate structures following Mat-
thiessen et al. (2010:145, emphasis in original), who describe univariate
structures as structures which are ‘generated as an iteration of the same
functional relationship’. Indeed, Matthiessen (1995:627) states that ‘[groups]
have a univariate structure with a Head, which is the only obligatory elem-
ent, and optional Modifiers’. This is a principle that is generally accepted
throughout the SFL literature but has never really been examined critically,
except perhaps in Fawcett (2010). Two examples of univariate structures are
given with the nominal word complex in (10) and the Vgp in (11).
(10) investment trust cash management account
(11) will have been eaten
In example (10), we find a series of five nouns, and in example (11) a series
of four verbs. This type of structure is therefore seen as serial and recursive,
involving an interdependency which can be accounted for through the
logical metafunction in terms of logical relations or modification (see
Martin 1996). However, the example in (10) is not a complete Ngp: it is
missing something. Even if we consider this as a word complex, as the Head
element of the Ngp, it is not the only obligatory element. It must be
grounded as an instance (see Langacker 2016).
Actual instances of Ngps with similar word complexes are given in (12)
and (13). In both cases, it should be clear that the rankshifting of the person’s
and with a 3 per cent annual rate of return prevents a description of these
expressions as word complexes.
(12) the person’s income management account (EnTenTen13, SketchEngine)
(13) a cash management account with a 3 per cent annual rate of return (EnTen-
Ten13, SketchEngine)
As mentioned above, multivariate structures are said to differ from univari-
ate ones in that the structure is ‘a configuration of elements each having a
distinct function with respect to the whole’ (Matthiessen et al. 2010:145,
emphasis in original). This configuration relates to constituency as discussed
above in the sense that there is a kind of generic structure involved which is
configurational and non-recursive. This type of structure is accounted for by
the experiential metafunction. In this sense, we can describe the Ngp as a
configuration of the following elements: Deictic + Numerative + Epithet +
Classifier + Thing. Halliday and Matthiessen (2014:437, emphasis in original)
describe multivariate structure as follows:
Groups have developed their own multivariate constituent structures with
functional configurations . . . Here the elements are (i) distinct in function,
(ii) realized by distinct classes, and (iii) more or less fixed in sequence.
A configuration of such a kind has to be represented as a multivariate
structure. Treating the group simply as a ‘word complex’ does not account
for all these various aspects of its meaning. It is for this reason that we
recognize the group as a distinct rank in the grammar.
These three points can be seen as properties of multivariate structures, i.e.

their elements have their own function, their elements belong to different
classes, and their overall structure is relatively fixed. If we consider the Vgp,
it is less clear that it has any similar kind of multivariate structure.
Using these two types of structure to distinguish groups from phrases is
centred on one key claim and that is that groups have both structures while
phrases have only one. This is explained by McDonald (2017:251) as follows:
The group incorporates two principles of structure: univariate, in which a

single kind of functional relationship is seen as multiplied recursively, for
example modification involving a Head and one or more Modifiers; and multi-
variate, in which a number of different functional roles can be recognised.
It is often argued that the PP is the only unit to have only a multivariate
structure, i.e. no univariate structure. However, this is also true of the clause;
it is a multivariate structure. In this section, we will consider the criteria of
variate structure in terms of how it applies to the group units described above.
Halliday and Matthiessen (2014:390) maintain that the Ngp is a ‘multi-
variate structure: a configuration of elements each having a distinct func-
tion with respect to the whole’, which is illustrated in Table 5.12. When
compared to the Vgp, as shown in Table 5.13, it seems that the experiential
configuration includes elements with very little distinction in terms of
class, and that the arrangement appears more similar to the iterative
structure of example (10). Quiroz (2017:304–5) points out that ‘a univariate
structural interpretation does not map onto the multivariate organisation
of the verbal group in any self-evident way’. Fontaine (2017a) has argued
against a univariate analysis of the Ngp, although according to different
criteria. If the Ngp is more multivariate in nature, the Vgp seems more
univariate. Halliday and Matthiessen (2014:398) explain that ‘the verbal
group is also structured logically, but in a way that is quite different from,
Table 5.12 Nominal group (adapted from Halliday and

those two splendid old electric trains
Deictic Numerative Epithet1 Epithet2 Classifier Thing
Table 5.13 Verbal group (adapted from Halliday and Matthiessen

2014:397)
couldn’t have been going to be being eaten
Finite Auxiliary1 Auxiliary2 Auxiliary3 Auxiliary4 Auxiliary5 Event
and has no parallel in, the nominal group. The logical structure of the
verbal group realizes the system of tense.’ The nature of this unit is not
clear and open to debate.
Furthermore, with the exception of the Vgp, all group units have the
potential to include elements that are rankshifted. These rankshifted elem-
ents are said to have no logical relationship with the Head element of the
unit, e.g. the Qualifier without sugar in the Ngp a cake without sugar does not
have a functional relationship to the Head cake. It is this feature which casts
some doubt on the univariate structure of groups. Halliday and Matthiessen
(2014:425) take the position that ‘prepositional phrases are phrases, not
groups; they have no logical structure as Head and Modifier, and cannot be
reduced to a single element’. We can conclude from this that phrases do not
have univariate structure (although there is only one such unit at group
rank). Given that most groups also have this as potential (except, as already
stated, the Vgp), we would have to consider whether we need a nominal
group for nominal units without rankshifted elements and a phrase for one
that do, i.e. nominal group and nominal phrase, as is the case for prepos-
itions (i.e. prepositional group and prepositional phrase). A similar case
could be made for adverbial units at group rank, for example, comparative
adverbs such as more quickly than he can, which is more phrase-like than
group-like. However, it is much easier to take the position that any unit
with the potential for rankshifted elements has a more configurational
nature to it and therefore has no univariate structure. Doing so, however,
has important theoretical consequences, since it would suggest that, by
definition, the Vgp is a group, and that all other units are in fact phrases.
5.3.3 Notion of Rank Scale

With groups, ‘rankshifted units can serve as Postmodifiers in Ngps and
Advgps (but not in verbal ones) but they are not an obligatory part of the
structure as they are in prepositional phrases’ (Matthiessen 1995:627). We
have already seen that the Ngp has the potential to include rankshifted
units, i.e. to express Qualifiers as well as some Deictics and Epithets;
however, the Vgp does not have this potential in English. In the description
of the PP in Section 5.2, it was clear that rankshifted units are a key feature
of this unit.9 In this section, we will consider the extent to which the
optional/obligatory status of rankshifted units is useful to distinguish
between groups and phrases.
Before discussing how each unit ‘behaves’ in terms of the rank scale, we
will briefly outline Halliday’s five principles of the rank scale (taken from
Halliday and Matthiessen 2014:9–10, emphases in original):
9
The role of rankshifting is debatable here but without a clear status of lexical representation within SFL theory, along
with a robust debate about the rank scale, e.g. ‘be willing/keen/eager to do; be afraid/scared to do’ (Halliday and
Matthiessen 2014:427 note 17), but see Tucker (1998). Further discussion of this is beyond the scope of this chapter.
(a) There is a scale of rank in the grammar of every language. That of

English (which is typical of many) can be represented as:
clause
phrase/group
word
morpheme
(b) Each consists of one or more units of the rank next below. For example,
Come! is a clause consisting of one group consisting of one word
consisting of one morpheme.
(c) Units of every rank may form complexes: not only clause complexes
but also phrase complexes, group complexes, word complexes and even
morpheme complexes may be generated by the same grammatical
resources.
(d) There is the potential for rank shift, whereby a unit of one rank may
be down-ranked (downgraded) to function in the structure of a unit of
its own rank or of a rank below. Most commonly, though not uniquely,
a clause may be down-ranked to function in the structure of a group.
(e) Under certain circumstances it is possible for one unit to be enclosed
within another; not as a constituent of it, but simply in such a way as to
split the other one into two discrete parts.
In theory, all group units adhere to the first principle, which situates the
primary classes of unit in terms of constituency. Principles (b) and (d) will
be discussed below, whereas (c) and (e) will not, since the potential to form a
unit complex (e.g. through coordination) and the potential to interrupt a
unit have no significant bearing on the nature of the class of unit. As we
will show, however, the requirement for a group or phrase to consist of the
rank below (principle (b)) and the potential for rankshifting (principle (d))
are relevant to the classification of intermediary units.
In terms of constituency, principle (b) states that a given unit must
contain at least one unit of the rank below it. This principle holds for all
group units, and it is difficult to ignore the role that word class (as per
Figure 5.1) plays in determining the nature of the unit at this rank, even
though originally class is meant to be related to the function served in the
unit above. However, as we have seen for the Ngp, there are some difficul-
ties as concerns Ngps that have an adjective as Head. If, for example, we
were to eliminate the logical structure at the group unit rank (Fontaine
2017a) and to uncouple adjectives from the nominal word class, then we
would find that the Ngp adheres in a more parallel way to the other
primary classes. In other words, the current anomaly in the verbal primary
class, which includes both verbs and prepositions, connected respectively to
a Vgp and a PP, would be addressed in the nominal primary class by
including a noun (nominal) group, an adjective group, a determiner group,
and a numeral group. This is perhaps unnecessarily complex, and it is not
suggested here as a position that should be adopted. However, this raises
important questions. The point being made here is that this is an area that
requires attention and development in the theory.
Principle (d) states that rankshifted units can serve as postmodifiers in
Ngps and Advgps (although not in verbal ones), but that they are not an
obligatory part of the structure as they are in PPs. However, as we have
seen above, in the case of Advgps with comparative adverbs, the rankshifted
postmodifier is obligatory, as in the PP. If we accept an adjective group as
Tucker (1998, 2017) does, this also applies to comparative adjectives (e.g.
kinder than most people).
This distinguishes Vgps from all other units at group rank, since they
cannot have a postmodifier and therefore the potential is not there.
Another distinguishing feature relates to the fact that Vgps cannot be
rankshifted. As Matthiessen (1995:715) points out, ‘Unlike nominal groups,
adverbial groups and prepositional phrases, verbal groups serve a single set
of functions in the clause and they cannot be rankshifted.’
Hence, these two principles of the rank scale lead us to ask whether there
is any need to maintain a distinction between phrase and group at group
rank, and whether the Vgp belongs as a unit at this rank. If we adhere to
the rank scale principles, we find that many of the assumptions about
groups and phrases are challenged, and the picture that emerges is quite
different from what might have been expected.
Having completed the analysis of the three principle differences between
groups and phrases, we will now consider their function in the unit above
and the unit below (or same rank) in order to get a full sense of groups and
phrases as a unit of rank.
5.3.4 Evaluation of the Group Units

Having now considered the three main criteria, we reach the following descrip-
tion. It appears that none of these units maintains all three criteria. There is
room for debate in the details. The only unit that stands out as clearly different
is the Vgp. While each criterion discussed above is helpful for better under-
standing the nature of each unit, as a set, the picture is less clear. A summary of
the outcome of the three criteria as applied to each unit is given in Table 5.14
and Table 5.15. Here we have added a fourth criterion which questions whether
Table 5.14 Experiential structure at and below the clause
Class of Function in experiential Function in experiential structure below the

group unit structure of clause clause
Nominal Participant, Qualifier in the Ngp, certain types of determiner

Circumstance (Deictic, Numerative, etc.), Range in PP
Verbal Process None (no potential for this)
Adverbial Circumstance Intensifier in Advgp (see Temperer in CG)
Prepositional Circumstance, Qualifier in the Ngp, Standard in Advgp (and
Participant Adjgp) as Scope or Finisher in CG
Table 5.15 Comparison of units following three criteria
Does the unit

Are all word have both
elements of the univariate and
unit of the same multivariate Are rankshifted units Can the unit be reduced to
Class of unit class? structures? optional? a single element?
Ngp No, not all are No, only Yes No, if you assume phoricity
nominal multivariate must be indicated (e.g.
*boy is nice vs. boys are
nice, the boy is nice), i.e.
it must be a grounded
instance of a type (see
Langacker 2016).
Vgp Yes, all are verbal Yes, both, but No rankshifting Can only be reduced to one
this is highly potential element in non-finite
debatable clauses, otherwise both
the Finite and Event
elements are necessary.
Advgp Yes, all are Both for simple No, for comparative Yes
adverbial adverbials, adverbials,
but only rankshifting is
multivariate obligatory
for
comparatives
PP Yes, all are ‘verbal’ No, only This depends on No, cannot be reduced to a
multivariate assumptions about single element. In SFL
prepositions, but intransitive prepositions
generally are considered adverbial.
rankshifting is
obligatory
Adjgp (if we No, if adjectives No, only No, for comparative Yes
accept are classed as multivariate adjectives,
that there nominals, but rankshifting is
is a need yes, if classed obligatory
for this as adverbials
unit)
the unit can be reduced to a single element, i.e. a single word. As the tables
show, it is very difficult to clearly distinguish between phrases and groups.
Many of the features that are meant to account for PPs also apply to groups,
with the exception of the Vgp. In fact, when all units and criteria are con-
sidered, the Vgp is the only unit to stand out from the others so distinctly.
This chapter set out to consider the nature of units at the rank between
clause and word (i.e. the group rank) and to examine the theoretical reasons
for distinguishing two different units at this rank. What this has revealed is
that depending on the criteria used, comparing units can reveal a variety of
outcomes. There are very clearly different types of unit at the group rank,
but whether the nature of the units is substantially different remains a
question. We wanted to find out whether it is theoretically justified to

maintain two types, i.e. group and phrase. Based on our investigation, one
clear result shows through and that is that the Vgp is significantly different
from the other group units. It cannot be rankshifted nor does it have the
potential to include rankshifted elements as part of its unit. Its organizing
principle seems to prefer univariate relations, and it is the only group to
have this strong preference.
Discussions in the SFL literature related to structure can be somewhat
challenging at times. The reliance on univariate and multivariate concepts
for defining units is not justified. It might be more productive to think of
relations rather than structures when it comes to discussions related to
univariate-ness and multivariate-ness. This would allow us to account for
the principle organizing system of units. The clause, for example, like the
Ngp, is a configuration of elements, but the Vgp seems far less like this and
far more serial in nature.
In a sense, it makes little difference what these units are called, i.e.
whether group or phrase. Each has its own internal structure and func-
tional elements. Given that the term ‘group’ is so prevalent in SFL theories,
and given that there does not seem to be any justifiable reason for differen-
tiating between groups and phrases, we suggest to use the term ‘group’ for
all the different types of group units at the rank between clause and word,
including the prepositional phrase.
References
Berry, M. 1975. An Introduction to Systemic Linguistics, Vol. 1: Structures and

Berry, M. 2017. Stratum, Delicacy, Realisation and Rank. In T. Bartlett and
G. O’Grady, eds., The Routledge Handbook of Systemic Functional Linguistics.
London: Routledge. 42–55.
Caffarel-Cayron, A. 2017. The Verbal Group in French. In T. Bartlett and
Davidse, K. 2004. The Interaction of Identification and Quantification in
English Determiners. In M. Achard and S. Kemmer, eds., Language, Culture
and Mind. Stanford: CSLI Publications. 507–33.
Fawcett, R. 1980. Cognitive Linguistics and Social Interaction: Towards an Inte-
grated Model of a Systemic Functional Grammar and the Other Components of an
Interacting Mind. Heidelberg: Julius Groos and Exeter University.
Fawcett, R. 2000a. In Place of Halliday’s ‘Verbal Group’, Part 1: Evidence
from the Problems of Halliday’s Representations and the Relative Simpli-
city of the Proposed Alternative. Word 51(2): 157–203.
Fawcett, R. 2000b. In Place of Halliday’s ‘Verbal Group’, Part 2: Evidence
from Generation, Semantics and Interruptability. Word 51(3): 327–75.
Fawcett, R. 2010. A Theory of Syntax for Systemic Functional Linguistics. Amster-

Fontaine, L. 2013. Analysing English Grammar: A Systemic-functional Introduc-
Fontaine, L. 2017a. The English Nominal Group: The Centrality of the Thing
Element. In T. Bartlett and G. O’Grady, eds., The Routledge Handbook of
Systemic Functional Linguistics. London: Routledge. 267–83.
Fontaine, L. 2017b. On Prepositions and Particles: A Case for Lexical Repre-
sentation in Systemic Functional Linguistics. Word 63(2): 115–35.
Fries, P. H. 1977. English Predications of Comparison. In R. DiPietro and
E. Blansitt, eds., The Third LACUS Forum 1976. Columbia: Hornbeam Press.
545–56.
Ghesquière, L. 2014. The Directionality of (Inter)subjectification in the English
Noun Phrase: Pathways of Change. Berlin: Mouton de Gruyter.
Halliday, M. A. K. 1961. Categories of the Theory of Grammar. Word 17(2):
241–92.
Halliday, M. A. K. and C. M. I. M Matthiessen 2014. Introduction to Functional
Grammar. 4th ed. London: Edward Arnold.
Kilgarriff, A. et al. 2014. The Sketch Engine: Ten Years On. Lexicography 1:
1–30.
Langacker, R. 2016. Nominal Structure in Cognitive Grammar: The Lublin Lec-
tures. Edited by A. Głaz, H. Kowalewski, and P. Łozowski. Lublin: Marie
Curie-Skłodowska University Press.
Martin, J. R. 1992. English Text: System and Structure. Amsterdam: John Benjamins.
Martin, J. R. 1996. Types of Structure: Deconstructing Notions of Constitu-
ency in Clause and Text. In E. H. Hovy and D. R. Scott, eds., Computational
and Conversational Discourse: Burning Issues – An Interdisciplinary Account.
Heidelberg: Springer. 39–66.
Tokyo: International Language Sciences Publishers.
Matthiessen, C. M. I. M., K. Teruya, and M. Lam. 2010. Key Terms in Systemic
Functional Linguistics. London: Continuum.
McDonald, E. 2017. Form and Function in Groups. In T. Bartlett and
Morley, D. 2000. Syntax in Functional Grammar: An Introduction to Lexicogram-
mar in Systemic Linguistics. London: Continuum.
Quiroz, B. 2017. The Verbal Group. In T. Bartlett and G. O’Grady, eds., The
Routledge Handbook of Systemic Functional Linguistics. London: Routledge.
301–18.
Tucker, G. 1998. The Lexicogrammar of Adjectives: A Systemic Functional
Approach to Lexis. London: Cassell Academic.
Tucker, G. 2017. The Adjectival Group. In T. Bartlett and G. O’Grady, eds.,
The Routledge Handbook of Systemic Functional Linguistics. London: Routledge.
284–300.
6
Context and Register
Wendy L. Bowcher
6.1 Introduction
This chapter presents an overview of the concepts of context and register;

their history in Systemic Functional Linguistic (SFL) theory, the relation
between the two concepts, and the most influential perspectives and
models developed in this area by SFL researchers.
6.2 Pre-SFL to SFL Theory: A Brief History of the

Concepts of Context and Register
6.2.1 Context
In the early stages of the development of SFL theory, ‘context’ referred to
the semantic level of language; there was phonology, grammar and lexis,
and context (see Halliday 1961; Ellis 1966; Gregory 1967). At that time, the
extra-linguistic environment was known as ‘situation’ (Halliday 1961; Halli-
day et al. 1964). Gregory makes the distinction between situation and
context in this way:
By context is understood the correlation of formally described linguistic

features, grouping of such features within texts and abstracted from them,
with those situational features themselves constantly recurrent and relevant
to the understanding of language events. Situation is an aspect of the descrip-
tion of language events, not a level of language or linguistics. Context is seen
as a level of language, as its concern is with certain patterns and pattern
correlations which are part of . . . linguistic behavior.
(Gregory 1967:178)
I am grateful to Edward McDonald for comments on an earlier draft of this chapter. I alone am responsible for any
shortcomings.
Context and Register 143
During the 1960s, as the ideas of what would become SFL theory began to
take shape, context, in the sense of semantics, came to be called ‘meaning’
or simply ‘semantics’, and situation came to be referred to as ‘context of
situation’. The changes, however, were not merely terminological, as we
will see, but resonated with the changing shape of Halliday’s thinking on
the architecture of language.
The coinage of the term ‘context of situation’ is attributed to Malinowski,
whose ethnographic work in the Trobriand Island communities led him to
realize that the concept of the context of an utterance needed to encompass
the linguistic, situational, and cultural context of a language (Malinowski
1923:306). Malinowski’s research complemented other work at the time,
such as Ogden and Richards’ (1923) discussion of the relation between signs
and meaning, and their suggestion that a description of ‘sign situations’
should come from observation of a corpus of instances rather than individ-
ual or ‘exceptional’ cases (Ogden and Richards 1923:19) – a scientific, not an
‘intuitive’ approach (Ogden and Richards 1923:20). Within this scholarly
climate, Malinowski made important claims regarding the relation between
meaning, language, and situation: ‘Language is essentially rooted in the
reality of the culture, the tribal life and customs of a people, and . . . it
cannot be explained without constant reference to these broader contexts
of verbal utterance’ (Malinowski 1923:305). However, being an anthropolo-
gist and not a linguistic theoretician, he did not develop the concept of
‘context of situation’ within an explicit theory of language.
Firth, a linguist who worked with Malinowski in the 1930s when Mal-
inowski ‘was especially interested in discussing problems of languages’
(Firth 1950:43) stated clearly that, while the term context of situation was
‘first widely used in English by Malinowski’, it became a ‘key concept in the
technique of the London group’1 for the study of language (Firth 1950:42).
Firth differentiated Malinowski’s concept of context of situation from his
own; Malinowski’s concept was essentially a material idea: ‘an ordered
series of events’ (Firth 1950:43). For Firth, however, the concept of context
of situation was ‘best used as a suitable schematic construct to apply to
language events . . . a group of related categories at a different level from
grammatical categories but rather of the same abstract nature’ (Firth
1950:43) through which the meanings of language in use could be inter-
preted. From the start, Firth could see the connection between this abstract
concept and language use and how the concept might become part of the
description and analysis of language. For example, he presents the
following utterance:
‘Ahng gunna gi’ wun fer Ber’’

(I’m going to get one for Bert)
1
The London Group or London School refers to those linguistic scholars in Britain taught and influenced by J. R. Firth.
These scholars were also often referred to as ‘neo-Firthians’.
144 WENDY L. BOWCHER
and asks:
What is the minimum number of participants? Three? Four? Where might it

happen? In a pub? Where is Bert? Outside? Or playing darts? What are the
relevant objects? What is the effect of the sentence? ‘Obvious!’ you say. So is
the convenience of the schematic construct called ‘context of situation’. It
makes sure of the sociological component.
(Firth 1950:44)
Because Firth saw language as inextricably linked with the sociality of

living, the study of language had to take into account ‘man’s active partici-
pation in the world’ (Firth 1957:2). However, Firth was careful to avoid
including in his schema ‘the description of mental processes or meaning in
the thoughts of the participants’ and ‘any consideration of intention, pur-
port or purpose’ (Firth 1957:9). Such processes were not ‘ignored’ per se,
rather,
[a]s we know so little about mind and as our study is essentially social, I shall
cease to respect the duality of mind and body, thought and word, and be
satisfied with the whole man, thinking and acting as a whole, in association
with his fellows.
(Firth 1957:2)
With this view of language, context, and ‘the whole man’, the kind of lan-
guage that was to be studied from Firth’s point of view was ‘actual language
text’. Further, the primary sets of relations for enquiry were ‘the interior
relations connected with the text itself’, which include the syntagmatic and
paradigmatic relations within language elements; the situational relations,
which include the ‘interior relations within the context of situation, the focal
constituent for the linguist being the text’; the ‘analytic relations set up
between parts of the text’; and ‘special constituents, items, objects, persons
or events within the situation’ (Firth 1957:5). As for the interior relations of
the context of situation, Firth (1957:9) proposed the following:
(a) The participants: persons, personalities, and relevant features of these.

(i) The verbal action of the participants.
(ii) The non-verbal action of the participants.
(b) The relevant objects and non-verbal and non-personal events.
(c) The effect of the verbal action.
Context of situation was clearly connected, in a scientific sense, with what

was observable in the way in which language was used. Text, moreover, was
considered a constituent of the context of situation: ‘The placing of a text as a
constituent in a context of situation contributes to the statement of mean-
ing since situations are set up to recognize use’ (Firth 1957:11, emphasis in
the original). This stance extended to the meaning of a word; a word was
considered to be intimately tied to its use within a context of situation, thus
setting up a correlation with a specific way of speaking and a specific set of
words within a given context – ‘collocations are actual words in habitual

company’ (Firth 1957:14). Indeed, what emerges from Firth’s writings is
that context of situation was invoked as a technique for making ‘statements
of meaning’. That is, context of situation could be used as a means of
understanding and accounting for the what, the how, and the why of a
speech event (see Hasan 2014:3), and was an applicable concept in the study
of language activity. This applicability was demonstrated by Halliday (1959/
1974) who included an analysis of the context of situation based on Firth’s
schema in his analysis of a vernacular text in early Mandarin Chinese.
Ellis (1966:79) has suggested that ‘context of situation’ is ‘one of Firth’s
decisive contributions to linguistic theory’, but he notes that because Firth
‘left this concept in many ways unelaborated’ there were problems which
needed to be considered in order to develop linguistic ‘categories, as power-
ful and as general as possible, to relate the level of form with that of
situation’ (Ellis 1966:79). There is much in Ellis’ work that points to the
originative scholarly thinking on the concept of context of situation in the
late 1950s and early 1960s (and of course, not just within the London
School, although that is our focus here), and which also points forward to
the concept as it has been eventually theorized in the SFL model of lan-
guage. For example, he discusses the concepts of ‘potential’ and ‘instance’,
citing McIntosh (1961) who introduced the terms ‘potential’ and ‘actual’,
and the notion of ‘scale of delicacy’ in relation to linguistic analysis, nota-
tion, and semantic focus (see also Halliday 1961). The issue of ‘relevancy’,
which has to do with minimizing intuition and maximizing the generality
and intersubjectivity of contextual and registerial descriptions, was also an
important theme of discussion (see Catford 1965; Ellis and Ure 1974; Greg-
ory 1967; also raised by Ogden and Richards 1923). Generally, context of
situation became a key explanatory feature of language in use and general-
izable across languages, largely inspired by the observation that the lan-
guage people use differs in relation to its conditions of use (see Ellis 1965).
6.2.2 Register
The recognition that language varies according to how it is used played a
significant role in forming the concept of ‘register’ – the term ‘register’ as
used in linguistics being attributed to Reid (1956). It goes without saying
that one does not speak the same with one’s mother as with one’s friends or
husband, nor does a sports commentator speak the same way when com-
mentating a game as when ordering a meal in a restaurant. Firth high-
lighted this in his advice to the Air Ministry in suggesting the type of
Japanese that would be most appropriate to learn in order to take up an
effective position against them during World War II:
When I was consulted by the Air Ministry on the outbreak of war with
Japan, I welcomed the opportunity of service for the Royal Air Force
because I saw at once that the operating reconnaissance and fighter air-
craft by the Japanese could be studied by applying the concept of the
limited situational contexts of war, the operative language of which we
needed to know urgently and quickly. We were not going to meet the
Japanese socially, but only in such contexts of fighting as required some
form of spoken Japanese.
(Firth 1950:43–4)
Language varies according to the occasions on which it is being used, and

speakers of a language typically have no difficulty recognizing this fact and
managing their ‘repertoires’ of language use accordingly.
The relationship between register and context of situation became an
important focus of discussion and research in a number of published papers
in the mid twentieth century (e.g. Ellis 1966; Ellis and Ure 1974; Halliday
et al. 1964), and the value and applicability of the term soon became
widespread, particularly in the field of language teaching (e.g. Halliday
et al. 1964; White 1974; Carter 1978; Ure 1982). Halliday et al. (1964), for
instance, referred to register as ‘varieties of language’ that ‘cover the total
range of our language activity’, with the further claim that ‘it is only by
reference to the various situations, and situation types, in which language
is used that we can understand its functioning and its effectiveness’ (Halli-
day et al. 1964: 89). Ellis (1966) proposed a range of different technical
terms covering what he saw as subcategories of register: ‘register’, a lin-
guistic category, and ‘division of idiolect’, distinguished by formal features
and correlating with types of situations; ‘register range’, or the total reper-
tory of registers that a person may use; ‘register choice’, or the specific
register a person chooses out of his or her repertory of registers; and
‘register-features’, or the features of the language spoken in a specific
situation.
Register is thus defined as linguistic variation that is ‘use-based’ in
contrast to (or perhaps complementary to) variation that is ‘user-based’,
such as dialectal variation. This is not to say that a register cannot point
to the identity of the user within the situation. Consider, for example,
that using the register of sports commentary typically points to the
speaker as being a ‘sports commentator’ (although not always), but the
major difference between these two types of variation is that register
primarily points to the identity of the situation, whereas dialectal vari-
ation points to the identity of the speaker. Nevertheless, both register and
dialect make contact within the situation; speaking one’s original dialect
may be considered disadvantageous in certain situations, such as job
interviews, and so some speakers may choose a more standard dialect
(see Godley and Escher 2012) in those situations. Speakers are generally
aware that they need to speak with relevance to what is taking place and
with whom they are interacting, which involves speaking the register of
that situation.
Because registers are language varieties, they are linguistically defined.
That is, ‘it is by their formal properties that registers are defined’ such that
if two samples of language activity from what, on non-linguistic grounds, could

be considered different situation-types show no differences in grammar or
lexis, they are assigned to one and the same register: for the purpose of the
description of the language there is only one situation-type here, not two.
(Halliday et al. 1964:89)
6.2.3 Register in Relation to Context

The relationship of register (linguistic variation) to context of situation (situ-
ational variation) is of paramount importance in SFL theory. But in the
developmental days of SFL theory, how to theoretically account for this
relationship was much disputed, with debates often centering on ‘relevancy’,
the most useful and descriptive parameters of context of situation, and the
relation between register and the nature of the language system itself.
As to the description of situational features, various schemas were sug-
gested. Ellis (1966), for instance, suggests several features of the situation:
role (participant roles); field (type of activity); formality (informal or
formal); mode (spoken or written); and thesis, or the ‘event . . . to which
the utterance refers’ (Ellis 1966:84). Halliday et al. (1964:90–2) suggest
‘Field’ or ‘what is going on: to the area of operation of the language
activity’; ‘Mode’, or ‘the role played by the language activity in the situ-
ation’, such as whether the language is spoken or written; and ‘Style’
‘which refers to the relations among the participants’. Gregory (1967)
weighed in on the debate with a critique of the various attempts at categor-
izing the situational features and with a proposal of his own categories:
Field of Discourse; Mode of Discourse; and Tenor of Discourse. While
Gregory preferred the term ‘diatypic variety’ to ‘register’, his introduction
of the term ‘tenor’ in place of ‘style’ (the term used in Halliday et al. 1964)
has been standard in SFL ever since (see Gregory 1967:195).
With regard to relevancy, we have already noted that in Firth’s descrip-
tion of context of situation he refers to ‘relevant features’ and ‘relevant
objects’ etc. But just how to determine what is relevant in terms of situ-
ational variables was important to unlocking the systematic relationship
between language and context so that this relationship could be analyzed
more objectively. In response to the issue of ‘relevancy’, Gregory contends:
Those situational elements which are potentially contextually relevant to

given linguistic forms or groups and complexes of forms are discovered, or
‘invented’, by commutation, by changing, as Catford (1965:36) noted, situ-
ational features and observing what textual changes take place, by changing
an item or items in the text and observing what situational change occurs.
This entails careful and continuous contrasts amongst the records, substan-
tial and situational, of related series of language events.
(Gregory 1967:179)
These ‘careful and continuous contrasts’ are indeed important for a scien-
tific approach to understanding the relationship between language and
context, but Halliday took things a step further than this. Halliday’s under-
standing of language as social activity led to important theoretical tenets
associated with the concepts of context of situation and register in SFL
theory. One of these is the stratal nature of language and the inclusion of
context as a stratum. This is a legacy of Firth’s description of language into
three strata (Firth 1957; although see also Lamb (1966), whose stratifica-
tional grammar was influential in Halliday’s thinking): graphology/phon-
ology, grammar, semantics, and his proposal for the extra-linguistic level of
context of situation. Because of his view that language is social activity,
Halliday has not considered context to be an isolated concept, but one that
is fully integrated into the description of language:
The linguistic system has evolved in social contexts . . . the system is a

meaning potential, which is actualized in the form of text . . . [and] the
situation [is] embodied or enshrined in the text not piecemeal, but in a way
which reflects the systematic relation between the semantic structure and
the social environment.
(Halliday 1977:199)
Another theoretical tenet is the concept of ‘realization’. Realization is an

‘interstratal relationship’ (Halliday 1992:20) and is modelled as a ‘metaredun-
dancy’ relation. That is, with reference to the strata of meaning, wording, and
sound, meaning is not just realized by wording and then wording is realized
by sound, like a chain of ‘this, then that’. Rather, ‘meaning is realized by the
realization of wording in sound’ or from the reverse perspective, ‘sounding
realizes the realization of meaning in wording’ (Halliday 1992:24; see also
Taverniers, this volume). Hasan refers to this as a ‘realization/activation’
relationship, where, for instance, ‘contextual features activate . . . meaning-
wording’ and meaning-wordings realize contextual features (Hasan 2014:45).
Of key importance in the model of SFL is Halliday’s hypothesis that text is
‘a semantic unit’ and a ‘continuous process of semantic choice’ realized in
lexicogrammatical choices (Halliday 1977:193–5). The semantic system is
organized under three main categories: ideational (experiential and logical),
interpersonal, and textual meanings (see Taverniers, this volume). These
meanings pattern in specific ways across texts and are the means through
which the social situation is ‘embodied’ in the text (Halliday 1977). The seman-
tic system itself is thus an interface between the social system (the various
contexts of situation) and the grammatical system of the language. In situa-
tions of language in use, certain linguistic choices are favoured or ‘at risk’.
Halliday has demonstrated that this relationship is patterned and non-random,
and not absolute or categorical, and although his early work expressed this
relation in a rather ‘deterministic’ way (e.g. Halliday 1977), it is clear that the
idea of ‘at risk’ or ‘favoured’ options does not suggest a one-to-one relation
between the features of context and the activated linguistic features:
The patterns of determination that we find between the context of situation

and the text are a general characteristic of the whole complex that is formed
by a text and its environment. We shall not expect to be able to show that the
options embodied in one or another particular sentence are determined by the field,
tenor and mode of the situation. The principle is that each of these elements in
the semiotic structure of the situation activates the corresponding compon-
ent in the semantic system, creating in the process a semantic configuration,
a grouping of favoured and foregrounded options from the total meaning
potential that is typically associated with the situation types in question. This
semantic configuration is what we understand by the ‘register’: it defines the
variety . . . that the particular text is an instance of. The concept of register is
the necessary mediating concept that enables us to establish the continuity
between a text and its sociosemiotic environment.
(Halliday 1977:203, emphasis added)
In English, Halliday demonstrates that ideational meaning is systematic-

ally, although not entirely, realized through the grammatical system of
transitivity, with logical meaning realized through systems for clause
complexing, interpersonal meaning largely through the mood system, and
textual meaning primarily through systems of theme and information as
well as through cohesive devices. Thus, from context through the strata of
language and vice versa, Halliday has shown an activation-construal relation.
Meaning resides not at one level (e.g. semantics) but is a property of all levels
of language and into context (see Halliday 1985; Firth 1957; Berry, this
volume; Butt, this volume; Taverniers, this volume; Webster, this volume).
Thus, ‘relevancy’, and accounting for what is relevant in terms of fea-
tures of the context of situation in relation to register, as it turns out,
requires a systematic, realization/activation model of language and the
social system, and this is evident in Halliday’s finely tuned systemic-
functional architecture of language.
Crucial in the process of realization is that of instantiation, an ‘intrastratal
relation’ wherein observation and description of language are moved along a
scale from potential (the system) to instance (the text). That is, it is the same
phenomenon that is being examined, but it is being examined from a differ-
ent ‘depth of vision’. Halliday explains this by making an analogy with the
relationship between climate and weather (Halliday 1992). Further, it is
through instantiation that the language system remains ‘metastable’: ‘per-
sist[ing] only through constantly changing by interpenetration with [the]
environment’ (Halliday 1992:26). It is this metastability that allows for a high
degree of predictability in situations of language use. That is,
There is no situation in which the meanings are not to a certain extent

prescribed for us. There is always some feature of which we can say, ‘This
is typically associated with this or that use of language’. Even the most
informal spontaneous conversation has its strategies and styles of meaning.
(Halliday 1985:40; cf. Cloran 1987)
Because registers are varieties of language according to how language is

being used in a situation, the instantiation of a register is a configuration of
linguistic meanings, or ‘a continuous process of semantic choice’ (Halliday
1977:195). The SFL concept of text as a ‘semantic’ unit is important in the
overall understanding of the relationship between context and register.
Halliday explains:
A text has a generic structure, is internally cohesive, and constitutes the

relevant environment for the selection in the ‘textual’ systems of the gram-
mar. But its unity as a text is likely to be displayed in patterns of ideational
and interpersonal meaning as well. A text is the product of its environment, and
it functions in that environment.
(Halliday 1977:195, emphasis added; also see Hasan 1985)
This observation is echoed throughout SFL theory. Text is not simply a

language event but a ‘sociological event’ (Halliday 1978:139). It is ‘naturally
occurring language use . . . having a social function, and possessing the
attributes of texture and structure’ (Hasan 2014:4). Moreover, because a text
represents a specific register, its unity is a characteristic of that register. That
is, while a text may involve different ‘movements’, collectively these move-
ments hang together in a way that is characteristic of a specific register.
Thus, how context of situation and register are related has to do with those
features of the language system that are likely to be ‘at risk’ in a given
situation. Halliday has pursued a line of inquiry which aims to specify those
aspects of the context of situation which affect or ‘rule’ (Halliday 1977:19)
what speakers choose from the options that make up the semantic system,
the metafunctions. His theory models ‘the systematic relationship between
language and the environment’ (Halliday 1977:19) by proposing that the
context of situation is a semiotic construct and ‘an instance, or instantiation,
of meanings that make up the social system’ (Halliday 1977:19). These are
the ‘conditions’ (to use an earlier term) under which linguistic choices are
made. Variation in the conditions correlates with variation in what is said. It
follows, then, that register, being variation in language use, is modelled as a
‘semantic configuration’ and located at a different stratum than context of
situation. That is, register is located
at the semantic level, not above it. Shifting in register means re-ordering the
probabilities at the semantic level . . . whereas the categories of field, mode
and tenor belong one level up. These are the features of the context of
situation; and this is an interface. But the register itself I would see as being
linguistic; it is a setting of probabilities in the semantics.
(Thibault 1987:610, emphases in original)
With regard to text and its relation to context of situation and to register,
Halliday makes the following claim:
The text is a continuous process. There is a constantly shifting relation

between a text and its environment, both paradigmatic and syntagmatic:
the syntagmatic environment, the ‘context of situation’ (which includes the
semantic context – and which for this reason we interpret as a semiotic
construct), can be treated as a constant for the text as a whole, but is in fact
constantly changing, each part serving in turn as environment for the next.
And the ongoing text creating process continually modifies the system that
engenders it, which is the paradigmatic environment of the text.
(Halliday 1977:198)
Halliday’s concepts of register and context (including context of situation)

and their place in SFL theory are shown in Figure 6.1.
SYSTEM INSTANCE
CONTEXT sociocultural specific context

environment of situation
activation/construal
contexts of context of
REALIZATION:
situation situation type
LANGUAGE language semantics registers register type text (semantic unit)

system lexico-
grammar
phonology
Figure 6.1 Register and context in SFL theory (adapted from Halliday 1999:8)
In Figure 6.1 we can see two basic features of SFL theory: the hierarchy of
strata (Context – Language) and the cline of instantiation (System –
Instance). As already noted, the strata are related through the relation of
realization/construal: context is realized in language and language con-
strues context, and this is indicated by the vertical lines. The cline of
instantiation (the horizontal axes) depicts varying perspectives on the same
phenomenon: a specific language event represents a selection from the
language system; a specific context of situation represents a selection from
the sociocultural context. Between the system and instance ends of the
cline are ‘types’ or categories of contexts of situation correlating with types
of texts or registers. Instantiation and realization are fundamental prin-
ciples explaining how language changes and is also maintained, that is, how
it is an ‘open dynamic system’ in that each instantiation (whether that be
of context or language) ‘resets’ the ‘overall probabilities’ of the respective
systems (Halliday 1992:27).
6.3 Context and Register: Developments and Perspectives
Within SFL theory, there have been several developments and variations of
Halliday’s conception of context and register. This section briefly describes
the most influential of these.
6.3.1 Hasan: The Essentialness of Context

Hasan has consistently demonstrated and validated the SFL centrality of
context for an understanding of language. Whether in analyzing lexico-
grammatical choices, semantic systems, contextual parameters, or cultural
consequences of language, she has repeatedly argued for keeping context in
view, no matter what the analytical goal, if one is to understand the
workings of language: ‘Seeing language as a form of human social action
does not preclude attempts to understand the logic of its form and vice versa’
(Hasan 2005:68, emphasis in original). She has bemoaned the idea that
context is often used merely to
illuminate the nature of parole, [but] not of langue . . . [or] . . . a good device
for ‘mopping up’ some of the problems that inhere in an essentially
extra-social view of language. . .where context becomes a mundane ‘real-
ity’ to be taken for granted, while language becomes a mysterious mental
organ, and grammar a body of knowledge encrypted in the human brain
at birth.
(Hasan 2001:3, italics in the original)
Rather, Hasan has argued that context is central to the ontogenesis, phylo-
genesis, and logogenesis of language (see Hasan 2001), and is part of ‘a
productive principle, reflection on which [has] enabled SFL to offer a scien-
tific description of “how language works”’ (Hasan 2005:55–6). In sum,
Hasan’s work speaks to her view that ‘there can be no language without
context’ (Hasan 2001:8).
Overall, Hasan’s contribution has added detail, precision, and depth
to Halliday’s conception of context and register, not least because she
has questioned many of the assumptions and terminology associated
with these two concepts (see Lukin 2016 for a lengthy discussion of
Hasan’s contribution to the development and description of context).
An example of her questioning includes an early discussion on the
concept of relevancy. Although Halliday theorized the probabilistic
correlation between features of language and features of context, it
was in probing the idea of relevancy and its relation to context and
register that Hasan has proposed that relevant context, or the context
of situation, is embedded in a material situational setting (MSS): ‘Situ-
ation-type is an abstraction from the totality of material situational
setting’ (Hasan 1973:275; also see Hasan 1981:110, Hasan 1995:219).
Her later work has honed this idea: those features of the social situation
which are ‘illuminated’ in the text (Hasan 1995:219, 2005:61) realize
the relevant contextual parameters of field, tenor, and mode, the fea-
tures of the context of situation. The MSS, on the other hand, is not part
of the context of situation, but acts as ‘a dormant source for affecting
the verbal goings on’ (Hasan 1981:110). Features of the MSS enter the
description of the context of situation when they are directly refer-
enced, no matter how minor that reference may be. Hasan has proposed
that both the MSS and the context of situation could be encapsulated in
the overarching terms Action, Reflection, and Contact (ARC): ‘I think of
ARC as relevant to any form of joint social practice, whether this
involves language or not. By contrast, field of discourse, tenor of dis-
course and mode of discourse are . . . specifically discourse related’
(Hasan 2001:7; also see Hasan 2014:11–13).
The concept of Material Situational Setting has been linked with various
other concepts in the theory, including that of institutionalization, which
relates to the degree of negotiation possible within a situation due to the

convergence (or not) of different semiotic systems through which a situ-
ation typically unfolds. Hasan argues:
Social processes which are institutionalized would logically be multiply

coded semiotically . . . And what, from the point of view of verbal coding,
may be seen as just a material situational setting, can from the point of view
of the social process be seen as situation semiotically coded through a series
of distinct codes.
(Hasan 1981:116)
Bowcher (1999) builds on this idea by outlining a set of questions to probe

the context of a text to assist in ‘measuring’ the degree of institutional-
ization in a given context. For instance, in a court of law, layout of the
room, seating arrangements, functionally designated furniture (e.g. the
witness box), placement of certain individuals, forms of dress, ritualized
procedures and actions, formulaic expressions, etc., multiply code social
roles and relations, activities, and the order of those activities. There is little
room for individual negotiation of these features of the situation, thus
indicating a highly institutionalized context.
Hasan early on flagged the connection between MSS and context-
dependent and context-independent text (Hasan 1973), an issue which she
returns to in Hasan (1999), and which is also taken up in detail by Cloran
(1994, 1999), who shows how a single text may move in and out of a
stronger or lesser connection with the material environment in which it
takes place. Cloran categorizes these different ‘moves’ in terms of rhet-
orical units. A ‘rhetorical unit’ (RU) analysis is different from rhetorical
structure theory (RST) analysis (see Mann and Thompson 1988; Mann et al.
1992) and ‘takes as its point of departure Hasan’s message semantics’
(Cloran 1999:196–7; also see Low and Fung, this volume). It involves analyz-
ing the way in which messages group together to form distinct configur-
ations of semantic features, an RU being a grouping of semantic features in
a text. The semantic structure of a text is thus shown to have a hierarchy: a
text consists of one or more RUs, and an RU consists of one or more
messages (Cloran 1999:197).
At a more local level, Hasan has interrogated specific terminology. An
example of this is her discussion of the term ‘activity’:
In ordinary life, the word has many meanings and each seems clear in its
‘context’, but what exactly did it mean in the description of field of dis-
course? Here it seemed to have multiple values: it was not clear if the word
referred to precisely the same phenomenon in its various appearances, such
as ‘social activity’, ‘relevant activity’, ‘language activity’. There were also
‘descriptive references’ such as ‘what is going on’, and ‘the area of the
operation of the language activity’. Sometimes ‘the whole activity’ was said
to consist of two kinds of ‘activities’, a ‘language activity’ which ‘assisted’
‘the whole event’, in which case it would seem that the ‘whole event’ was to
consist of both ‘language activity’ and some other kind of ‘activity’ which
was not linguistic. At other times, the ‘whole of the relevant activity’ could
be accounted for ‘practically’ by ‘language activity’.
(Hasan 2014:6)
Such querying has led to more rigorous accounts of, for instance, the
concept of action in relation to field and mode (see Hasan 1999, 2014;
Bowcher 2013, 2014), but there is still more work to be done in this regard
as evidenced by Hasan’s comment: ‘I have long felt the need to explore the
differences between “act”, “action” and “activity”: this would bring greater
order in understanding field as the parameter concerned with “doing” of
some kind’ (Hasan 2014:51 note k).
Hasan’s conception of context and register as ‘configurations’ of features
has been an important influence on how researchers now research and
model the relationship between context and text. With reference to context
of situation, Hasan prefers the term ‘configuration’ to ‘combination’
because the features of context do not simply ‘combine’; ‘rather, contextual
configuration is like a chemical solution, where each factor affects the
meanings of the others’, thus claiming an ‘interdependence between the
three parameters’ (Hasan 1995:231).
The concept of contextual configuration (CC) being ‘an account of the
significant attributes of [a given] social activity’ (Hasan 1985:56) is central to
understanding two key features of the semantic nature of text: its structure
and texture. Texture refers to the relations among the meanings of a text
and is determined through an analysis of a text’s ‘cohesive harmony’,
which relates to the cohesive ties among elements of a text (see Taboada,
this volume). Text structure is termed ‘generic structure potential’ (GSP),2
as it is used
to describe the structure of not only a specific text type but also a range of
other related text types. Each member of that range of text types will have some
structural properties in common with other members: no individual text
type will have the same structural shape as any other, and none will be
entirely different. The entire range of such text types will constitute a single
register family.
(Hasan 2014:9–10, emphases in original)
A GSP has both obligatory and optional elements. The obligatory elements
are defining in terms of the text’s register, while the optional elements
indicate variation among texts belonging to a given register. Hasan has
demonstrated how the CC can be used to ascertain obligatory and optional
elements of a text as well as the sequence of these elements and their
iteration (Hasan 1985:56). While both GSP and texture reside at the seman-
tic level of language, GSP is said to act as a link between the texture of a text
2
GSP was originally called ‘generalised structure potential’ (see Hasan 1978, 2014:51 note i).
and the relevant context of a text (Hasan 1985:99). With regard to the
relation between texture and context, Hasan comments that ‘situation type,
at a high degree of specificity, is relevant to texture; you could see it as the
motivating force of texture. But by the same token, the facts of texture
construe the very detailed aspects of the situation in which the text came to
life’ (Hasan 1985:115). Hasan also observes that cohesive chains may ‘dis-
play a close relationship to the structural movement of the text’ (Hasan
1985:115). This has been demonstrated in work such as Butt et al. (2010),
Cloran (1999) (in relation to Rhetorical Units and Material Situation), and
Lukin (2010) (see Khoo 2016 for a survey of work on Cohesive Harmony).
Hasan’s GSP was innovative, and perhaps a concept before its time, as it
has not been fully understood within SFL. Some critics have suggested it is a
static or ‘synoptic’ representation of text structure (see Hasan 1995), but the
concept of GSP offers a means of accounting for structure in terms of both
variation and stability across registers, or text types, and in relation to
features of context of situation. Hasan’s argument has always been that
‘the elements of text structure cannot be defined by reference to the rank
status or sequential ordering of the lexicogrammatical units which have
the function of realizing these elements’ (Hasan 1978:229). Rather,
the controls upon the structural make-up of a text are not linguistic in
origin . . . instead, the control is contextual: the nearest non-linguistic ana-
logue of a text is not a logico-mathematical formula, but a non-verbal social
event. A text is a social event whose primary mode of unfolding is linguistic.
(Hasan 1978:229)
Ultimately, a GSP statement ‘explicitly signals those features whose selec-

tion would be the realization of some systematic variation across the
derived structures: the derived structures do not vary accidentally; they
vary with predictable perturbations in the configuration of the underlying
context’ (Hasan 2014:10). Various scholars have utilized Hasan’s concept of
GSP including Cloran (2016), in relation to Rhetorical Units, and more
recently Bortoluzzi (2010), Bowcher (2015), Bowcher and Liang (2016), and
Cheong (2004), in relation to the structure of multimodal texts.
Other ideas associated with the concept of contextual configuration
include ‘interdependence’ and ‘permeability’. With reference to context,
interdependence refers to the fact that what one does with whom and in
what manner are connected. Interdependence is significant in the process
of realization as it means that there is no one-to-one relation between
contextual variables and features of the language, i.e. field is not singularly
realized in the experiential metafunction, etc. (Hasan 1995). However,
Hasan points out there are ‘default’ linguistic realizations of the contextual
features, and that language, being an open-dynamic system, necessitates
‘the possibility of departures from the highly probable’ (Hasan 2014:8).
Hasan has been critical of those SFL scholars unsatisfied with the probabil-
istic nature of the realization relation between the contextual variables and
the semantic systems (see Hasan 1995). She points out that the ‘predicting’
power of the contextual parameters with the metafunctions is a reflection
of the nature of the relationship between social variables and linguistic
choices, a mode of explanation wholly acceptable in sociolinguistics, where
‘predictions about linguistic features correlating with situational ones such
as age, gender, geographical and/or social provenance, have typically been
stated in probabilistic terms’ (Hasan 2014:7; see also Hasan 1995, 1999).
With regard to the concept of ‘permeability’, Hasan posits that this has to
do with the ‘conditioned environments’ and the ‘regularity of a set of
relations’ that show how meanings are ‘expanded’ in the sense that the
meaning ‘space’ of certain linguistic categories is permeated by the mean-
ings of another category, and complementary to this, certain linguistic
categories ‘may abandon’ their meaning space and permeate that of other
categories (Hasan 2016:passim). Permeability exists between categories at
the same stratum of language, with ‘the diagnosis of permeability [being]
assisted by inter-stratal relation’ (Hasan 2016:374). Realization is thus
‘needed to recognize the relation of permeability’ (Hasan 2016:374), but
realization is of a different kind of relation. The concept of permeability is
thus different from both realization and interdependency.3
Another important contribution made by Hasan to the study of context
and register is her work on the paradigmatic representation of context.
For this she uses system networks, the modus operandi for representing
the paradigmatic nature of different features at different language strata
(see Webster, this volume). Hasan has argued that ‘the design of the
system network is well suited to contextual parameters’ (Hasan
2014:14) as much as it is to representing choices at other strata. This
does not mean this form of representation is necessarily straightforward
(see Bowcher 2014 for a discussion of some of the issues): networks can
get highly complex, and the different levels of abstraction require a
different ‘value’ assigned to the descriptive choices. Essentially, however,
system networks are ‘a form of argument’ and ‘are a consistent means of
checking what is the better motivated proposal in linguistic description
[in that] [t]he network either accounts for the linguistic variation and its
consequences, choice by choice, or it does not’ (Butt 2001:1825, emphasis
in original). While there has been considerable progress in the develop-
ment of system networks for each of the contextual variables of field,
tenor, and mode, including work by Berry (2016), Bowcher (2007, 2013,
2014), Butt (2004), and Hasan (1999, 2009, 2014), there is still much work
that needs to be done in specifying their features and applying them to
the study of context and register. An example of a system network for
contextual field is shown in Figure 6.2.
3
See Hasan (2016) for a discussion and demonstration of her concept of permeability, and Miller and Bayley (2016) for
work on the concept of hybridity.
praccal
ACTION
conceptual
natural (Sensible)
irrealis (Intelligible)
specialized
FIELD SPHERE OF ACTION
quodian
instuonal
individualized
bounded
spao-temporal
locaon connuing
PERFORMANCE OF ACTION immediate

longitudinal
goal orientaon overt

unconscious
constant
variable
Figure 6.2 System network for field (from Bowcher 2014:203)
6.3.2 Matthiessen: Registerial Cartography

The importance of register and context in the study of language is under-
scored in the following comment by Matthiessen:
Languages are aggregates of registers, and they evolve through registers.

Registers emerge as adaptations to new contextual pressures on
languages . . . and they may fade away as contextual conditions change: the
registerial make-ups of languages keep evolving, changing the character of
languages in the course of evolution.
(Matthiessen 2014a:7; also see Matthiessen and Teruya 2016:212)
Matthiessen and colleagues are developing a context-based typology of

registers, described as ‘a large-scale registerial cartography for a wide range
of languages’ (Matthiessen 2015:2). The research aims to
examine, describe and theorize registers according to Halliday’s trinocular

vision . . . supplementing the view ‘from above’ – from contexts, with the
views ‘from below’ – from lexicogrammar and phonology (or graphology),
and ‘from roundabout’ – from the level of semantics itself, the level at which
the variation takes place in the first instance (in terms of the ‘meanings at
risk’ in different contexts).
(Matthiessen 2014a:8, emphasis in original)
This work on mapping registers derives directly from Jean Ure’s unpub-
lished work on register classification, which focused primarily on field and
mode (Matthiessen 2015:3–5). The starting point for classifying registers is
context, since registers are categorized by reference to their contexts of
situation (Matthiessen 2015). Thus, features from all parameters of context
of situation are relevant to categorizing a register, although so far in this
project, it is the contextual variable of field that has been given the most
attention.
Since field concerns the ‘nature of the activity’, Matthiessen uses the
concept of activity, or social process, as his starting point. He distinguishes
three broad types of processes which he defines in the following way:
(a) Semiotic processes (i.e. ‘meaning’ processes – semiotic processes consti-
tutive of context, manifested through social processes)
(b) Semiotic processes potentially leading to social processes (i.e. ‘meaning’
leading to ‘doing’)
(c) Social processes (i.e. ‘doing’ processes – social processes constitutive of
context, semiotic processes facilitating (i.e. ‘meaning’ facilitating
‘doing’)).
(Matthiessen 2014b:170–1)
These three superordinate categories capture a range of contexts: from

those which are entirely constituted by language through to those which
are almost entirely materially construed but in which language plays a
facilitative role. Within these three broad categories are eight primary
fields of activity: ‘expounding’, ‘reporting’, ‘recreating’, ‘sharing’, and
‘exploring’ fall within category (a); ‘recommending’ and ‘enabling’ fall
within (b); and ‘doing’ falls within (c). Each of these primary fields can be
subdivided to produce more delicate descriptions of activity types.
Figure 6.3 shows the eight primary fields of activity (in the inner circle)
with further subcategorization into secondary fields (in the outer circle).
The three broad categories, however, are not shown. The figure itself
represents a ‘typology’ of fields of activity.
In Figure 6.3, we can see that each primary field of activity can be further
subdivided into more delicate choices such as in the case of ‘recommend-
ing’ into ‘promoting’ or ‘advising’. As more analyses are conducted, greater
degrees of delicacy of these socio-semiotic processes will emerge. For
instance, more delicate distinctions for the category of ‘expounding’ are
given in Figure 6.4 (also see Matthiessen 2015:10).
Figure 6.3 Matthiessen’s map of the different ‘fields of activity’ (from Matthiessen 2014a:11)
Figure 6.4 More delicate distinctions for the category of ‘expounding’
In registerial mapping, as in any account of register, contextual descrip-

tions go hand in hand with analyses at the semantic and lexicogrammatical
level. However, Matthiessen also makes use of rhetorical structure theory
(RST) (Mann and Thompson 1988; Mann et al. 1992) in analyzing the logico-
semantic relations between parts of a text. RST is a descriptive framework
for analyzing text structure in terms of the patterns of relations that hold
between parts of a text, the hierarchical arrangement of text elements, and
the communicative role played by text structure. Matthiessen adapts RST
for a more systems-focused view of text relations. He proposes a system
termed logico-semantic relation. This system has three primary
systems: nuclearity, logico-semantic type, and orientation. The
system of nuclearity is derived directly from RST and concerns the
distinction between ‘relations linking the text segments as equal in

status . . . or as unequal’ (Matthiessen 2014a:15); logico-semantic type
presents choices between projection and expansion (see Halliday and Mat-
thiessen 2014; Berry, this volume); and orientation presents a choice
between ‘linking two text segments as representations of experience
(“external”) or as interaction moves (“internal”)’ (Matthiessen 2014a:15;
also see Halliday and Hasan 1976). Analyses involve shunting across strata
to build a picture of the linguistic and contextual features of different
registers in order to demonstrate and describe how they differ or overlap
with each other (Matthiessen 2015).
There are various publications in which the registerial cartography pro-
ject is described and exemplified. For instance, Matthiessen (2014b) pre-
sents a useful description and exemplar analysis utilizing the registerial
mapping techniques in relation to developing an SFL ‘appliable discourse
analysis’ (ADA), and Matthiessen and Teruya (2016) work through several
different texts which display different types of registerial indeterminacy,
such as ‘blends’, ‘ambiguity’, and ‘overlap’. Matthiessen (2015) presents
analyses of several text types in order to illustrate the nature of the analyses
involved in registerial mapping. He also presents a discussion of how
registerial mapping can be applied in the fields of educational linguistics
and in healthcare communication studies. In the latter area, Matthiessen
argues that the mapping of healthcare discourses, particularly those con-
cerned with ‘communication in emergency (or accident and emergency)
departments in large hospitals’ (Matthiessen 2015:44) can be used to ‘iden-
tify the registers that a patient is likely to have to engage with both within
and outside institutions of healthcare . . . [and] trace patient journeys
through a hospital department’ (Matthiessen 2015:45).
Matthiessen’s work on registerial cartography engages directly with
research on genre typology and topology conducted by Martin and col-
leagues (see Section 6.3.3). Matthiessen acknowledges ‘many connections
with the very rich and detailed work on “genre agnation”, typically
“genres” of writing . . . within the “genre model” (Matthiessen 2014b:173).
6.3.3 Martin: Context: Genre and Register

In modelling the relationship between context and language, Martin (1992)
utilizes the Hjelmslevian concepts of ‘connotative’ and ‘denotative’ semi-
otic and ‘planes’ of analysis (Hjelmslev 1961). Context is modelled as a
connotative semiotic, a semiotic system which is abstract and requires a
different semiotic system through which it is expressed, this latter system
being thus denotative. In Martin’s model, the connotative plane is stratified
into genre (context of culture) and register (context of situation). Together,
these make up the ‘content plane’. Language is then modelled as the
‘expression plane’, the denotative system. Thus, register in Martin’s model
is not treated as a text type, as we have seen in Halliday’s, Hasan’s, or
Matthiessen’s conception of register. Rather, register is located at the

stratum of context and is ‘constituted by the contextual variables field,
tenor and mode . . . [and] is the name of the metafunctionally organized
connotative semiotic between language and genre . . . a semiotic system in
its own right’ (Martin 1992:502).
The contextual plane of genre is proposed as a means of accounting for
the goal or purpose of a social process in the sense that analysis at this level
focuses ‘on making explicit just which combinations of field, tenor and
mode variables a culture enables, and how these are mapped out as staged,
goal-oriented social processes’ (Eggins and Martin 2012:175). The term
‘genre’ replaced Martin’s earlier term ‘functional tenor’, which had been
borrowed from Gregory (1967; see Martin 1999). In Martin’s early work,
functional tenor was treated as a variable different from the other context-
ual variables of field, tenor, and mode, and was considered useful in
describing the social purpose of a text (Martin 1999:28). Martin argues that
separating functional tenor (genre) from the other contextual variables had
‘the advantage of consolidating Halliday’s suggestion that field was natur-
ally related to ideational meaning, tenor to interpersonal meaning and
mode to textual meaning’ (Martin 1999:27; see also Martin 1992:505).
An important difference between genre and register in Martin’s model is
that whereas register is metafunctionally organized, i.e. is organized into three
parameters of field, tenor, and mode, which correlate with the semantic
systems of ideational, interpersonal, and textual meanings, genre is not meta-
functionally organized. Martin argues that this means that ‘texts can be classi-
fied in ways which cut across metafunctional components in language’ (Martin
1992:505) in the sense that a text can be classified as a specific genre-type, with
a specific schematic structure, and within the structural components of the
text different values of field, tenor, and mode may be realized at different
stages in the genre. Thus, while register is a ‘metafunctionally diversified’
‘reading’ of context, genre provides a ‘metafunctionally transcendent’ ‘read-
ing’ (Martin 2012a:276); it is a plane above register. Further, genre becomes a
‘pattern of register patterns’ in that it ‘shapes’ register ‘by conditioning the
way in which field, mode and tenor are recurrently mapped onto one another
in a given culture’ (Martin 2012d:64). Martin’s stratified connotative semiotic
model of context with language as denotative semiotic is shown in Figure 6.5.
Martin has also proposed the term ‘macro-genre’ to refer to texts which
contain more than one genre. In (Martin 2012b) he uses a secondary school
student’s written geography report entitled ‘Endangered Species’ to illus-
trate how several genres may be drawn on to produce one macro-genre. An
interesting direction for research into macro-genres is the exploration of
not only how different genres within a macro-genre relate to each other,
but ‘into how many and what kinds of macro-genre’ different genres may
occur (Martin 2012e:313).
The relation among the analytical planes as shown in Figure 6.5 is one of
realization, with the relation of realization shown by the double-headed
connotative semiotic
stratified context plane expression form
tenor
genre field
mode
discourse phonology/
lexicogrammar
semantics graphology
stratified content form expression form
denotative semiotic
Figure 6.5 Martin’s model of stratified context plane (connotative semiotic) with
language as expression form (denotative semiotic) (from Martin 1999:40)
arrows. Martin explains realization as ‘a scale of abstraction involving the

recoding of one level of meaning as another’ (Martin 2008:32). Thus, for
example, lexicogrammar realizes the more abstract level of discourse
semantics, and the denotative semiotic of language as expression form
realizes the connotative semiotic of context, the latter being a more
abstract plane. Instantiation is also modelled as a scale, but in this case it
is a scale of ‘generalization’ (Martin 2008:32) and can be represented as an
‘instantiation hierarchy’ (see Figure 6.6).
System (generalized meaning potential)

Genre/register (semantic sub-potential)
Text type (generalized actual)
Text (affording instance)
Reading (subjectified meaning)
Figure 6.6 Martin’s instantiation hierarchy (Martin 2008:33)
On the right-hand side of Figure 6.6 are several key terms. ‘Generalized
meaning potential’ refers to the totality of choices possible within a culture.
Within this system are ‘sub-potentials’, such as genres and registers. The
term ‘generalized actual’ refers to the text type, which is also a potential, in
that various individual texts may be classified as a certain type, but unlike
genre and register, which are ‘extra-linguistic’ levels, this term refers to the
categorization of a group of actual texts. The term ‘affording instance’
attempts to capture the idea that the semantic and lexicogrammatical
choices of a text motivate a certain interpretation or reading of a text,
and the notion of ‘subjectified reading’ refers to the way a text is actually
read and understood, deriving from the ‘meaning potential afforded by
individual texts’ (Martin 2008:33). Thus, the hierarchy represents a
narrowing of perspective from potential (most generalized) to instance,
the latter being ‘the reading of a particular text’ (Martin 2008:33).
Martin’s work includes comparing and contrasting related genres (genre

agnation) to explore the ways in which genres are alike (a typological
perspective) and how and to what extent they are similar or different (a
topological perspective). That is, a typological analysis categorizes texts as
belonging to one generic type or another (Martin 2012a:262). The basis for
categorization depends on what is seen as ‘foregrounded’ in a text; it is
‘privileging one dimension of texture over another as more or less critical
for categorization’ (Martin and Rose 2008:130). For instance, in developing a
typology of history texts, Martin and Rose start with the basic categorical
opposition of texts that are ‘field timed’ (following a chronological flow)
and those which are ‘text timed’ (following a logogeneric/textual flow).
A topological analysis, on the other hand, ‘approaches genre agnation as a
matter of degree, arranging texts on clines with respect to their similarities
and differences’ (Martin 2012a:262). For instance, in order to ascertain
similarities and differences among texts, analysis might focus on such
features as choices in mood, expansion, or tense (see Martin
2012a:263–4). Martin has shown the value of this kind of analysis in
educational contexts where ‘topological analysis has been used . . . to map
learner pathways for generic development’ (Martin 2012a:264), particularly
in primary and secondary literacy curricula.
Martin’s concept of genre has been widely applied within various fields of
education including literacy development in mother-tongue and second-
language contexts and cross-disciplinary language and curriculum develop-
ment (see Martin 2012c; Byrnes, this volume; Mickan, this volume).
6.4 Context: Networks and Scales
Recent work on context includes advances in the development of system

networks for the contextual parameters of field, tenor, and mode. For
instance, working with networks developed by Butt (2004) and Hasan
(1999, 2009), Bowcher (2013, 2014) has developed the field and mode
networks in an attempt to ‘focus the Field network more on the nature of
the activity in terms of its experiential elements and at the same time . . .
[focus] the Mode network more closely on features to do with the modalities
of expression and the degree to which these modalities may be deployed in
a situation’ (Bowcher 2014:199; also see Figure 6.2 in Section 6.3.1). Part of
her work has centred on the argument that the primary choices within the
networks ‘should reflect the core defining features of the contextual par-
ameters’ and her system networks reflect this stance (Bowcher 2014:177).
She has also considered the contextual choice of material action proposed
by Hasan (1999), and whether this choice should be located in the network
for field or for mode, with Bowcher (2014) contending that it is better
located in mode. Her arguments concerning material action include issues
surrounding the contexts of multimodal texts (Bowcher 2007, 2013, 2014),
as well as those surrounding the concepts of relevancy (what falls within

relevant context) and trace (how do features of different contexts leave their
trace (or not) in the language of the text) (Bowcher 2013). These latter issues
are also included in Bartlett’s (2013, 2016, 2017) work on context.
Bartlett has explored the relationship between context of situation and
the ‘environment’ within which a context is embedded, arguing that we
need to look ‘beyond the text to the semiotic histories that are embedded in
the wider environment’ (Bartlett 2013:346) and thus incorporate into our
understanding of the relationship between context and text sociological
and ethnographic detail. Part of his argument rests on the observation that
context of situation cannot always be ‘read off the text’ (Bartlett 2017:317).
Rather, certain choices of language, or twists and turns in a text, may
reflect cultural knowledge and understandings rather than ‘illuminate’
any specific contextual feature (Bartlett 2013:348). In order to explore the
relations between context and language, Bartlett takes up the concepts of
first- and second-order context (see Halliday 1977), where, for example, the
first-order field might be a game of football and the second-order field a
discussion of that game; first-order tenor relations would be those not
defined by language but by the social system, such as mother and child,
whereas second-order tenor relations come into being within a given situ-
ation, such as questioner-respondent (Halliday 1977:201–2; Matthiessen
2009). Bartlett has also questioned certain relationships said to pertain
between context and language. For instance, the relations of ‘construal’
and ‘activation’, as in the following formulation: language construes con-
text and context activates language (see Hasan 2014). He contends that
these relations are not truly converse relations for all aspects of context
and language. For instance, language does not ‘construe’ the channel of
communication (Bartlett 2016).
Taking all these issues into account, Bartlett argues for a multiscalar
model of context consisting of ‘environment’ – ‘everything that surrounds
the situation, including social and individual histories, as well as material
features of the setting’ – and ‘semiotic context’, which he defines as ‘the
second-order reality that is construed by, and which can be read off, the text
itself’, and which he labels ‘sctx’ (Bartlett 2017:385). He also proposes four
concepts which he suggests better describe the relations between context
and text: ‘activation’ (the influence of any feature of the environment on
any aspect of a text whether that be covert or overt); ‘construal’ (the way in
which contextual features can be explicitly ‘read off’ the text); ‘correlation’
(the ‘tendencies’ for certain language features to ‘co-occur’ with certain
environmental features); and ‘indexicality’ (‘ways of speaking’ associated
with socially established activities or groups of people) (see Bartlett
2017:385–6).
Berry (2016) is also concerned with the notion of ‘construal’ and with not
only features of context that can be read off a text but which may influence
linguistic choices in some way. She proposes a model of context that takes
into account ‘pre-text’ context and ‘via-text’ context features. Pre-text con-
textual features may be within the general parameters of field, tenor, and
mode. For instance, prior to a telephone conversation, there are the pre-text
mode features of ‘spoken’ and ‘not-co-present’ already in place. Social
distance features might also be in place prior to a text coming into exist-
ence, and hence be pre-text tenor features, but during the text, such
features may change, such as when a friend who is also one’s boss switches
from friendly chat to outlining some job requirements needed within the
next day or two. Such changes to the social status quo would be considered
via-text contextual features. Berry has developed system networks to
account for these kinds of features (see Berry 2016).
This chapter has provided only an outline of the concepts of context and
register, their development, models, and current research. Context and
register are cornerstones of SFL theory (Lukin et al. 2011) and offer a rich
and varied way into studying language and its relation to society. Readers
are encouraged to consult the reference list to gain a deeper understanding
of the concepts, their theoretical and research value, and their analytical
utility.
References
Bartlett, T. 2013. ‘I’ll Manage the Context’: Context, Environment and

the Potential for Institutional Change. In L. Fontaine, T. Bartlett, and
Bartlett, T. 2016. Multiscalar Modelling of Context: Some Questions Raised
by the Category of Mode. In W. L. Bowcher and J. Y. Liang, eds., Society in
Language, Language in Society: Essays in Honour of Ruqaiya Hasan. Basing-
stoke: Palgrave Macmillan. 166–83.
Bartlett, T. 2017. Context in Systemic Functional Linguistics: Towards
Scalar Supervenience? In T. Bartlett and G. O’Grady, eds., The Routledge
Handbook of Systemic Functional Linguistics. London: Routledge. 375–90.
Berry, M. 2016. On Describing Contexts of Situation. In W. L. Bowcher and
J. Y. Liang, eds., Society in Language, Language in Society: Essays in Honour of
Ruqaiya Hasan. Basingstoke: Palgrave Macmillan. 184–205.
Bortoluzzi, M. 2010. Energy and Its Double: A Case Study in Critical Multi-
modal Discourse Analysis. In W. Swain, ed., Thresholds and Potentialities of
Systemic Functional Linguistics: Applications to Other Disciplines, Specialised
Discourses on Languages Other than English. Trieste: Edizione Universitarie
Trieste. 158–81.
Bowcher, W. L. 1999. Investigating Institutionalization in Context. In

M. Ghadessy, ed., Text and Context in Functional Linguistics. Amsterdam:
Bowcher, W. L. 2007. Field and Multimodal Texts. In R. Hasan, C. M. I. M.
Matthiessen, and J. J. Webster, eds., Continuing Discourse on Language,
Volume 2. Sheffield: Equinox. 619–46.
Bowcher, W. L. 2013. Material Action as Choice in Field. In L. Fontaine,
T. Bartlett, and G. O’Grady, eds., Systemic Functional Linguistics: Exploring
Choice. Cambridge: Cambridge University Press. 318–41.
Bowcher, W. L. 2014. Issues for Developing Unified Systems for Contextual
Field and Mode. Functions of Language 21(2): 176–209.
Bowcher, W. L. 2015. Structure and Multimodal Texts. In J. Wildfeurer, ed.,
Building Bridges for Multimodal Research: International Perspectives on Theories
and Practices of Multimodal Analysis. Frankfurt am Main: Peter Lang. 167–89.
Bowcher, W. L. and J. Y. Liang. 2016. GSP and Multimodal Texts. In W. L.
Bowcher and J. Y. Liang, eds., Society in Language, Language in Society: Essays
in Honour of Ruqaiya Hasan. Basingstoke: Palgrave Macmillan. 251–74.
Butt, D. G. 2001. Firth, Halliday and the Development of Systemic Func-
tional Theory. In S. Auroux, E. F. K. Koerner, H.-J. Niederehe, and
K. Versteegh, eds., History of the Language Sciences. Berlin: Walter de Gruy-
ter. 1806–38.
Butt, D. G. 2004. Parameters of Context: On Establishing the Similarities and
Differences between Social Processes. Unpublished mimeo. Sydney: Macquarie
University.
Butt, D. G., A. R. Moore, C. Henderson-Brooks, R. Meares, and J. Haliburn.
2010. Dissociation, Relatedness and ‘Cohesive Harmony’: A Linguistic
Measure of Degrees of ‘Fragmentation’. Linguistics and the Human Sciences
3(3): 263–93.
Carter, R. 1978. Register, Styles and Teaching Some Aspects of the Language
of Literature. Educational Review 30(3): 227–36.
Catford, J. C. 1965. A Linguistic Theory of Translation. London: Oxford Univer-
sity Press.
Cheong, Y. Y. 2004. The Construal of Ideational Meaning in Print Advertise-
ments. In K. O’Halloran, ed., Multimodal Discourse Analysis: Systemic-
functional Perspectives. London: Continuum. 163–95.
Cloran, C. 1987. Negotiating New Contexts in Conversation. Occasional
Papers in Systemic Linguistics 1: 85–110.
Cloran, C. 1994. Rhetorical Units and Decontextualisation: An Enquiry into Some
Relations of Context, Meaning and Grammar. Nottingham: University of
Nottingham.
Cloran, C. 1999 Context, Material Situation and Text. In M. Ghadessy, ed., Text
and Context in Functional Linguistics. Amsterdam: John Benjamins. 177–217.
Cloran, C. 2016. Construing Instructional Contexts. In W. L. Bowcher and
J. Y. Liang, eds., Society in Language, Language in Society: Essays in Honour of
Ruqaiya Hasan. Basingstoke: Palgrave Macmillan. 275–99.
Eggins, S. and J. R. Martin. 2012. Genres and Registers of Discourse. In

Z. Wang, ed., Genre Studies: Collected Works of J. R. Martin, Volume 3. Shang-
hai: Shanghai Jiaotong University Press.
Ellis, J. 1965. Linguistic Sociology and Institutional Linguistics. Linguistics
19: 5–20.
Ellis, J. 1966. On Contextual Meaning. In C. E. Bazell, J. A. Catford, M. A. K.
Halliday, and R. H. Robins, eds., In Memory of J. R. Firth. London: Longmans.
79–95.
Ellis, J. and J. Ure. 1974. The Contrastive Analysis of Language Registers. Unpub-
lished mimeo.
Firth, J. R. 1950. Personality and Language in Society. The Sociological Review:
Journal of the Institute of Sociology 42(2): 37–52.
Firth, J. R. 1957. A Synopsis of Linguistic Theory 1930–55. In J. R. Firth et al.,
eds., Studies in Linguistic Analysis: Special Volume of the Philological Society.
Oxford: Blackwell. 1–32.
Godley, A. and A. Escher. 2012. Bidialectal African American Adolescents’
Beliefs about Spoken Language Expectations in English Classrooms. Jour-
nal of Adolescent and Adult Literacy 55(8): 704–13.
Gregory, M. 1967. Aspects of Varieties Differentiation. Journal of Linguistics
3(2): 177–274.
Halliday, M. A. K. 1959/1974. The Language of the Chinese ‘Secret History of the
Mongols’. Oxford: Basil Blackwell.
241–92.
Halliday, M. A. K. 1977. Text as Semantic Choice in Social Contexts. In T. A.
van Dijk and J. S. Petofi, eds., Grammars and Descriptions. Berlin: Walter de
Gruyter. 176–225.
Halliday, M. A. K. 1985. Part A. In M. A. K. Halliday and R. Hasan, Language,
Context, and Text: Aspects of Language in a Social-semiotic Perspective. Geelong:
Deakin University Press. 1–49.
Halliday, M. A. K. 1999. The Notion of ‘Context’ in Language Education. In
M. Ghadessy, ed., Text and Context in Functional Linguistics. Amsterdam:
Halliday, M. A. K., A. McIntosh, and P. Strevens. 1964. The Linguistic Sciences
and Language Teaching. London: Longmans.
Hasan, R. 1973. Code, Register and Social Dialect. In B. Bernstein, ed., Class,
Codes and Control, Volume 2: Applied Studies towards a Sociology of Language.
Hasan, R. 1978. Text in the Systemic-functional Model. In W. U. Dressler,

ed., Current Trends in Textlinguistics. Berlin: Walter de Gruyter. 228–46.
Hasan, R. 1981. What’s Going on: A Dynamic View of Context in Language.
In J. E. Copeland and P. W. David, eds., The Seventh LACUS Forum 1980.
Columbia: Hornbeam Press. 106–21.
Hasan, R. 1985. Part B. In M. A. K. Halliday and R. Hasan, Language, Context,
and Text: Aspects of Language in a Social-semiotic Perspective. Geelong: Deakin
Hasan, R. 1995. The Conception of Context in Text. In P. H. Fries and
M. Gregory, eds., Discourse in Society: Systemic Functional Perspectives. Meaning
and Choice in Language: Studies for Michael Halliday. Norwood: Ablex.
Hasan, R. 1999. Speaking with Reference to Context. In M. Ghadessy, ed.,
Text and Context in Functional Linguistics. Amsterdam: John Benjamins.
219–328.
Hasan, R. 2001. Wherefore Context? The Place of Context in the System and
Process of Language. In R. Shaozeng, W. Guthrie, and I. W. Ronald Fong,
eds., Grammar and Discourse: Proceedings of the International Conference on
Discourse Analysis. PRC Macau: Publications Centre, University of Macau.
1–30.
Hasan, R. 2005. Language and Society in a Systemic Functional Perspective.
In R. Hasan, C. M I. M. Matthiessen, and J. J. Webster, eds., Continuing
Discourse on Language: A Functional Perspective, Volume 1. Sheffield: Equinox.
55–80.
Hasan, R. 2009. The Place of Context in a Systemic Functional Model. In
M. A. K. Halliday and J. J. Webster, eds., Continuum Companion to Systemic
Hasan, R. 2016. In the Nature of Language: Reflections on Permeability and
Hybridity. In D. R Miller and P. Bayley, eds., Hybridity in Systemic Functional
Linguistics: Grammar, Text and Discursive Context. Sheffield: Equinox.
337–83.
Hjelmslev, L. 1961. Prolegomena to a Theory of Language. Madison: University
of Wisconsin Press.
Khoo, K. 2016. ‘Threads of Continuity’ and Interaction: Coherence, Texture
and Cohesive Harmony. In W. L. Bowcher and J. Y. Liang, eds., Society in
Lamb, S. 1966. Outline of Stratificational Grammar. Washington, DC: George-
town University Press.
Lukin, A. 2010. ‘News’ and ‘Register’: A Preliminary Investigation. In
A. Mahboob and N. K. Knight, eds., Appliable Linguistics. London: Con-
tinuum. 92–113.
Lukin, A. 2016. Language and Society, Context and Text: The Contributions
of Ruqaiya Hasan. In W. L. Bowcher and J. Y. Liang, eds., Society in

Lukin, A., A. R. Moore, M. Herke, R. Wegener, and C. Wu. 2011. Halliday’s
Model of Register Revisited and Explored. Linguistics and the Human Sciences
4(2): 187–213.
Malinowski, B. 1923. The Problem of Meaning in Primitive Languages.
Supplement 1. In C. K. Ogden and I. A. Richards, eds., The Meaning of
Meaning. New York: Harcourt, Brace & World.
Mann, W. C and S. A. Thompson 1988. Rhetorical Structure Theory: Toward
a Functional Theory of Text Organization. Text 8(3): 243–81.
eds., Discourse Description: Diverse Linguistic Analyses of a Fund-raising Text.
Benjamins.
Martin, J. R. 1999. Modelling Context: A Crooked Path of Progress in
Contextual Linguistics. In M. Ghadessy, ed., Text and Context in Functional
Linguistics. Amsterdam: John Benjamins. 25–61.
Martin, J. R. 2008. Tenderness: Realisation and Instantiation in a Botswanan
Town. In N. Nørgaard, ed., Systemic Functional Linguistics in Use. Odense:
University of Southern Denmark. 30–62.
Martin, J. R. 2012a. A Context for Genre: Modelling Social Processes in
Functional Linguistics. In Z. Wang, ed., Genre Studies: Collected Works of
J. R. Martin, Volume 3. Shanghai: Shanghai Jiaotong University Press.
248–77.
Martin, J. R. 2012b. From Little Things Big Things Grow: Ecogenesis in
School Geography. In Z. Wang, ed., Genre Studies: Collected Works of J. R.
Martin, Volume 3. Shanghai: Shanghai Jiaotong University Press. 278–302.
Martin, J. R. 2012c. Language in Education: Collected Works of J. R. Martin,
Volume 7. Edited by Z. Wang. Shanghai: Shanghai Jiaotong University
Press.
Martin, J. R. 2012d. Language, Register and Genre. In Z. Wang, ed., Register
Studies: Collected Works of J. R. Martin, Volume 4. Shanghai: Shanghai Jiao-
tong University Press. 47–68.
Martin, J. R. 2012e. A Universe of Meaning: How Many Practices? In
Z. Wang, ed., Genre Studies: Collected Works of J. R. Martin, Volume 3. Shang-
hai: Shanghai Jiaotong University Press. 303–13.
Martin, J. R. and D. Rose. 2008. Genre Relations: Mapping Culture. Sheffield:
Equinox.
Matthiessen, C. M. I. M. 2009. Ideas and New Directions. In M. A. K. Halliday
and J. J. Webster, eds., Continuum Companion to Systemic Functional Linguis-
tics. London: Continuum. 12–58.
Matthiessen, C. M. I. M. 2014a. Registerial Cartography: Context-based
Mapping of Text Types and Their Rhetorical-relational Organization.
Proceedings of the 28th Pacific Asia Conference on Language, Information and

Computation. Unpublished conference paper.
Matthiessen, C. M. I. M. 2014b. Appliable Discourse Analysis. In Y. Fang and
J. J. Webster, eds., Developing Systemic Functional Linguistics: Theory and
Application. Sheffield: Equinox. 138–208.
Matthiessen, C. M. I. M. 2015. Register in the Round: Registerial Cartog-
Matthiessen, C. M. I. M. and K. Teruya. 2016. Registerial Hybridity: Indeter-
minacy among Fields of Activity. In D. R. Miller and P. Bayley, eds.,
Hybridity in Systemic Functional Linguistics: Grammar, Text and Discursive
Context. Sheffield: Equinox. 205–39.
McIntosh, A. 1961. ‘Graphology’ and Meaning. Archivum Linguisticum 13:
107–20.
Miller, D. R. and P. Bayley, eds. 2016. Hybridity in Systemic Functional Linguis-
tics: Grammar, Text and Discursive Context. Sheffield: Equinox.
Ogden, C. K. and I. A. Richards. 1923. The Meaning of Meaning. New York:
Harcourt, Brace & World.
Reid, T. B. W. 1956. Linguistics, Structuralism and Philology. Archivum
Linguisticum 8: 28–37.
Thibault, P. J. 1987. An Interview with Michael Halliday. In R. Steele and
T. Threadgold, eds., Language Topics: Essays in Honour of Michael Halliday.
Ure, J. 1982. Introduction: Approaches to the Study of Register Range.
International Journal of the Sociology of Language 35: 5–33.
White, R. 1974. The Concept of Register and TESL. TESOL Quarterly 8(4):
401–16.
7
Intonation
Wendy L. Bowcher and Meena Debashish
In memory of William S. Greaves (1935–2014).
7.1 Introduction
Intonation has always been a part of the description of language in Systemic

Functional Linguistics (SFL), but it has received less attention than many
other areas of the theory. Work that is substantively devoted to the descrip-
tion of intonation primarily includes Halliday’s (1967, 1970) early publica-
tions, Tench (1996),1 and Halliday and Greaves (2008). There are also two
collections focusing on Systemic Phonology: Tench (1992b) and Bowcher
and Smith (2014b). This chapter draws from all these volumes and other
individual publications to present an overview of the SFL model of inton-
ation and a detailed description of the English tone groups. Our aim is to
provide a usable description of English intonation with occasional reference
to recent research, areas under debate, and possible future directions.
7.2 Background
Intonation is the main feature of phonology that is emphasized in SFL theory.

The SFL description of intonation derives largely from the work of Firth, who
distinguished ‘prosodic systems from phonematic systems’ (Firth 1948:128)
thus placing equal emphasis on syntagmatic and paradigmatic features. His
concept of ‘prosody’ was different from that of ‘suprasegmental features’
as developed in (mostly) American structuralist traditions in that Firth
considered sounds and phonological features in context thus revealing a
1
Although aligning with much of the SFL intonation framework, there are some differences between Tench’s (1996)
description of intonation in English and Halliday’s.
172 WENDY L. BOWCHER AND MEENA DEBASHISH
“top-down” analysis of language’ (Couper-Kuhlen 1986:4). Further, Firth con-

sidered the structuralist focus on ‘segmentation’ of sounds as ‘artificial and
actually misrepresent[ing] the phonology of the language’ (Tench 1992a:5).
Rather, he aimed to integrate into the study of phonology such features as the
‘interrelation of syllables’ in terms of length, stress, tone quality, nasality, and
voice quality (Firth 1948:138), and the study of phonology more fully into a
theory with the investigation of meaning at its core.
Firth’s prosodic analysis was non-universalist and polysystemic: non-
universalist in that the phonological systems of different languages were
described in their own terms, acknowledging that ‘phonetic features which
in one language are treated as prosodic may not be so treated, or may be so
with reference to different structures, in other languages’ (Robins
1970:194); and polysystemic in that the approach ‘treats language as a
complex set of interacting systems, each with its own characteristic prop-
erties’ (Ogden 2012: para 3). With regard to his polysystemic approach,
Firth (1948:151) says:
The phonological structure of the sentence and the words which comprise it
are to be expressed as a plurality of systems of interrelated phonematic and
prosodic categories. Such systems and categories are not necessarily linear
and certainly cannot bear direct relations to successive fractions or segments
of the time-track instances of speech. By their very nature they are abstrac-
tions from such time-track items. Their order and interrelations are not
chronological.
Meaning was central to Firth’s ideas on language (see Butt, this volume) and
to the study of phonology:
The meaning of any particular instance of everyday speech is intimately

interlocked not only with an environment of particular sights and sounds,
but deeply embedded in the living processes of persons maintaining them-
selves in society . . . the dominating interest of the immediate situation, the
urge to diffuse or communicate human experience, the intimate sounds,
these are the origins of speech.
(Firth 1968:13)
Such an approach interprets the sounds of speech as realizing ‘worded

meaning within context of situation’ (Bowcher and Smith 2014a:9).
Meaning is likewise central to SFL theory. From a phonological viewpoint,
spoken language is understood as a ‘succession of melodies’ (Halliday
1970:6) which collaborate with other linguistic choices and resonate with
grammatical, semantic, and contextual features during the act of meaning.
With regard to the relationship between intonation and grammar, Halliday
(1967:10) points out:
It is not enough to treat the intonation systems as if they merely carr[y] a set of
emotional nuances superimposed on the grammatical and lexical items and
categories. . . . English intonation contrasts are grammatical: they are exploited
Intonation 173
in the grammar of the language. The systems expounded by intonation are just
as much grammatical as are those, such as tense, number and mood, expounded
by other means . . . Therefore, in the description of the grammar of spoken
English, ‘intonational’ and ‘non-intonational’ systems figure side by side.
With regard to meaning, intonation is studied from the point of view that
‘[i]f you change the intonation of a sentence you change its meaning . . .
Intonation is one of the many kinds of resources that are available in the
language for making meaningful distinctions’ (Halliday 1970:21). More-
over, its significance in the act of communicating is indispensable, as Tench
(1996:151) comments:
Intonation features at every point in the process of communication; it is

impossible to account for any kind of linguistic communication without it.
From the process of reception by the addressee to interpretation and evalu-
ation, intonation is recognised, processed and taken into account. Even in
writing, and then reading, intonation plays a part.
Over the course of the development of SFL theory, there has been a shift in
the description of intonation in relation to other linguistic categories. For
instance, whereas earlier volumes introducing the SFL model of grammar
described intonation as ‘beside the clause’ (see IFG1 and IFG2),2 later
volumes (IFG3 and IFG4)3 more fully integrate intonation into descriptions
of various features of the grammar, reflecting Halliday’s early assertion
that “intonational” systems operate at many different places in the gram-
mar’, they are not independent but are ‘incorporated throughout the
description wherever appropriate’ (Halliday 1967:10–11).
The different pitch movements – falling, rising, or combinations of
these – contribute to melodic variation in language, and each language
has its own set of tones, a system of tone choices, which contributes to
the meaning-making system of the language. Moreover, as with other
systems and levels of analysis in SFL theory, intonation(al) systems are
interdependent with other systems. This more encompassing view of the
role of intonation in the description of language reflects the SFL multi-
faceted architecture of language. The next section outlines the place of
intonation in the SFL architecture of language.
7.3 Intonation in the SFL Architecture of Language
Within the SFL architecture of language, there are several organizational

dimensions from which language is modelled and studied. They include the
following:
2
IFG1 and IFG2 refer to Halliday (1985a) and (1994) respectively.
3
IFG3 and IFG4 refer to Halliday and Matthiessen (2004) and (2014) respectively.
• The hierarchy of strata related through realization;

• The hierarchy of rank related through constituency;
• The cline of instantiation which is a relation between system and text;
• The axes of paradigmatic and syntagmatic, the former displaying rela-
tions of delicacy and the latter of rank;
• The metafunctional organization of language.
Intonation is discussed in relation to each of these dimensions.
7.3.1 The Hierarchy of Strata Related through Realization

In SFL theory, language is modelled as systems at several strata. These strata
are context, meaning, lexicogrammar, phonology, and phonetics, which are
related through realization.4 The stratum of phonetics has a different
relationship in terms of realization than the other strata because phonetics
refers to the potential resources from which ‘sounded’ meanings are made.
A pitch contour, for instance, is a phonetic feature, but when deployed in a
meaningful way it becomes a part of the phonological description of a
language. In SFL, phonology and lexicogrammar are located at different
strata, and often graphology is shown as the ‘written equivalent’ of phon-
ology. This is true only in so far as they are located at the same stratum, but
it is important to keep in mind that each of these modes has specific
grammatical and semantic consequences and resonances (see Davies 2014;
Fawcett 2014). Halliday distinguishes between intonational systems and
intonation systems. Intonational systems are located at the stratum of
lexicogrammar and include the system of information distribution,
and intonation systems at the stratum of phonology and include such
systems as tonicity and tone.
At the stratum of grammar, spoken English is analyzed in terms of
information units, and at the stratum of sound (phonology) in terms of
tone units, or tone groups. These units are related to each other through
realization: a tone unit realizes an information unit. While realization can
be considered between strata – lexicogrammar realizes semantics – it is
important to keep in mind that in the act of meaning through languaging,
realization is a multi-stratal, simultaneous operation, and not a determin-
istic or ‘one-by-one’ mechanism that takes place between units in one
stratum and those in another, one after the other (see Halliday 1992). In
the overall operation of realization, context redounds with the redundancy
of the semantics with the redundancy of the lexicogrammar with the
phonology, where redound means ‘realizes and is realized by’ (Halliday
1992; see also Taverniers, this volume). The relation of realization is pre-
sented in Figure 7.1 with tone unit and information unit shown.
4
The Cardiff model of language is a bi-stratal model. For a discussion of this in relation to intonation, see Fawcett (2014:
331–4).
Intonation 175
Context (Semancs (Lexicogrammar Phonology ))

informaon unit tone unit
Figure 7.1 The relation of realization among the strata of language with tone unit
and information unit shown [note: the downward slanting arrows mean ‘is realised by’]
(cf. Halliday 1992:24)
However, to say that a tone unit realizes an information unit appears

contradictory to the statement that there is no ‘one-to-one’ relation
between units in one stratum and those in another. This apparent contra-
diction is explained in the following way:
The information unit is a unit of the lexicogrammar; so it functions in the

construction of meaning – it faces the semantics, so to speak. The tone unit,
on the other hand, is a unit of the phonology; it functions in the organization
of speech sound.
(Halliday and Greaves 2008:41)
Thus in terms of grammar, the information unit organizes speech in terms

of the flow of messages, whereas the tone unit organizes speech in terms of
the flow of sound or melodies. So the location of the information unit at the
stratum of grammar and of the tone unit at the stratum of phonology is a
means of modelling these two different perspectives on the same phenom-
enon, that of speech. However, the resource of sound and the more abstract
organizational role of grammar are not the same kind of phenomena.
Therefore, the boundaries of the tone unit and the information unit do
not always coincide exactly. This is illustrated in the following excerpt,
which is analyzed for tone groups. The excerpt also illustrates some of the
conventions used for displaying intonation analyses: a double forward slash
for the boundary of the tone unit, a single forward slash for a foot bound-
ary, a caret5 used to indicate a silent Ictus.
// 4 ᴧ by the /time the /Great /Central was /built the // 1+ trains could /
manage the /gradients /much more /easily and the //13 Great /Central
/line //. . .
(excerpt from Halliday 1970:127)
Table 7.1 indicates that the first and second tone units encompass more
‘wording’ than the information units they realize due to the tail of the tone
melody extending over a few words belonging to the subsequent grammat-
ical units of clauses or phrases. This wording is shown in italicized text in
Table 7.1. When these tone units are represented as information units, the
‘tails’ are shifted into the information unit, which brings the information
5
The use of the caret ᴧ varies across the SFL literature. Some authors prefer to use a raised caret ^ (e.g. Halliday and
Matthiessen 2014; O’Grady 2017), while others (Halliday 1970; Halliday and Greaves 2008) prefer to use the low
caret ᴧ. In this chapter we use the low caret so as to distinguish this from the raised caret which typically means ‘is
followed by’.
Table 7.1 Tone units and information units (tone unit boundaries marked
with double forward slashes)
Type of Unit Utterance
Tone Unit //by the time the Great Central was built the//
Information Unit by the time the Great Central was built
Tone Unit ___//trains could manage the gradients much more easily and the//
Information Unit the trains could manage the gradients much more easily
Tone Unit ______//Great Central line//
Information Unit and the Great Central line
unit more in line with the grammatically discrete elements of clauses,

phrases, groups, or words (see Smith 2008:62 note 1, 96, Appendix 3).
The coinciding of an information unit with a clause is considered to be an
unmarked relation, or ‘neutral tonality’, with deviations from this being
marked (Halliday and Greaves 2008: 142). However, in Fawcett’s generative
model of intonation, the concept of neutral tonality is not taken up. Rather,
the approach is modelled on the lines that where there is a new semantic
(syntactic) unit in a sentence, the choice is whether or not that semantic
unit should be ‘given a separate information unit’ (Fawcett 2014:328).
7.3.2 The Hierarchy of Rank Related through Constituency

The hierarchy of rank is a means of modelling the constituents of each
stratum. For intonation, the hierarchy of rank primarily applies to the
stratum of phonology. For English, the phonological units are tone unit
(melodic line), foot, syllable, and phoneme. These may differ in other
languages. For instance, McGregor (1992) suggests that for Gooniyandi they
are tone unit, word, syllable, and phoneme. In English, a tone group
consists of one or more feet, a foot one or more syllables, and a syllable
one or more phonemes. Figure 7.2 shows the constituents of the tone unit.
melody // TONE UNIT // *The tone unit or melodic line is the highest phonological rank.
melodic line *Tone unit boundary shown by double forward slashes.
rhythm // foot / foot / foot / foot // *A tone unit or melodic line of speech may consist of one or more feet.
*A foot boundary is shown by a single forward slash.
*Each foot contains one or more syllables.

// S2 / / S1 / S1 S2 S3 // *Salient syllables are located immediately to
the right of the single forward slash.
*The first syllable in the tone unit may be silent.
arculaon *A foot may also be silent.
*The caret is used to denote a silent ictus.
*The tonic syllable is underlined.
// phph/ /phphph /phph phph phph // *A syllable is constuted by one or more phonemes.
Figure 7.2 The constituents of the tone unit showing some analytical conventions
As already noted, the tone unit at the phonological stratum realizes the
information unit in the grammatical stratum. At the grammatical stratum
Intonation 177
the information unit is the only rank recognized in Halliday’s model of

systemic phonology and is on par with the clause (see Halliday and Matthies-
sen 2014:115), albeit of a different realizational phenomenon. Matthiessen
(1995:603) suggests that this is because the rank scale ‘is “taken over” from
phonology’. This becomes clear in considering the syntagmatic axis in relation
to the intonation and intonational systems (see Section 7.3.4). However, Smith
(2008) proposes a unit at a lower rank than the information unit which he calls
the information group. An information group is realized by a foot, and the
Ictus in the foot realizes a Prominent. The associated systems are informa-
tion grouping, a system at the rank of information group which is below
the information unit and located in the system of information distribu-
tion, and information prominence, a system at the rank of Prominent,
which is below the rank of information unit and located in the system of
information focus. Smith identifies instances of rank shift between infor-
mation unit and information group. One example of this is the choice of a
foot for each word in a lexicogrammatical group, such as a nominal group.
He explains that this kind of choice construes ‘a textual prosody’, ‘which
heightens the attention to the elements of this group’ (Smith 2008:110). This
is popularly (i.e. often in the mass media) represented in written form by
using a period after each word in a group or a clause:
Every. Single. Piece. (word level)

I Just. Don’t. Know. (clause level)
Smith explains that the selection of a number of Prominents as well as the

Focus (realized by the tonic) in a nominal group or clause ‘may be ascribed
to the “zooming/focussing” power of this system’ (Smith 2008:111).
There is also the suggestion of a unit above the tone unit – a phonological
paragraph (a paraphone), that Halliday raised himself in an article on
grammatical categories (Halliday 1961; see also Iwamoto 2014; Tench
1996). However, this is not included in the description of rank but is
discussed in relation to the textual metafunction (see Section 7.3.5).
7.3.3 The Cline of Instantiation: Relations between

System and Instance
System networks are a hallmark of SFL representations of choices available
at each rank of linguistic analysis and represent sets of choices that are
potentially available in any instance of language in use.
At the rank of foot are the systems of foot composition and ictus
state. The foot is a rhythmic unit, and the choices available range from one
or more syllables. Halliday (1970:1) says:
The first syllable in the foot is always salient. The salient syllable carries the
beat . . . [and is] followed by one or more non-salient, or weak syllables . . . A foot
may begin with a silent beat, without the rhythm being disrupted or lost.
The first syllable in the foot whether silent or spoken is called the Ictus.
Further, this might be the only choice, and thus the foot would be a ‘simple
foot’ which could be silent or spoken. On the other hand, there may be any
number of syllables in a foot, and those syllables following the Ictus,
whether one or more, are known collectively as Remiss. A foot of this latter
type is called a compound foot. The system network for choices in foot
composition and ictus state is shown in Figure 7.3.
At the rank of tone unit, the three main systems of intonation are
tonality, tonicity, and tone (see Figure 7.4). Tonality refers to choices
available for the organization of a discourse into tone units; tonicity refers
to the selection and assignment of prominence within the tone unit; and
tone refers to the choices of pitch movement. Halliday (1963a, 1963b,
1967, 1970) based these systems of English intonation on various samples
of spoken English, the largest consisting of ‘just under 2,000 tone groups’
(Halliday 1967:9).
simple
FOOT
COMPOSITION
compound
+Remiss
Remiss: syllable1-n
Ictus^Remiss
foot
+Ictus
filled
ICTUS Ictus:syllable1-n
STATE
empty
Ictus:silent
Figure 7.3 System network for choices in FOOT COMPOSITION and ICTUS STATE (from Halliday
and Matthiessen 2014:18)
TONALITY
INTONATION TONICITY
SYSTEMS
TONE
Figure 7.4 Intonation systems
Intonation 179
Consider, for instance, the following excerpt from a recording of a story

read for children:
//1 Mr. /Fox//1 ᴧ was /strolling /through the /woods// 3 when he /

noticed// 1 ᴧ a /plump /hen// 1 sitting //1 ᴧ on the /branch of a /tree//
(extract from Storynory n.d.)
In this sample of reading aloud, we can see a high number of tone groups
across one or two clauses. For example, the second clause ‘when he noticed a
plump hen sitting on the branch of a tree’ is spoken in four tone groups. The
choice of reading the sentence into four tone groups represents selections
from the system of tonality, and the selection of tonic prominences (shown
here by bolded text) reflects choices from the system of tonicity. The choice
of tone groups themselves – tones 3, 1, 1, and 1 – represents choices from the
system of tone.
System networks have also been developed for features at the level of
segmental and word phonology, such as Young’s (1992) work on English
consonant clusters and Tench’s (2014) work on English word phonology. The
concept of system, however, is understood somewhat differently when it is
applied at the levels of word, syllable, and phonemes. In discussing his systems
of consonant clusters, Young explains that while the concept of system net-
work implies sets of choices of meanings, ‘by the time we get . . . [to] segmental
phonology all the meaning choices have been made long ago and everything
now to be selected is predetermined’ (Young 1992:58). Tench reiterates this
view: ‘System at the level of word (and also at the level of groups/phrases) is
rather the specifications of what the speakers of a language recognize as
having been established in, or “chosen” by, the language to represent its
words’ (Tench 2014:274). Tench has developed system networks outlining
the possible syllable structures, peaks, margins, strong and weak vowels,
and syllable initial consonants in English (standard southern English pronun-
ciation) (see Tench 2014). An interesting point made by Young is that even
though system networks (for consonant clusters) display predetermined sets
of choices, ‘they probably have a more positive role to play in decoding, and
they certainly need to be built into a model of English which accounts for the
ability of speakers to add to their vocabulary (for example, by means of foreign
loanwords) words that conform to the phonology of English’ (Young 1992:58).
While system networks represent choices, Fawcett (2014) argues that the
typical SFL system networks for intonation are largely descriptive frame-
works useful for analyzing language instances, or output (2014:325).
Fawcett (2014) thus proposes a generative model of intonation and punctu-
ation. He sets out several concepts and realization rules related to ‘intonation
components’ and ‘punctuation components’ needed for a comprehensive
model of grammar that can be used for generating English text. We do not
take up this argument here, but note that Fawcett (2014) provides a careful
and detailed proposal for ‘construct[ing] a generative systemic functional
grammar of intonation and punctuation for English’ (Fawcett 2014:396).
An instance of language represents choices that have been made from

various systems (amongst those from the other systems of language).
Between system and instance are registers or text types. Different registers
display different patterns of choices within the three systems of tonality,
tonicity, and tone. These patterns of choices identify an instance as
belonging to a specific register and play a role in forming the linguistic basis
for comparing and contrasting different registers and text instances. While
all options in the linguistic system are available to a speaker on any one
occasion, there are constraints on what is or is not ‘at risk’ of being selected
when one speaks, and these at risk choices define the various registers.
There are many studies scattered throughout the literature indicating the
role of intonation in identifying different registers – some of these from an SFL
perspective and others from within other theoretical paradigms (e.g. Bowcher
1998; Crystal and Davy 1969; Johns-Lewis 1986; Kuiper 1996; Smith 2008;
Tench 1988, 1996, 2014). Along with choices from within the intonation
systems, a spoken register is characterized by other systematic and identifi-
able choices in sound, rhythm, and sound quality. Tench uses the term
‘prosodic composition’ to refer to ‘the choices, preferences, proportions, and
omissions of specific features of prosodic substance, including voice quality,
pace of utterance, rhythmicality, and loudness, that play an essential role in
the distinctive “sound” of a particular genre’ (Tench 2014:273). Speakers of a
language can identify a range of registers by their individual prosodic compos-
ition and do so quite accurately. That is, they know when they hear a sports
commentary, a news report, or an argument whether or not they hear the
details of the wording of such spoken registers. Van Leeuwen’s (1992) work on
the rhythmic patterns found in different types of radio broadcast registers
adds an important dimension to understanding the concept of prosodic com-
position in relation to registers of spoken language. An interesting argument
proposed by van Leeuwen (1992:250) is that specific patterns of accent and
juncture are ‘motivated, not by the linguistic system of English, but by the . . .
norms and values of [the] social institutions’ in which the speech occurs.
There is much scope for developing descriptions of various spoken regis-
ters and also for describing the configurations of linguistics features such as
grammatical choices, patterns in cohesive ties, and lexical choices alongside
phonological patterns. So far, only a few studies focus on the relationship
between intonation choices and other linguistic features. These include
Lukin’s (2014) multidimensional study of a televised news report; Bow-
cher’s (2003, 2004) work on radio sports commentary; Bowcher and Zhu’s
(2014) study of native and non-native English speakers reading aloud a
children’s story; and, of particular note, Smith’s (2008) research into the
configuration of a range of features in several register varieties, including
casual conversation, talk during a surgical operation, interviews from a
television programme, and telephone sales.
Other research into intonation within the domain of register studies
includes Caldwell’s (2014) analysis of rap and sung performances and
Banks’ (2014) analysis of the pronunciation of the past verb form (-ed) in
Intonation 181
classical choral music, such as Handel’s Messiah. These studies broaden the
scope of the study of intonation beyond spoken language registers.
Thus, while some inroads into identifying the phonological character of
different registers have been made, further research using the SFL model of
intonation to compare and contrast choices across the various intonation
systems in different situations would help to establish a more comprehen-
sive description of different registers in language by adding a phonological
dimension to register variation and identification.
7.3.4 The Paradigmatic Axis and Syntagmatic Structures

The paradigmatic axis is concerned with choice: What are the options avail-
able or what are the potential choices that can be made within a given system?
As already noted, the system of intonation has three sub-systems: tonal-
ity, tonicity, and tone. In the process of speaking, selections are made
from these systems. Figure 7.4 in Section 7.3.3 shows that these systems are
simultaneous systems and hence interdependent, in that choices in one
affect each of the other systems (simultaneity of choice is indicated by the
use of curly brackets in a system network; either–or choices by square
bracketing). For instance, the system of tonality has to do with the range
of choices available for chunking information in spoken language, and
within each chunk is a tonic syllable (tonicity). Each tone unit is character-
ized by a specific melodic shape, or tone (tone). The choice of tone type is
activated by the contextual and semantic features of the situation. Selections
across these systems are made as a discourse unfolds. Further, choices in
each of these systems play a role in realizing choices in the system of infor-
mation at the stratum of lexicogrammar and at the semantic level (for the
latter see Section 7.3.5). For instance, the element assigned as tonic (ton-
icity) realizes the choices in the system of information focus in the
information system. The distribution of tone units across a discourse
(tonality) realizes choices in the system of information unit and specif-
ically information distribution. And a tone unit (tone) realizes an
information unit, although the boundaries of each are not necessarily one-
to-one as illustrated in Section 7.3.1.
As for the system of tone, according to Halliday, English has five simple
tones, i.e. tone units with a single tonic, which are referred to by using the
numerals 1 to 5, and two compound tones, i.e. tone units with a major and
minor tonic syllable, which are referred to by using the numerals 13 (one-
three) and 53 (five-three). Each single tone is realized with a single pitch
movement: fall, rise, level rise, falling-rising, or rising-falling, and the com-
pound tones as a major and a minor pitch movement: a fall followed by a
slight rise or a rise-fall followed by a slight rise. The pitch movements in the
tonic segment(s) define the pitch contour of the tone unit, and are called the
primary tones. Thus, the paradigmatic choices available for the utterance
But I thought everyone was required to contribute something can be ascertained by
asking: How many tonic prominences does the context activate? [Tonicity/
Tonality]; which syllables are assigned tonic prominence? [Tonicity]; and

which tone groups are activated by the context? [Tone]
In answer to the first two questions, we can posit that every syllable could
potentially be assigned tonic prominence, but the syllables that are actually
assigned prominence depend on the contextual environment and certain
English pronunciation conventions to do with word accent; for instance, it
is highly unlikely that bute in contribute would be given tonic prominence
unless, perhaps, someone was trying to point out a pronunciation error of
someone else. So, in a context where someone previously thought that
everyone was required to contribute something to raising funds for an
event but then finds out that only certain people (including themselves)
are required to do so, the utterance would likely be spoken in the following
way:
//5 ᴧ but /I thought /everyone was re/quired to con/tribute /something//
The focus in this context must necessarily be on ‘everyone’ (the tonic

syllable being ‘ev’ in ‘everyone’), and the element of surprise (or perhaps
irritation) is expressed through the use of Tone 5 as shown.
On the paradigmatic axis is the relationship of ‘delicacy’. Delicacy is the
principle of moving from general to more specific and applies to any system
network where there is more than one subsystem. The degree of delicacy is
represented by how many subsystems extend to the right of a system
network. In the system network of foot composition (see Figure 7.3),
for instance, there is little distance between the entry point and the final
choice of [simple]. That is, there are few degrees of delicacy involved. In
other systems there may be many more degrees of delicacy, such as in the
system of tonic composition (see Figure 7.5).
In the system of intonation, delicacy relates largely to the choice of
tone group and the kinds of meaningful distinctions that are made through
height of melodic shape and length of utterance over which the tone group
extends, as well as other features such as nasalization or affectations in
pitch (i.e. those not related to tone group but to overall voice quality).
Within SFL theory, the tone group features that are systematized are the
shape and the range in the height/depth of the pitch contour in the tonic
and the pretonic segments of the tone group.
Whereas the paradigmatic axis is a vertical relation, the syntagmatic axis
is a horizontal relation and has to do with the sequencing of structures
derived from the paradigmatic choices. Take, for instance, the system of
foot composition. This system is located at the rank of foot where there
is a salient syllable which functions as ‘Ictus’, and weak syllable(s) (if there
is/are any) as ‘Remiss’ (Halliday 1967:12). A simple foot is realized by an
Ictus syllable. In a compound foot, the Ictus (whether silent or spoken) is
followed by Remiss. Thus, the structure is:
Ictus (^ Remiss)
Intonation 183
The terms Ictus and Remiss are functional terms, with the Ictus syllables
playing a role in the rhythm of the discourse; these syllables tend to occur
at fairly equal intervals of time in continuous spoken English. This affects
the syllables which may occur within the feet from one Ictus syllable to
another; that is, in order to maintain tempo, the Remiss or non-salient
syllables (when present) tend to be ‘squashed’ through contraction or
weakening of vowels, particularly when there are multiple syllables.
With regard to the tone unit, paradigmatically there are several choices
of single and compound tones. The tone groups, as carrying the melodic
shape of the language, may be characterized by primary and ‘accompani-
ment’ pitch movements (Halliday and Greaves 2008). In this sense, the
‘functioning elements’ in the tone group are the Tonic and the Pretonic,
‘which supply the framework within which the speaker’s variations in
pitch and loudness are perceived and interpreted by the listener’ (Halliday
and Greaves 2008:42). The major pitch movement of the tone unit is
initiated on an Ictus syllable. This constitutes the tonic syllable, and the foot
in which the tonic syllable is located is called the tonic foot. The tonic foot
may be followed by one or more feet, which continue the pitch movement
initiated in the tonic syllable. For a Pretonic segment to be present, there
must be at least one complete foot prior to the tonic foot and not connected
with the previous tone group. The Pretonic has its own set of pitch contour
patterns, but the pitch movement in the Tonic is defining as far as tone
group choice goes. In other words, the pitch movements in the Pretonic are
determined by those in the Tonic, and each Tonic pitch movement has its
own set of Pretonic pitch movements. Thus, the tone unit structure is
described as having an obligatory Tonic which is optionally preceded by a
Pretonic. In terms of syntagmatic structure, when a Pretonic is present, the
tone unit is realized as Pretonic ^ Tonic. Figure 7.5 displays the tone group
choices and the choices of [with pretonic] or [without pretonic].
Figure 7.5 indicates that a tone unit is realized (&) with a tonic element.
The choices in the system of tonic composition are simple or compound.
In the system of pretonic are the choices [with pretonic] or [without
pretonic]. Moving to the right of the figure we find more delicate choices
available for the tones of English. Here we can see the various secondary
tones that are possible – the indirect secondary tones are choices deriving
from the [with pretonic] system, and the direct secondary tones deriving
from the simple and compound tone systems. The system network also
indicates that there are choices of both indirect and direct secondary tones
for certain tone groups. For example, a narrow Tone 1 [1-] may be spoken on
an even pretonic [.1]. These variations in secondary tones are described in
Section 7.4.2 of this chapter.
We can illustrate the syntagmatic structure of tone units and the paradig-
matic choices at a primary level of delicacy using a clause complex from Halli-
day (1970:120): ‘Not always was the kangaroo as now we do behold him but a
different animal with four short legs.’ This is analyzed in the following way:
without pretonic
pretonic
with pretonic
+ Pretonic:
Pretonic: foot1-n
Pretonic ^ Tonic
.1 even
-1 bouncing
1+ wide …1 listing
tone 1 fall 1. medium [tone 1]
1- narrow
.2 high
tone 2 (high rise/- 2. straight [tone 2] -2 low
high fall-high rise)
SIMPLE
simple PRIMARY 2- broken
TONE tone 3 low rise .3 mid
[tone 3] -3 low
tone group
+Tonic; tone 4 (rise-) fall-rise 4. high
Tonic: foot1-n 4- low
TONIC tone 5 (fall-) rise-fall 5. high

COMPOSITION 5- low
tone 13 fall plus low rise
COMPOUND
compound
PRIMARY
+Tonic 2; TONE tone 53 (fall-) rise-fall
Tonic 2: foot1-n plus low rise
Tonic^Tonic 2
Figure 7.5 Choices in the TONE U NIT system showing the more delicate choices available in the pretonic and system of TONE [note: in this
diagram Tonic 2 refers to a Minor Tonic] (from Halliday and Matthiessen 2014:18)
Intonation 185
Table 7.2 Distribution of tone, tonic, and pretonic elements in the excerpt.
(NB: MT = minor tonic)
Tone No. of feet in

choice Pretonic Tonic the tone group
13 // ᴧ not /always was the /kangaroo as 5

/now we do be /hold
him but a // MT
1 //different /animal with // 2
1 //four / short /legs // 3
// 13 ᴧ not /always was the /Kangaroo as /now we do be /hold him but

a //1 different /animal with //1 four /short /legs //
The clause complex is spoken on three tone units. The third tone group is
realized with both a pretonic (/four /short) and tonic (/ legs//), therefore the
choice is ‘+pretonic’. The first and the second tone units are both realized
with only tonic segments, i.e. ‘-pretonic’, but with differences between the
two. In the first tone unit, the tonic segment is preceded by an incomplete
foot (// ˄ not /) which does not constitute a pretonic segment. In the first tone
group there are five feet, in the second there are two, and in the third there
are three. Each foot begins with an Ictus syllable, or salient syllable, but in
the first foot in the first tone unit there is a silent Ictus indicated by the caret
symbol. Generally, monosyllabic lexical words and the accented syllables of
polysyllabic words tend to take salience in a tone unit while the monosyllabic
grammatical or function words tend to be non-salient. We can see that the
example reflects this pattern of salience. The rhythm (or beat) is carried by
these salient or Ictus syllables. The sequence of Tone Unit Structures in this
excerpt is [Tonic^Minor Tonic] ^ [Tonic] ^ [Pretonic ^ Tonic], and at the
grammatical level this would mean that there are three quanta of infor-
mation and thus three information foci, with the first quantum of infor-
mation comprising a major and minor information focus. The distribution of
tone, tonic, and pretonic segments for the excerpt is displayed in Table 7.2.
Each tone unit, whether it be of one or more feet, has a distinct pitch
contour and contains a point of prominence known as the tonic syllable.
Tonic prominence is always assigned to a salient syllable in a foot within
the tone group in response to the relevant activating feature(s) of the
context of situation and the co-text. This assignment of tonic prominence
realizes the functional element of New in an information unit, which is a
culmination point or pulse. This relates to the textual function of the
information unit; its role in organizing discourse in terms of the status of
information – information is presented as ‘Given’ or it is presented as
‘New’. Ascertaining where the New element begins is somewhat indeter-
minate because New is realized by sound not written words, and New is a
culmination point of a range of prosodic features. Furthermore, because the
boundary of the New is indeterminate, ascertaining what might be con-

sidered Given – at least in terms of information prior to the culminating
point of the New – can also be indeterminate.
The definitions of Given and New are typically presented as ‘information
that is presented by the speaker as recoverable (Given) or not recoverable
(New) to the listener’ (Halliday and Matthiessen 2014:118). However, there is
more to these functional elements of the information unit. The determin-
ation of what falls within the New and what within the Given involves both
phonological and semantic considerations. Drawing on ideas from Prince
(1981), O’Grady (2014) discusses the issue of ‘what counts as New’ and
suggests the possibility of two types of ‘New’: New1 is where the tonic is
assigned to items that are freshly introduced in discourse, and New2 is where
the tonic is assigned to items to fulfil a speaker’s ‘communicative goals’ and
‘to direct’ the hearers’ attention to these goals ‘irrespective of whether the
particular lexical items are very much in the air’ (O’Grady 2014:49).
Although Given typically precedes New, there are times when Given may
follow the New. Thus, in terms of syntagmatic structure, the structure of
Given and New in the information unit is (Given ^) New (^ Given), where
parentheses indicate optionality.
7.3.5 The Metafunctional Organization

In SFL ‘function’ is considered to be ‘a fundamental property of language’
(Halliday 1985b:17). That is, the functions of language are not just about the
‘uses’ to which language is put in everyday life, but are ‘the very foundation
to the organization of language itself’ (Halliday 1985b:17), and it is for this
reason that the functions are labelled ‘metafunctions’ (Halliday and Mat-
thiessen 2014:31). Adult language is modelled as multifunctional – as
construing human experience and enacting social relations. These are the
ideational and interpersonal functions respectively. Along with these is the
textual function which is the ‘enabling’ function of language. This function
‘relates to the construction of text’ in that ‘construing experience and
enacting interpersonal relations . . . depend on being able to build up
sequences of discourse, organizing the discursive flow, and creating cohe-
sion and continuity as it moves along’ (Halliday and Matthiessen
2014:30–1). The functions of language are the interface for the socio-
semiotic domains of experience, and together construe the features of
context of situation: Field (the nature of the activity), Tenor (the nature of
the participants involved in the activity), and Mode (the nature of the
channel of communication).
The systems of tone, tonality, and tonicity play a role in the con-
strual of certain kinds of meanings related with different functions of
language. For instance, with regard to the interpersonal function, tone
choice can accord with or alter the speech function typically associated
with certain grammatical structures, such as when a rising tone (Tone 2) is
Intonation 187
Table 7.3 Choices in the system of key and their typical meanings (see
Halliday and Matthiessen 2014:169)
Grammatical Mood Tone Choice Typical Meaning
Declarative Tone 1 unmarked certainty/neutral

Tone 2 sense of protest or
indignation
Tone 3 tentativeness
Tone 4 sense of reservation
Tone 5 insistence
WH-interrogative Tone 1 unmarked neutral
Tone 2 tentativeness
Tone 2 (tonic on WH-element) echo question
Yes/no interrogative Tone 2 unmarked neutral
Tone 1 peremptory question
Imperative clause Tone 1 unmarked in positive command
Tone 3 unmarked in negative invitation
Tone 13 marked polarity (tonic pleading request
is on do/don’t)
Tone 4 plea
used on a declarative clause; instead of giving information, this choice

serves to enact the speech function of demanding information. The choices
of tone when interpreted in relation to Mood choice may be considered
marked or unmarked, and together these represent a system known as key.
These are displayed in Table 7.3.
The choice of tone, and the more delicate choices within the system of
tone, such as degrees in the height of the fall or rise in pitch, play a role in
construing other kinds of interpersonal meanings such as intensity of
feeling, degrees of involvement, or sense of commitment to what is being
said (see Halliday 1970; Halliday and Greaves 2008).
With regard to the ideational function, and specifically the logical func-
tion, choices in the system of tone play a role in construing the ‘connec-
tions’ between bits of information. For example, a relationship of
coordination is construed through the sequence of Tone 3 (low rise)
followed by Tone 1 (fall), and of subordination by Tone 4 (fall-rise rounded)
followed by Tone 1 (fall). When these tone sequence choices parallel the
grammatically construed choices, such as where a relation of coordination
is grammaticalized by the use of the conjunction and, and that of subordin-
ation by the subordinator whereas, the phonological realization is said to
be unmarked. However, the two systems do not always parallel each
other. A Tone 3 might be used on a grammatically coded subordinate
clause, and a Tone 4 on a coordinating clause: such choices are considered
marked (see Halliday and Greaves 2008:129–35; Halliday and Matthiessen
2014:553–4).
The system of tonality also comes into play in relation to the logical
function. For example, a projecting clause complex may not be intonation-
ally construed through the use of two tone groups but through presenting
Figure 7.6 Unmarked Theme/Rheme and Given/New functional elements in a clause
the clause complex as one quantum of information through the use of one
tone group only, as in:
//1 ᴧ he /said there was /nobody /left//
O’Grady (2017:155) suggests that although this choice is considered ‘the

unmarked realisation’, he is careful to add that in fact there has been ‘no
extensive corpus investigation of this claim’ and thus the claim remains
‘unverified’ (O’Grady 2017:155). This suggests an interesting area for future
investigation. In contrast with these kinds of clause complexes, as noted
already, one clause might be spoken on several tone groups, as in the
example of reading aloud in Section 7.3.3, thus organizing the talk into
several quanta of information.
Section 7.3.4 included a brief discussion on the information unit in
relation to the textual metafunction; its role in organizing discourse in
terms of the status of information. There is a complementary relation
between the information unit and other textual systems such as cohesion
and the clause structures of Theme and Rheme. For instance, often, and
indeed in regards to the unmarked case, New falls within the Rheme of the
clause and Given within the Theme. Each of these functional elements
construes a culmination of information, with Theme being considered
speaker-oriented (it is the point of departure of the message), and New as
listener-oriented (it is what the speaker wants the listener to attend to in
the situation). The textual metafunction, however, enables a variety of ways
in which to package experience. For instance, Matthiessen (1995:607)
observes that:
The assignment of New illustrates how the textual metafunction may work
independently of the hierarchic organisation generated by the experiential
one. In particular, the element New of the information unit is not restricted
to focus (1) on an element of structure selected from within a single clause;
nor (2) on an element of clause structure.
Figure 7.6 shows the unmarked mapping of Theme/Rheme and Given/

New functions in a clause, and Figure 7.7, which uses the sentence from the
children’s story ‘How Love and Peace Came to the Woods’ presented in
Intonation 189
//1 Mr. / Fox//1 was /strolling /through the / woods//

Mr Fox was strolling through the woods
Theme Rheme
New New
// 3 when he / noticed// 1 a /plump / hen// //1 sitting //1 on the /branch of a / tree//
when he noticed a plump hen sitting on the branch of a tree
Theme Rheme
Given New New New New
Figure 7.7 Analysis of marked information unit distribution
Section 7.3.3 shows several instances of marked New. Bowcher (2003, 2004)
describes the relationship between Theme and New choices in excerpts
from radio sports commentaries, and specifically in play-by-play talk of a
Rugby League game. She finds that in this register there is a complementary
relation between information carried by the Theme and that carried by the
New, with informational peaks falling predominantly on players either as
Participants or as Circumstances (destinations of the ball), and that marked
informational prominence is not common in this type of talk. Zhu (2014)
focuses on the relationship between Theme and Information choices in a
BBC news reading. Lukin (2014) also analyzes the relationship between New
and Theme in her study of a news report of the ‘Coalition’s’ war with Iraq in
2003, finding that ‘in the choice of “person” and “place”, the system of IF
[information focus] is reinforcing patterns established via the system of
theme’ (Lukin 2014:65).
Choices from the systems of tonality and tonicity play a role in the
degree to which the boundaries of information units coincide with those of
clauses and what is assigned focus by the speaker. Various researchers
highlight registerial patterns in this regard. For instance, Lukin (2014:63)
notes that there is a ‘higher ratio of tone units to the grammatical unit of
clause’ in broadcast news (see van Leeuwen 1992; Smith 2008). However,
such a claim is not particularly revealing, considering that other registers
also exhibit a high number of tone units per clause, such as reading aloud
children’s stories (e.g. Bowcher and Zhu 2014; Halliday 1970), and indeed,
Lukin effectively acknowledges the too-general nature of her claim when she
asks, ‘Is there a metafunctionally significant pattern in what is selected for
focus by the location of intonational focus?’ This kind of question is of more
value in identifying registerial differences in tonality, tonicity, and
information distribution, and Lukin’s analysis indicates that in her
news broadcast data choices construe largely textual and interpersonal
meanings. Other register-focused findings include Smith’s (2008) work
which includes a description of the way that intonation choices (amongst
others) play a role in shifting the focus of talk taking place in a surgery from
experiential to interpersonal meanings, and Bowcher and Zhu’s (2014:17)
study points out that even though a children’s story ‘has a low lexical density
and is grammatically simple, when it is read aloud, it is assigned a high
“informational” density through the resources of the intonation system’.
Of relevance to intonation choices in relation to the textual function is the
higher unit of paraphones, or phonological ‘paragraphs’. Tench (1996:24)
argues that ‘phonological paragraphs’ typically begin with a ‘high pitch on
the onset syllable of the initial intonation unit’ and that the pitch gradually
falls until the final tone unit, wherein ‘the depth of fall in the final unit is
the lowest in the whole paragraph’. He also observes that the tempo tends to
slow down in the final intonation unit and that ‘there is a longer pause than
is normally allowed between intonation units’ between phonological para-
graphs (Tench 1996:24; see also Tench 2014:272–3). Iwamoto (2014) picks up
the idea of phonological paragraphs, calling them ‘paraphones’ after Halli-
day (1961:253 note 30) and proposes that ‘paraphoning’ is a textual process
across all strata of the language system and is semogenic in nature. He
argues that a paraphone is a semantic unit, whose boundaries are realized
by specific phonetic cues, such as pitch levels, and he hypothesizes that
paraphoning differs across registers in that ‘speaker[s] select one way [of
paraphoning] over others according to the context of situation to create
distinctions in meaning’ (Iwamoto 2014:143). Iwamoto’s work leaves open
an enticing area for further research into register variation.
The textual metafunction plays a key role in the construal of the contextual
parameter of mode. We noted in Section 2 Tench’s (1996:151) comment on the
importance of intonation in reading and writing, and several SFL studies have
focused on this very point. These include Davies’ (1989, 1992, 1994a, 1994b,
2014) extensive work on the relation between cohesion, information structure
in written and spoken text, punctuation, and intonation systems for effective
understanding and reading aloud of written text. There is also Cummings’
(2000, 2001, 2014) research into the interpretation of the intonation patterns
relevant to written text, and Bowcher and Zhu’s (2014) study of native and
non-native English speakers reading a children’s story. The relationship
between spoken and written language features is also of critical importance
in Fawcett’s (2014) generative model of English intonation and punctuation.
The next section of this chapter describes in more detail the tones in
English as modelled in Systemic Functional Linguistics. Our aim is to
highlight the basic shapes of the tones of English, as it is the tone group
that is the core around which other choices in the system of intonation
and the system of information operate.
7.4 The Tones of English
While this section sets out a description of the tones of English, some experi-
ence in listening to the different tones should be gained prior to conducting
one’s own analysis. A good place to gain experience in hearing the tones is
Halliday and Greaves (2008). Additionally, Greaves (2014) is an interactive
Intonation 191
chapter on the SFL system of intonation and includes practice in rhythm and
hearing tones, and in analyzing spoken utterances using Praat.6
According to the SFL model, there are seven primary tones in English, within
which five are simple tones and two are compound tones. We describe each of
the primary tones in turn before turning to the secondary tones.
7.4.1 The Primary Tones

7.4.1.1 Tone 1
Tone 1 is realized with a falling pitch contour in the tonic syllable. This tone
begins at a mid or mid high pitch level and continues downward till the end
of the tone group.
Figure 7.8 Tone 1 – the tonic and pretonic pitch contours
If there is a pretonic segment, the corresponding pitch contour in the pretonic

is level at mid or mid high pitch, as is exemplified in the following examples.
I’ve finished my work. // 1 ᴧ I’ve /finished my /work //
Why didn’t you finish your work? // 1 why didn’t you /finish your /work //
Finish your work! // 1 finish your /work//
7.4.1.2 Tone 2
Tone 2 is a sharp rising pitch contour from low or mid low, and it covers a
wide pitch range. The pretonic pitch contour is either high level (as in
Figure 7.9b), or steps down from high or mid high to the point/pitch level
from where the tonic pitch movement begins (as in Figure 7.9a).
Figure 7.9a Tone 2 – the sharp rising tonic with a step down pretonic
Figure 7.9b Tone 2 – the sharp rising tonic with a high level pretonic
6
See www.equinoxpub.com/systemic-phonology-files for the accompanying sound files to Greaves (2014).
Will you please finish your // 2 ᴧ will you / please /finish your / work //(step down
work? pretonic)
Have you finished your // 2 ᴧ have you / finished your / work //(high level
work? pretonic)
7.4.1.3 Tone 3
Tone 3 is a level rise from about low or mid low to mid pitch. If there is a
pretonic segment, it has a level contour.
perhaps. // 3 ᴧ per/haps // (no pretonic)

It (the exam) won’t be so hard. // 3 ᴧ it / won’t be so / hard //
7.4.1.4 Tone 4
Tone 4 is a fall-rise pitch contour with more force on the falling movement.
This tone exhibits a rise-fall hook onset before the main falling-rising pitch
movement; the pitch first rises from mid to about mid high before execut-
ing the fall-rise movement, which is a key feature of the tone. The fall
covers a wide pitch range, and the rise is almost to the same level as the
beginning of the fall.
The pretonic contour steps down from high pitch to around mid pitch
level, i.e. to the level from where the rise-fall hook onset of Tone 4 begins.
He finished his work, but . . . // 4 ᴧ he / finished his / work but // . . . //
If you don’t finish your work in time // 4 ᴧ if you / don’t / finish your / work in / time //. . .
7.4.1.5 Tone 5
Tone 5 exhibits a pitch movement which is in the opposite direction to Tone
4. This is a rise-fall pitch contour with more force on the rising movement.
This tone also exhibits a hook onset which has a fall-rise movement. The
pitch in the tonic syllable first falls from mid-high to mid and then rises to
cover a wide range before executing the fall.
Intonation 193
The pretonic steps up from about mid level to mid high pitch level, i.e.
almost to the level from where the fall-rise hook onset for Tone 5 begins.
I’d never seen such a hullabaloo // 5 ᴧ I’d /never /seen such a /hullaba/loo//
7.4.1.6 Compound Tones: Tone 13 and Tone 53

In addition to the five simple tones there are two compound tone units, i.e.
tone units with two tonic syllables – one major and the other minor. These
are considered fusions of two tones rather than two successive tones in that
the pretonic, if present, is located before the first tone only. That is, there is
no intervening pretonic in the Tone 1!3 or the Tone 5!3 sequence: the
pretonic would be before the Tone 1 or Tone 5 only. There is some debate as
to whether compound tones should rather be considered as sequences of
two tone groups (see O’Grady 2017:152 for discussion of this issue; also see
Tench 1990). Of particular interest in this discussion is O’Grady’s (2017:152)
comment that ‘the criterion of information structure posits that a tone
group equates to an information unit’ and that ‘this is compromised by
positing two foci, albeit of different status, within a compound tone group’.
However, the issue of what defines a compound tone group is akin to that of
compound nouns, viz. where to draw the line as to what constitutes two
words or a compound word (e.g. breast feeding, breastfeeding; ready-made,
readymade). In such cases, there are two pieces of information drawn
together into the compound word. An information unit realized by a
compound tone group is essentially doing the same thing. For instance,
Halliday (1970:88) explains that the minor element (the rise) in a Tone 13 is
typically either
(1) an adjunct or dependent clause, or sometimes a co-ordinate clause (i.e.

subsidiary new information); (2) a ‘partial utterance’, such as a vocative,
speaker’s comment or other ancillary matter; (3) the displaced subject in a
substitution clause; or (4) some other element which is not new (e.g.
repeating part of a question) but to which the speaker wants to give some
prominence.
Thus, it seems reasonable to have an information unit realized by a com-

pound tone group with major and minor tonics. Furthermore, the inton-
ation system and the grammatical system work side by side. There is
an interplay among the choices made, with each pointing to the different
functions associated with spoken and written language. There is thus
an apparent need for more empirical investigations into the issues
surrounding compound tone groups, which is beyond the scope of this

chapter. Our description of the compound tones here follows that of Halli-
day (1970) and Halliday and Greaves (2008).
7.4.1.7 Tone 13 (Tone One Three)

Tone 13 has a falling contour from either mid or mid high followed by a low
rising contour of Tone 3 to about mid low level.
The pretonic is level at mid or mid-high, and it is defined by the Tone

1 part of this compound tone.
//13 Give my re/gards to your /parents /next time you /see them//
(example from Halliday 1970:88)
7.4.1.8 Tone 53 (Tone Five Three)

This tone exhibits a combination of the rise-fall contour of Tone 5 followed
by a low rising contour of Tone 3.
The pretonic pitch steps up from about mid to mid high.
//53 ᴧ I’d /rather /like one if you /feel you can /spare it//
(example from Halliday 1970:93)
7.4.2 The Secondary Tones

As explained in the previous section, the primary tone system indicates a
choice among seven tones: five simple and two compound tones. Section
7.4.1 also described the neutral pitch contour for each primary tone, in the
tonic segment, and the typical contour in the pretonic segment. However,
there are possible variations in the pitch contours of both the tonic segment
and the pretonic segment for each tone. While Table 7.3 outlined the
typical meanings of the primary tones associated with grammatical mood
(the system of key), the variations that are described in this section
Intonation 195
contribute to more subtle meanings within each general meaning of the

tones and realize a system of secondary tones, with a distinction between
direct secondary tones – those where specific variations occur in the tonic
segment – and indirect secondary tones – those in the pretonic segment
which follow a specific melodic shape. These more delicate options were
mentioned in Section 7.3.4 with the system of primary and secondary tones
shown in Figure 7.5. In this section, we describe the possible secondary
tones for each primary tone.
7.4.2.1 Direct Secondary Tones of Tone 1

Tone 1 has three systemic pitch choices in the tonic segment, or three direct
secondary tones identified conventionally with a period, a plus sign, and a
minus sign: [1.], [1+], and [1-]. Tone [1.] is a fall from about mid or mid high
level. Tone [1+] is a fall from high to low, and Tone [1-] is a fall from mid low
to low.
Figure 7.15 Direct secondary tones of Tone 1
// 1. why don’t you / clean your / room//

// 1+ why don’t you / clean your / room//
// 1- why don’t you / clean your / room//
7.4.2.2 Indirect Secondary Tones of Tone 1

There are three distinct pitch contours possible in the pretonic segment of Tone
1: [.1] ‘even’, [-1] ‘bouncing’ or ‘uneven’, and [. . .1] ‘listing’. And these express
indirect secondary tones of Tone 1 (the fall is only shown once in Figure 7.16).
Figure 7.16 Indirect secondary tones of Tone 1
// .1. why don’t you / get up and / clean your / room//

// -1. why don’t you / get up and / clean your / room//
// . . .1. one / two / three / four //
The neutral type [.1] is also referred to as the ‘even’ type, as the pitch
contour is more or less level/even: at mid pitch level for [1.], steps up from mid
low to high for tone [1+], and steps down from mid high to mid low for [1-].
Figure 7.17: Indirect (neutral) secondary tone [.1] ([.1.], [.1+], and [.1-] respectively)
The [-1] ‘uneven’ pretonic has a dipping or bouncing contour from around
mid to a fairly high pitch in each foot. The [. . .1] pretonic has a level rising
contour in each foot.
7.4.2.3 Tone 1: Tone [.1.] vs. Tone [-1.] (Attitudinal)

The use of [-1.] instead of [.1.] brings in a sharp attitudinal ‘intensive’ or
argumentative meaning, as in
// .1. why don’t you / get up and / clean your / room//

// -1. why don’t you / get up and / clean your / room//
In the second, the swinging movement in each foot adds to the intensity of
the meaning of the sentence. This pretonic occurs more naturally with a
‘strong’ [1+] tonic, making it more forceful.
7.4.2.4 Tone 1: Tone [.1.] vs. Tone [. . .1.] (Semantic)

The level rising contour in each foot of [. . .1] is used to enumerate the items
in a list occurring before a final tonic syllable, or focus of information, as in
// . . .1. one / two / three / four //
where each pretonic foot ends with a slight rise.

The contrast in meaning between a [.1] and a [. . .1] pretonic can be seen in
examples such as
// .1 red / white and / blue / jackets // (‘jackets with a mixture of red,

white and blue colours’)
// . . .1 red / white and / blue / jackets // (‘red jackets, white jackets and
blue jackets’)

The direct secondary system of Tone 2 gives two choices: [2.] a sharp rise
from low pitch or [2] a sharp fall rise, falling from high to cover a wide pitch
range and then rising.
Figure 7.18 Direct secondary tones of Tone 2
Intonation 197
// 2. ᴧ did you /finish your as /signment //

// 2 ᴧ did you /finish your as /signment //

The indirect secondary system also has two choices: either a high level [.2]
or a low level [-2] pitch, as shown in Figure 7.19.
Figure 7.19 Indirect secondary tones of Tone 2
These indirect secondary tones [.2] and [-2] can combine with either of the
direct secondary tones, [2.] or [2], but the combination of [-2] with [2]
appears to be rare.
7.4.2.7 Tone .2. vs. Tone -2. (Attitudinal)

// .2. ᴧ did you /finish your as /signment? //: ‘neutral’
// -2. ᴧ did you /finish your as /signment? //: ‘surprised’
The high level pretonic with a jump down in pitch to a sharp rise in
the tonic [.2.] is the most unmarked way of realizing the yes-no interroga-
tive. When a low level pretonic combines with a neutral tonic [-2.],
the yes-no interrogative acquires an additional meaning of being more
‘involved’.
7.4.2.8 Tone .2. vs. Tone .2 (Semantic)

// .2. ᴧ did you /finish your as /signment? //: ‘neutral’
// .2 ᴧ did you /finish your as /signment? //: ‘specific focus on
assignment’
In Tone [2.], there is a straightforward rise in the tonic syllable which is

often on the last lexical word, indicating that the query is with reference to
the entire information unit. This is the neutral tone for a Yes-No interroga-
tive. On the other hand, the marked fall-rise variant, Tone [2], draws the
attention of the listener to a specific point of query in the information unit.

Tone 3 has variations only in the pretonic segment, i.e. the indirect
type. There are two secondary tones: one is level at mid pitch, [.3], and
the other is level at low pitch, [-3]. So, the low level rising tone in the tonic
segment can occur with a level pretonic contour either at mid or
low pitch.
Figure 7.20 Indirect secondary tones of Tone 3 ([.3] and [-3] respectively)
7.4.2.10 Tone .3 and Tone -3 (Attitudinal)

While the mid level pretonic [.3] conveys the neutral meaning of a level
rising tone, in this case, ‘reassuring’, the low pitched pretonic [-3] gives a
marked meaning of being ‘noncommittal’ or ‘indifferent’.
// .3 ᴧ it (the exam) / won’t be so /hard // ‘reassuring’

// -3 ᴧ it (the exam) / won’t be so /hard // ‘unconcerned’
7.4.2.11 Direct and Indirect Secondary Tones of Tone 4

There are two direct secondary tones for Tone 4: [4.] and [4]; the indirect
secondary tones are determined by the pitch contour of the direct
secondary tones.
Figure 7.21 Direct and indirect secondary tones of Tone 4 ([4.] and [4] respectively)
//4. ᴧ it’s a / bit /dangerous// (‘I can’t help being worried; . . .’)
(from Halliday 1970:110)
Tone [4.] is the neutral one with a fall-rise pitch contour from mid high, and it
covers a wide pitch range. The pretonic for this tone is a step down contour
from high to about mid pitch level. Tone [4] is the marked variant with the
fall-rise pitched lower. The preceding pretonic contour exhibits a fall-rise
pitch movement in each foot, which seems to be imitating the fall-rise
movement of the tonic segment. This series of low pitched fall-rise move-
ments in the pretonic adds to the intensity of the tone.
//4 not unless he’s /willing to a/pologise // (‘I might see him if he does’)
(Halliday 1970:111)
7.4.2.12 Tone 4. and Tone 4 (Attitudinal)

// 4 ᴧ he’s / finished his as /signment //
// 4 ᴧ he’s / finished his as /signment //
The Tone [4] variant makes the meaning of fall-rise pitch more intense, and
is accompanied by a distinct voice quality. For instance, if, in a
Intonation 199
particular context (such as indicated in the examples), Tone 4. is used to

indicate ‘reservation’, the use of Tone [4] instead makes it a more personal
or intense reservation.

Tone 5 has two direct secondary tones: [5.] and [5]. Similar to Tone 4, the
pretonic contours are fixed for each direct secondary tone.
Figure 7.22 Direct and indirect secondary tones of Tone 5 ([5.] and [5] respectively)
Tone [5.] is the neutral tone having a rise from about mid to mid high or high
and then a fall to a lower pitch level. The pitch in the pretonic segment exhibits
a step up from about mid low to about mid high. Tone [5] is a low pitched rise-
fall contour at mid low/mid pitch level. This is preceded by a pretonic, with
each foot exhibiting a step down rise-fall movement.
//5. ᴧ I /can’t be/lieve they would /ever have /thought that a/bout her//
//5 ᴧ I can’t be/lieve they could /be so /stupid//
7.4.2.14 Tone 5. and Tone 5 (Attitudinal)

// 5. ᴧ the / soup was / very / tasty //
// 5 ᴧ the / soup was / very / tasty //
The meaning of the neutral Tone [5.] is related to the rise-fall prosody, i.e.
‘there was some doubt, but all is fine’. On the other hand, Tone [5], which is
lower pitched and usually accompanied by a breathy voice quality, is used
to indicate ‘awe’, and sometimes, depending on the context, ‘sarcasm’ or
‘disappointment’, as in the examples.
7.4.2.15 Tone 13 and Tone 53: Secondary Tones

As mentioned earlier, Tone 13 and Tone 53 are compound tones, and the
pretonic, if present, occurs only before the first tonic segment. The pitch
contour in the pretonic indicates the direct secondary tones of Tone 1 or
Tone 5 in the compound tones.
7.5 Conclusion
This chapter has provided an overview of intonation within the SFL architec-
ture of language with some mention along the way of specific research that
has been conducted. The second half of the chapter outlined the melodic
shapes and general meanings of the primary and secondary tone groups in
English. Our aim has been to provide a description useful for understanding
the place of intonation in the SFL architecture of language and for conduct-
ing some basic analyses of spoken English, albeit recognizing that practice in
listening to the tone groups would be essential for undertaking such an
analysis. Further, while it is impossible to do justice to all of the areas of
research that have been developed for intonation within the SFL framework
in a chapter of this size, we hope that we have provided sufficient back-
ground to offer readers ideas for possible research directions.
References
Banks, D. 2014. A Note for –ed: Comments on the Treatment of –ed in

Handel’s Messiah. In W. L. Bowcher and B. A. Smith, eds., Systemic
Phonology: Recent Studies in English. Sheffield: Equinox. 221–34.
Bowcher, W. L. 1998. Intonation in Radio Sports Commentating: Towards
an Analysis and Interpretation. In Y. Nagahara, ed., Kobunnokairuinikan-
suru Kijutsutekioyobi Rirontekikenkyuu. Japanese Ministry of Education
Research Grant no. 07451097.
Bowcher, W. L. 2003. Creating Informational Waves: Theme and New
Choices in Play-by-play Radio Sports Commentary. In M. Amano, ed.,
Creation and Practical Use of Language Texts: Proceedings of the Second Inter-
national Conference for the Integrated Text Science. Nagoya: Graduate School
of Letters, Nagoya University. 111–22.
Bowcher, W. L. 2004. Theme and New in Play-by-play Radio Sports Com-
mentating. In D. Banks, ed., Text and Texture: Systemic Functional Viewpoints
on the Nature and Structure of Text. Paris: L’Harmattan. 455–93.
Bowcher, W. L. and B. A. Smith. 2014a. Introduction. In W. L. Bowcher and
B. A. Smith, eds., Systemic Phonology: Recent Studies in English. Sheffield:
Equinox. 1–24.
Bowcher, W. L. and B. A. Smith, eds. 2014b. Systemic Phonology: Recent Studies
in English. Sheffield: Equinox.
Bowcher, W. L. and S. Zhu. 2014. Meaningful Reading: Intonation Choices
by Native and Non-native English Speakers Reading The Giving Tree.
Unpublished manuscript. 1–36.
Caldwell, D. 2014. A Comparative Analysis of the Rap and the Sung Voice:
Perspectives from Systemic Phonology, Social Semiotics and Music Stud-
ies. In W. L. Bowcher and B. A. Smith, eds., Systemic Phonology: Recent
Studies in English. Sheffield: Equinox. 235–63.
Couper-Kuhlen, E. 1986. An Introduction to English Prosody. London: Edward
Arnold.
Crystal, D. and D. Davy. 1969. Investigating English Style. London: Longman.
Cummings, M. 2000. The Inference of Given Information in Written Text.
In E. Ventola, ed., Discourse and Community: Doing Functional Linguistics.
Tübingen: Gunter Narr Verlag. 331–53.
Intonation 201
Cummings, M. 2001. Intuitive and Quantitative Analyses of Given/New in

Texts. In J. de Villiers and R. J. Stainton, eds., Communication in Linguistics,
Volume 1: Papers in Honour of Michael Gregory. Toronto: GREF Publishers.
61–94.
Cummings, M. 2014. The Spoken Interpretation of Written Text. In W. L.
Bowcher and B. A. Smith, eds., Systemic Phonology: Recent Studies in English.
Sheffield: Equinox. 199–217.
Davies, M. 1989. Prosodic and Nonprosodic Cohesion in Speech and
Writing. Word 40(1–2): 255–62.
Davies, M. 1992. Prosodic Cohesion in a Systemic Perspective. In P. Tench,
ed., Studies in Systemic Phonology. London: Pinter. 206–30.
Davies, M. 1994a. Intonation IS Visible in Written English. In S. Čmejrková,
F. Daneš, and E. Havlová, eds., Writing vs Speaking: Language, Text, Discourse,
Communication. Tübingen: Gunter Narr Verlag. 199–204.
Davies, M. 1994b. I’m Sorry I’ll Read that Again: Information Structure in
Writing. In S. Čmejrková and E. Štícha, eds., The Syntax of Sentence and
Text: A Festschrift for Frantisek Daneš. Tübingen: Gunter Narr Verlag.
75–89.
Davies, M. 2014. The Black Hole in Graphology. In W. L. Bowcher and B. A.
Smith, eds., Systemic Phonology: Recent Studies in English. Sheffield: Equinox.
153–98.
Fawcett, R. 2014. The Meanings and Forms of Intonation and Punctuation
in English: The Concepts Required for an Explicit Model. In W. L. Bow-
cher and B. A. Smith, eds., Systemic Phonology: Recent Studies in English.
Firth, J. R. 1948. Sounds and Prosodies. Transactions of the Philological Society
47(1): 127–52.
Firth, J. R. 1968. Linguistic Analysis as a Study of Meaning. In F. R. Palmer,
ed., Selected Papers of J. R. Firth 1952–59. London: Longmans. 12–26.
Greaves, W. S. 2014. Locating the Limerick ‘Wall Street Irene’ and the
Sonnet ‘On His Blindness’ in the Semiotic Space between the Body as
Signal Generator/Receiver and the Body as Social Interactant. In W. L.
Bowcher and B. A. Smith, eds., Systemic Phonology: Recent Studies in English.
Halliday, M. A. K. 1961. Categories of the Theory of Grammar. Word 17(3): 242–92.
Halliday, M. A. K. 1963a. Intonation in English Grammar. Transactions of the
Philological Society 62(1): 143–69.
Halliday, M. A. K. 1963b. The Tones of English. Archivum Linguisticum 15(1):
1–28.
Halliday, M. A. K. 1967. Intonation and Grammar in British English. The Hague:
Mouton de Gruyter.
Halliday, M. A. K. 1970. A Course in Spoken English: Intonation. London: Oxford
University Press.
Halliday, M. A. K. 1985a. Introduction to Functional Grammar. London: Edward
Arnold.
Halliday, M. A. K. 1985b. Part A. In M. A. K. Halliday and R. Hasan, Language,

Context, and Text: Aspects of Language in a Social-semiotic Perspective. Geelong:
Deakin University Press.
Halliday, M. A. K. 1994. Introduction to Functional Grammar. 2nd ed. London:
Edward Arnold.
Halliday, M. A. K. and W. S. Greaves. 2008. Intonation in the Grammar of
English. Sheffield: Equinox.
Grammar. 3rd ed. London: Arnold.
Iwamoto, K. 2014. A Multistratal Approach to Paragraph-like Organisation
in Lectures. In W. L. Bowcher and B. A. Smith, eds., Systemic Phonology:
Recent Studies in English. Sheffield: Equinox. 116–49.
Johns-Lewis, C. 1986. Prosodic Differentiation of Discourse Modes. In
C. Johns-Lewis, ed., Intonation in Discourse. London: Croom Helm. 199–219.
Kuiper, K. 1996. Smooth Talkers: The Linguistic Performance of Auctioneers and
Sportscasters. Mahwah: Lawrence Erlbaum.
Lukin, A. 2014. Creating a Parallel Universe: Mode and the Textual Meta-
function in the Study of One News Story. In W. L. Bowcher and B. A.
Smith, eds., Systemic Phonology. Sheffield: Equinox. 53–90.
Tokyo: International Language Science Publishers.
McGregor, W. B. 1992. Towards a Systemic Account of Gooniyandi Segmen-
tal Phonology. In P. Tench, ed., Studies in Systemic Phonology. London:
Pinter. 19–43.
Ogden, R. 2012. Firthian Prosodic Analysis. Firthian Phonology Archive. Avail-
able online at: https://sites.google.com/site/firthianarchive/fpa. (Last
accessed 27/07/2017.)
O’Grady, G. 2014. An Investigation of How Intonation Helps Signal Infor-
mation Structure. In W. L. Bowcher and B. A. Smith, eds., Systemic
Phonology. Sheffield: Equinox. 27–52.
O’Grady, G. 2017. Intonation and Systemic Functional Linguistics: The Way
Forward. In T. Bartlett and G. O’Grady, eds., The Routledge Handbook of
Prince, E. 1981. Towards a Taxonomy of Given-New Information. In P. Cole,
ed., Radical Pragmatics. New York: Academic Press. 223–56.
Robins, R. H. 1970. Aspects of Prosodic Analysis. In F. R. Palmer, ed., Prosodic
Analysis. London: Oxford University Press. 104–11.
Smith, B. A. 2008. Intonational Systems and Register: A Multidimensional Explor-
ation. PhD Thesis, Macquarie University. Available online at:
www.isfla.org/Systemics/Print/Theses/SmithBradPhD.pdf. (Last accessed
27/07/2017.)
Intonation 203
Storynory, n.d. How Love and Peace Came to the Woods. Available online at:
www.storynory.com/2017/04/22/love-peace-came-woods. (Last accessed
27/07/2017.)
Tench, P. 1988. The Stylistic Potential of Intonation. In N. Coupland, ed.,
Styles of Discourse. London: Croom Helm. 50–84.
Tench, P. 1990. The Roles of Intonation in English Discourse. Frankfurt am Main:
Peter Lang.
Tench, P. 1992a. From Prosodic Analysis to Systemic Phonology. In
P. Tench, ed., Studies in Systemic Phonology. London: Pinter. 1–17.
Tench, P., ed. 1992b. Studies in Systemic Phonology. London: Pinter.
Tench, P. 1996. The Intonation Systems of English. London: Cassell.
Tench, P. 2014. Towards a Systemic Presentation of the Word Phonology of
English. In W. L. Bowcher and B. A. Smith, eds., Systemic Phonology: Recent
Studies in English. Sheffield: Equinox. 267–93.
van Leeuwen, T. 1992. Rhythm and Social Context: Accent and Juncture in
the Speech of Professional Radio Announcers. In P. Tench, ed., Studies in
Systemic Phonology. London: Pinter. 231–62.
Young, D. 1992. English Consonant Clusters: A Systemic Approach. In
P. Tench, ed., Studies in Systemic Phonology. London: Pinter. 44–69.
Zhu, S. 2014. Intonation: Signal of Information Peaks. In W. L. Bowcher and
B. A. Smith, eds., Systemic Phonology: Recent Studies in English. Sheffield:
Equinox. 91–115.
8
Continuing Issues in SFL
Mick O’Donnell
8.1 Introduction
This chapter will discuss various issues that are not fully resolved within
Systemic Functional Linguistics (SFL). Two main issues will be addressed:
• To what degree are the grammatical categories of SFL determined on

notional grounds (mirroring extra-linguistic organization), rather than
on regularities of form?
• Where does genre belong in relation to other components of the model?
Both of these issues are still under debate within the community, often
leading to divergent approaches, and, if we are not aware of the underlying
differences, result in misunderstanding of the arguments others are
making.
This chapter will refer extensively to the four editions of ‘Introduction to
Functional Grammar’, the first two by Halliday (Halliday 1985, 1994), and
the last two revised by Matthiessen (Halliday and Matthiessen 2004, 2014).
To simplify references, I will refer to these as IFG1, IFG2, IFG3, and IFG4.
8.2 How ‘Semantic’ is the Grammar?
In this section, I will explore the criteria used to define grammatical

categories in SFL, in particular, the degree to which grammatical categories
are based on notional grounds rather than in terms of grammatical react-
ance. Prior discussions related to this point can be found in Butler (2003),
My thanks to Tom Bartlett, Margaret Berry, Lise Fontaine, Jim Martin, and Geoff Thompson for comments on this work.
While Butler (2003) is only lightly cited in this work, it has strongly influenced my way of thinking about functional
grammar discussed in Section 8.2.
Downloaded from https://www.cambridge.org/core. Newcastle University, on 27 Aug 2019 at 06:51:59, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316337936.010
Continuing Issues in SFL 205
but see also Hudson (1971); O’Donnell et al. (2008); Fawcett (2009); Tucker
(2014); Gwilliams and Fontaine (2015).
8.2.1 Linguistic Levels as Arbitrary Points of Abstraction

Linguistic descriptions of discourse can range from descriptions of the
physical manifestation of the discourse (the marks placed on the page,
the sounds produced by the speaker) up to descriptions of the cultural,
social, and material context in which the discourse is produced.
The realm of most linguistic work is, however, in the space between these
two more manifested end-points. Between these points, linguists typically
choose to produce a number of levels of analysis, each more abstract than
the last. On the physical manifestation side, linguists move away from the
purely phonetic description of the discourse, focusing instead on a more
abstract phonemic description. More abstractly, words are identified, and
within them, morphemic elements. Even more abstractly, the grouping of
words into larger units is explored. Moving further, some linguists will
explore the way in which these groupings of words relate to what the
speaker/writer is trying to achieve through the production of these words.
In constructing these levels of analysis, the linguist is to a degree free to
determine where in the space between physical manifestation and context
the levels may go. Some choose to traverse the space in small steps, positing
eight or so levels of linguistic analysis. Others may choose to skip across the
divide in large steps. Others still, afraid of deep water, ignore the task of
spanning the divide, and instead focus on describing the physical manifest-
ation, ignoring context- and use-related aspects of language.
The most typical levels of analysis in linguistic models focus on firstly,
the organization of sounds (phonetics or phonemics) and secondly, the
organization of words into larger units (syntax or grammar). Often, a
number of levels above this will be posited, exploring the deep water far
from the physical manifestation (semantics, pragmatics, rhetoric, etc.).
This section is concerned with the grammatical level of description,
looking to the variation in abstraction that is possible between different
approaches. A basic mechanism for determining grammatical categories
involves grouping together units of text that fill the same textual
context, and assigning them a class. Using an over-simplified example,
all words that can fill the gap in ‘a ______ is needed’ could be classed as
‘noun’, and more specifically, ‘singular noun’. Grammars can then be
written describing the valid classes of words and word groups, the
sequences of these classes, and their interrelation in terms of dependency
or constituency. Classes are based on the ‘mutual substitutability’ of
elements.
Many approaches, particularly those following Bloomfield, believed in
the autonomy of syntax, that ‘syntactic phenomena are essentially inde-
pendent of the conventional semantic, pragmatic and discoursal functions
206 MICK O’DONNELL
of those phenomena’ (Butler 2003:6). In these approaches, one would con-

struct the grammar of a language by exploring the patterns of words
independent of their meaning. One could take a body of text, replace each
word by a number, and work towards discovering the grammar, free of the
confusing distraction of the meanings of the words.
8.2.2 Functional Grammars

This however is not the only approach to building grammars. Functional
linguists work on the assumption that the system of grammar is ‘so intim-
ately bound up with external motivating factors that it makes no sense to
try to describe it without reference to those factors’ (Butler 2003:12). In
Halliday’s words:
The particular form taken by the grammatical system of language is closely

related to the social and personal needs that language is required to serve.
(Halliday 1970:142)
The relation between the meaning and the wording is not, however, an
arbitrary one; the form of the grammar relates naturally to the meanings
that are being encoded. A functional grammar is designed to bring this out;
it is a study of wording, but one that interprets the wording by reference to
what it means.
(IFG1:xvii)
A systemic grammar is one of the class of functional grammars, which
means (among other things) that it is semantically motivated, or ‘natural’.
In contradistinction to formal grammars, which are autonomous, and there-
fore semantically arbitrary, in a systemic grammar every category (and
‘category’ is used here in the general sense of an organizing theoretical
concept, not in the narrower sense of ‘class’ as in formal grammars) is based
on meaning: it has a semantic, as well as a formal, lexicogrammatical
reactance.
(Halliday and Matthiessen 2006:3–4)
Halliday argues that this intimate relation between grammar and its con-
text of use is not accidental, but rather the result of language (including
grammar) having evolved in its use:
Language has evolved to satisfy human needs; and the way it is organized is
functional with respect to these needs – it is not arbitrary. A functional
grammar is essentially a ‘natural’ grammar, in the sense that everything
in it can be explained, ultimately, by reference to how language is used.
(IFG2:xiii)
The concept of the social function of language is central to the interpretation
of language as a system. The internal organisation of language is not acci-
dental; it embodies the functions that language has evolved to serve in the
life of social man.
(Halliday 1973:44)
Or in Martin’s words:
[Functional grammar] explores grammar as being shaped by, at the same

time as playing a significant role in shaping, the way we get on with
our lives.
(Martin et al. 2010:1)
This is not to say that grammars can be defined entirely in terms of how
words are used. This would lead to what is called a ‘notional grammar’:
grammar defined entirely in terms of meaning.1 In such an approach, a
‘verb’ might be defined as ‘a word that expresses an event’, and ‘noun’ as ‘a
word that expresses an entity’.
For Halliday, a functional grammar has to relate outwards to the mean-
ings it realizes, and also account generatively for the range of forms that
realize these meanings. He stresses that our grammatical organization
cannot be divorced from the need to account for structural patterning:
All the categories employed must be clearly ‘there’ in the grammar of the
language. They are not set up simply to label differences in meaning. In
other words, we do not argue: ‘these two sets of examples differ in meaning;
therefore they must be systematically distinct in the grammar’. They may
be; but if there is no lexicogrammatical reflex of the distinction, they are not.
(IFG1:xx)
Thus, in a systemic grammar, categories are determined both externally (to

capture similarities in meaning/use) and structurally (to capture similar-
ities in lexicogrammatical realization).
By IFG3, these criteria were further developed into the idea of ‘trinocu-
larity’, that grammatical concepts are defined from above (the semantics),
from the same level (grammatical form), and also from below (phonology in
this case):
We cannot expect to understand the grammar just by looking at it from its

own level; we also look into it ‘from above’ and ‘from below’, taking a
trinocular perspective. But since the view from these different angles is
often conflicting, the description will inevitably be a form of compromise.
All linguistic description involves such compromise; the difference between
a systemic description and one in terms of traditional school grammar is that
in the school grammars the compromise was random and unprincipled,
whereas in a systemic grammar it is systematic and theoretically motivated.
(IFG3:31)
Halliday and Matthiessen raise the point here that, when building a gram-
matical description, evidence from the three viewpoints may conflict, and
the model builder needs to choose what importance they give to each
1
David Rose, in the Sysfling discussion list (09/04/2011), suggested that the term ‘notional grammar’ is appropriate for
grammars which ignore grammatical reactance totally, while ‘functional grammar’ is appropriate for a grammar based
on the identification of recurring grammatical structures which have distinct semantic functions.
Table 8.1 Two approaches to structural analysis
Structural analysis Functional analysis

Premod. Head Postmod. Quantifier ... Thing Qualifier
seven apples seven apples

a handful of apples a handful of apples
a man of means a man of means
source of evidence. To exemplify this problem, Table 8.1 shows three

nominal groups, and presents two alternative ways of structurally analyz-
ing them (also see Fontaine and Schönthal, this volume). The leftmost
analysis, a more structural approach, takes structural similarity as the most
important principle (‘from its own level’). In addition, if these phrases were
Subject in a sentence, the finite verb would usually agree with the noun
denominated as ‘Head’.
The functional approach places lower emphasis on the seeming struc-
tural similarity, focusing instead on what the elements are doing semantic-
ally (‘from above’): both of the first two examples are talking about apples,
and seven and a handful of are functioning to specify how many apples are
involved. In a similar vein, functionalists would claim that with a cup of
water, we are not talking about a cup, but rather about a quantity of water.
So when we say he threw a cup of water over his brother, we usually understand
that the cup was not in fact thrown, just the water.
8.2.3 Weighing Formal and Notional Evidence

So far in this paper, functional grammars have been defined just as those
which take into account evidence from both meaning (above) and from
form (aside/below). Halliday and Matthiessen, in the quotation above, point
out that functional grammars need to reach compromises between these
different sources of evidence. One point that needs to be made clear is that,
even within SFL, the principles for making these compromises can vary: SFL
linguists vary in the degree of importance they give to structural and
meaningful criteria in determining grammatical categories.
Halliday gives a clue as to where he places priority:
The fact that this is a ‘functional’ grammar means that it is based on

meaning; but the fact that it is a ‘grammar’ means that it is an interpretation
of linguistic forms. Every distinction that is recognized in the grammar . . .
makes some contribution to the form of the wording. Often it will be a very
indirect one, but it will be somewhere in the picture.
(IFG1:xx)
The fact that the required grammatical reactance can be very indirect
shows that priority is given to similarity of meaning when deciding on
grammatical categories, rather than to structural regularity. This is made

explicit in IFG3:
Being a ‘functional grammar’ means that priority is given to the view ‘from
above’; that is, grammar is seen as a resource for making meaning – it is a
‘semanticky’ kind of grammar. But the focus of attention is still on the
grammar itself.
(IFG3:31)
Rather than talking about the relative importance of the criteria, it may be
better to talk of starting point. In IFG2 (xiv), Halliday contrasts the traditional
approach of starting with modelling the word forms (morphology), building
a syntax on top of that, and only then asking what these forms mean, with
the SFL approach, which starts by interpreting a language as a system of
meanings, and then explores how those meanings can be realized as forms.
A possible critique to this approach is that there are many ways to
organize a language in terms of meaning, and only a subset of these will
allow a simple mapping onto forms. What we ideally want is a meaningful
organization of language which has the strongest correlation with regular-
ities of form. We cannot do this by exploring meaning in isolation from the
forms that realize them.
This is the reverse of the criticism levelled by functionalists and cognitiv-
ists against Bloomfield and his successors, who tried to construct an
autonomous syntax without considering meaning (Tomlin 1990; Newmeyer
1991:62; Halliday in Martin 2013:164). The full answer is that, in the
construction of a language model, we need to consider meanings and forms
at the same time.
Structuralists often apply the principle of Occam’s Razor: the best gram-
matical description is that which uses the least rules to describe the phe-
nomena at hand. The principle can also be applied to the construction of
functional grammars: the best description is the briefest which represents
the meaningful aspects of language use and from which all forms can be
generated (with the mapping of meaning onto form included in the size of
the description).
Unfortunately, SFL grammars (or semantic specifications) are rarely pre-
sented with both system network and realization statements (Hudson 1971
and Matthiessen 1995 being good exceptions), and thus commonality of
meaning may play a bigger role in grammar construction than it should.
Halliday’s verbal processes (see Berry, this volume) offer a good area
through which to demonstrate the problems of grammatical classification.
Exactly what constitutes a verbal process is often debated within the com-
munity, and the four versions of IFG have shifted on the issue over time.
There is a common belief within the community that the test to identify a
verbal process is that there must be projection in clausal form (or at least
potential for clausal projection). This test would result in He said he was going
being classified as verbal, while He talked about the weather would not.
However, all versions of IFG include at least two classes of verbal process:
the projecting kind just mentioned, and one which involves a Target (I’m
always praising you to my friends). As there is no structural similarity between
these two types of verbal process which would motivate their grouping, we
must assume they are being grouped totally on notional grounds: they both
express a verbal action.2
We might propose putting targeted verbal processes aside, and say that
the clausal projection criterion applies to the remaining verbal processes.
However, it seems that the SFL community as a whole is divided as to what
to do with processes which involve verbal action, but where clausal projec-
tion is not involved. This involves verbs such as talk, and grumble, and
includes cases where no Matter is specified (We talked for hours), and also
where Matter is specified (He talked about his hometown).
An online survey was conducted in 2004 to test how a range of SFL
practitioners coded various clauses in terms of process types (O’Donnell
et al. 2008). Seventy-five respondents coded thirty-two difficult clauses. The
survey revealed a spread of coding styles, ranging from heavy dependence
on structural criteria, to those who coded largely on the semantics of the
clause. In respect to We talked for hours, 60 per cent of coders placed it as
behavioural, and 35 per cent followed notional criteria, coding it as verbal
(the remaining 5 per cent coded it as material, some indicating they did not
use the behavioural category).
In regards to He talked about his hometown, a surprising result was that
15 per cent of those who had coded the previous sentence as behavioural
swung over to verbal for this case. This suggests that these coders do
require presence of the verbal product to code as verbal, but do not go so
far as to require clausal projection (a similar pattern was shown in the
coding of mental processes). This variation in the coding community dem-
onstrates that the nature and degree of structural reactance needed varies
across the community. A further study reported in Gwilliams and Fontaine
(2015) confirmed these results.
Part of the disparity in coding verbal processes may stem from the
treatment of this area in the four versions of IFG. In IFG1, verbs like talk
were not covered explicitly in the section on verbal processes, although a
later section on Range classified She speaks German and Don’t talk nonsense! as
verbal processes (IFG1:133). Behavioural processes are said to be intermedi-
ate between material and mental processes, which seems to exclude the talk
verbs from this category.
IFG2 however expands behavioural processes to include a ‘near verbal’
category, which includes talk, grumble, and chatter (IFG2:139). The possibility
of Matter with these verbs is explicitly mentioned, so He talked about the
2
Tom Bartlett (personal communication) prefers to phrase the semantic label for verbal processes as ‘transfer of
information from one person (or semiotic object) to another’. This would leave We talked about the weather out of
verbal processes, while leaving He called me a bastard in.
Table 8.2 Types of verbal processes (IFG4:305)
TYPE Examples of verbs
activity targeting praise, flatter, commend, compliment, congratulate; insult,

abuse, slander, blame, criticize, chide, censure, pillory, rebuke
talking speak, talk
semiosis (neutral say, tell; go, be like
quoting)
indicating tell (sb that), report, announce, notify, explain, argue, convince
(that), persuade (sb that), promise (that)
ask (sb whether), question, enquire (whether)
imperating tell (sb to do), ask (sb to do), order, command, require, promise,
threaten, persuade (sb to do), convince (sb to do), entreat,
implore, beg
weather would be classified as behavioural. The talk example was removed

from the Range examples.
In terms of verbal processes, IFG2 says that
verbal processes do display distinctive patterns of their own. Besides being

able to project . . ., they accommodate three further participant functions in
addition to Sayer.
(IFG2:141)
The targeting type of process is again mentioned, with a statement that this
subtype of verbal process does not easily project reported speech. A list of
verbs taking a (nominal) Verbiage includes some which cannot easily pro-
ject: He described the apartment, or He outlined his plan.
In IFG3, representing a revision of IFG2 by Matthiessen, we see a change
back towards more notional coding. To talk to that priest about Kukal is said to
be verbal (IFG3:252). However, there seems to be some inconsistency here,
as grumbled about the food (IFG3:251) is said to be behavioural. I believe this
was a state of transition from Halliday’s original more structural orienta-
tion towards Matthiessen’s more notional orientation.
In IFG4, the talk processes are fully instantiated as a subtype of verbal
process, as shown in Table 8.2. However, behavioural processes still include
verbs such as chatter, grumble, and talk, which appears to be an inconsist-
ency, with grumbled about the food explicitly mentioned as behavioural
(IFG4:302).
Thompson (2015) believes this is not just a problem for verbal processes,
but general across transitivity classification:
Halliday (1994: xix) has consistently argued that ‘all the categories employed
must be clearly “there” in the grammar of the language. They are not set up
simply to label differences in meaning’; and in the case of transitivity certain
key grammatical criteria for categorization (such as preferred tense/aspect,
and the potential to project) have been elaborated (e.g. Halliday 1994:
115–16). However, it has proved difficult to implement the principle of
‘clearly “there” in the grammar’ in all cases: the grammatical criteria by
which one process type can be differentiated from another are not always
precisely definable. As a result, analysts may, implicitly or explicitly, find
themselves forced to fall back on purely semantic criteria.
(Thompson 2015:21–2)
The above discussion has tried to show that in Halliday’s grammar, the
notional criteria dominate over secondary, indirect, grammatical reactance.
Some in the community however stress the importance of the grammatical:
Process types are entirely grammatical categories. The names of the process
types are just aide-memoires that capture only their most common notional
features; they are not useable as criteria for defining on notional grounds.
(Tom Bartlett 2011, sys-func discussion list)
I would strongly encourage holding onto grammatical reactances when
reasoning about process type. These are the grounding strength of our SFL
approach to case relations, compared with work in other models. We should
be enriching our argumentation based on reactances . . . We need to push on
to tackle the challenge of finding distinctive reactances for process types as
we move from language to language – and NOT abandon the criterial argu-
mentation the reactances afford. It is very worrying to think that the power
of the SFL approach (its revelation of the meaningful ways in which lan-
guages construe reality) might become its undoing via a collapse into notion-
alism, or accommodation of notionalist ‘reasoning’ (sic) alongside reasoning
based on grammatical criteria.
(Jim Martin 2011, sys-func discussion list)
To summarize the discussion so far, we can distinguish three types of

grammars:
• notional grammars, based on semantic concerns, ignoring syntactic

reactance;
• formal grammars, based on structural concerns, ignoring semantic
factors;
• functional grammars, which take both meaning and structural issues
into account.
Approaches to functional grammars, however, can vary widely, from, on

one side, structurally oriented functional grammars which favour struc-
tural criteria over semantic ones, to notionally oriented functional gram-
mars, which favour the semantic over the structural.
8.2.4 Complex Relations between Notional Situation and

Grammatical Form
One problem for notionally oriented functional grammars is that there is
often a complex relationship between semantic representations and the forms
that express them. One example is in the relation between the speech function
systems and their grammatical expression. The ‘demand information’ speech
function can be realized grammatically in terms of interrogative syntax (Did

you like the movie?) or via declarative syntax with interrogative intonation (You
liked the movie?). In this area of the grammar (Mood) at least, Halliday’s gram-
mar has given more weight to the structural organization of language,
grouping clauses based on similarity of form, rather than on similarity of
speech function (a semantic concern).
However, when dealing with Transitivity, things are not so clear. Fawcett
(2009:214) points out that ‘there is NOT a simple one-to-one relationship
between the realms of experience and the types of Process’ (his ‘realms of
experience’ correspond to a language-external experiential representation).
He points out that different notional situations can be realized through
distinct grammatical realizations. Using my own examples in (1) to (4), a
situation of a man being preoccupied about his tomatoes could be expressed
in various ways (the process categories given are arguable, but based on the
syntactic similarity to less contentious clauses):
(1) He is very worried about his tomatoes (relational attributive)

(2) He worries about his tomatoes (behavioural)
(3) He has concerns about his tomatoes (relational possessive)
(4) He thinks his tomatoes might die (mental)
The point here is that, given the different possible expressions of a notional
situation, using notional criteria to classify process types does not seem
promising.
Gwilliams and Fontaine (2015) look at the problem from the other side:
because of ideational grammatical metaphor, similar clausal expressions
can be used for distinct notional situations, for instance, ‘Ivy touched Fred
with a stick’ (representing a notionally material action) and ‘Ivy touched
Fred with her words’ (representing a notionally mental action).
They note that such ambiguous cases give rise to two distinct problems.
Firstly, because these examples allow for two analyses, they introduce the
potential for inconsistent coding amongst analysts, given that some may
favour semantic criteria, and others, syntactic criteria. They point out
however that this can be avoided by explicit direction as to the coding
criteria.
The second problem they think is more important:
If a process can be interpreted in more than one way, being constrained to a

single classification may lead to an analytic interpretation that does not truly
reflect the semiotics of the message, going against the primary objective
of SFL.
(Gwilliams and Fontaine 2015:3)
The solution they propose is to always allow for two analyses of clauses: a
surface analysis based on syntactic tests, and a deep analysis based on
notional grounds (although in most cases, these would be the same). They
suggest that most of the contentious clauses encountered in process type

coding are exactly those where the deep coding differs from the surface.
There are other cases of ambiguity not due to grammatical metaphor, but
rather the result of words covering a range of notional space. For instance,
look at the clauses in examples (5) and (6):
(5) We agreed with each other that I was right.

(6) I agree with the Prime Minister that something should be done.
The first example implies verbal discussion, so would be coded notionally as

a verbal process. The second example is a little more difficult: it looks like a
statement of a mental state rather than of an explicit verbalization, so
would be coded, in many contexts, as a mental process (as the agreement
is not a verbalization but rather a statement of concord of ideas). So, two
clauses with essentially the same sentence structure are coded differently
with reference to the type of situation being referred to. Structural criteria
cannot help us decide on process type, and in these cases, we need to ask
ourselves what kind of (notional) activity is being represented.
Gwilliams and Fontaine (2015:8) argue that ‘semantic information is a
kind of subjective distractor’ from proper coding of transitivity, and that
one should instead follow Fawcett’s approach, basing process type coding
on the presence of structural elements (Participant roles = PRs):
In analyzing Process types and PRs, it doesn’t help to use the realm of
experience as a guide. And the analyst who has been forewarned of this
problem is less likely to fall into the trap of skipping the stage of applying the
tests for the Participant Roles, when trying to establish the Process type of a
clause.
(Fawcett 2009:215)
The point of this discussion has been to show that the issue of relative
importance of notional vs. structural criteria in the grammar is still an
open issue. Surveys of coding practice show practitioners range from more
to less notional in their coding of process type, and the four versions of IFG
are themselves in flux as to the importance of notional vs. structural
criteria.
8.3 Current Issues Related to Genre
The issues of what ‘genre’ is, and where it belongs in relation to the other
components of the linguistic model, have long been debated within SFL, and
the debate continues today between different parts of the community. This
section will discuss some of the main issues in this area.
One note on terminology: Halliday uses the term ‘register’ to refer to the
set of linguistic features that realizes a particular configuration of situ-
ational features. Martin, on the other hand, uses the term ‘register’ to refer
to the configuration of situational features (e.g. Martin 1992). In the discus-

sion below, I will follow Halliday’s use rather than Martin’s, even when
discussing Martin’s work.
8.3.1 How Do We Define Genres?

A first issue involves exactly how genres are defined. Paltridge (1996) put
forward two principal approaches to defining genres:
• In terms external to the text, generally the purpose of the text, e.g. to
persuade, to educate, to entertain, etc.
• In terms internal to the text, most typically in terms of common schematic
structures, or linguistic styles, e.g. editorial, narrative, anecdote,
report, etc.
‘Genre’ has been used in various places in both of these senses. Lee (2001:38)
describes the first approach, which makes a distinction between ‘genres’
(defined on external criteria) and ‘text types’ (defined on internal criteria):
A genre, in this view, is defined as a category assigned on the basis of external

criteria such as intended audience, purpose, and activity type, that is, it
refers to a conventional, culturally recognised grouping of texts based on
properties other than lexical or grammatical (co-)occurrence features, which
are, instead, the internal (linguistic) criteria forming the basis of text type
categories.
Biber (1988:170) is a prominent proponent of this approach:
Genre categories are determined on the basis of external criteria relating to

the speaker’s purpose and topic; they are assigned on the basis of use rather
than on the basis of form.
Swales (1990:58) also follows this approach:
A genre comprises a class of communicative events, the members of which

share some set of communicative purposes. These purposes are recognised
by the expert members of the parent discourse community, and thereby
constitute the rationale for the genre.
In earlier works within SFL, ‘genre’ was not directly covered by the model,
although aspects relatable to external definitions of genre were mentioned,
placed within Context of Situation. This includes Halliday’s inclusion of
‘purpose’ and ‘rhetorical mode’, Ure and Ellis’ ‘role’, and Gregory and
Carroll’s ‘functional tenor’. These aspects will be discussed further below.
More recently, the internal definition of genre has been more prominent
within SFL: genres being defined as groupings of text with common text
structures. Hasan (1978:229), for instance, clearly takes this approach:
The generic membership of the text is determined by reference to the

structural formula to which the actual structure can be shown to belong.
Martin (and Martin and Rose) also indicate the use of internal criteria:
Genre networks would thus be formulated on the basis of similarities and

differences between text structures.
(Martin 1992:505)
Recurrent global patterns were recognized as genres, and given names.
(Martin and Rose 2008:7)
Martin (personal communication) notes that the external/internal distinc-

tion is complicated when applied to his model, as genre is realized through
constraints on register, and register could be seen as external to the linguis-
tic system. One could thus say that genres are recognized in terms of
patterns external to language. He comments however that as register is
itself a connotative semiotic system, it may be extra-linguistic, but it is not
extra-semiotic.
Paltridge (1996:238) argues that Martin often labelled genres with cat-
egories more frequently associated with external, purpose-based defin-
itions, such as poems, lectures, seminars, recipes, etc. Martin however
responds that he avoids such labels
precisely because they are common sense everyday labels which aren’t
names of recurrent patterns of meaning (e.g. a poem can be almost any
genre in my terms – anecdote, report, description, narrative, procedure
etc. – where genre is a recurrent pattern of meaning).
(Martin 2015, personal communication)
Martin’s key definition of genre as ‘a staged, goal-oriented, purposeful

activity in which speakers engage as members of our culture’ (Martin
1984:25) includes internal aspects (‘staged’, ‘activity’) but also includes
external aspects (‘goal-oriented’, ‘purposeful’). The internal/external defin-
ition of genre is also reflected in the following formulation from Martin and
Rose (2003:7):
We use the term genre in this book to refer to different types of texts that
enact various types of social contexts.
One possible interpretation parallels the arguments in Section 2 in relation

to the form/function co-evolution: our social needs and the activity
sequences we use to satisfy them have evolved in tandem, with the
consequence that each text type is an inseparable fusion between an
externally defined purpose and a linguistic means of achieving that
purpose. Taking this approach, one could argue that if one discovers a
set of texts which share the same generic structuring, then the functional
assumption would be that these texts also reflect externally defined
commonalities: similarity of purpose, etc. There is no explicit discussion
of this assumption in the literature, but such an underlying assumption
might explain the mix of internal and external elements in Martin’s
approach.
In summary of this section, while the definition of genre in terms of

shared text structure seems to dominate within SFL today, it is not the only
approach.
8.3.2 How are Genres Realized?

In some approaches, the linguistic realization of a ‘genre’ is explored in
terms of the recurrent stylistic qualities of the texts that belong to the
genre. In other words, a genre can be defined as a set of expected linguistic
patterns, whether lexical, grammatical, or semantic. Genre in this
approach is very relatable to Halliday’s notion of Register. Alternatively
(or additionally), a genre can be realized in terms of requiring recurrent
schematic staging of the texts in the genre. Within SFL, this is often called
‘generic structure’.
SFL models are perhaps most differentiated in the degree to which they
explicitly handle the notion of generic structure. In many early approaches,
the term ‘genre’ was avoided as a technical term, and instead, terms such as
‘purpose’, ‘purposive role’, ‘language use’, or ‘functional tenor’ were used,
all of which were categories of the Context of Situation, and were realized
in terms of stylistic patterning of the text, via Register. In these approaches,
generic staging was not usually mentioned. In later work, Martin and
Hasan do explicitly address genre and generic structure, and in their works,
genre is defined in terms of commonality of text structure.
It remains to be discussed whether genre in their models is also realized
through overall generic styling of the text. In the work of Hasan, it seems
that the stylistic realization of a genre is treated as an aspect of register. For
example, Hasan (1978:241) says that ‘[t]he terms “register” and “genre” as
used here are then interchangeable’. In Martin’s model, since genre selec-
tions are realized as constraints on register, the overall stylistic commonal-
ity of a genre can also be seen as handled via register, although genre,
where genre choices are made, is placed stratally above register, where the
stylistic consequences are managed.
8.3.3 Where Does ‘Genre’ Belong in the SFL Model

In this section, we consider where SFL approaches place ‘genre’ in relation
to the level of Context of Situation (which is known as ‘Register’ in
Martin’s model).
8.3.3.1 Genre as a Component of Context of Situation

As mentioned above, in earlier SFL models (and still sometimes today),
concerns similar to genre were placed as components of the Context
of Situation. Different practitioners placed these concerns in different
components of the situation, whether Field, Tenor, or Mode (see Martin
1999).
As a component of Field: Halliday’s earliest model of Context of Situation

included ‘purpose’ as an aspect of Field:
THE FIELD: Here we include the subject matter; and also the type of situation
in which language is used, including the purpose – e.g., didactic or explana-
tory, for information, for action, consolation or self-satisfaction.
(Halliday 1965:14)
Matthiessen (2015) reports on unpublished work by Jean Ure, where she is

said to include under field the category ‘field of activity’, which was subdiv-
ided into eight areas: expounding, reporting, recreating, sharing, doing,
enabling, recommending, and exploring. Matthiessen (2015:6) also
follows this approach, although renaming the parameter as ‘socio-semiotic
process’:
The field of activity is ‘what’s going on’ in context . . . The activity is either
primarily social or primarily semiotic – i.e. either primarily a process of
interactive behaviour or one of exchanging meaning. To capture this, I have
called this parameter SOCIO-SEMIOTIC PROCESS.
Later in the paper, he relates his socio-semiotic processes to the concept of

‘genre’ as used by Martin and others.
Hasan (1999) includes ‘verbal action’ under Field, which is further speci-
fied with features such as ‘informing’, ‘narrating’, ‘instructing’, etc. (her
earlier works had placed much of this under ‘rhetorical mode’,
discussed below).
As a component of Tenor: the model of Gregory and Carroll (1978) includes
two subcomponents under Tenor: ‘functional tenor’ and ‘personal tenor’.
‘Functional tenor’ is close to ‘purpose’, being described as
the category used to describe what language is being used for in the situ-
ation. Is the speaker trying to persuade? to exhort? to discipline?
(Gregory and Carroll 1978:53)
Functional tenor was placed under tenor because it relates in effect to how
the interactants relate to each other (as the persuader and persuaded, etc.),
very external criteria. Note however that Gregory and Carroll state that
‘genre’ covers more than just functional tenor, involving also field, personal
tenor, and mode:
We prefer to characterise genre in terms of all the dimensions of language

variety. Most significant literary genres, such as epic, ode, lyric, sonnet,
tragedy, farce and comedy, have author/reader expectations as regards not
only the medium relationship involved but also as regards the purposive
roles/on-going social activities, and the personal and functional addressee
relationships which are at risk, and so field and personal and functional
tenors are likewise relevant to their description. Literary genres can be seen
as individual kinds of marked registers within literature.
(Gregory and Carroll 1978:44–5)
As a component of Mode: under Halliday’s later model, the closest thing to

‘genre’ is included as an element of Mode, called ‘rhetorical mode’ (Halliday
1978:63; Halliday and Hasan 1989:143; Halliday and Matthiessen 2006:332).
IFG4 defines rhetorical mode as follows:
Rhetorical mode encompasses a number of rhetorical categories concerned

with the contribution of the text to the situation it operates in: informative,
didactic, persuasive, exhortatory, pragmatic, and so on.
(IFG4:38)
IFG4 notes that some rhetorical modes are oriented towards the Field of the
text (e.g. informative, didactic, explanatory and explicatory contexts), and
others towards the Tenor of the text (e.g. persuasive, exhortatory, horta-
tory, polemic contexts).
As a fourth component: Ellis and Ure (Ellis 1965; Ellis and Ure 1969; Ure and
Ellis 1977) have two distinct components in place of Halliday’s Tenor:
‘formality’ and ‘role’. Role corresponds roughly to genre, being defined as
‘the dimension correlating with the social or other role of the utterance or
text, e.g., conversation, literature, technical writing, etc.’ (Ellis 1965:13).
8.3.3.2 ‘Genre’ as Realization of Context of Situation

Hasan, like Halliday, does not often use the term ‘genre’ directly. Where she
does, she states that, for her, ‘the term “genre” is a short form for the more
elaborate phrase “genre-specific semantic potential” (Halliday and Hasan
1989:108). Here she is talking about the stylistic realization of genres.
She considers the category of genre as superfluous, given that the associ-
ation between a genre and the semantic patterns that realize it is already
covered by the notion of register: ‘For most material purposes register and
genre are synonymous’ (Hasan 1978:230) and ‘The terms “register” and
“genre” as used here are then interchangeable’ (Hasan 1978:241). Since, in
her model, register stratally realizes Context of Situation, Hasan thus places
genre as below the Context of Situation.
Here though, she is referring to genre as stylistic realization in terms of
patterns of linguistic selection, and not to generic structure. For her, ‘gen-
eric structure’ is distinct from ‘genre’. She places generic structure as a
realization of selections of the context of situation. More details of her
approach will be given below.
8.3.3.3 ‘Genre’ Stratified above Context of Situation

It should be apparent from the discussion above that ‘genre’ in the SFL
model does not clearly sit within any one of the situational categories, and
has been, at different times, placed in all three of the components. Add-
itionally, even when placed in one component, it is still shown to have
influence on the other situational components.
When Martin was first teaching a ‘Functional Varieties’ course at the
University of Sydney, he started off using Gregory’s ‘Functional Tenor’
approach, but students found this difficult, as they had earlier been exposed
to Halliday’s Rhetorical Mode approach (see Martin 2014). Class discussions
explored the cross-component implications of genre, and two class members,
Guenter Plum and Joan Rothery, suggested ‘positioning functional tenor as a
deeper variable, since the purpose of a text influenced all of interpersonal,
ideational and textual meaning’ (Martin 2014:12). The eventual result of this
discussion, under Martin’s leadership, was to rename the deeper variable as
‘Genre’, recognizing it as a stratum separate from Context of Situation.
‘Personal Tenor’ could thus be renamed as simply ‘Tenor’.
Genre in Martin’s approach is described in terms of both a system
network (defining genres and their variants), and a layer of structure, such
that choices in the genre network determine which schematic elements are
realized in the text. In his model, the register of each stage is determined
via interstratal realization:
As part of the realisation process, generic choices would preselect field, mode
and tenor options associated with particular elements of text structure.
(Martin 1992:505)
For Martin it seems, Genre only interfaces with Context of Situation (which
he calls Register), and does not interface directly with lower strata:
Genre is a pattern of register patterns, register a pattern of discourse seman-

tics ones, which are in turn a pattern of lexicogrammatical ones, in turn a
pattern of phonological ones.
(Martin 2014:14)
Register [is] the expression plane of genre.
(Martin 2014:13)
If this is so, then there must be some variables in the register which can pass
on the linguistic constraints of the genre to the lower stratum. The linguistic
patterns which were previously activated by functional tenor still need to be
activated by some variables in the Context of Situation (Martin’s Register
layer), and these variables need to be activated by the choice of Genre. The
alternative is to allow selections in the Genre stratum to directly interface
with each of the strata below: limiting the allowable contextual configur-
ations, in addition to activating linguistic possibilities. Both approaches are
viable, although each one has strong consequences for linguistic modelling,
and there should be a clear statement as to which approach is being followed.
Martin’s model of Genre over Register is fairly widely accepted within the
educational side of SFL. His approach is not uncritically accepted however,
particularly in regards to Hasan and those who follow her model (e.g. see
Hasan 1995). She argues that by putting Genre outside the semiotic space,
human interaction is de-humanized:
My own view is that the stratification of genre and register, the collapsing of
the social and the verbal, at both these planes, . . . has a highly deleterious
effect: It moves the whole issue of text structure and its activation from
active, feeling, reacting participants co-engaged in some interaction to given
forms of talk that represent the ways things are done in our culture, as if the
culture is unchanging and as if the participants are simply pre-programmed.
(Hasan 1995:283)
Martin responds that his Genre is in fact a semiotic system, open to human
choice:
A culture for me is a system of genres . . . genres do meaning, just like

everything else in semiosis does . . . genres don’t realise the social, they are
my theory of the social: we live genres.
(Martin 2015, personal communication)
Lukin et al. (2011:189) put forward a different argument, that placing genre
and register together provides a simpler analytical tool:
As a central conceptual tool that does not stratify the relation of genre and
register, Halliday’s notion of register helps us recognize – or at least frame
and test – the idea that recognized social situations might sometimes be the
same register, or identify and evaluate the register differences in what are
normally counted as ‘the same’ social activities: it is a model well suited to
calibrating the shuffling and reshuffling of cultural space-time and its
boundaries.
Martin (1999:505), on the other hand, argues that placing genre within
context of situation is just not feasible:
It seems to us impossible to associate the accomplishment of genres as stated

goal-oriented social process with any one metafunction (ideational, interper-
sonal or textual) or correlating register variable (field, tenor or mode). For us
genre redounds simultaneously with field, tenor and mode, and thus with
ideational, interpersonal and textual meanings.
8.3.4 Where Does Generic Structure Belong in the Model?

For Martin, the issue of where generic structure belongs in his model is
clear: selections from the Genre network are realized (in part) by creating a
schematic structure at the Genre level (Martin 1992:505). Each schematic
element then constrains the Context of Situation (his Register) by setting
appropriate Field, Tenor, and Mode selections. For him then, generic struc-
ture determines contextual features, while for Halliday and Hasan, context-
ual features determine generic structure.
For Halliday (1978:134), ‘Generic structure is outside the linguistic
system; it is language as the projection of a higher level semiotic structure’,
but ‘it can be brought within the general framework of the concept of
register.’ Register for Halliday is not a stratum of the model, but rather a
relationship between the Context of Situation and the linguistic strata: a
register is the set of linguistic choices that recurrently occur in the texts
produced in a given situation type. So, by placing generic structure within
the framework of register, generic structures are thus linguistic patterns
that realize particular situation types.
One of the most likely places for generic structure in the linguistic model
is on the Semantic stratum. The closest thing to genre in Halliday’s later
model is ‘rhetorical mode’, which is a component of Mode. And because
Halliday often states that Mode is in most cases realized through the
Textual metafunction, we might infer that, for Halliday, generic structure
is a part of the Textual component of the Semantic stratum, along with
Information Structure and Thematic Structure.
For Hasan also, generic structure is a realization of the Context of Situ-
ation: the structure of a text is determined by the selection of features from
the Context of Situation network, which
can predict the obligatory and the optional elements of a text’s structure as
well as their sequence vis-à-vis each other and the possibility of their iteration.
(Halliday and Hasan 1989:56)
She stresses that one cannot expect elements of the text structure to be
determined by individual situational features, but that they are determined
by the configuration of features selected from Field, Tenor, and Mode (what
she calls a ‘contextual configuration’, or a CC):
We need the notion of CC for talking about the structure of the text because
it is the specific features of a CC . . . that permit statements about the text’s
structure. We cannot work from the general notion of, say, ‘field’ since it is
not possible to claim, for example, that field always leads to the appearance
of this or that element.
Thus, in her model, generic structure is the realization of the Context of

Situation. What is not clear from her work is whether this is intra-stratal or
inter-stratal realization. If the first, then text structure would be seen as
structure at the Context of Situation stratum. If the second, text structure
would be stratally below Context, which would then place text structure on
the Semantic stratum (as the Hallidayan model does not posit any stratum
between Context and Semantics).
There is evidence (although not very strong) that Hasan places generic
structure stratally below Context of Situation, with register (and thus
genre) linking them:
In the SF model the concept of register is a ready-made link between context

and generic structure.
Matthiessen (2015:10) follows Hasan, saying that generic structure is a

realization of situational features:
Certain systemic terms in the systems of the socio-semiotic process (field of

activity) network . . . have realization statements associated with them. These
realization statements specify fragments of the structure of the situation
type . . . For example, ‘non-sequential’ is realized by the presence of the
contextual element of Phenomenon Identification (‘non-sequential’ & +
Phenomenon Identification) . . . In other words, the kind of potential that is
built into Hasan (1978, 1984) specifications of generic structure potential
(GSP) is represented systemically here – which is, of course, in line with her
theory of context.
The phrase ‘structure of the situation type’ implies that he takes generic
structure to be a structure level of the context of situation (context of
situation thus having system and structure specifications). He confirms this
two pages later:
The realization statements in Figure 6 are inter-axial but intra-stratal: that is, they
relate paradigmatic order to syntagmatic order within the stratum of context.
(Matthiessen 2015:12)
Contextual features can alternatively be realized directly as registerial

constraints:
But contextual realization statements can, of course, also refer to patterns

below the stratum of context – inter-stratal realization statements. If con-
textual elements are realized linguistically, these patterns are semantic; that
is, contextual elements are realized by patterns of meaning, as shown by
Hasan, Ruqaiya (1984).
(Matthiessen 2015:12)
8.3.5 How are Variations in Language across Generic Structure

Explained?
One problem for these approaches is to explain how patterns of linguistic
choices are not constant across a text as a whole, but rather, change as the
text shifts from stage to stage. Bateman (2008:185) explains this point:
when we describe the linguistic details of texts in close detail, it is rare that an
entire text exhibits precisely the same range of stylistic options. More often we
can locate particular phases or segments of a text showing a relatively homo-
genous range of stylistic options and other segments of the same text that
show different options being taken up. Therefore, a single linguistic text, or
linguistic event, may appeal to several distinct registers while it is unfolding
and yet still be seen as a coherent example of a single ‘type’ of text. . . . Since
texts need not be homogeneous, simple ‘labels’ for registers or genres are
rarely appropriate. . . . Each stage can take on a distinctive register.
For Martin’s model, this is not a problem, as each schematic element can
be related directly to the register selections appropriate for that section.
He says:
Making genre rather than register variables responsible for generating sche-
matic structure makes it easier to handle changes in experiential, interper-
sonal and textual meanings from [one] stage to another in a text . . .
Underlying register, genre can be used to predict these changes, stage by
stage, while at the same time accounting for a text’s overall coherence.
(Martin 1992:506)
For the Halliday and Hasan models, this is more of a problem. For them,
generic structure is created by the selected contextual configuration. But
there is no mechanism to allow for the element of the generic structure to
turn back and change the contextual configuration which would allow
variation in linguistic selection, although Martin (personal communication)
suggests that Hasan’s (2015) ‘ITERATION’ systems, which allow re-entry into
the Field network using a recursive system, might work for this. However, in
that article, the recursive system was applied to modelling situations with
multiple fields, not to modelling the staging of texts/interactions.
8.3.6 Alternative Approaches: Phasal Analysis and

Dynamic Context
As discussed just above, in models where context is seen as constant over
the text as a whole, it is difficult to account for the differing linguistic
choices that occur over stages of a text. A solution to this problem is to
allow for the context of situation to shift as the text unfolds.
There has been various discussions of dynamicity in SFL over the years,
mostly in respect to dialogic interaction, (e.g. see Hasan 1981; Ventola 1983,
1987; Halliday 1984; Martin 1985; Cloran 1987; O’Donnell 1990, 1999).
Hasan (1981:118), for instance, says:
When the context is co-operatively negotiated, the text and context evolve
approximately concurrently, each successive message functioning as an
input to the interactants’ definition of what is being achieved.
These words suggest that the Context of Situation is not constant over a
given text or interaction, but can change as the text unfolds. Generic
staging in a text can thus be seen as the result of a sequence of shifts in
the context of situation, a change in what the participants are trying to
achieve at each point of the text (e.g. from motivating a study to detailing
that study). Each shift in Context of Situation is associated with a shift in
the register of the text.
One approach that takes this assumption most seriously is that of Phasal
Analysis (Gregory and Malcom 1981; Malcolm 2010), which allows for
phasal shifts in register throughout a text:
Phase characterizes those stretches of text where there is a significant

measure of consistency in what is being selected ideationally, interpersonally
and textually . . . phase can be thought of as a delicate statement of register.
(Gregory and Malcolm 1981:8)
Stillar (1992:105), another proponent of Phasal Analysis, expresses the idea

that the apparent staging of a text can be viewed as the product of shifts in
the context of situation:
Discourse is necessarily the linguistic reflection of dynamic shifts in our

recurrent relations to experience, interaction and medium – that is, interde-
pendent ‘parts’ of a discourse can be viewed as the micro-registerial realiza-
tion of instantial situations.
Phases differ from stages in that phases can overlap; for instance, we might
have a phase of consistent ideational selections (e.g. talking about Lego),
containing shifts in the interpersonal selections (change from monologue to
dialogue).
My own work on dynamic modelling of interaction (O’Donnell 1990,
1999; O’Donnell and Sefton 1995) takes a similar approach to the idea of
context as dynamically mutable, but that work was not trying to explain
the registerial shifts over stages. O’Donnell (2012) looked more deeply at
dynamic shifts in tenor over a text.
In dynamic approaches like these, we might do away with modelling
generic structure as such, seeing the apparent staging as the result of the
dynamic shifts in the Context of Situation, the interactants’ notion of what
is going on. As the context shifts throughout the interaction/text, as a
result, the register shifts as well. We do not need to posit text structure
intermediate between context and text.
We are left however with the problem of modelling the process of how
shifts in the context of situation take place, both in dialogue and in written
text. Cloran (1987) offers an interesting discussion of how contextual shifts
in interaction can be negotiated by the participants. Much work is however
needed to apply this dynamic context perspective to describe registerial
shifts in written texts, such as are usually explained by generic structure.
8.3.7 Summary
The problem of where Genre belongs in the SFL model stems from the
seemingly circular relation between Genre and Context of Situation. On
the one hand, we can say that the Context of Situation determines whether
a given Genre (or generic structure) is appropriate or not, and more deeply,
particular variants of a generic structure potential may be activated or
deactivated by particular contextual features. On the other hand, each
element of a generic structure is associated with distinct language patterns,
so, to this extent, the stages of the genre determine the register used
within them.
In the traditional SFL approach, the Context of Situation, and thus Regis-
ter, is seen as something constant over a text as a whole. And given the need
to account for staging of language patterns over a text, the traditional
approach thus needs to account for this staging outside of context. Hasan
has taken one approach, placing the staging as a realization of the Context
of Situation. But taking that approach makes it difficult to account for the
distinct micro-registers of language use in each stage. Martin has chosen to
place the notion of genre, and thus of staging, above that of Context of
Situation, and thus allows each generic stage to reflect distinct registerial
patterns. An alternative approach avoids the problem by rejecting the
assumption that the Context of Situation is static, allowing for micro-shifts
in register to result from a dynamically shifting context.
This chapter has explored two areas of interest within SFL that are still
unresolved, that of the degree of notionalism in determining grammatical
categories, and also the exact role and nature of genre within the model.
Both of these areas lead to active discussions on the SFL discussion lists, at
conferences, and in publications. Often these discussions contain confu-
sions where participants see the issue through the lens of their own
assumptions, not aware of the underlying issues that lead their very words
to mean different things to different readers. It is the hope of the author
that, by bringing these underlying issues to the surface, future discussion
will be less distracted by mistaken interpretations, and driven more
through mutual understanding of the different sides of the issues.
References
Bateman, J. 2008. Multimodality and Genre: A Foundation for the Systematic

Analysis of Multimodal Documents. Basingstoke: Palgrave Macmillan.
Biber, D. 1988. Variation across Speech and Writing. Cambridge: Cambridge
University Press.
Butler, C. 2003. Structure and Function: A Guide to Three Major Structural-
Functional Theories, Part 1: Approaches to the Simplex Clause. Amsterdam:
John Benjamins.
Cloran, C. 1987. Negotiating New Contexts in Conversation. Occasional
Papers in Systemic Linguistics 1: 85–110.
Ellis, J. 1965. Linguistic Society and Institutional Linguistics. Linguistics
3(19): 5–20.
Ellis, J. and J. Ure. 1969. Language Varieties: Register. In A. R. Meetham, ed.,
Encyclopedia of Linguistics: Information and Control. Oxford: Pergamon.
251–9.
Fawcett, R. 2009. Seven Problems to Beware of When Analyzing Processes
and Participant Roles in Texts. In S. Slembrouck, M. Taverniers, and
M. van Herreweghe, eds., Will to Well: Studies in Linguistics, Offered to
Anne-Marie Simon-Vandenbergen. Ghent: Academia Press. 209–24.
Gregory, M. and S. Carroll. 1978. Language and Situation: Language Varieties
and Their Social Contexts. London: Routledge.
Gregory, M. and K. Malcolm. 1981. Generic Situation and Discourse Phase:
An Approach to the Analysis of Children’s Talk. Unpublished mimeo. Toronto.
Gwilliams, L. and L. Fontaine. 2015. Indeterminacy in Process Type Classifi-

cation. Functional Linguistics 2(8): 1–19.
Halliday, M. A. K. 1965. Speech and Situation. English in Education 2(A2):
14–17.
Halliday, M. A. K. 1970. Language Structure and Language Function. In
J. Lyons, ed., New Horizons in Linguistics. Harmondsworth: Penguin. 140–65.
Halliday, M. A. K. 1973. Explorations in the Functions of Language. London:
Edward Arnold.
A Systemic Functional Interpretation of the Nature and Ontogenesis of
Dialogue. In R. Fawcett, M. A. K. Halliday, S. Lamb, and A. Makkai, eds.,
The Semiotics of Culture and Language, Vol 2: Language and Other Semiotic
Systems of Culture. London: Pinter. 3–35.
Halliday, M. A. K. 1985. Introduction to Functional Grammar. London: Edward
Arnold.
Edward Arnold.
Halliday, M. A. K. and R. Hasan. 1989. Language, Context and Text: Aspects of
Language in a Social-semiotic Perspective. 2nd ed. Oxford: Oxford University
Press.
Grammar. 3rd ed. London: Edward Arnold.
through Meaning: A Language-based Approach to Cognition. London: Continuum.
Hasan, R. 1978. Text in the Systemic-functional Model. In W. Dressler, ed.,
Current Trends in Text Linguistics. Berlin: Mouton de Gruyter. 228–46.
Hasan, R. 1981. What’s Going On: A Dynamic View of Context. Seventh
LACUS Forum. Columbia: Hornbeam Press.
Hasan, R. 1995. The Conception of Context in Text. In P. Fries and
M. Gregory, eds., Discourse in Society: Systemic Functional Perspectives.
New York: Ablex. 183–284.
219–328.
Hudson, R. A. 1971. English Complex Sentences. Amsterdam: North Holland.
Lee, D. Y. W. 2001. Genres, Registers, Text Types, Domains, and Styles
Clarifying the Concepts and Navigating a Path through the BNC Jungle.
Language Learning and Technology 5(3): 37–72. Available online at: https://
llt.msu.edu/vol5num3/lee. (Last accessed 15/05/2017.)

4(2): 187–213.
Malcolm, K. 2010. Phasal Analysis: Analyzing Discourse through Communication
Linguistics. London: Bloomsbury.
Martin, J. R. 1984. Types of Writing in Infants and Primary School. In
L. Unsworth, ed., Reading, Writing, Spelling: Proceedings of the Fifth Macarthur
Reading/Language Symposium. Sydney: Macarthur Institute of Higher
Education. 34–55.
Martin, J. R. 1985. Process and Text: Two Aspects of Human Semiosis. In
J. Benson and W. Greaves, eds., Systemic Perspectives on Discourse, Vol. 1.
Norwood: Ablex. 248–74.
Benjamins.
Martin, J. R., ed. 2013. Interviews with M. A. K. Halliday: Language Turned Back
on Himself. London: Bloomsbury.
Martin, J. R. 2014. Evolving Systemic Functional Linguistics: Beyond the
Clause. Functional Linguistics 1(3): 1–24.
Martin, J. R., C. M. I. M. Matthiessen, and C. Painter. 2010. Deploying
Functional Grammar. Beijing: Commercial Press.
Martin J. R. and D. Rose. 2003. Working with Discourse: Meaning beyond the
Clause. London: Continuum.
Martin J. R. and D. Rose. 2008. Genre Relations: Mapping Culture. Sheffield:
Equinox.
Newmeyer, F. J. 1991. Functional Explanation in Linguistics and the Origins
of Language. Language and Communication 11(1–2): 3–28.
O’Donnell, M. 1990. A Dynamic Model of Exchange. Word 41(3): 293–328.
O’Donnell, M. 1999. Context in Dynamic Modelling. In M. Ghadessy, ed.,
63–99.
O’Donnell, M. 2012. Tenor in a Dynamic Model of Context. Paper presented at
Register and Context 2012. Macquarie University, 6–8 February 2012. Avail-
able online at: www.wagsoft.com/Presentations/ODONNELL-MAcquarie-
Tenor2012.pdf. (Last accessed 15/05/2017.)
O’Donnell, M. and P. Sefton. 1995. Modelling Telephonic Interaction:
A Dynamic Approach. Interface: Journal of Applied Linguistics 10(1): 63–78.
O’Donnell, M., M. Zappavigna, and C. Whitelaw. 2008. A Survey of Process
Type Classification over Difficult Cases. In C. Jones and E. Ventola, eds.,
From Language to Multimodality: New Developments in the Study of Ideational

Meaning. London: Continuum. 47–64.
Paltridge, B. 1996. Genre, Text Type, and the Language Learning Classroom.
ELT Journal 50(3): 237–43.
Stillar, G. 1992. Phasal Analysis and Multiple Inheritance: An Appeal for
Clarity. Carlton Papers in Applied Language Studies 9: 104–28.
Swales, J. M. 1990. Genre Analysis: English in Academic and Research Settings.
Cambridge: Cambridge University Press.
Thompson, G. 2015. Pattern Grammar and Transitivity Analysis. In
N. Groom, M. Charles, and S. John, eds., Corpora, Grammar and Discourse.
Tomlin, R. 1990. Functionalism in Second Language Acquisition. Studies in
Second Language Acquisition 11: 155–77.
Tucker, G. 2014. Process Types and Their Classification. In K. Kunz, E. Teich,
S. Hansen-Schirra, S. Neumann, and P. Daut, eds., Caught in the Middle:
Language Use and Translation. A Festschrift for Erich Steiner on the Occasion of
his 60th Birthday. Saarbrücken: universaar, University of the Saarland.
401–16.
Ure, J. N. and J. Ellis. 1977. Register in Descriptive Linguistics and Linguistic
Sociology. In O. Uribe-Villegas, ed., Issues in Sociolinguistics. The Hague:
Mouton de Gruyter. 197–243.
Ventola, E. 1983. The Dynamics of Genre. Nottingham Linguistic Circular
13:103–23.
Ventola, E. 1987. The Structure of Social Interaction: A Systemic Approach to the
Semiotics of Service Encounters. London: Pinter.
9
The Cardiff Model
of Functional Syntax
Anke Schulz and Lise Fontaine
9.1 Introduction
Within the theory of Systemic Functional Linguistics (SFL), there is, what
Halliday refers to as the ‘powerhouse’ of the theory, the ‘central processing
unit’ where ‘meanings are created’ (Halliday 1994:15). Despite this central
role in the theory, there have been relatively few developments specifically
related to the lexicogrammar since the 1980s and 1990s. There has been,
however, one concerted effort to promote debate in this area and to suggest
theoretical developments to the grammar by a team of scholars at Cardiff
University (e.g. Fawcett 1980, 2000a, 2008a, 2017; Tucker 1998, 2017; Tench
1990, 1996, 2017). This chapter provides an overview of the Cardiff
approach to syntax within the SFL framework. What we might refer to as
the ‘standard’ model of SFL1 is described in various chapters in this volume,
notably, Berry, Butt, Fontaine and Schönthal, Taverniers, and Webster, and
is represented in various other chapters throughout the rest of the volume.
There is also a very useful comparison between the more ‘standard’ model
and the Cardiff model in Butler (this volume), but see also Butler (2003a,
2003b) and Butler and Gonzálvez-García (2014). While such comparisons
are valid and important, the focus in this chapter is not on comparing the
two models, although where appropriate, important issues are mentioned.
The Cardiff model has its basis in SFL theory and, in particular, in
Halliday’s earlier work. While many of the principles are shared between
the two models, Butler (2003a:153) points out that ‘there are in the Cardiff
account some important differences in the underlying goals, as well as
extensions and simplifications of the grammar itself’. There have been
several key concerns that have driven the model, and while a full
1
For ease of reference, the term ‘standard model’ will be used to refer to the more widely known model of
grammar, e.g. Halliday and Matthiessen (2014) or any of the editions of IFG, Introduction to Functional Grammar
(e.g. Halliday 1994 or Halliday and Matthiessen 2004), also sometimes referred to by some as the Sydney grammar.
The Cardiff Model of Functional Syntax 231
consideration of all of them is beyond the scope of this chapter, we will

focus on two in particular. One relates to how an SFL theory can or should
account for a single representation (e.g. a clause) which expresses various
different structures (e.g. mood, transitivity); this concerns the relationship
between meaning and form, or the concept of instantiation in SFL. Another
relates to the nature of choice and the place of semantics in the system
network (see Fawcett 2013). The aim of this chapter is to examine these two
concerns in terms of how they contribute to the Cardiff model of functional
syntax as it applies to the English language.
The chapter is organized as follows. In Section 9.2, we situate the Cardiff
model in its historical roots, describing briefly how it developed. Section 9.3
provides the foundation for the functional syntax developed in the Cardiff
model by outlining its main features, including the role of planners, system
networks, and probabilities. Following this, in Section 9.4, we give a brief
overview of the main grammatical units along with a discussion of the key
concepts of filling, componence, and exponence. This also includes
examples of the way in which transitivity and participant roles work in
the Cardiff model and how the various strands of meaning are expressed in
a single representation. In Section 9.5, we outline the Cardiff approach to
clause analysis using a simple example. Finally, the chapter is summarized
in Section 9.6.
9.2 A Brief History of the Cardiff Model
All SFL theory can be said to stem primarily from Halliday’s (1961) most
important and influential article, ‘Categories of the Theory of Grammar’.
Butler (2003a:153) acknowledges, as Fawcett himself does, that there
are more similarities than differences between the two models.
While the Cardiff model is very clearly rooted in Halliday’s early work,
there are points of divergence that have shaped the path leading to the
Cardiff model. Fawcett (2010:93) explains the different pathways as
follows: developments in SFL theory, from early work in the 1960s, were
shaped by Halliday’s involvement with the Penman project (see Matthies-
sen and Bateman 1991) and all the work that Halliday has done since then,
whereas the Cardiff model shares the same roots but diverges slightly,
being influenced by Hudson (1971) and work by Fawcett (1973, 1980).
Fawcett considers Halliday’s 1970 paper, ‘Language as Choice in Social
Contexts’, as resembling most the Cardiff Grammar (Fawcett 2010:93).
The Cardiff model was also shaped by a computational implementation,
the COMMUNAL project (e.g. Fawcett et al. 1993). Butler and Gonzálvez-
García (2014:49) suggest that, because of this, the model ‘offers a high
level of explicitness’. However, as Fawcett states (2008a:13), both models
share ‘the same historical roots and they still share essentially the same
basic concepts’.
232 ANKE SCHULZ AND LISE FONTAINE
Over the course of its development, Halliday’s language model has

developed with an increasing focus on semantic phenomena (see O’Donnell,
this volume). In fact, Fawcett (2010:10) claims that the more recent devel-
opments give no detailed account of the syntax and that the article
‘Categories of the Theory of Grammar’ (Halliday 1961) is the best and
the only account of syntax given by Halliday or his colleagues. Fawcett’s
first series of publications on this topic, ‘Some Proposals for Systemic
Syntax, Parts 1–3’ (1974, 1975, 1976), was developed by trying to apply
Halliday’s model to the analysis of text. The title of Fawcett’s 1980 book,
Cognitive Linguistics and Social Interaction: Towards an Integrated Model of a
Systemic Functional Grammar and the Other Components of an Interacting Mind,
is an explicit indication of the different direction Fawcett was taking in
developing the theory, i.e. that both social interaction and cognition
were important to linguistic modelling.
9.3 Overview of the Model/Features of the Model
The theoretical framework of the Cardiff model has been described in many
different publications but notably in Fawcett (1980, 2000a, 2008a), Tucker
(1998), and Neale (2002), as well as in many articles and book chapters. It
has been and is being developed around the world and in different lan-
guages, for example, in work on Chinese (He 2014), German (Schulz 2008,
2015), and Japanese (Funamoto 2014).
For Fawcett (2000a:34), the basic relationship between meaning and form
in any sign system can be described in terms of realization (i.e. meaning is
realized by form). The relationship between meaning and form is illustrated
in Figure 9.1, where the system networks are a components of the gram-
mar, representing the semantic options available to speakers. The output
of the networks is a set of selection expressions which then becomes the
input to the realization rules; the realization rules and the potential struc-
tures are another component of the grammar, also expressed as potential.2
In this sense, as Butler (2003a:185) explains, ‘The level of form is also
regarded as having a potential, consisting of realisation rules’. The output
from this component is a layer of richly labelled tree structures. As shown
in the diagram in Figure 9.1, there is a loop enabling this process to
continue, where it is possible for a realization rule to state a re-entry rule,
for example, when an element of a unit is ‘filled’ by another unit (this will
be made clear in Section 9.4).
2
The relationship between meaning potential and instance is described by the concept of instantiation, which
operates in a different dimension from that of realization. Although this is an important distinction, it will not be
discussed in this chapter (see e.g. Wegener 2011).
Figure 9.1 The main components of a systemic functional grammar (e.g. Fawcett 2000a:36)
9.3.1 Cognitive and Social Aspects of the Cardiff Model

For Halliday, language is primarily a social semiotic system; speakers use
language to interact with others in culturally determined ways. In the
Cardiff Grammar, although this view is shared, the theory is developed
within a cognitive approach to language. Butler and Gonzálvez-García
(2014) found that both models were very close together as compared to
other theories in the wider functional-cognitive context since they were
readily distinguishable from all other theories in their study, but they also
found that the differences between them were largely due to the more
cognitive orientation of the Cardiff model, e.g. it attempts in some way to
model language production and understanding processes.
9.3.2 The Main Components of the Model

The main components of the model are those that have been developed for
the computational implementation of the theory as part of the COMMUNAL
project. The computational model has been crucial in testing the language
model, and although it is in some way attempting to model human cogni-
tion, it does not claim to provide an accurate description of cognitive
behaviour.
The main components of the generative (language production) aspect of
the model are given in Figure 9.2. This diagram has been described in
differing degrees of detail in the following publications: Fawcett (1980,
1993, 2013) and Fawcett et al. (1993). For the purposes of this presentation,
only the production components are being considered, and so this simpli-
fied diagram leaves out detail concerning the language understanding
components (see Fawcett (2013) for the complete diagram).
The topmost components of the model concern the cognitive and socio-
linguistic processes: the processes required to plan what to say and to guide
the selection of lexical and grammatical units so that the output (e.g.
sentence or clause) matches the speaker’s goals and intentions. The lower
components, specifically the sentence planner, relate most directly to trad-
itional linguistics (e.g. syntax and morphology). With input from the higher
Figure 9.2 Partial view of the main generative components of the COMMUNAL
computer model
components, the sentence planner produces the best formal and semantic
representation.
9.3.2.1 The Overall Planner

The overall planner, the first and highest component, is a kind of ‘overseer’
which directs discourse planning and guides the system as a whole. This is
the component that plans the propositional content in consultation with
the belief system; in other words, this is where the earliest decisions are
made in terms of what is going to be said or what is going to be talked about
for a given proposition. As we can see from Figure 9.2, there is a two-way
flow of information between the two; i.e. a relationship of consultancy
between the overall planner and the belief system. The main difference
between a planning component and the belief system is that the latter is
static (although updatable); it contains objects. Planning components have a
role to play in the entire process of language generation. The planners
consult the belief system for various reasons to assist in the planning
process, i.e. a kind of decision-making.
The belief system is at the heart of both language generation and under-
standing. Although this is not shown here, various components of language
understanding also need to consult the belief system. According to Fawcett
(1994:78), it is object-oriented ‘in the sense that it consists of a vast number
of specific objects and generic objects’. It is not a system in the dynamic
sense of something that is operational; however, it is assumed that the
objects it contains are organized systemically. As Fawcett explains

(1990:164), the belief system includes ‘general and specific beliefs about
(“knowledge of”) situations and things in some domain; specific beliefs
about the content of the preceding discourse, about various aspects of the
current social situation, about the addressee and his beliefs of all types, his
attitudes, his goals and plans’. No theoretical distinction is being made here
between knowledge, belief, and information, including general and specific
beliefs about social situations, about the addressee, about the performer-
addressee relationship. It also contains beliefs about register and the con-
tent of any relevant prior discourse. Finally, it holds ontological relations
(e.g. knowledge about lexical relations).
The output from the first stages of planning is a basic logical form (Lin
1993), which would specify, for example, the type of process involved, how
many objects are involved, and the participant roles involved. For example,
the logical form for a clause such as (1), taken from Fontaine (2008), would
be event(event1[agent=object1, process=work]), which is basically describ-
ing the speaker’s intention to say something about something working.
(1) The chemo pills she takes are working
This basic logical form is then the input into the next stage of planning,
which is called the microplanner. The microplanner is a component that
handles various algorithms that guide the choices (the selection of options)
in the system networks. The system networks represent the networks of
systems of semantic options, not decision trees. This is a very important
component, yet very little attention has been given to this area in SFL. It is
broadly accepted in natural language generation that such a component
is necessary. However, there are still many unanswered questions as to how
it should work and what parts of the generation process belong in the
microplanner and what parts belong elsewhere. The microplanner is simply
a set of algorithms that determine the selection of options in the system
networks; for example, the selection of Theme or verb tense.
9.3.2.2 The Discourse Planner

The output from the microplanner feeds into the Discourse Planner, and, in
through an enriched logical form, it feeds into the component for predeter-
mination rules, which in turn is the input to the system networks in the
lexicogrammar. With the output of the algorithms in the microplanner, it
modifies planning from a higher level into plans that fit the more local
discourse constraints of genre and exchange structure grammars. In a sense,
it is this component that ensures the clause being generated will make sense
or fit in with the ongoing discourse (Lin et al. 1993). The output is a
discourse structure representation which, together with the enriched logical
form, serves as the input into the system networks. At this point then, a
good number of systemic selections have been made, for example, Theme,
time/tense, the semantic requirements of any referring expressions.
9.3.2.3 The Sentence Planner

The final component considered here is the sentence planner. It contains a
component for the lexicogrammar, which in turn contains two further
components: the semantic system networks component and the realization
component (see Figure 9.1). The semantic system networks are first tra-
versed, and the output from this is a semantic representation (a collection
of the semantic features selected, which is called a selection expression).
This in turn results in a set of relevant realization rules. The output of
the realization component is a rich formal representation: a fully labelled
tree diagram representing semantic and formal elements (ignoring here the
actual process of realization as speech). The lexicogrammar is therefore the
component that generates the clause and produces the formal representa-
tion. The role of the system networks is to integrate the various decisions or
selections made in the various higher components, for example, the micro-
planner. It is not a decision-making component but rather an integration
and production mechanism.
9.3.2.4 The System Networks

Both the Sydney and Cardiff system networks rely on the same basic
notation and presentation. However, in the Cardiff model, there is ‘only
one level of networks in the specification of the potential of a language at
the level of meaning, rather than two (semantic, lexicogrammatical) in the
Sydney grammar’ (Butler and Gonzálvez-García 2014:49).
The realization rules are integral to the system network since they
determine the relationship between the semantic options selected by the
speaker and the formal representation realized (as speech or text, etc.).
They are an essential part of the lexicogrammar even if they are rarely
made explicit in most SFL writing. These rules or statements are basically
instructions on how a particular meaning is realized. Tucker (1998:47)
identifies four components of the realization rules:
1. rule number;
2. network feature(s);
3. any conditions on the rules;
4. rule operations.
For example, in the small system network presented in Figure 9.3 below, if
the feature [situation] is selected (as it would need to be in order to
generate a clause), then as Fawcett (2008a:100) explains, the correspond-
ing realization rule is to insert a clause and within the clause to insert a
main verb (see Section 9.4). If the feature [information giver] is selected
then the realization rule specifies that the Subject must be positioned
before the Operator.
Realization rules may be simple or complex. For example, the realization
rule for the feature [thing] in the system network for thing, or referent-as-
thing, is given by Fawcett (1998) as the following:
Figure 9.3 A very small systemic functional grammar for the English clause (Fawcett
2008a:93)
60: thing:
if congruent_thing then ngp,
if minor_relationship_with_thing then pgp.
This example shows the rule number (60), the system feature ([thing]),
and the operation (insert unit, i.e. ‘unit insertion rule’). This rule handles
the difference between examples such as the woman and to the man as
in the woman gave the ticket to the man. In the first case, the woman, rule
60 would insert a nominal group. In the second case, it would insert a
prepositional group for to the man. A more complex rule will have condi-
tions such as in the following example, which covers the realization of the
system feature of [prediction] and [future time reference point] (Fawcett
2008b):
5: prediction or future-trp (time reference point):

if not negative then O < ‘will’,
if negative then O < ‘wo’.
Rule (5) applies when either [prediction] or [future time reference point]
has been selected. It also describes the conditions of realization depend-
ent on whether or not the system feature [negative] has also been
selected. If it has not been selected, the Operator will be expounded by
will, and if it has, the Operator will be expounded by wo. This will be
combined with the realization rule for [negative], which is expressed as
follows (and also applies when the feature [confirmation seeker] is
selected):
21: negative or confirmation-seeker:

O <+ ‘n’t’.
The final realization then for [prediction] and [negative] will be that the
Operator will be expounded by won’t, as in He won’t phone. It is therefore
the features of the system network and the realization rules which create
the formal representation. As Fawcett (2000a:149) explains,
The first stage is the selection of the features in the system network (i.e., the
creation of a selection expression). The realization rules then integrate the
various partial ‘strands of meaning’ that are represented by these features
into a single functional structure.
As the examples above show, the realization component has the function of
integrating the selected semantic features and operationalizing the realiza-
tion of the form.
It is also possible for a realization rule to specify another semantic
feature that must be selected on a future pass or on the same pass
through the network. In the former, a preference is given for a par-
ticular feature on a future pass. This is one way in which preselection
occurs, where a realization rule preselects other features. According to
Tucker (1998:46), ‘The preselection of one feature from a system is
equivalent to stating that in the relevant context there is no systemic
choice, and consequently no choice in meaning or in structural
realization.’
The first pass through the networks deals with the clause ([situation]),
whereas the second pass typically will serve to fill an element of the
clause such as Subject (see Section 9.4 for the concept of filling). Subject
is most commonly filled by a nominal group ([thing]) and the selection of
[thing] would have been preselected from a realization rule determined
by the selection of certain semantic features on the first pass. Preselec-
tion may also occur if the planner has passed down an instruction to
select a particular feature (e.g. validity assessment). Some pass rules
cover relationships between features which can be thought of as a kind
of co-dependency. For example, certain process types are far more likely
to prefer progressive aspect (e.g. Material or action type processes) and
others are far more likely to disprefer progressive aspect (e.g. Mental
processes). In other words, action type processes in the present tense are
more likely to be progressive than are mental type processes (consider
John is bouncing the ball versus John bounces the ball and John sees the bird
versus John is seeing the bird).
9.3.3 Probabilities
The role of the probabilities in the system networks is basically two-fold.
One is that they reflect tendencies or the frequency distribution of the fea-
tures in a given system. The other is that they can be set (for example, by a
realization rule) to preselect (effectively remove the ‘choice’ from the system)
or exclude a particular systemic feature. In the generation of a referring
expression, for example, the selection of the feature [thing] is preset to
100 per cent. Figure 9.4 illustrates a simplified system network for the ‘infor-
mation’ sub-network of the Mood system network, where the semantic fea-
tures are shown together with their associated probabilities and examples of
the realization for each feature. As Fawcett (2008b) explains, ‘Probabilities . . .
change from one context of situation to another, and furthermore . . . they
are often overridden by the requirements of the performer’s current commu-
nicative purposes.’ Therefore, the probability associated with a given seman-
tic feature is not fixed but will change depending on many other variables.
9.3.4 Traversing the System Network

The selection of a given feature may lead to another system (or set of
systems) or it may lead to a realization rule (as illustrated in Figure 9.3
and Figure 9.4). Systems may be entered in parallel, as is the case in
Figure 9.3, where there are four systems in parallel (shown by ‘{‘). In theory,
they are entered simultaneously.
Figure 9.4 A simplified system network for the ‘information’ sub-network of MOOD
(Fawcett 2008a:157)
Table 9.1 Strands of meaning in Cardiff SFG (from

Fawcett 2008a)
expressed in the unit of:
strand of meaning Clause
experiential TRANSITIVITY
CIRCUMSTANCES
CONTROL & DISPOSITION
TIME
logical relations COORDINATION
SUBORDINATION
EXTERNAL LOGICAL RELATIONS
interpersonal MOOD
negativity POLARITY
validity BASIC & AUXILIARY VALIDITY
ADJUNCTIVAL VALIDITY
affective AFFECTIVE ADJUNCTS
thematic SUBJECT THEME
NON-S PR AS MARKED THEME
ADJUNCT THEME & INTEGRATION
ENHANCED THEMES
informational RECOVERABILITY
UNMARKED NEWNESS
CONTRASTIVE NEWNESS
INFORMATION STATUS
These system networks are not detailed enough to illustrate the process
of traversing the network, but the full system network is too large to
present here. Instead, as a brief example, we will consider some of the
semantic features that would be selected in order to generate a simple
clause such as I drove the car (invented example), using the system networks
in the computational implementation of the Cardiff Grammar as presented
in Fawcett, Tucker, and Lin 1993. The first selection is the feature [situation]
(other features are possible, such as [thing], but [situation] is the semantic
feature for the clause). This then leads to eight parallel systems,3 corres-
ponding to the eight major strands of meaning in the Cardiff model
(see Table 9.1): transitivity, time specification, situation coordin-
ation, subordination type, mood, polarity, logical circumstance,
and information focus. Each of these systems presents the semantic
options available within that system, and each will lead to further systems.
Selections are made for each system, and this process is continued until a
terminal feature is reached for all relevant systems.
Then in the transitivity system (see Neale 2002, 2017), a process type
must be selected, and in our case it would be an action process type,
specified as driving. In addition to this and as part of the transitivity system,
subject-theme is selected, which must be considered here in order to
3
Names of major systems are given in small caps, features are given in square brackets.
identify subject-verb agreement. This selects which participant is meant

to be fused or conflated with the Subject element (i.e. I drove the car versus
The car was driven by me) – see conflation in the next section. Further
selections are made here in terms of voice and the number of participants
expressed. The time specification system represents circumstances of
time, but there are none in our example. The situation coordination
system identifies whether the clause will be simplex or complex. Our
example is a simplex. The subordination type is preset to ‘main’ because
this is the first pass of the system networks and what is being generated is
the main clause. In the mood system, a selection is made between [infor-
mation] or [directive]. In this example, [information] would be selected as
this clause is an information giver. In the polarity system, [positive] would be
selected (as opposed to I didn’t drive the car). In the logical circumstance
system, since there are no circumstance roles involved in the generation of
this clause, the [no logical circumstance specified] feature would be
selected. Finally, the information focus system represents the meanings
associated with contrastive newness, and since no element in the clause is
to be marked as contrastively new, the feature [no element marked as
contrastively new] would be selected. This first pass through the system
network results in a selection expression that determines the realization
rules that apply. At this point, what has been generated so far can be
expressed in notation as follows:
P
|Cl ! S/Theme/Agent
M < ‘drive’
C/Affected
P
This notation states that a sentence ( ) is generated and filled with a
clause (Cl). The clause has three elements: Subject (S), Main Verb (M), and
Complement (C). The Subject is conflated with Theme and Agent functions
(S/Theme/Agent), and the Complement is conflated with the role of
Affected (C/Affected). It also states that the Main Verb is expounded by
the item drive (M < ‘drive’). This terminology will be explained in the next
section. The realization rules have been omitted here to save space, but
they would be similar in nature to the examples shown above. Further-
more, there would be a realization rule associated with the verb drive
which would ensure the correct morphology of the verb, given the seman-
tic options selected in the system network. Further passes are necessary
in order to complete the target clause (I drove the car). This involves
re-entering the system network to generate the two participants (I and
the car). This will be done following a realization rule from the transitiv-
ity system which would state that the network must be re-entered prefer-
ring the [thing] feature. This preselection leads to the generation of the
referring expressions which will express the participant roles, following a
similar process.
This brief example has been used to give readers an idea of the steps
involved and the relationship between the system network and realization.
These are the important concepts which underpin the Cardiff approach to
syntax.
9.4 Cardiff Theory of Functional Syntax
It has been claimed by Fawcett (2000a:1) that current systemic functional

theory has no theory of syntax with the exception of his book and Halli-
day’s (1961) ‘Categories of the Theory of Grammar’, which was indeed a
theory of syntax.
This section presents the syntactic framework within the Cardiff model.
According to Fawcett (2007:8), there are four key concepts which contribute
to the syntax of any unit in the Cardiff Grammar. These are
1. the system network that defines the language’s meaning potential,

2. the selection expression of features chosen from it on any one traversal
of it,
3. the realization rules that these trigger, and
4. the structure that is their output in any one instance, consisting of
syntax and items.
In the Cardiff Grammar, the transition from system network to syntax or

formal presentation is seamless in the sense that the output from the
system network is a set of selection expressions together with the relevant
realization rules which determine the formal realization of the linguistic
expression being generated. The structure of the clause in the Cardiff
Grammar differs from the standard SFL description. One particular differ-
ence, as noted by Butler (2003a:158), is that ‘the structural functions Sub-
ject, Complement and Adjunct are regarded, in the Cardiff Grammar, as
elements of syntactic structure, and so do not themselves belong to any
particular metafunction’. Furthermore grammatical units are not organ-
ized by nature in terms of ‘rank’ as they are in the standard model of
grammar in SFL. As Fawcett (2000a:233) explains, ‘The concept of the “rank
scale” has no practical role to play in either the theoretical-generative
version of the Cardiff Grammar or in the text-descriptive version that is
used for analyzing texts’ (see Fontaine and Schönthal, this volume, on the
rank scale). Instead, by abandoning the rank scale, the Cardiff model is able
to prioritize a more probabilistic account of grammatical units through the
concept of ‘filling’, which accounts better for the fluid nature of grammar,
at least in English (see Butler 2002:74).
The main components of the Cardiff Grammar are items, units, and
elements. The term item refers to ‘word base forms and inflectional mor-
phemes’ (Tucker 1998:48n). Unit refers to any grammatical unit which is
composed of more than one component. Unlike the standard SFL approach
Figure 9.5 The basic categories of syntax (Fawcett 2008a:74)
Figure 9.6 The basic relationships of syntax (Fawcett 2008a:75)
to units (see Fontaine and Schönthal, this volume), the Cardiff model
defines them by their internal structure (see ‘type’ vs. ‘class’ in Halliday’s
early writing; also see Berry 1975; Fontaine and Schönthal, this volume),
i.e. the components that constitute the unit. These components are
referred to as functional elements, which form the basic categories of
the functional syntax in the Cardiff Grammar. They are illustrated in
Figure 9.5. The relationships among the various components are shown
in Figure 9.6. The relationships among them are explained by Fawcett
(2008a:76) as follows:
1. a unit is composed of one or more elements;

2. some elements are directly expounded by items;
3. other elements are filled by a unit (or units);
4. any such unit is itself composed of one or more elements;
5. and so on, up to a maximum depth of around seven units (plus or
minus two) until finally
6. the lowest element is expounded by an item.
Therefore, componence is the relationship between a unit and its elements.

Filling describes the relationship between an element and the unit realizing
that element (e.g. Subject is typically filled by a nominal group) (see
Figure 9.5). When an element is not filled by a unit but rather expounded by
an item, the relationship is that of exponence (see Figure 9.6).
Whereas for Halliday and Matthiessen (2004:54) the elements of the
clause are word groups and phrases such as nominal group, verbal group,
prepositional phrase, and adverbial group, in the Cardiff Grammar, the
elements of the clause are functional elements which are either filled by a
unit of structure or expounded by an item. For example, Subject is most
likely to be filled by a nominal group, and the Main Verb is most likely to be
expounded by some lexical verb.
According to Butler (2002:75–6), it is ‘the relationships of filling and
exponence (that) make possible the realization of recursion in the gram-
mar. Recursion is located in the systemic component of the grammar,
leading to realization rules which specify that the system network must
be re-entered.’ It is therefore the re-entry into the system network that
specifies the recursion and specifically filling that captures this process.
There is another important distinction which is inherent in the repre-
sentation of the structure of the clause. This is the role of the formal
structure of the clause. In the Cardiff Grammar, the three main strands
of meaning are integrated into the overall model to such a point that they
have no formal role in the grammar. For Halliday, Subject, for example, is
not an element of structure, although this was its original proposal in
earlier work (notably Halliday 1961). In the standard SFL model, Subject,
Predicator, Complement, and Adjunct are treated as ‘the “secondary” struc-
ture of interpersonal meaning’ (Fawcett 2000b:154).
The full list of the strands of meaning is given in Table 9.1, where each
strand of meaning is associated with the unit which expresses it. The main
elements of the clause are given in Table 9.2. Most of the work done on
these strands of meaning specifically is only available in the form of unpub-
lished reports, but see Huang (2017), who has worked extensively on the-
matic meanings; Neale (2002, 2017) on transitivity; and of course many of
the publications listed in the references by Fawcett and also by Tucker.
Table 9.2 The main elements of the clause in the Cardiff Grammar
Element Example
Subject (S) John, in John drove the car

Operator (O) does, in John does drive the car
Auxiliary Verb, several types (X) been, in John has been driving the car
Infinitive Element (I) to, in John wants to drive the car
Main Verb (M) drove, in John drove the car
Main Verb Extension (MEx) asleep, in John fell asleep
Complement (C) the car, in John drove the car
Adjunct, many types (A) yesterday, in John drove the car yesterday
Linker (&) but, in But John drove the car
Binder (B) because, in Because John drove the car
Formulaic Element (F) Thanks
Let Element (L) let’s, in Let’s eat out
Negator (N) not, in John did not drive the car
The remaining units will now be described briefly in turn. As suggested

above, there is no verbal group in the Cardiff model, since all verbal items are
treated as elements of the clause (see ‘exponence’ and ‘filling’). The reasons
are detailed in Fawcett (2000a, 2000b, 2000c) and discussed in Quiroz (2017).
9.4.1 The Nominal Group

The nominal group has as its components five types of element: deter-
miners (various types), selector, modifier, qualifier, and head. The full list
of the elements of the nominal group, with examples, is as given in Table 9.3
(see also Fontaine 2012, 2017).
The order of occurrence of these elements is more or less fixed as follows,
where ‘&’ is a Linker and is an element of all groups (since all groups can be
conjoined):
(&) (rd) (v) (pd) (qd) (v) (sd or od) (v) (dd) (m)* h (q)*
Elements in parenthesis indicate that the element is optional and elements

marked with an asterisk (*) indicate that the element may be repeated (e.g. a
nominal group may have more than one modifier or qualifier).
One common structure of the nominal group is shown in Figure 9.7. The
configuration here contains the most common elements, although
depending on the type of text, qualifiers are generally quite rare (see e.g.
Quirk et al. 1985; Biber and Clark 2002; Fontaine 2008). As stated above, all
elements are either filled by another unit or they are expounded by an item.
In the case of the example in Figure 9.7, the deictic determiner and the head
are both expounded by an item, the and castle respectively. The modifier and
qualifier are each filled by another unit: the modifier is filled by a quality
group (old), and the qualifier is filled with a prepositional group (in the centre
of the city), as we can see in Figure 9.8.
Table 9.3 Examples taken from Fawcett (2007) of individual elements of the
nominal group [note: element being sampled is in italics; head is
underscored; each example may be composed of more than one element]
Element of the nominal group: Example:
representational determiner (rd) a recording of her voice

selector (v) (always of), in various positions a recording of her voice
partitive determiner (pd) the back of the house
fractionative determiner (fd) half the population
quantifying determiner (qd) two cultures
superlative determiner (sd) the fastest of the runners
ordinative determiner (od) the first of the runners
qualifier-introducing determiner (qid) those of her family who are living here
typic determiner (td) a different brand of oil
deictic determiner (dd) the castle
modifier (m) the old castle
head (h) the old castle
qualifier (q) the old castle in the centre of the city
e.g. the old castle in the centre of the city
Figure 9.7 A common configuration of the nominal group in the Cardiff Grammar
e.g. the old castle in the centre of the city
Figure 9.8 Filling and exponence in the nominal group
We will complete the syntactic representation for this example once we

have described the quality group and the prepositional group.
There is a distinction to be made between determiners, on the one hand,
and modifiers and qualifiers, on the other hand. In very general terms,
determiners function to identify the referent through a process called ‘selec-
tion’, and modifiers and qualifiers function to (sub-)classify the referent
(Fawcett 2007). The treatment of determiners in the Cardiff Grammar has
been worked out in detail, as can be seen from the various types of deter-
miners treated as elements of the nominal group. As Fawcett (2007:8)
explains, ‘Modifiers also serve one (typically) of three broad functions: in
principle, all modifiers can be used to classify the referent, but some simply
depict (see “defining” and “nondefining”) and some are affective (nice, nasty).’
9.4.2 The Prepositional Group

The prepositional group (pgp) is a unit having a preposition as its head
together with two other elements: completive (cv), which is an obligatory
element, and an optional prepositional temperer (pt). The completive is
always filled with either a nominal group or a clause. In Figure 9.9 it is
shown filled with a nominal group. The preposition (as the head of the
group) is typically expounded by an item (e.g. in). An example of a fully
labelled tree diagram of the prepositional group can be seen in Figure 9.10.
9.4.3 The Quality Group

The quality group (qlgp) is perhaps the most developed unit in the Cardiff
Grammar, having been described extensively in Tucker (1998), and argu-
ably more developed as a unit in the Cardiff Grammar than in any other
account of its syntax. This group expresses the syntax and functions associ-
ated with both adjectives and adverbs. It is called the quality group as it
covers the range of qualities used to describe referent things and referent
Key
pgp: Prepositional Group
tp: prepositional temperer
p: preposition
cv: completive
Figure 9.9 Diagram of the structure of the prepositional group
Figure 9.10 Example of the syntax of the prepositional group
Figure 9.11 Full tree diagram showing a nominal group, a quality group, and a
prepositional group
situations. Consequently, it is most commonly found filling the modifier

element in the nominal group (for adjectives), as seen in Figure 9.11, but
also found filling the Complement element in relational clauses or filling
the adjunct element in the clause (for adverbs). The pivotal element of the
quality group is the apex, which is always expounded by an adjective or an
adverb depending on whether the quality group is functioning relative to a
referent thing or a referent situation. The full syntax of the quality group is
illustrated in Figure 9.12.
Examples of quality groups with an adjective expounding the apex are
shown in Table 9.4, and examples of an adverb expounding the apex are
given in Table 9.5.
Table 9.4 Examples of the quality group with an adjective

expounding the apex (from Tucker 1998)
Ex te ta a td f sc
much angrier than ever with him
so clever
politically stupid
strong enough
Table 9.5 Examples of the quality group with an adverb

expounding the apex
Ex te ta A td f sc
very carefully
more carefully than ever
Figure 9.12 Full structure of the quality group

Key:
ex = extent; d = determiners (quality deictic determiner, quality quantifying determiner);
t = temperers (te = emphasizing temperer; ta = adjunctival temperer); a = apex;
td = degree temperer; sc = scope elements (may be repeated [sc1] [sc2]); f = finisher
(can occur either before or after the scope elements).
9.4.4 The Quantity Group

The quantity group expresses the quantities of things, situations, qualities,
and even quantities themselves (Fawcett 2000a:207). The pivotal element of
this group is called amount (am), and in addition to this, there are two other
elements: adjustor and quantity finisher, as shown in Figure 9.13. As with
other groups, these elements may be filled by another unit (often a nominal
group in the case of the quantity group, as in numbers) or expounded by
an item. Examples of the realization of quantity group are illustrated in
Table 9.6.
9.4.5 The Genitive Cluster

The final structure described in this presentation of the Cardiff Grammar is
the genitive cluster. It does not follow the same structure as the groups
discussed above. The most common role of the genitive cluster is to fill the
element of deictic determiner in the nominal group. The generic structure of
the genitive cluster is given in Figure 9.14, with examples given in Table 9.7.
Table 9.6 Examples of the quantity group (from

Fawcett 2000a:307)
ad am qtf
sixty
about sixty
more than I had expected
Table 9.7 Examples of the

realization of the genitive cluster
(from Fawcett 2000a:307)
po g o
my
the man ’s
his own
his very own
Key:
ad = adjustor; am = amount; qtf = quantity finisher
Figure 9.13 The structure of the quantity group
Key:
po = possessor; g = genitive element;
o = own element
Figure 9.14 Diagram showing the generic structure of the deictic determiner
This concludes the description of the various grammatical structures in

the Cardiff Grammar. One final area is left to be discussed, and this is the
view of transitivity as it relates to Participant Roles in the clause with
respect to experiential meaning.
9.4.6 Transitivity in the Cardiff Grammar

Experiential meaning is the strand of meaning which represents events and
objects; it represents the speaker’s experience of the world. One of the main
types of experiential meaning is the system of transitivity. It is this system

that defines processes and participants. An overview of the classification
of main processes and participants is given in Table 9.8 (for a full list of
processes see Neale 2002, 2017).
Although the basic relationships of the syntax developed in the Cardiff
Grammar were given earlier in this section, one relationship has not yet
been discussed, and it is important to do so since it explains how the
various functions come together in a multifunctional view of language.
Specifically, it needs to be made clear how the functions exemplified in
Table 9.8 map onto the structure of the clause as shown in Figure 9.15.
According to Fawcett (2008a:84), conflation is the technical term for the
process whereby elements are fused together. Often conflation occurs
between elements representing meanings from two different strands of
meaning (Fawcett 2008a:84), illustrated in Figure 9.15 by using a trun-
cated version of the example given in Figure 9.5 above, Ivy will wash her
Table 9.8 Main processes and participant roles in the Cardiff Grammar
Process ([Pro]) Participant Roles Example
Action Agent [Ag] I [Ag] bought [Pro] a new game [Af]

Affected [Af]
Relational Carrier [Ca] That [Ca] is [Pro] a great game [At]
Attribute [At] I [Ca] have [Pro] two dogs [Pos]
Possessed [Pos] My mother [Ca] is [Pro] in Canada [Loc]
Location [Loc]
Mental Perceiver [Perc] John [Cog] knows [Pro] the answer [Ph]
Cognizant [Cog] John [Perc] saw [Pro] the moose [Ph]
Emoter [Em] John [Em] loves his new job [Ph]
Phenomenon [Ph]
Environmental (no Participant Roles, only Process) It rained [Pro] today
Figure 9.15 Example of conflation between elements
hair. In this example, the Subject is conflated with Agent and the Comple-
ment is conflated with Affected. Depending on the process type, Subject
will most often be conflated with some Participant Role (although not
always), while Complement is always conflated with a Participant Role.
Adjuncts may be conflated with a Circumstance Role. It is through confla-
tion that the relationship between syntax and semantics is made explicit
in the Cardiff Grammar.
It is also possible for an element to be conflated with more than a single
Participant Role. Both the standard SFL model and Cardiff SFG include
compound Participant Roles where a single element of structure expresses
a compound role in transitivity, although this is not generally accounted for
by conflation as it is in the Cardiff model. An example of this is given in (2),
where one referent expresses a compound role. The key to the abbreviations
is as follows: [S]=Subject, [M]=Main Verb, [C]=Complement, [Ag]=Agent,
[Af]=Affected, [Ca]=Carrier, and [At]=Attribute. Compound roles are hyphen-
ated. Consequently, in (2), Ivan expresses the conflated roles of Affected
and Carrier.
(2) The war [S/Ag] made [M] Ivan [C/Af-Ca] very rich [C/At] (from Fawcett
forthcoming)
9.5 Analysis of the Clause in the Cardiff Grammar
In this section we analyze a simple clause in order to relate some of the

concepts discussed above to clause analysis.
According to Fawcett (2010:305), the clause fills the following elements
typically as follows, with relative frequencies given in parentheses: Sen-
tence (85 per cent); Complement (7 per cent); Adjunct (4 per cent); qualifier
in the nominal group (2 per cent); finisher in the quality group (0.5 per
cent); and very infrequently other elements such as completive in the
prepositional group. It is worth noting that the term ‘sentence’, for Fawcett
(e.g. 1980, 2000a), is not a unit but rather refers to all ‘types of minimal
linguistic elements in a discourse’ (Fawcett 1980:222).
The entry point into unfolding the grammatical analysis of any clause in
SFL is through the verbal items expressing the process. Fawcett (2008a:212)
uses a re-expression test, the process test, to identify the Participant Roles
(PR) in a clause: ‘In this Process of xxx-ing, we expect to find someone/
something xxx-ing (someone/something) ((to or from someone/something/
somewhere))’ (see also Fontaine 2012 for detailed guidance on using this
test). Essentially a PR is a semantic role that is ‘expected’ by the Process.
Any constituent that the Process does not expect is not a PR but an Adjunct.
This approach is very similar to argument structure or verb valency. For
example, if we compare the verbs have, put, and bake as used in examples (3)
to (7), the outcome of the process test would predict the following
transitivity configuration for each: in (3), two PRs, I and a dog (someone has
something); in (4), one PR, I (someone is having a bath); in (5), three PRs, I,
my keys, and in the desk (someone is putting something somewhere); in (6),
two PRs, I and a cake (someone is baking something); and in (7), we might
debate whether the process test can be used definitively to work out the
‘expected’ PRs, since it is not immediately obvious that bake as a process
requires a third PR (see Fawcett 2009). While the test is not perfect, it does
help provide a framework with which to build the analysis. However, the
process test can only be applied if the main verb has been identified (see
steps 3 to 5 below).4
(3) I have a dog

(4) I am having a bath
(5) I put my keys in the desk
(6) I baked a cake
(7) I baked John a cake
The Cardiff model has built up a set of guidelines that help to identify the
elements of a clause (see Fawcett 2008a:208–31). Here, we provide a brief,
slightly adapted version of the guidelines to illustrate the main principles in
the syntactic analysis of the clause.
Guidelines for analyzing clause syntax:
1. It is easiest to begin the analysis with the clause as information-giver,

i.e. in declarative structure. If it is in another form, re-express it for
the purpose of analysis.
2. Check for a Linker [&], e.g. coordinating conjunctions and, or, but or a
Binder [B], i.e. subordinating conjunctions such as if, because, that, when,
if any.
3. Find the Main Verb [M], which is the lexical verb. The Process [Pro] is
always conflated with the Main Verb. There may also be a Main Verb
Extension [MEx], for example, a phrasal verb, e.g. ran_[M] up_[Mex].
4. Use the process test to identify the expected PRs.
5. Check for any Auxiliary Verbs [X] to the left of the Main Verb, i.e.
forms of be, have, or do or one of the modal auxiliaries, e.g. is_[X] eating_
[M], has_[X] eaten_[M], might_[X] eat_[M].
6. If the clause under investigation has an [X], look for an Auxiliary
Extension [XEx], e.g., be_[X] going_[XEx] or Infinitive Element to [I],
e.g., ought_[X] to_[I].
7. Check for an Operator [O]. The Operator is conflated with an auxiliary
verb or lexical forms of be, which have the potential for forming an
4
There is insufficient space here to explain how to identify whether or not the main verb is extended: in e.g. I’m looking
up the answer, is this a process of ‘looking’ or a process of ‘looking up’? See Fawcett (2009; forthcoming).
interrogative clause, e.g. is_[O/X] eating_[M], has_[O/X] eaten_[M], might_

[O/X] eat_[M].
8. Check if there is a Negator [N], i.e. the word not, e.g. might_[O/X] not_[N]
eat_[M].
9. Find the Subject [S]. In a declarative clause, the S precedes the Auxil-
iary and Main Verb. A wh-item may be the Subject, e.g. She_[S] might_
[O/X] not_[N] eat_[M]. If the Subject has been ellipted, note this by
adding (S) to the analysis.
10. Find any Complements [C]. Complements are identified as units that
are expected by the process, given the process test. A wh-item may be a
Complement, e.g. She_[S] ate_[M] the cake_[C]. If the Complement has
been ellipted, note this by adding (C) to the analysis.
11. Check for any Adjuncts [A]. A wh-item may be an Adjunct. Adjuncts
are optional elements of the clause, whereas Subject and Comple-
ments are conflated with participant roles, which are expected (see
the Process test below), depending on the Main verb.
12. Check if there is a Let Element [L], e.g. Let’s_[L] go!
13. Check if there is a Vocative [V], which is used to address people
directly, e.g. Darling_[V], let’s go.
14. Find the Ender [E], e.g. in written text, a full stop, or in spoken text,
rising intonation.
These steps will be used to analyze the clause given in (8), first by establish-
ing the structural configuration of the clause and then by discussing the
full functional account within the Cardiff model.
(8) my Mom would visit Texas from time to time (Source: Ententen13,
SketchEngine)
In our example (see (8) above), the clause is in the declarative structure (step
1); it has no Linker or Binder (step 2); the main verb is visit (step 3); and there
is a modal auxiliary verb would (step 5). Having identified the main verb, the
process test can be applied (step 4), which provides the basic syntax of the
clause; in a process of visiting we expect someone to be visiting someone/
somewhere, i.e. two PRs. Considering steps 6 and 7, we can conclude that
there are not auxiliary extensions since there is only one auxiliary verb in
this clause, and that the auxiliary is conflated with the Operator since it has
the potential to express mood. There is no Negator element (step 8), and we
can identify the subject (step 9) because the Operator (would) would be to the
left of the subject in an interrogative, e.g. Would my mother visit Texas from
time to time? Having done the process test in step 4, we can identify a
Complement, i.e. the non-subject PR, to complete step 10. So far then, we
have a two-participant process with the basic syntax as follows: S O/X M C,
i.e. Subject, Operator/Auxiliary, Main verb, and Complement. There is one
unit unaccounted for, and so in step 11 we assume that from time to time is
an Adjunct. None of the elements listed in steps 12 to 14 apply, therefore
Figure 9.16 Syntax for example (7)
Figure 9.17 Analysis of the syntax and semantics of example (7) in the Cardiff Grammar
(based on Fawcett 2010:148)
Key: Cl = Clause; S = Subject; Ag = Agent; O = Operator; M = Main verb; C =
Complement; Af = affected; A = adjunct.
our final syntax is given in Figure 9.16. While this gives us the basic
structure, we do not yet have a full account of the semantics nor of the
nature of the structural units which fill these elements of the clause. If we
recall the concepts of componence, filling, and exponence as discussed
above (see also Figure 9.6), we can say that this clause is composed of five
elements. The subject, complement, and adjunct elements are filled by
units of structure, in this case by a nominal group for both the subject
and complement and by a prepositional group for the adjunct. The
remaining two elements, operator and main verb, are each expounded, by
a modal auxiliary verb and a lexical verb respectively. However, as shown in
Table 9.1, there are eight strands of meaning yet to be analyzed.
While there is no space here to discuss in detail how the semantic
analysis is handled in the Cardiff model, we will simply briefly highlight
a few important points about it. Figure 9.17 illustrates the analysis of the
syntax and the semantics. In a sense this diagram brings together the key
concepts of the Cardiff model. As Fawcett (2010:148) explains, ‘A clause is
regarded as the realization, in a single, integrated structure, of the various
types of meaning that are modelled in the system networks
of TRANSITIVITY, MOOD, THEME, and so on.’ What is being conflated is not

structures but rather elements, e.g. Agent with Subject. The structural
units serve to express the various meanings, and the conflation of the
various strands of meaning is handled by realization rules; i.e. the realiza-
tion rules are not only structure-building but also have an integration
function which enables the clause, as a single structure, to represent mul-
tiple meanings.
9.6 Summary
In this chapter we have shown how the Cardiff model developed from
Halliday’s early work, diverging from Halliday’s standard model due to
different influences and concerns; however, the core concepts are clearly
shared. The priority given in the Cardiff model to conflating meanings
rather than structures has informed the approach it has developed, and
we have shown in this chapter how that works. In addition to this, the role
of realization is made explicit in the Cardiff model through the relationship
between the system networks, selection expressions, and realization rules
through to the structural representation of these meanings. Here we have
demonstrated how these two key aspects of the model have shaped its
development in terms of how it applies to the English language.
Limits of space have not permitted a more extensive presentation of the
Cardiff model, and that was never the aim of the chapter. We set out to give
some insight into the model and interested readers can continue to read
about it by following up with the various publications referred to in this
chapter.
References
Berry, M. 1975. An Introduction to Systemic Linguistics, Volume 1: Structures and

Biber, D. and V. Clark. 2002. Historical Shifts in Modification Patterns with
Complex Noun Phrase Structures: How Long Can You Go without a Verb?
In T. Fanego, M. J. Lopez-Couso, and J. Perez-Guerra, eds., English Historical
Syntax and Morphology. Amsterdam: John Benjamins. 43–66.
Butler, C. 2002. On Being True to Form: Models of Syntax in Systemic
Functional Grammar. Functions of Language 9(1): 61–86.
Butler, C. 2003a. Structure and Function: An Introduction to Three Major
Structural-functional Theories, Volume 1: Approaches to the Simplex Clause.
Amsterdam: John Benjamins.
Butler, C. 2003b. Structure and Function: An Introduction to Three Major
Structural-functional Theories, Volume 2: From Clause to Discourse and Beyond.
Butler, C. and F. Gonzálvez-García. 2014. Exploring Functional-cognitive Space.

Fawcett, R. 1973. Generating a Sentence in Systemic Functional Grammar.
Reprinted in M. A. K. Halliday and J. R. Martin, eds. 1981. Readings in
Systemic Linguistics. London: Batsford.
Fawcett, R. 1974. Some Proposals for Systemic Syntax, Part 1. MALS Journal
1: 1–15.
2(1): 43–68.
2(2): 35–68.
Fawcett, R. 1980. Cognitive Linguistics and Social Interaction: Towards an Inte-
grated Model of a Systemic Functional Grammar and the Other Components of an
Interacting Mind. Heidelberg: Julius Groos.
Fawcett, R. 1990. The Computer Generation of Speech with Semantically
and Discoursally Motivated Intonation. In Proceedings of 5th International
Workshop on Natural Language Generation, Pittsburgh.
Fawcett, R. 1993. Language as Program: A Reassessment of the Nature of
Descriptive Linguistics. Language Sciences 14(4): 623–57.
Fawcett, R. 1994. On Moving on on Ontologies. In Proceedings of the Seventh
International Workshop on Natural Language Generation. New Mexico State
University, Las Cruces, New Mexico, USA.
Fawcett, R. 1998. System Network for Thing. Version 5. Unpublished manu-
script, Cardiff University.
Fawcett, R. 2000a. A Theory of Syntax for Systemic Functional Linguistics.
Fawcett, R. 2000b. In Place of Halliday’s ‘Verbal Group’, Part 1: Evidence
from the Problems of Halliday’s Representations and the Relative Simpli-
city of the Proposed Alternative. Word 51(2): 157–203.
Fawcett, R. 2000c. In Place of Halliday’s ‘Verbal Group’, Part 2:
Evidence from Generation, Semantics and Interruptability. Word 51(3):
327–75.
Fawcett, R. 2007. Modelling ‘Selection’ between Referents in the English
Nominal Group: An Essay in Scientific Inquiry in Linguistics. In C. S.
Butler, R. Hidalgo Downing, and J. Lavid, eds., Functional Perspectives on
Grammar and Discourse: Papers in Honour of Professor Angela Downing.
Fawcett, R. 2008a. Invitation to Systemic Functional Linguistics: The Cardiff
Grammar as an Extension and Simplification of Halliday’s Systemic Functional
Grammar. Sheffield: Equinox.
Fawcett, R. 2008b. Systemic Functional Grammar as a Formal Model of Language:
A Micro-grammar for Some Central Elements of the English Clause. Unpublished
manuscript, Cardiff University.
Fawcett, R. 2009. Seven Problems to Beware of When Analyzing Processes
and Participant Roles in Texts. In S. Slembrouck, M. Taverniers, and
M. van Herreweghe, eds., From ‘Will’ to ‘Well’: Studies in Linguistics Offered to

Anne-Marie Simon-Vandenbergen. Ghent: Academia Press. 209–24.
Fawcett, R. 2010. A Theory of Syntax for Systemic Functional Linguistics.
Fawcett, R. 2013. Choice and Choosing in Systemic-Functional Grammar. In
L. Fontaine, T. Bartlett, and G. O’Grady, eds., Systemic Functional Linguistics:
Exploring Choice. Cambridge: Cambridge University Press. 115–34.
Fawcett, R. 2017. From Meaning to Form in the Cardiff Model of Language
and its Use. In T. Bartlett and G. O’Grady, eds., The Routledge Handbook
of Systemic Functional Linguistics. London: Routledge. 56–76.
Fawcett, R. forthcoming. Functional Syntax Handbook: Analyzing English at the
Level of Form. Sheffield: Equinox.
Fawcett, R., G. Tucker, and Y. Lin. 1993. How a Systemic Functional
Grammar Works: The Role of Realization in Realization. In H. Horacek
and M. Zock, eds., New Concepts in Natural Language Generation. London:
Pinter. 114–86.
Fontaine, L. 2008. A Systemic Functional Approach to Referring Expressions:
Reconsidering Post-modification in the Nominal Group. PhD Thesis, Cardiff
University.
Fontaine, L. 2012. Analyzing English Grammar: A Systemic-functional Introduc-
Fontaine, L. 2017. The English Nominal Group: The Centrality of the Thing
Element. In T. Bartlett and G. O’Grady, eds., The Routledge Handbook of
Funamoto, H. 2014. The Interface between Culture and Mind: A Systemic
Functional Account of Nominality. 北陸大学紀要第 38 号: 51–80.
241–92.
Arnold.
Halliday, M. A. K. and C. M. I. M. Matthiessen. 2004. An Introduction to
Functional Grammar. 3rd ed. London: Hodder.
He, W. 2014. Bi-functional Constituent Constructions in Modern
Mandarin Chinese: A Cardiff Grammar Approach. Language Sciences 42:
43–59.
Huang, G. 2017. Theme in the Cardiff Grammar. In T. Bartlett and
Hudson, R. 1971. English Complex Sentences: An Introduction to Systemic Gram-
mar. Amsterdam: North-Holland.
Lin, Y. 1993. Aspects of a Linguistic Approach to Logical Form and
Reasoning: COMMUNAL Report No 32. Unpublished manuscript, Cardiff
University.
Lin, Y., R. Fawcett, and B. Davies. 1993. GENEDIS: The Discourse Generator
in COMMUNAL. In A. Sloman, D. Hogg, G. Humphreys, A. Ramsay, and
D. Partridge, eds., Prospects for Artificial Intelligence: Proceedings of AISB 93,
the 9th Biennial Conference of the Society for the Study of Artificial Intelligence
and the Simulation of Behaviour. Amsterdam: IOS Press. 148–57.
Matthiessen, C. M. I. M. and J. Bateman. 1991. Text Generation and
Systemic-functional Linguistics: Experiences from English and Japanese. London:
Pinter.
Neale, A. 2002. More Delicate TRANSITIVITY: Extending the PROCESS TYPE
System Networks for English to Include Full Semantic Classifications. PhD Thesis,
Cardiff University.
Neale, A. 2017. Transitivity in the Cardiff Grammar. In T. Bartlett and
Quirk, R., S. Greenbaum, S. Leech, and J. Svartvik. 1985. A Comprehensive
Grammar of the English Language. London: Longman.
Quiroz, B. 2017. The Verbal Group. In T. Bartlett and G. O’Grady, eds.,
301–18.
Schulz, A. 2008. Tense, Modality and Polarity: The Finite Verbal Group in
English and German Newsgroup Texts. In N. Norgaard, ed., Systemic
Functional Linguistics in Use. Odense: OWPLC 29. 697–716.
Schulz, A. 2015. Me, Myself and I: A Corpus-based, Contrastive Study of English
and German Computer-mediated Communication from a Systemic Functional
Perspective. PhD Thesis, Technische Universität Darmstadt.
Tench, P. 1990. The Roles of Intonation in English Discourse. Bern: Peter Lang.
Tench, P. 1996. The Intonation Systems of English. London: Cassell.
Tench, P. 2017. The Phoneme and Word Phonology in SFL. In T. Bartlett and
Tucker, G. 1998. The Lexicogrammar of Adjectives: A Systemic Functional
Approach to Lexis. London: Cassell.
Tucker, G. 2017. The Adjectival Group. In T. Bartlett and G. O’Grady, eds.,
284–300.
Wegener, R. 2011. Parameters of Context: From Theory to Model and Application.
PhD Thesis, Macquarie University.
10
SFL in Context
Christopher S. Butler
10.1 Introduction: The Aim of This Chapter
This chapter situates SFL in what has been called functional-cognitive space,
a multidimensional space based on a wide range of properties, in which
various functionally and/or cognitively oriented and/or constructionist
approaches to language can be plotted (Butler and Gonzálvez-García 2005,
2014; Gonzálvez-García and Butler 2006). The aim is to highlight significant
similarities and differences between SFL and other functional/cognitive/
constructionist models.1
10.2 The Framework for Analysis
The discussion will be based on the detailed comparison of sixteen functional

and/or cognitive and/or constructionist models presented in Butler and Gon-
zálvez-García (2014). The models in question, in addition to SFL, were Func-
tional Discourse Grammar (henceforth FDG – Hengeveld and Mackenzie);
Role and Reference Grammar (RRG – Van Valin); the work of Talmy Givón,
Interactional Linguistics and its antecedents in Emergent Grammar (EG+ –
Bybee, Ford, Hopper, Thompson); Word Grammar (WG – Hudson); the Col-
umbia School (CS – Diver, Huffman, Davis, and others); Cognitive Grammar
(CG – Langacker); Sign-based Construction Grammar (SBCG – Sag, Michaelis);
Cognitive Construction Grammar (CCG – Goldberg); Embodied Construction
Grammar (ECG – Bergen, Chang); Frame-semantic Construction Grammar
(FSCG – Boas); Radical Construction Grammar (RCG – Croft); the collostruc-
tional approach (CLS – Gries, Stefanowitsch); and the Lexical Constructional
Model (LCM – Mairal Usón and Ruiz de Mendoza Ibáñez). The Parallel
1
Considerations of space permit only a brief summary of this work here: for a detailed argument, backed up by
copious references to the literature, readers are referred to Butler and Gonzálvez-García (2014).
260 CHRISTOPHER S. BUTLER
Architecture (PA) model of Jackendoff was also added, since it to some

extent bridges the gap between formalist and functional/cognitive/construc-
tionist models.2 The Minimalist Program (MP) is also discussed at certain
points in the book, though only for purposes of comparison.
The analysis in Butler and Gonzálvez-García (2014) is based on two kinds
of information source: a questionnaire completed by twenty-nine experts in
the various models, and the authors’ own reading of the relevant literature.
The questionnaire respondents were asked to rate each of fifty-eight features
for their importance in relation to ‘their’ model of language, on an ordinal
scale with four choices: ‘Not true’, ‘True to a limited extent’, ‘True to a
considerable extent’, ‘Completely true and central’. In the discussion of
respondents’ ratings in the sections which follow, ratings of ‘Not true’ or
‘True to a limited extent’ will be referred to as negative, those of ‘True to a
considerable extent’ and ‘Completely true and central’ as positive.
They were also invited to add comments to their ratings. The questionnaire
ratings were analyzed by means of three statistical techniques –
correlation, multidimensional scaling, and hierarchical cluster analysis –
in order to detect similarities and differences between models across the
fifty-eight features. After detailed analysis of these ratings, respondents’
comments, and the relevant literature, the authors allocated a second set
of ratings for each of the models except MP, basing their decisions on all the
evidence available, and giving just a binary (+/-) score for each feature. These
ratings were then analyzed by means of the same statistical techniques.
The questionnaire items were arranged in six groups, according to what
kind of information they were intended to elicit: (i) communication and
motivation; (ii) coverage of the model; (iii) the kinds of data used for
description; (iv) explanatory factors; (v) the form of the grammar; and
(vi) applications of the model. These categories will be used as the frame-
work for the discussion later in the present chapter.
10.3 The Overall Picture
10.3.1 The SFL Questionnaire Responses in Relation to Those

for Other Models
Four questionnaires were completed by experts in SFL: two from prominent
figures working with the Sydney model and two from equally central
scholars working with the Cardiff approach.3 Using a conservative measure
of correlation, the ratings for the two Sydney approach questionnaires
2
For information on these approaches to language see the brief profiles of each in Butler and Gonzálvez-García
(2014: Chapter 2) and the key references given there.
3
The term ‘Sydney approach/model’ is used here as a convenient label for the model of SFL put forward by Halliday
and his colleagues, while ‘Cardiff approach/model’ is used to refer to the model advocated by Fawcett and his
co-workers. My view is that although the two approaches share a large number of aims and assumptions, the Cardiff
model should not be seen simply as a minor offshoot from what some consider to be ‘mainstream’ SFL, but deserves
to be considered on a par with the Sydney model.
SFL in Context 261
Figure 10.1 Two-dimensional multidimensional scaling plot of relationships among the

questionnaire responses
correlated significantly with each other and with those for one of the
Cardiff questionnaires, but not with the ratings for either the second
Cardiff questionnaire or those for any other model.4 Neither did the ratings
for the second Cardiff questionnaire correlate significantly with those for
any other model. Using a less conservative technique, but one which is not
backed up by rigorous significance testing, all four sets of SFL ratings
emerged as correlated with each other, and similarities with the LCM
ratings were also revealed, as well as a negative correlation with one of
the MP questionnaires. These results already suggest that SFL is very dis-
tinctive in relation to the other models studied in our research, and we will
see below that this conclusion is corroborated through other techniques.
In Figure 10.1, taken from Butler and Gonzálvez-García (2014:188), is
shown the map produced by the multidimensional scaling programme
PROXSCAL which forms part of the IBM SPSS Statistics package.5,6 This
analyzes the distances between the sets of questionnaire responses and pro-
duces a visual representation of their relationships. Figure 10.1 shows an
analysis along just two dimensions, though higher numbers of dimensions
4
Lack of space precludes discussion of the methodology used in the statistical analyses. For details, the reader is referred
to Butler and Gonzálvez-García (2014: Chapter 4).
5
I am grateful to John Benjamins Publishing Company for permission to reproduce the diagrams labelled here as
Figures 10.1 to 10.4, taken from Butler and Gonzálvez-García (2014).
6
The questionnaires labelled as EG (Emergent Grammar), UBT (Usage-Based Theory), and EXT (Exemplar Theory) are
those which are brought together as EG+ in the discussion and as EG_PLUS in the final analysis illustrated in Figures 10.3
and 10.4.
Figure 10.2 Hierarchical clustering dendrogram for the questionnaire data
are also possible. It can be seen that the four SFL questionnaires cluster
together in the bottom left quadrant of the plot, with negative values on both
dimensions of the analysis. Dimension 1 is clearly concerned with the extent
to which a model resembles formalist approaches: the questionnaires with
positive values on this dimension (MP2, MP1, FDG3, RRG2, RRG1, SBCG,
FDG1, FDG2, in descending order) are for the most formally oriented models,
while the model with the most strongly negative value (the Emergent Gram-
mar of Hopper: EG) is radically functional, and the others with negative
values, apart from SFL, are cognitively based.7 Dimension 2 of the plot is
harder to interpret linguistically (for further analysis see Butler and Gonzál-
vez-García 2014:189).
7
MP1 and MP2 refer to the two MP questionnaires, FDG1, FDG2, and FDG3 to the three FDG questionnaires, and
so on.
SFL in Context 263
Figure 10.2, taken from Butler and Gonzálvez-García (2014:193), shows

the dendrogram produced by the hierarchical clustering programme avail-
able in IBM SPSS Statistics. It can be seen that the two Sydney SFL question-
naires and the first Cardiff SFL questionnaire cluster quite tightly at the
left-hand end of the diagram, and that this cluster is then joined by the
second Cardiff questionnaire. The whole SFL cluster then forms a very
much looser grouping with a large cluster which itself consists of two parts:
(i) a group of cognitively oriented models and (ii) SBCG, CLS2, and PA. This
supercluster then amalgamates with a small one containing the two CS
questionnaires and EG. The large cluster so formed is quite distinct from
the one containing the questionnaires for MP, FDG, and RRG, towards the
formal end of the spectrum.
10.3.2 The Final Ratings for SFL in Relation to Those

for Other Models
The correlation analyses of the final dichotomous ratings allocated for each
of the sixteen models show that the SFL ratings do not correlate signifi-
cantly with those for any other model, either on the more conservative or
on the less conservative measure. The same is true for WG and CS. Again,
then, SFL turns out to be quite distinctive in its pattern of features.
Figure 10.3, taken from Butler and Gonzálvez-García (2014:476), shows the
two-dimensional multidimensional scaling plot for the final ratings data.
Figure 10.3 Multidimensional scaling plot for the final ratings data
As with the questionnaire ratings, dimension 1 is concerned with the degree

of closeness to or distance from formalist models. The two models with
the most negative ratings on this dimension are FDG and RRG, which are
the most formally inclined of all the functionalist models; the third most
negative model, SBCG, is unlike other construction-based models such as
CCG, ECG, FSCG, and RCG in not being strongly cognitive in nature, and
retains some formalist characteristics. SFL, on the other hand, occupies
the most positive position on dimension 1, followed by WG, which is ultim-
ately derived from an early version of systemic linguistics, though now
differing from SFL in a number of important ways, including its strongly
cognitivist orientation. As with the analysis of the questionnaire responses,
dimension 2 is more difficult to interpret linguistically (for detailed discus-
sion see Butler and Gonzálvez-García 2014:479).
In Figure 10.4, taken from Butler and Gonzálvez-García (2014:481), we
see the hierarchical clustering dendrogram for the final ratings data. This
presents an extremely clear picture of the isolation of SFL from the rest of
the models: it lies totally outside the main groupings. In the following
section I will discuss SFL in relation to the various questionnaire items
and also present the results of our analysis of the other fifteen models for
purposes of comparison.
Figure 10.4 Hierarchical clustering dendrogram for the final ratings
SFL in Context 265
10.4 Analysis of the Six Groups of Questionnaire Items

in Relation to SFL
10.4.1 Communication and Motivation

The first section of the questionnaire was concerned with fundamental
features of functional approaches to language, and here SFL does not stand
out from the other functionally and/or cognitively oriented models.
The four SFL questionnaires agreed that the communicative function of
language is fundamental in accounting for why languages are as they are,
that the language system is not arbitrary or self-contained, but shaped by
external factors, and that grammar (in the narrow sense of syntax) is
also not arbitrary or self-contained, but moulded by its role in the realiza-
tion of meaning. There is good evidence in the literature for these claims
(see e.g. Halliday 1985:xiii, xvii).
10.4.2 Coverage
We now turn to what aspects of the study of language the model is intended
to cover. An important distinction is that between approaches which aim to
offer as full a model as possible of language as a whole, and those which
aim primarily to offer simply a model of grammar in the wide sense, to
include semantics, pragmatics, morphosyntax, and phonology. The four
SFL questionnaires were unanimous in agreeing that the model goes
beyond the grammar, and this is amply borne out by the voluminous
literature on, for example, context, register, and genre, as well as on the
‘appliability’ of descriptions based on theoretical constructs. The only
models which were given a negative rating for this feature in the final
analysis were FDG, RRG, and CS, all of which were deemed to concentrate
largely on the grammar itself.
A related question is whether the model aims to account for all system-
atic phenomena within its chosen domain, or just a ‘core’ as in formalist
approaches. SFL certainly does not restrict itself to a putative ‘core’, and this
is reflected not only in the unanimous questionnaire responses, but also in
the impressive breadth of coverage evidenced even by standard works such
as Halliday and Matthiessen (2014) and Fawcett (2000, 2008). Halliday
(2009:73) comments that ‘SFL places a high value on comprehensiveness
in description’.
A further important distinction between approaches is whether they
build in models of the processes by which speakers and writers produce
and understand language, as well as of the systematic patterns found in
language. This item in the questionnaire resulted in a split between the
SFL respondents, the Sydney approach experts giving a negative response,
the Cardiff ones a positive score. The SFL literature clearly shows that those
working within the Sydney approach do not consider themselves to be
modelling the minds of speakers and hearers, but rather the options
available within a social semiotic framework. On the other hand, the

Cardiff model is indeed concerned with ‘cognitive modelling’ (see Fawcett
in press).
The next two items in the questionnaire were concerned with whether
the areas of categorization and construal are important for the model in
question, the latter referring to the different ways in which a situation may
be conceived and presented. All SFL respondents agreed that both areas are
indeed of great importance for their model, and this claim is amply sup-
ported by the SFL literature. The same final positive rating was given for a
number of other models too, including a sizeable group of cognitive/-
constructionist models (EG+, CG, CCG, ECG, FSCG, RCG, LCM) and also PA.
However, there is an important difference between SFL and the cognitive
models here. Whereas the latter see categorization and construal as cogni-
tive operations, SFL treats them in terms of options in a complex set of
system networks which constitute a social semiotic.8 It is important to
realize that categorization by means of system networks is not confined
to the functioning of language to describe the world around us and within
us, but also extends to interpersonal and textual spheres (see below). One
particularly important aspect of construal in SFL is the phenomenon of
grammatical metaphor (see Halliday and Matthiessen 1999: Chapter 6; 2014:
Chapter 10; also see Taverniers, this volume), through which meanings
which are most straightforwardly expressed in one particular form (e.g. a
clause) can be repackaged in another form (e.g. a nominal group, in the
process of nominalization).9
Also investigated in the questionnaire was the question of whether
particular approaches to language are designed to account for the proper-
ties of whole texts/discourses, as well as of the smaller units which con-
tribute to them. Again, all the SFL respondents gave a positive answer
here. In the discussion of this area in Butler and Gonzálvez-García (2014:
Chapter 6), we divided the original question into two parts, in response to
misunderstandings on the part of some respondents: firstly, whether
particular models account for the structure and functioning of extended
stretches of discourse, this being what the original question was intended
to probe; and secondly, whether a model deals with phenomena which
extend beyond the boundaries of individual clauses and their combin-
ations. In the case of SFL, the answer to both is decidedly positive. SFL is
one of the most textually oriented of all the models covered in our study
(see e.g. Halliday 1994:xxvii). Both the Sydney and the Cardiff variants
build in models of text/discourse structure, and detailed attention is also
given, particularly in the Sydney model, to relationships, such as those of
8
I have argued elsewhere (Butler 2013) that these two orientations are not mutually exclusive, i.e. that both cognitive
and social considerations are important when considering matters of categorization and construal.
9
SFL uses the term ‘nominal group’ in place of the more usual ‘noun phrase’ (also see Fontaine and Schönthal, this
volume).
SFL in Context 267
cohesion, which extend beyond the boundaries of clauses and clause

complexes (see Halliday and Matthiessen 2014: Chapter 9). In the final
analysis of models in Butler and Gonzálvez-García (2014), the only model
other than SFL which was considered to pay detailed attention to extended
stretches of discourse was EG+, while all models except WG, ECG, CLS, and
PA were rated as dealing with phenomena which go beyond the boundar-
ies of clauses and their combinations.
The next two questions in our questionnaire dealt with non-representational
properties of language, and more particularly with interpersonal phenom-
ena such as speech acts and those concerned with the structuring of infor-
mation, for instance, in terms of categories of topic and focus, or theme and
rheme. The responses of the experts were again uniformly positive. One of
the most important dimensions in the architecture of the Sydney SFL model
is that of metafunction, which relates to the manner in which the highly
generalized functions of language in everyday life are actually built into
the structure and functioning of language itself. A basic distinction here
is between the ideational metafunction, which is concerned with how we
make sense of the world outside and inside us, and the interpersonal meta-
function, which deals with how we establish and enact social relationships.
The ideational can be split into two subfunctions: the experiential, to do with
the relationship between language and the world of experience, and the
logical, concerned with logico-semantic relationships such as modification,
coordination, and subordination. The final metafunction is the textual, a set
of resources which enable us to create coherent texts out of ideational
content and the associated interpersonal relations. Importantly, the three
main metafunctions are seen as equally important in the analysis of
linguistic units.
Central to the ideational (more narrowly, experiential) metafunction is
an analysis of transitivity relations in the clause, concerned with the type
of process (e.g. material processes of doing/happening; mental processes of
perception, cognition, and the like; relational processes of being, having,
etc.), the participants in the process, and the attendant circumstances
(Halliday and Matthiessen 2014: Chapter 5). The interpersonal metafunc-
tion deals with mood in the clause (declarative, interrogative, imperative,
exclamative) and with the relationship between mood and the semantic
level of speech function (giving or demanding information or goods
and services), as well as with modality (Halliday and Matthiessen 2014:
Chapter 4; see also Martin 1992: Chapter 2). The textual metafunction is
concerned with the distribution of information in the clause, in terms of
theme (the starting point for the message) and rheme, and also with the
division of information into that which is ‘given’ (assumed to be common to
addressor and addressee) and that which is presented by the addressor as
‘new’ for the addressee (Halliday and Matthiessen 2014: Chapter 3). The
metafunctions are also operative to some degree in the analysis of clause
constituents (groups/phrases).
In the Cardiff model, no fewer than eight major strands of meaning (experi-
ential, interpersonal, thematic, logical relations, polarity, validity assess-
ment, affective, and informational) are recognized at the semantic level
(Fawcett 2008:245).
The final ratings for the questionnaire item concerned with interpersonal
phenomena were equally distributed between positive and negative scores:
for FDG, Givón, EG+, WG, CG, RCG, and LCM, in addition to SFL, this area is
important, while the remaining models pay comparatively little attention
to it. For the importance of information structuring, FDG, RRG, EG+, SBCG,
CCG, LCM, and PA, in addition to SFL, were given positive ratings, the other
models negative ones.
Finally in this section, the questionnaire asked whether detailed atten-
tion was given to the non-discreteness of language. Both Sydney SFL
respondents gave positive answers, as did one of the Cardiff respondents.
The literature shows that Sydney SFL, in particular, recognizes the ‘messi-
ness’ of linguistic categories, but does not deal with it through the concept
of prototype, as in more cognitively oriented models (see Halliday 2003a:2,
28). One important aspect of indeterminacy in SFL is the use of the cline, a
continuum which permits potentially infinite gradation (Halliday
1961:249). This concept is important in, for example, the arrangement of
systems in ordering of delicacy, from quite general ones to the left of the
network to more detailed and specific ones to the right. It is also involved in
the dimension of instantiation, which relates the grammar to other aspects
of language and is modelled as a cline with the whole potential of a
language at one end and specific instances of language use at the other.
A further mechanism for dealing with non-discreteness is to supplement
the typology inherent in system networks with a topological perspective
which operates with the idea of a flexible, multidimensional space in which
categories can be located (Martin and Matthiessen 1991). Added to these
mechanisms is the importance of probability: probabilities are assigned to
choices from systems, and differ with the variety of language under scru-
tiny. This is true for both the Cardiff and the Sydney approaches (see
Fawcett 2008:18–19, 122; Matthiessen 2014). Of the other models investi-
gated in our study, only FDG, RRG, CS, SBCG, and CLS were seen as not
giving detailed attention to non-discreteness. It will be apparent that apart
from SFL, the cognitively oriented models are the ones which pay most
attention to this area. CLS is the exception, because although cognitive in its
orientation, its scope is largely limited to the interplay of grammatical and
lexical phenomena.
10.4.3 The Database for Description

We now move on to a set of nine questions concerned with the kinds of data
used in the models under study. Given that functional and/or cognitive
approaches are very much concerned with language as a communicative
SFL in Context 269
tool, it might be expected that the database for description would include
samples of attested naturally occurring language use. We saw earlier that
SFL pays a great deal of attention to the structuring of texts, and it is
therefore not surprising that all four SFL respondents agreed that samples
of authentic text, such as corpora, text collections, or individual texts, do
indeed form an important part of the data used (see Halliday and Matthies-
sen 2014:51–3, 69–74). In the final analysis of the models, taking into
account the literature as well as questionnaire findings, only RRG, WG,
ECG, and PA were rated negatively for this feature.
The next three questions were concerned with the range of languages
from which data are taken and the approach to language typology favoured
by the model. The SFL respondents’ ratings were mixed for the question
concerned with whether, during the development of the approach, applic-
ability to the whole range of language types found in the world was a major
consideration: one Sydney and one Cardiff respondent gave positive ratings,
the other two ratings being negative. However, all respondents agreed that
data from a range of languages were not normally used in arguing for
specific theoretical constructions, but rather that there was a preference
in SFL for describing single languages in some detail and only then making
comparisons. These opinions are amply confirmed by the SFL literature: see
e.g. Caffarel et al. (2004). In the final table of ratings, only SFL, CS, SBCG,
CLS, and LCM were given negative scores for the importance of applicability
to all language types. Ratings for the other two questions were predictably
totally complementary: FDG, RRG, Givón, EG+, CG, and RCG use data from a
wide range of languages in arguing for particular theoretical constructs,
whereas SFL, WG, CS, SBCG, CCG, ECG, FSCG, CLS, LCM, and PA tend to
describe particular languages in detail before making comparisons.
All SFL experts also agreed that data from extended stretches of discourse
are used as well as individual sentences or utterances, as expected from a
text-oriented model. The only other models which were rated positively in
the final analysis were Givón, EG+, CS, and ECG.
Three out of the four SFL respondents expressed the view that data from
sociolinguistic and sociological studies are not widely used in arguing for
particular constructs, the exception being one of the Cardiff linguists. This
is an interesting result in the light of the fact that Hallidayan linguistics
treats language as a social semiotic, and the Cardiff linguists aim to integrate
cognitive and sociological aspects of language and its use. The clue to this
apparent anomaly was given by one Sydney model respondent, who com-
mented that SFL is interested primarily in social practices as construed in
language, so that social matters are approached largely if not entirely
through the analysis of language itself. Of the sixteen models we investi-
gated, only WG and ECG were ultimately rated as making significant use of
the sociolinguistic and/or sociological literature.
The questionnaire also probed the use of data from different synchronic
varieties (dialects, registers) in arguing for particular theoretical constructs.
Here, all but one of the SFL respondents agreed that such data were of
importance, in conformity with the long history of engagement with lan-
guage varieties in this model (see Matthiessen 2007:538–40, 2009:29–43).
In the final analysis only one other model, WG, was rated positively for this
feature.
The question on the use of diachronic data was uniformly negatively
rated by the SFL informants, and this view is borne out by a study of the
literature. The models which were finally rated positively for this item were
FDG, Givón, EG+, WG, CG, RCG, and CLS.
Finally, in this section, we asked whether data from psycholinguistic and
psychological studies were used in arguing for specific theoretical points or
for providing a principled account for linguistic generalizations. This again
produced uniformly negative ratings from our SFL respondents. We have
seen that Sydney SFL does not see itself as modelling the mind. Cardiff SFL
is indeed concerned with cognitive modelling, but has made very little use
of the psycholinguistic or psychological literature. For discussion of this
issue, readers are referred to Butler (2008, 2013). The models which were
finally rated positively for this feature were FDG, Givón, EG+, WG, CCG,
ECG, CLS, and PA.
10.4.4 Explanation
We come now to a set of fourteen questions which are concerned with
aspects of explanatory connections between language and the factors which
are held to motivate its structure and functioning. These range over matters
related broadly to cognitive motivation, the relationship between language
and usage, iconicity, the relationship between the form of language and
discourse requirements, sociocultural factors, the question of language
universals, and the relationship between language and its acquisition.
We begin with a question which asks whether knowledge of language
(often called ‘competence’) is seen as intimately related to the use of
language (often termed ‘performance’). All four SFL respondents gave posi-
tive answers to this question. However, Hallidayan SFL does not conceptual-
ize the distinction as one between knowledge and use, but rather in terms
of an opposition between the potential afforded by the linguistic system
and the use which speakers and writers make of this potential in actual
texts (see the interview with Halliday in Parret 1974:84–5, reprinted in
Halliday 1978:38). We saw in Section 10.4.2 that this distinction is captured
in the concept of instantiation. In the final analysis of models in Butler and
Gonzálvez-García (2014), all were given a positive rating for this feature.
A further question related to knowledge of language asks whether this is
seen as different from other types of knowledge or not. Our respondents
were split evenly on this question, one Sydney and one Cardiff expert going
each way. The problem here is partly that because Hallidayan SFL thinks of
language not as knowledge but as a complex set of resources available to
SFL in Context 271
users, it is not natural for systemicists to think of language in relation to

other kinds of knowledge. There is, however, one fact which points towards
regarding language as different, namely, the claim, taken from scholars
such as Gerald Edelman and Terrence Deacon, that language has a special
relationship with the brain, and more specifically that the two co-evolved
(Halliday 1995, 2004; for further discussion see Butler 2013). In the final
analysis of models, the only one seen as claiming that linguistic knowledge
is different from other kinds of knowledge, apart from SFL, was RRG.
A further question addresses the issue of whether considerations of
human cognitive mechanisms such as attention, memory, etc. are import-
ant in motivating theoretical claims. Both Sydney respondents gave nega-
tive answers, and one Cardiff response was negative, the other positive.
Even in Cardiff SFL, there is little evidence from the literature for a positive
answer here. The final analysis of models gave negative ratings only to FDG,
SFL, and SBCG.
A related question was whether similarities and differences between
language and other cognitive systems are important. Here, the SFL responses
were uniformly negative. The focus in SFL is firmly on language, sometimes
in relation to other semiotic systems, but not with respect to other cognitive
systems. The other models which were given negative ratings for this
feature were FDG, CS, and SBCG. This feature is one that serves to differen-
tiate SFL from the cognitively oriented models in our study.
There followed two questions concerned with the relationship between
grammar and usage. The first asked whether the model claimed that the
grammar of an individual’s language is distilled from repeated events of
linguistic usage, through the entrenchment of frequently used cognitive
routines. This question produced a clean split between the Sydney model
experts, who responded negatively, and the Cardiff grammarians, who gave
a positive rating. This no doubt reflects the fact, noted earlier, that the
Cardiff model is very much concerned with cognitive modelling, whereas
the Sydney model is not. It is thus not surprising that the mention of
‘entrenchment of frequently used cognitive routines’ may have led to nega-
tive ratings by the Sydney experts, despite the fact that Hallidayan SFL is
very much concerned with the relationship between the linguistic system
and its use. This feature was given a positive overall final rating, with the
proviso that this is stronger for the Cardiff model. The interest in the
relationship between system and use is consistent with the uniformly posi-
tive ratings given for the questionnaire item on the importance of frequency
in establishing an individual’s grammar, as well as in language change (see
Halliday 2003a:27). The models which were rated negatively in the final
analysis for the first of these questions were FDG, RRG, CS, SBCG, and PA,
and for the second question the same list plus LCM. It is striking that these
two features group the more formally inclined models together with CS,
which regards frequency as irrelevant to the establishment of components
of grammatical structure, though important for diachronic change.
Only one of the SFL respondents (Cardiff ) reported that iconicity is an

important motivating factor, and a negative rating for this feature is cer-
tainly supported by perusal of the literature. The models rated as positive in
the final analysis were FDG, Givón, CS, CG, CCG, and RCG.
A further question inquired whether it is claimed that grammars emerge
largely from the requirements of discourse. Again, only one (Cardiff )
respondent gave a positive rating here, but also indicated that if reference
were being made to the Emergent Grammar proposals, the rating would be
lower. In the final analysis only Givón, EG+, and CG were given positive
ratings for this feature.
Both Sydney respondents and one of the Cardiff experts agreed that the
ways in which language is shaped by, and in turn shapes, sociocultural
contexts is a central object of study for SFL, and the one dissenting rating
was explained by the respondent in terms of the fact that as this was a
central concern for the Sydney model, the Cardiff approach had paid less
attention to it. The literature on the relationship between language and
social context in Hallidayan SFL is quite extensive: indeed, we can regard
this as one of the key aspects of the model (see Butler 2003:373–95; Halliday
2007; Halliday and Matthiessen 2014:32–42; Hasan 2005, 2009, 2014). Apart
from SFL, only EG+, WG, and RCG were given positive ratings for this item,
showing that this is clearly a minority concern in functional and cognitive
studies.
Two questions related to diachronic factors: grammaticalization (or
‘grammaticization’), i.e. the historical process by which items which were
originally lexical come to serve a grammatical function, and subjectiviza-
tion, the process through which items come to be used to indicate the
speaker’s/writer’s own position. Uniformly negative ratings were given by
all SFL respondents to both items. There is a very brief discussion of
grammaticalization in Halliday and Matthiessen (2014:67–9). FDG, Givón,
EG+, CG, RCG, and CLS were ultimately rated positively for grammaticaliza-
tion, but only EG+ and CG for subjectivization.
Two further questions were concerned with universals. One asked
whether universals with specific linguistic content are postulated as part
of an innate human endowment. SFL ratings were all negative here (see
Halliday 1994:xxxiii–xxxiv, 2002:12), as were the final ratings for all models
except PA. On the other hand, the SFL responses were uniformly positive for
the item which asked whether innate biological, including cognitive, cap-
acities and predispositions of human beings which are general rather than
specific to language are regarded as the most important factors (together
with exposure to the language) in language acquisition. Such a position is
consistent with SFL work on the stages of language acquisition in relation to
the body, including the brain (Halliday 2004). Only FDG, CS, SBCG, and RCG
were ultimately rated as negative for this item.
Finally in this section, respondents were asked whether the learnability
of the theoretical constructs proposed in the model has been empirically
SFL in Context 273
tested by investigation of language acquisition. We were at first surprised

to find that all SFL experts rated this negatively. There is a considerable
amount of work on first language acquisition within a SFL framework,
which all respondents would have been aware of (see Butler 2003:413–24;
Halliday 2003b; Painter 2009; Torr 2014). However, as one expert com-
mented, this work has not been specifically concerned with testing whether
the constructs of the model are learnable. This position contrasts strongly
with that in, for example, RRG, where there is a good deal of work on the
empirical testing of particular theoretical constructs with regard to the
plausibility of their operation in acquisition. In the final analysis, only
RRG, CCG, ECG, and CLS were rated positively for this feature.
10.4.5 The Form of the Grammar

Here we have a set of twenty questionnaire items which are concerned with
the characteristics of the grammar itself. We begin with the question of
whether the grammar is generative, in the sense that it provides an explicit
set of linked rules, resources, or principles which allow it to provide an
exhaustive description of the structure and interpretation of linguistic
expressions. Both Cardiff respondents and one of the Sydney experts replied
positively. The literature makes it abundantly clear that SFL is indeed
committed, at least in principle, to full generativity, through the inter-
action of systemic choice with realization rules/statements to generate
structures. Of the sixteen models in our study, only Givón, EG+, CS, CG,
and ECG were rated negatively for this feature.
With regard to whether elegance and simplicity are important factors in
designing the grammar, redundancy being avoided wherever possible, all
four respondents answered in the negative. Halliday (1994:xix) states that
SFL ‘is an extravagant theory, not a parsimonious one’. The models rated
positively for this feature in the final analysis were RRG, SBCG, ECG, FSCG,
RCG, CLS, LCM, and PA.
A further question asked whether the model, even if it has various layers
or levels, advocates a monostratal account of the syntax itself, rather than
postulating a separate underlying abstract level of structure which is
related to a more concrete surface level through processes akin to trans-
formations. All four respondents considered this to be the case for SFL: in
the Sydney model the lexicogrammar is related upwards to the semantics,
downwards to the phonology; in the Cardiff model there is a single level of
syntactic structure. In the final analysis all sixteen models were rated
positively for this feature.
Responses to the question of whether empty/invisible categories in the
grammar are strongly dispreferred were particularly interesting in view of
the replies from experts in other models. Three of our four respondents
gave negative ratings here, SFL being the only one of the sixteen models
which was rated negatively for this feature in the final analysis. The overall
negative rating can be explained by the fact, commented on by one respond-

ent, that covert categories (Whorf’s ‘cryptotypes’) play an important part in
SFL. Halliday and Matthiessen (1999:27), who cite Whorf, state that in their
account of ideational semantics ‘there are many examples of cryptotypes in
this sense’.
The following item was included in the questionnaire with SFL specifically
in mind: ‘It is claimed that the paradigmatic options offered by a language
are the central part of the grammar, syntagmatic structures being derived
from these paradigmatic options.’ This is a central tenet of SFL: although
form and meaning are seen as mutually defining, system networks represent
the meaning potential of the language, and so are more abstract than forms,
the latter being derived from systemic choices by means of realization rules/
statements (Halliday 2009:64–5). Unsurprisingly, all four SFL respondents
regarded this claim as absolutely central to the model. Experts in some other
models, however, found the question baffling until it was explained to them,
and one went so far as to admit that s/he did not know ‘what to make of this
Hallidayan jargon’. The central attention given to paradigmatic relations is
indeed a defining feature of SFL: only one other model, CS, was finally rated
positively for this feature, and then with some reservations. In CS, all
grammatical systems are structured in terms of paradigmatic semantic
oppositions which the language user chooses from, and the signal of that
choice is then incorporated into the syntagmatic structure of the utterance.
A further question probed whether the grammar deals with form
(defined as morphosyntax, or grammar in the narrow sense of the term)
at one particular level and with meaning at (an)other level(s). Three of the
four SFL respondents replied in the affirmative, one commenting that this
was absolutely central to the Cardiff model, the architecture of which is
built around the two levels of syntactic form and semantics, related by
means of realization rules. In Sydney SFL the picture is not quite so clear-
cut, since the level of lexicogrammar is already ‘pushed in the direction of the
semantics’ (Halliday 1994:xix). However, in recent years a separate level of
semantics ‘above’ the lexicogrammar has been proposed, the two being
linked in terms of realization (Halliday and Matthiessen 1999, 2014). In
the final analysis of models, a positive rating for this feature was also given
to FDG, RRG, Givón, WG, and PA.
The questionnaire also asked whether the model makes a distinction
between ‘semantic’ and ‘pragmatic’ types of meaning. Here, three respond-
ents gave a negative answer, but even the Cardiff expert who gave a positive
rating conceded that many kinds of meaning which are regarded as prag-
matic in some other models are incorporated into the semantics in SFL.
Halliday and Matthiessen (1999:12) make it clear that there is no separate
component of pragmatics in their model. The models which were rated as
making the distinction were FDG, RRG, Givón, WG, and SBCG.
In response to the question of whether (un)grammaticality is recognized,
all SFL experts gave a negative rating. In the Sydney model no distinction is
SFL in Context 275
made between grammaticality and acceptability (Halliday in Parret

1974:85, reprinted in Halliday 1978:38), whereas the Cardiff approach
regards probability as a more significant concept than grammaticality,
ungrammaticality being seen paradigmatically as the limiting case of the
range of probabilities assigned to choices from systems (see Fawcett
2008:19). SFL and CS were the only models given a negative rating for this
feature.
When asked whether formal (syntactic) patterns have real theoretical
status rather than being just epiphenomena which emerge from the use
of language in discourse, three of the four SFL respondents gave a positive
rating. This question was intended to ascertain whether particular models
espoused the ‘emergentist’ agenda proposed by Paul Hopper in his Emer-
gent Grammar. The literature confirms that no such claims are made in
SFL: indeed, EG+ was the only model to be rated positively for this feature in
the final analysis.
We also asked whether syntactic and lexical phenomena are seen as
forming a continuum rather than being distinct components of the model.
This item received positive ratings from all experts, and is indeed a central
plank of the SFL architecture (see Halliday 2009:73–4). The idea is that the
most general choices, towards the left of a system network, tend to be
realized grammatically, while the more delicate choices, towards the right
of the network, are more likely to be realized lexically (see Halliday and
Matthiessen 1999:5). However, the work of Tucker (1998) has demonstrated
that the view of lexis as ‘most delicate grammar’ is somewhat misleading if
pushed too far. With respect to this feature, the only models to be given a
negative final rating were FDG, RRG, Givón, CCG, and LCM.
The SFL responses were divided along Sydney vs. Cardiff lines for the
importance of collocation, the former giving negative ratings, the latter
positive ones. Collocation was certainly important in Halliday’s work in the
1960s and in work on lexical cohesion in the 1970s (Halliday and Hasan
1976), and although it is true that this area has received less attention from
Sydney model linguists since then, there are important proposals to reinter-
pret collocation as an aspect of wider grammatical patterning (see Martin
1992:309–21; Halliday and Matthiessen 2014:272). A similar approach is
taken by Tucker (2006), operating within the Cardiff model. Also important
in terms of collocation is Tucker’s (1996, 2001, 2007a, 2007b) work on
formulaic language within the Cardiff Grammar.
The next two items in the questionnaire were included with CS specific-
ally in mind, and were concerned with whether a single meaning is postu-
lated for each formal signal whenever possible, and whether a clear
distinction is made between the meaning of a form and the message, or
communicative output, which the form conveys, both being central fea-
tures of the CS model. In both cases, the Sydney model experts gave
negative answers, the Cardiff ones, positive ratings. The comments of
the experts indicated that both claims were truer of the Cardiff than
of the Sydney approach. However, in relation to the second feature, it

should be pointed out that the phenomenon of grammatical metaphor,
which is central to the Sydney enterprise, can be seen precisely in terms
of a distinction between ‘the meaning of a form’ and ‘the message
conveyed’ (see Section 10.4.2). In the final analysis, a positive rating for
the first feature was assigned to SFL, WG, CS, and CG, and for the second
feature to FDG, RRG, SFL, WG, CS, and ECG.
The questionnaire then asked whether the grammar makes central use of
constructions, in the narrow sense of a pairing of a form with a meaning/
function. This item caused considerable difficulty for some respondents to
our questionnaire. One of the Cardiff experts commented that if the defin-
ing characteristic of a construction is the association of a form with a
meaning, then all functional linguists should be considered construction
grammarians, and a similar remark was made by the WG respondent.
What these two experts clearly had in mind was that functional models
do associate meanings with forms. However, the manner in which they do
so is not necessarily that found in construction grammars, which involves
the packaging of form and meaning into a single entity, the construction.
With hindsight, we would have rephrased the definition of construction
as follows: a construction is a stored unit, of varying degrees of complexity,
to which is assigned a package of information, consisting of a form (syntac-
tic and possibly phonological information) and a meaning/function (seman-
tic and possibly pragmatic/discoursal information). In fact, all four SFL
respondents gave a negative rating for this feature.
Finally in this section, respondents were asked whether networks are
central to the model. Predictably, all SFL experts replied in the affirmative,
since system networks are at the heart of the model. All models except FDG,
RRG, and CS were rated positively for this feature, but it is important to
note that other models do not use networks in the same way as SFL: rather,
they are concerned mainly with hierarchies of constructions in
construction-based grammars.
10.4.6 Applications
The final section of the questionnaire consisted of three items relating to
the application of (descriptions derived from) the models. First, we enquired
whether applicability, for example to educational linguistics, stylistics, or
translation studies, is considered to be a major criterion for the success of
the model. Both Sydney model respondents and one Cardiff expert gave
positive ratings. Halliday makes it very clear that for him the value of a
theoretical approach resides in the use which can be made of it by those
with a practical interest in language: SFL is intended to be an ‘appliable’
model, and tries to answer questions asked by people who are not them-
selves linguists, but are nevertheless professionally engaged with language
(Halliday 1985:7, 2002:2, 2009:61). Halliday also sees SFL as a resource for
SFL in Context 277
intervening in language-related social processes. SFL has indeed been used

in a wide variety of applications, including educational linguistics, stylis-
tics, computational linguistics, translation, and clinical linguistics, as well
as in newer areas such as media linguistics, institutional linguistics,
and organizational linguistics (see Matthiessen 2009 for discussion). The
Cardiff model sees description and application as mutually dependent, in
the sense that application of a description can result in improvements to
the description itself – and indeed, to the theory on which it is based. Of the
sixteen models investigated in our study, WG, ECG, FSCG, and LCM were
also rated positively for the importance of applicability. In line with what
was said above, the SFL respondents were unanimous in reporting that
there have been both pedagogical and computational applications: for
surveys of pedagogical work see Christie and Unsworth (2005); Martin and
Rose (2005); Taylor Torsello and Baldry (2005); Williams (2005, 2014);
for accounts of computational work see Halliday (2005); Bateman and
O’Donnell (2014); O’Donnell and Bateman (2005); Teich (2009). For peda-
gogical work, WG, FSCG, and CLS were also rated positively in the final
analysis, while for computational applications RRG, ECG, FSCG, and LCM
also gained positive ratings.
10.5 Conclusions: Summary of Similarities and Differences

between SFL and Other Models
As we saw in Section 10.3, SFL stands out from other models in the way it
patterns overall with respect to the features discussed in detail in Butler
and Gonzálvez-García (2014). In this concluding section, I will highlight
some of the main points of similarity and difference between SFL and other
(groups of ) models.
Let us first turn to the features which SFL shares with most if not all of
the other fifteen models studied in Butler and Gonzálvez-García (2014). Like
all the other models, SFL rejects the autonomy of the language system, aims
in principle to investigate all systematic phenomena rather than just a
‘core’, regards knowledge of language as closely related to use, and offers
a ‘monostratal’ account of linguistic form in the sense that no underlying
formal level, linked to a ‘surface’ level by means of transformation-like
processes, is postulated. Like all other models except SBCG, SFL regards the
communicative function of language as central; like all but PA, it rejects the
autonomy of syntax; in common with all but EG+, it sees syntax as having
real theoretical status rather than being just an epiphenomenon of dis-
course; and similarly to all but PA, it does not postulate universals with
specific linguistic content. Like all but FDG and RRG, it is a model of
language rather than just of the grammar itself, does not make extensive
use of the sociolinguistic or sociological literature, and has little interest in
the process of subjectivization. Similarly to all except Givón, WG, and CLS,
SFL invokes the concept of construal; like all except FDG, RRG, and CS,
it makes use of networks, though in a different way from other models; and
in common with all but Givón, EG+, and CG, it does not claim that the
characteristics of grammars emerge largely from the requirements of dis-
course. Like all models except FDG, RRG, SBCG, and CS, it pays considerable
attention to categorization (though not in cognitive terms); it shares with
all but RRG, WG, ECG, and PA the use of data from attested samples of
language use; like all but FDG, CS, SBCG, and RCG, it recognizes the
importance of innate biological factors in language acquisition; in common
with all but RRG, CCG, ECG, and CLS, it has not tested the learnability, as
such, of the constructs it proposes; and like all but RRG, SBCG, LCM, and PA,
it does not place a high value on elegance and simplicity.
I will now present the features which SFL shares with a maximum of only
four other models, in other words, those for which SFL is either totally
distinctive or in a small group of models which are distinct from the rest.
Only one feature is totally distinctive, namely, the status of being negative
for the dispreferral of empty/invisible categories, due to the importance of
covert categories, or ‘cryptotypes’, in SFL. There are five features which SFL
shares with only one other model: accounting for the properties of whole
texts/discourses (with EG+); the use of data from synchronic varieties (with
WG); the centrality of paradigmatic relations (with CS); rejections of the
claim that linguistic knowledge is no different from other kinds of know-
ledge (with Givón); and the non-recognition of (un)grammaticality (with
CS). There is one feature which SFL shares with just two other models:
rejection of the claim that cognitive mechanisms motivate theoretical
claims (with FDG and SBCG). And there are five features which SFL shares
with three other models: the importance of relationships between language
and sociocultural context (with EG+, WG, and RCG); the postulation of a
single meaning for each formal signal where possible (with WG, CS, and
CG); the importance of pedagogical applications (with EG+, FSCG, and CLS);
the lack of importance of similarities and differences between language
and other cognitive systems (with FDG, CS, and SBCG); and the non-use of
the construction as a packaged pairing of a form with a meaning (with FDG,
Givón, and WG).
It is also of interest here to see which models, within a group of four or
fewer which are distinct from the rest, SFL has most in common with. SFL
shares four features with WG (three positive, one negative) and CS (two
positive, two negative), three with EG+ (all positive) and FDG (all negative),
two with Givón and SBCG (all negative in both cases), and one with RCG, CG,
FSCG, and CLS (all positive). The distribution here is interesting, in that SFL
has links not just with other centrally functional models such as CS and
FDG, but also with what we might describe as functional-cognitive (or
cognitive-functional!) approaches (Givón, EG+, WG), centrally cognitive
models (RCG, CG, FSCG, CLS) and even a non-cognitive, formally oriented
one (SBCG). This pattern is echoed in the more detailed analysis presented
SFL in Context 279
in Section 10.4, where we saw, for example, that SFL shares with Givón, EG
+, WG, CG, RCG, and LCM the property of paying considerable attention to
interpersonal meanings, and with FDG, RRG, EG+, SBCG, CCG, LCM, and PA
a concern with information structuring. The postulation of a continuum
between lexical and grammatical phenomena is also a feature which con-
nects SFL with a range of cognitive and functional-cognitive as well as
centrally functional models. Finally, we should note that there are cases
where SFL shares with cognitively oriented models an interest in a particu-
lar area, but the ways in which it deals with these areas are distinctive.
This is especially true of the linked areas of categorization and construal:
while cognitive models approach these phenomena from the viewpoint of
cognitive operations, SFL deals with them in terms of choices from system
networks which formalize paradigmatic options within a social semiotic
framework.
Work of the kind summarized in the present article demonstrates that
although SFL is distinctive in its pattern of features, there are nevertheless
some interesting links with a range of other approaches. In Butler (2013) it
is suggested that some of these links could be exploited, with advantages to
both sides, if scholars working in SFL and in other functionally and cogni-
tively oriented models paid more attention to each other’s work.
References
Bateman, J. and M. O’Donnell. 2014. Computational Linguistics: The Halli-

day Connection. In J. J. Webster, ed., The Bloomsbury Companion to M. A.
K. Halliday. London: Bloomsbury. 453–66.
Functional Theories, Part 2: From Clause to Discourse and Beyond. Amsterdam:
John Benjamins.
Butler, C. S. 2008. Cognitive Adequacy in Structural-functional Theories of
Language. Language Sciences 30: 1–30.
Butler, C. S. 2013. Systemic Functional Linguistics, Cognitive Linguistics
and Psycholinguistics: Opportunities for Dialogue. Functions of Language
20(2): 182–218.
Butler, C. S. and F. Gonzálvez-García. 2005. Situating FDG in Functional-
cognitive Space: An Initial Study. In J. Lachlan Mackenzie and M. de los
Ángeles Gómez-González, eds., Studies in Functional Discourse Grammar.
Bern: Peter Lang. 109–58.
Butler, C. S. and F. Gonzálvez-García. 2014. Exploring Functional-cognitive
Space. Amsterdam: John Benjamins.
Caffarel, A., J. R. Martin, and C. M. I. M. Matthiessen, eds. 2004. Language
Typology: A Functional Perspective. Amsterdam: John Benjamins.
Christie, F. and L. Unsworth. 2005. Developing Dimensions of an Educa-
tional Linguistics. In R. Hasan, C. M. I. M. Matthiessen, and J. J. Webster,
eds., Continuing Discourse on Language: A Functional Perspective, Volume 1.

Fawcett, R. P. 2000. A Theory of Syntax for Systemic Functional Linguistics.
Fawcett, R. P. 2008. Invitation to Systemic Functional Linguistics through the
Cardiff Grammar: An Extension and Simplification of Halliday’s Systemic Func-
tional Grammar. 3rd ed. Sheffield: Equinox.
Fawcett, R. P. in press. An Integrative Architecture for Systemic Functional
Linguistics and Other Theories of Language. Sheffield: Equinox.
Gonzálvez-García, F. and C. S. Butler. 2006. Mapping Functional-cognitive
Space. Annual Review of Cognitive Linguistics 4: 39–96.
241–92.
Halliday, M. A. K. 1985. An Introduction to Functional Grammar. London:
Edward Arnold.
Halliday, M. A. K. 1994. An Introduction to Functional Grammar. 2nd ed.
London: Edward Arnold.
Halliday, M. A. K. 1995. On Language in Relation to the Evolution of Human
Consciousness. In A. Sture, ed., Of Thoughts and Words: Proceedings of Nobel
Symposium 92: The Relation between Language and Mind. London: Imperial
College Press and the Nobel Foundation. 45–84.
Halliday, M. A. K. 2002. Introduction: A Personal Perspective. In J. J. Web-
ster, ed., On Grammar: Collected Works of M. A. K. Halliday, Volume 1. London:
Continuum. 1–14.
Halliday, M. A. K. 2003a. Introduction: On the ‘Architecture’ of Human
Language. In J. J. Webster, ed., On Language and Linguistics: Collected Works
of M. A. K. Halliday, Volume 3. London: Continuum. 1–29.
Halliday, M. A. K. 2003b. The Language of Early Childhood: Collected Works of
M. A. K. Halliday, Volume 4. Edited by J. J. Webster. London: Continuum.
Halliday, M. A. K. 2004. On Grammar as the Driving Force from Primary to
Higher-order Consciousness. In G. Williams and A. Lukin, eds., The Devel-
opment of Language: Functional Perspectives on Species and Individuals. London:
Continuum. 15–44.
Halliday, M. A. K. 2005. Computational and Quantitative Studies: Collected Works
of M. A. K. Halliday, Volume 6. Edited by J. J. Webster. London: Continuum.
Halliday, M. A. K. 2007. Language and Society: Collected Works of M. A.
K. Halliday, Volume 9. Edited by J. J. Webster. London: Continuum.
Halliday and J. J. Webster, eds., Continuum Companion to Systemic Functional
Linguistics. London: Continuum. 59–86.
SFL in Context 281

Hasan, R. 2005. Language and Society in a Systemic Functional Perspective.
In R. Hasan, C. M. I. M. Matthiessen, and J. J. Webster, eds., Continuing
55–80.
Hasan, R. 2009. The Place of Context in a Systemic Functional Model. In
Hasan, R. 2014. Systemic Functional Linguistics: Halliday and the Evolution
of a Social Semiotic. In J. J. Webster, ed., The Bloomsbury Companion to M. A.
K. Halliday. London: Bloomsbury. 101–36.
Benjamins.
Martin, J. R. and C. M. I. M. Matthiessen. 1991. Systemic Typology and
Topology. In F. Christie, ed., Literacy in Social Processes: Papers from the
Inaugural Systemic Functional Linguistics Conference, Deakin University, January
1990. Darwin: Centre for Studies of Language in Education, Northern
Territory University. 345–83.
Martin, J. R. and D. Rose. 2005. Designing Literacy Pedagogy. In R. Hasan,
C. M. I. M. Matthiessen, and J. J. Webster, eds., Continuing Discourse
on Language: A Functional Perspective, Volume 1. Sheffield: Equinox. 251–80.
Matthiessen, C. M. I. M. 2007. The ‘Architecture’ of Language According to
Systemic Functional Theory: Developments since the 1970s. In
R. Hasan, C. M. I. M. Matthiessen, and J. J. Webster, eds., Continuing
505–61.
Matthiessen, C. M. I. M. 2009. Ideas and New Directions. In M. A. K. Halliday
Matthiessen, C. M. I. M. 2014. Halliday’s Conception of Language as a
O’Donnell, M. and J. Bateman. 2005. SFL in Computational Contexts:
A Contemporary History. In R. Hasan, C. M. I. M. Matthiessen, and J. J.
Webster, eds., Continuing Discourse on Language: A Functional Perspective,
Painter, C. 2009. Language Development. In M. A. K. Halliday and J. J.
Webster, eds., Continuum Companion to Systemic Functional Linguistics.
Parret, H. 1974. Discussing Language. The Hague: Mouton.
Taylor Torsello, C. and A. Baldry. 2005. SFL in Text-based, Web-enhanced
Language Study. In R. Hasan, C. M. I. M. Matthiessen, and J. J. Webster,
eds., Continuing Discourse on Language: A Functional Perspective, Volume 1.
Teich, E. 2009. Linguistic Computing. In M. A. K. Halliday and J. J. Webster,

eds., Continuum Companion to Systemic Functional Linguistics. London:
Continuum. 113–27.
Torr, J. 2014. Language Development in Early Childhood: Learning How to
Mean. In J. J. Webster, ed., The Bloomsbury Companion to M. A. K. Halliday.
London: Bloomsbury. 242–56.
Tucker, G. H. 1996. ‘So Grammarians Haven’t the Faintest Idea: Reconciling
Lexis-oriented and Grammar-oriented Approaches to Language. In
R. Hasan, C. Cloran, and D. G. Butt, eds., Functional Descriptions: Theory in
Practice. Amsterdam: John Benjamins. 145–78.
Tucker, G. H. 1998. The Lexicogrammar of Adjectives: A Systemic-functional
Approach to Lexis. London: Cassell.
Tucker, G. H. 2001. ‘Getting our Heads around It’: Semantic and Syntactic
Tension in the Transitivity Analysis of Metaphorically-derived Multi-word
Verbs. Studi Italiani di Linguistica Teorica e Applicata 30(2): 303–13.
Tucker, G. H. 2006. Systemic Incorporation: On the Relationship
between Corpus and Systemic Functional Grammar. In G. Thompson
and S. Hunston, eds., System and Corpus: Exploring Connections. Sheffield:
Equinox. 81–102.
Tucker, G. H. 2007a. Between Lexis and Grammar: Towards a Systemic
Functional Approach to Phraseology. In R. Hasan, C. M. I. M. Matthiessen,
and J. J. Webster, eds., Continuing Discourse on Language: A Functional
Perspective, Volume 2. Sheffield: Equinox. 954–77.
Tucker, G. H. 2007b. ‘Sorry to Muddy the Waters’: Accounting for Speech
Act Formulae and Formulaic Variation in a Systemic Functional Model
of Language. In C. S. Butler, R. Hidalgo Downing, and J. Lavid, eds.,
Functional Perspectives on Grammar and Discourse: In Honour of Angela
Downing. Amsterdam: John Benjamins. 395–418.
Williams, G. 2005. Grammatics in Schools. In R. Hasan, C. M. I. M.
Matthiessen, and J. J. Webster, eds., Continuing Discourse on Language:
A Functional Perspective, Volume 1. Sheffield: Equinox. 281–310.
Williams, G. 2014. Halliday as an International Educator. In J. J. Webster,
ed., The Bloomsbury Companion to M. A. K. Halliday. London: Bloomsbury.
327–47.
11
Models of Discourse
in Systemic Functional
Linguistics
Tom Bartlett
11.1 What Is Discourse Analysis and How Does It Fit

into a Specific Linguistic Theory?
Ruqaiya Hasan (2009:36) has said of Systemic Functional Linguistics (SFL)

that ‘for a theory that introduces itself as a semiotic one it is woefully
neglectful of specifically sociolinguistic issues; its only substantial contri-
bution is in the field of discourse analysis, where it offers a framework for
the analysis of social context as well as for that of discourse’. This is an
interesting quotation, not least because it appears to suggest that offering a
framework for the analysis of discourse is a relatively tangential aspect of
sociolinguistic issues, while there is also a suggestion that the model for
the analysis of discourse and the model for the analysis of social context
are to date rather more separate than might be desired in ‘a theory that
introduces itself as a semiotic one’. In this chapter, I will discuss the
relationship between models of discourse and models of context in SFL as
a specific social semiotic theory of language, first through some introduc-
tory comments, but primarily in tracing the history of discourse analytical
models in SFL. The history I present will not be entirely chronological,
as transitions in different areas overlap, though it will attempt to capture –
or maybe impose – an overall trajectory of development. But this should not
be taken to suggest that discourse modelling in SFL has developed in linear
and telic fashion, driven entirely by its own internal logic. Different models
and alterations to existing models have arisen in response to a variety of
internal and external pressures – existing work in the field, the demands of
a particular research project, trends within the wider academy – and
different models were, and are, possible.
Many thanks to Jim Martin for comments on an earlier version, including help with references to Halliday and
Hasan’s early work and clarification of his own work and viewpoint. I have incorporated several of Jim’s points in the
chapter but have not been able to cover them all.
286 TOM BARTLETT
As a social semiotic theory, SFL models the relationship between lan-

guage and social activity as more than just one of representation
whereby language is used to verbalize pre-existing concepts or social
relations; rather, language and society have evolved together and are to
a very great extent one and the same thing. As Halliday (1978:4) has said,
‘Language is as it is because of the functions it has evolved to serve in
people’s lives’; to which can be added, and society is as it is because
of the social functions language has enabled it to perform. In this
co-evolutionary view, language does not just mirror society or act as an
intermediary between social life and the individual mind; from an SFL
perspective, to think is to mean and to act is, to a greater or lesser extent,
to language.
There are three aspects to this co-evolution of language and society:
ontogenesis, logogenesis, and phylogenesis (Martin 1999:49), which work
together in a complex chicken-and-egg fashion. Ontogenesis is the develop-
ment of language as a social phenomenon in the individual child, the
process of ‘learning how to mean’ (Halliday 1975), the internalization of
joint activity in the formation of the individual mind;1 logogenesis is the
development of meanings in ongoing discourse, the constant negotiation
and recalibration of experiential and interpersonal understandings in viva
voce; while phylogenesis is the process by which the language system as an
abstract whole is continually perturbed by such ongoing logogenesis and
the effects of this on the ontogenesis of the mass of individuals who, as a
distributed system, comprise society.
At the most basic level, then, discourse analysis is concerned with
logogenesis, with the creation of texts (written or spoken), the progres-
sion of meanings (both experiential and interpersonal) that they display,
and the textual devices which render them cohesive and which signal
their progression. Clearly, though, from a social semiotic perspective, a
model of discourse should also have a phylogenetic perspective, how
discourses (and hence society) change over time, as well as an ontogen-
etic perspective, how ongoing discourses, and the set of discourses in any
one period and place, socialize individuals within collectives; and to
complete the chicken-and-egg cycle, a theory of discourse needs to
account for how distinct experiences of socialization, in the individual
and in collectives of various forms, motivate the dynamics of specific
texts in context, the registers and codes that are abstracted from these
individual instances, and hence society as a meaning-making system.
From this perspective we can see why Hasan, as quoted above, hinted
at the shortcomings of a model of discourse analysis that is not fully
1
The mind is a contentious term in SFL, but, I think, unnecessarily so. Anxious to avoid a dualistic approach separating the
mind and the body, many authors have avoided the word and limited discussion to the brain and neural activity;
however, following Vygotsky (1978), seeing the mind as the socialized and individualized brain, far from being the
oxymoron it first appears, is a crucial conceptual pairing from an SFL perspective.
Models of Discourse in SFL 287
integrated with a model of context, just as for SFL a model of language as

a semiotic system divorced from the social is a theoretical absurdity. In
its attempts to theorize and capture the relationship between instance
and system, however, SFL has largely focused on the midway point of the
cline, modelling the relationship between broad variations in language
use within and between different sectors of society (codes) and according
to use (register) rather than on the interpretation of individual texts as
more narrowly contextualized discourse. Recently, however, Martin
(2010:16–22; also see Andersen et al. 2015:62–3) has called for a balan-
cing of this more phylogenetic perspective, a focus on variations in the
system, with a more logogenetic perspective focusing on the individual
instances of language that are both derived from and in turn perturb the
overall potential.
What unites the top-down systems approach and the bottom-up instance
approach is the fundamental SFL view of society as a semiotic construct
rather than as a linguistically mediated reality, with the corollary that a full
account of semiotic potential and a full account of context are two sides
of the same coin. From the top-down perspective, the social parameters of
context are built into the system, and linguistic behaviour is seen as both a
reflex and a construal of this variation; from the bottom-up view, context
can be read off an individual text through an analysis of the ideational,
interpersonal, and textual meanings construed within it. In its most extreme
form, the latter perspective claims that context is entirely construed by
language and other semiotic systems, and as a consequence a full account
of text is by definition an account of the context. This is the supervenient
model of context (Martin in Andersen et al. 2015:49), as opposed to the
circumvenient model of context in which, in its least refined form, the
context pre-exists and determines language. To interpret this opposition in
the metalanguage of SFL, we could say that, in the supervenient model,
language and context are in a Token/Value relationship – language represents,
encapsulates, or even is the context; whereas from the circumvenient perspec-
tive, we have a Carrier/Circumstantial Attribute relationship – language
is because of context.
In the following sections I will briefly outline various models of discourse
that have been developed within the SFL tradition over the last fifty years.
There is no room to provide a complete overview of any of these models,
but references are given to works where such an overview is provided;
the aim of the current chapter is rather to outline different models of
discourse analysis in SFL and the relationship between language and con-
text built into the various models. In addition to considering the orientation
of each model towards the circumvenient/supervenient distinction, I shall
also consider a third possibility, which I label scalar supervenience, an
approach which largely adopts the supervenient perspective, in which
language construes context, but insists on asking, ‘Why this context here
and now?’ (see also Bartlett 2013, 2017).
288 TOM BARTLETT
11.2 Cohesion and Text Analysis
In the beginning, or very nearly in the beginning, was Cohesion in English

(Halliday and Hasan 1976). Cohesion is the topic of a separate chapter in
this handbook (see Taboada, this volume), so I will not go into it here any
more than is necessary to illustrate the origins of discourse analysis in SFL
and to set up some of the themes to be explored in later sections. I say
Cohesion in English was ‘very nearly in the beginning’ as it grew out of earlier
papers by Michael Halliday on descriptive linguistics in literary studies (e.g.
Halliday 1964) and Ruqaiya Hasan’s work for the Nuffield Foundation in the
1960s. Some eight or nine years before Cohesion in English, Halliday
(c.1967–8) produced a paper entitled Cohesion for a now largely lost series
of Study Papers written for teachers taking part in the Schools Council
Programme in Linguistics and Language Teaching.2 The paper opens with
the following paragraph:
‘Cohesion’ refers to the way in which sentences of a text hang together. An

awareness of cohesion is one of the basic kinds of intuitive knowledge that
everyone has about their native language once they have mastered it. The
mature person can tell, if faced with a set of sentences in speech or in
writing, whether they make up a coherent whole or not. Such a coherent
whole is what is referred to in linguistics as a TEXT (spoken or written): we
could say therefore that the mature person is aware of ‘texture’ in his
language.
There are several points to note here, most notably the correlation between
cohesion and a text as a coherent unit and the idea that cohesion is ‘the way
in which sentences hang together’. In the following sections I will show
how this basic conception has been extended and elaborated in the work of
Hasan and others but rejected as a ‘grammar and glue approach’ by Martin
(Martin in Andersen et al. 2015:53).3
Developing the sketch set out in Halliday’s Study Paper, in Cohesion in
English Halliday and Hasan (1976) recognize four types of grammatical
cohesion as well as lexical cohesion between sentences (or better, between
clauses). The grammatical relations are reference, substitution, ellipsis, and
conjunction, and lexical relations include repetition, synonymy, and logical
relations such as hyponymy and meronymy, as well as collocational associ-
ation. It is the interweaving of these ties across stretches of language that
creates texture (as opposed to structure, see Sections 11.3.2 and 11.3.4) and
which are at the heart of the juggling act between continuity and develop-
ment that defines textuality in general and which charts the logogenetic
2
Halliday made this paper available for circulation, and my thanks to Martin Davies for sending me an electronic copy
of his typed-up version which also includes some of his own historical annotations.
3
Martin (personal communication) sees Hasan’s approach as a bridge between Halliday’s ‘grammar and glue’
approach and his own Discourse Semantics.
journey of specific texts. However, at this level of description the properties

described are entirely text-internal, and the connection to social life
remains to be uncovered. Widdowson (2007:8) draws the following distinc-
tion which, while not formulated within a specifically SFL framework, is a
generally applicable statement:
Unless it is activated by [a] contextual connection, the text is inert. . . .

Discourse in this view is the pragmatic process of meaning negotiation. Text
is its product.4
It would therefore be fair to say that Cohesion in English provides a model of

text analysis which is largely if not entirely divorced from the social condi-
tions of its production and uptake, rather than a discourse analysis, which
has to take such factors into account in some way or other; however, over the
next forty years Hasan was to incorporate the ideas developed in Cohesion in
English into a broader model of text and context.
11.3 The Hasanian Tradition: Texture, Structure, and the

Parameters of Context
Over a long and distinguished career, Ruqaiya Hasan worked in assiduous

detail on a variety of issues concerning the relationship between language and
society, and it is only in piecing her voluminous writings together that a full
discourse analytical model can be constructed.5 I can only provide the briefest
sketch of this work here along with an outline of how it all hangs together.
11.3.1 Message Semantics

Within SFL theory (or at least in the approaches of Hasan and Martin), a
distinction is made between the semantics and lexicogrammar as two dis-
tinct strata within the meaning-making potential of the language system.
Thus, the lexicogrammatical stratum comprises meaningful oppositions
such as declarative, interrogative, and imperative within the system of
MOOD, with the distinctions realized through the inclusion and sequencing
of the Subject and Finite elements. These three MOODS are said to have their
own meanings which remain constant through the different uses to which
they are put in terms of the semantics, which is a higher, more abstract level
combining the building blocks of the lexicogrammar to create the meanings
we exchange in discourse. Thus, within the semantic stratum the discourse
function of demand information (what might be called a ‘question’ in lay
4
The word pragmatic might raise a few SFL hackles here, but I think it can be glossed for the more context-appropriate
‘socially situated’ in this instance. I have deliberately cited Widdowson as a friendly critic of SFL whose observations
need to be addressed, even if not accepted.
5
See Hasan (2016), which brings many of the most significant papers together, and Lukin (2015) for a very useful
overview.
290 TOM BARTLETT
metalanguage) is congruently but not always realized by an interrogative at

the lexicogrammatical stratum. Hasan et al. (2007:713), for example, present
twelve discursively contrastive ways of demanding information, such as ask,
check, and probe, which each combine elements of clause grammar and
intonation in their own specific way. Similar networks (theoretically) exist
for both ideational and textual meaning, and in this way every message (the
smallest semantic unit capable of realizing a move in discourse, congruently
realized by a clause in the lexicogrammar) can be described in terms of the
three metafunctions separately and in conjunction (though it is the inter-
personal aspects of message semantics that have been by far the most
developed). The semantics of a particular text, as distributed across the
messages which realize it, represent that text’s register, which is defined as
variation according to use (Halliday and Hasan 1989:41) and hence a bridge
between text and context. I will pick up on the relationship between context
and register in Section 11.3.3 after taking a brief excursus to provide one
more piece for the discursive jigsaw puzzle. For an example of discourse
analysis based on message semantics, see Lukin (2012).
11.3.2 Cohesive Harmony

In the extract from Halliday’s Study Paper, reference was made to the
assumed correlation between cohesion (as a linguistic artefact of a text
in isolation6) and coherence (as a relationship between text and reader).
However, it is well accepted that cohesion per se is not a guarantee of
coherence, while non-cohesive texts can indeed be coherent thanks to the
bridging assumptions brought to bear by a reader’s real-world knowledge
and expectations. Starting from this ambiguous position, Hasan (1984) set
out to find a linguistic measure for extra-linguistic coherence in terms of a
more elaborated model of textual cohesion, initially as a means of investi-
gating the evaluation of children’s writing in primary schools. The model
she developed was based on the concept of cohesive harmony (see Martin
1992:417–33 for an introduction), a kind of metacohesion based not only
on the ratio of relevant tokens (lexical items entering into lexical strings) to
peripheral tokens (those which do not), but also and more tellingly on the
ratio of central tokens (relevant tokens which interact with each other in
repeated transitivity relations) to non-central ones (but see Martin 1992:370
for a critique of Hasan’s reliance on clause-rank transitivity relations,
which I will return to below). In a paper testing the correlation between
cohesive harmony and perceived cohesion using techniques from psycho-
therapy, Butt et al. (2007:263) conclude:
6
This is of course not quite true, as cohesion is necessarily dependent not only on the reader of a text having certain
linguistic capabilities, but also on a degree of real-world knowledge (e.g. it is necessary to know that ‘house’ and
‘kitchen’ are in a whole-part relationship, and so are superordinate and meronym respectively in linguistic terms, in
order to identify the cohesive link).
While cohesive devices do not in themselves, or in their aggregate, corres-

pond to the degree of coherence in a text, the measure of ‘cohesive harmony’
developed by Hasan (1984) does constitute an index of the ‘textness’ of
extended discourse, including interactive exchanges.
Taken together, Hasan’s message semantics, her approach to cohesive

harmony, and the general concept of cohesion represent all three metafunc-
tions, at least in terms of textual relations. It is through the concept of
register, to which we now turn, that Hasan correlates such textual features
to features of social organization and differentiation. See Scott (2010) for a
diachronic study of war reportage using cohesive harmony analysis.
11.3.3 Context and Register

Moving further up in the architecture of SFL, just as message semantics are
realized by a configuration of features from the lexicogrammar, so a spe-
cific context of situation is typified by the meanings that are at risk within it.
That is to say that out of all the systemic options available at the stratum of
semantics, a specific context will activate particular subsets, and, con-
versely, the use of such subsets of meaning is said to construe such a
context. The subset of meanings at risk within, or strongly associated with,
a particular context of situation is referred to as register, which is variation
according to use, as opposed to dialect, which is variation according to user.
Register is therefore a characteristic of whole texts rather than individual
messages, and while each individual message combines ideational, inter-
personal, and textual meanings to its own ends, at the textual level patterns
within these areas of meaning can be discerned which between them are a
function of the field (the nature of the activity), tenor (the personal relations),
and mode (rhetorical function and means of transmission) of the context of
situation. A strong but not absolute correlation is posited between field and
ideational semantics, tenor and interpersonal semantics, and mode and
textual semantics. This is known as the Context-Metafunction Hook-Up
Hypothesis or, more recently (Hasan 2014), as the Context-Metafunction
Resonance Hypothesis. A more nuanced version of the relationship between
context and semantics is Hasan’s (1995:231) concept of Contextual Config-
uration, the notion that a specific combination of the variables of field,
tenor, and mode is sufficient to describe the context of situation such that
correlations can be established between the context so defined and the
language in use, the register. For an example of discourse analysis based
on contextual and registerial variables, see Lukin et al. (2011).
11.3.4 Generic Structure Potential

The patterning within specific registers of ideational, interpersonal, and
textual meanings is what gives a particular text its texture, as suggested
above. But, following Hasan (in Halliday and Hasan 1989:52), texture is only
292 TOM BARTLETT
one form of patterning, the other being the overall structure of the text in
terms of the appearance and sequencing of the different stages, compulsory
or optional, that it moves through in order to achieve its social purpose. For
Hasan (1996:53) the GSP represents ‘the total potential of structures for a
genre, while the schematic structure of any one instance of the G would
represent a particular configuration permitted by the GSP itself’. For example,
Hasan (1996:54) suggests the GSP in Figure 11.1 for the Nursery Tale genre.
Figure 11.1 GSP of a Fairy Tale (Hasan 1996:54)
In this notation, the caret (^) represents obligatory sequencing of elem-

ents (if both are present), while the raised dot (•) means that the elements
on either side can appear in either order. Elements in brackets are optional,
while those in angle brackets (< >) can be included or interspersed in the
following element. The superscript arrow signals that the element can be
repeated indefinitely. Square brackets indicate the limits of any rule so that
in the above notation the Placement (if present) can either precede the
Initiating Event or be included within it (and no later element).
The terms Initiating Event and Sequent Event in the GSP sketched out above
may seem rather vague, but they are only as vague as is necessary to
capture the range of variation possible at this point in superordinate terms,
and this apparent vagueness is overcome in two possible ways. Firstly, in
different genres the different elements may be specified with greater preci-
sion so as to reflect the rather more constrained meaning potential at that
point, so that GSP of a shopping transaction, for example, would appear as
in Figure 11.2 (Hasan 1996:56).
Figure 11.2 GSP of a Shopping Transaction (Hasan 1996:56)
This GSP has the added complication that while Sale Request and Sale
Compliance necessarily appear in that order, they interact as a whole unit
with other elements of structure (as signalled by the curly brackets).
Secondly, in all cases the meaning potential for each stage is specified in
terms of ‘its crucial semantic attributes’ (Hasan 1996:58) and from there to
the range of lexicogrammatical patterns which potentially realize these.
Returning to the Nursery Tale, one of the crucial semantic attributes of the
element Placement is ‘person particularisation’, which will be realized
lexicogrammatically by ‘indefinite modification . . . of an animate/quasi
animate noun as Thing’ (Hasan 1996:62) – in other words, the Placement
will include a nominal group such as three little pigs or a beautiful princess.
The full description of a genre thus includes both the structure potential
of the necessary and optional elements and the further specification of
the semantic and lexicogrammatical attributes of these. And with regard
to the context of situation which these genres realize, it can be stated that
it is the least delicate options within the context that generate the overall
GSP, while more delicate features will generate the semantic details of
each element and hence the cohesive texture that runs throughout the
discrete elements like the coloured threads in a tapestry. Thus, the context-
ual features shopkeeper and customer in a transaction would generate the
structure in Figure 11.2, which would hold for all sales transactions
(as generically defined), while the inclusion of further details within the
contextual variable of field might specify that there will be continued
reference to clothes, fabrics, sizes, and colouring across the text, and more
delicate features within the contextual variable of tenor, such as familiarity
or status, would generate more specific ways of making enquiries and
requests. In this way the realization relationship between context and text
is able to capture both important generalizations across instances of a
genre type and the specific features of individual instances of that genre.
For a fuller discussion of texture and structure, see Halliday and Hasan
(1989: Part B).
One potential problem with this approach is Hasan’s (1995:219) con-
cept of relevant context, glossed as those aspects of the non-linguistic
environment that are made relevant through language. In this formula-
tion there is by definition a correlation between text and context, and
this has led to discussions as to whether this pairing is the essence of a
supervenient and non-essentialist conception of context or a circularity
which ignores, or at least marginalizes, how extra-textual features affect
text in less visible ways than direct inscription.7 I will return to these
issues in Section 11.7, but see Bartlett (2017) and Moore (2017) for
exchanges on this point.
11.3.5 Rhetorical Units

At roughly the same time as research into the language of repeated and
recognizable social activities was motivating the development of the con-
cepts of register and genre, a rather different approach to the analysis of
units above the clause was emerging from the applied work of Hasan
(1989) and her collaborator Carmel Cloran (1994, 2010), focusing on regu-
lar variation in semantic patterning in mother–child talk in families from
different socioeconomic backgrounds. This research was firmly rooted in
Bernstein’s work on socialization and sought answers to a specific social
problem (see my point in Section 11.1 regarding the contingent nature of
7
Recent work on multimodality opens this question up a little, as the borderline between material context and
semiotic system becomes increasingly fuzzy. This is an area which would merit further theoretical discussion.
294 TOM BARTLETT
developments): why children from working-class families were less suc-

cessful at school than their middle-class counterparts. Hasan and Cloran’s
work recategorized the two socioeconomic groups as low autonomy pro-
fessions (LAP) and high autonomy professions (HAP) according to their
degree of self-regulation in the workplace, and Cloran’s research con-
sidered the degree to which talk within the different families was either
directly related to the immediate task in hand or moved more freely
across a range of less immediate or even hypothetical topics. This categor-
ization was based on Bernstein’s much misunderstood notion of restricted
and elaborated codes and was seen as a cline from more to less heavily
contextualized language. The different categories were based on the
degree of remove of the talk from the material context in terms of both
space and time, and stretches of talk representing each category, or
Rhetorical Unit (RU), could extend for an indefinite number of clauses.
Rhetorical Units are thus seen as textual units comprising one or more
messages (see Section 11.3.1). At the most contextualized (ancillary) end of
the cline of contextualization are Actions,8 talk which regulates the
ongoing activity of the participants, and Commentaries, talk which
describes this activity, as in examples (1) and (2) respectively:
(1) Here, you sit in Nana’s seat.

(2) Do you want some passionfruit?
At the most decontextualized (constitutive) end of the cline are Generaliza-

tions and Conjectures, as in examples (3) and (4) respectively:
(3) Passion fruits usually come when it’s warm.

(4) You might fall over, if you do, and spill my tea.
The full schema of RU categories and an indication of their place on the

cline of contextualization are shown in Table 11.1 and Figure 11.3.
Based on this categorization, Cloran (2010) concludes that families from
LAP backgrounds tended to use more contextualized language than their
HAP counterparts. Taking the analysis a step further, and drawing on
Halliday and Hasan’s concept of a text as a cohesive stretch of talk or
writing of any length, Cloran (2010) goes on to describe the linkage between
RUs not only in terms of their semantic continuity, but also in terms of
whether new RUs are either embedded in the previous ones or expansions
of them. The distinction depends on whether the semantic content of a
previous RU is picked up on in the Theme or Rheme of the first clause of the
new RU. If it is picked up in the Theme, the relationship is said to be one of
embedding, as the second RU would seem to serve a function within the
overall purpose of the matrix RU, as in example (5).
8
Note the capital letters to show that these are technical metalinguistic terms, the names of which refer to similar lay
metalinguistic categories.
Table 11.1 Classification of rhetorical units (Cloran 2010)

now
here
event proposal proposion

concurrent prior forecast
non-habitual habitual non-hypothecal hypothecal
central enty
interactant acon commentary reflecon volional non-
volional
co-present observaon recount plan predicon conjecture
person/object
absent report account predicon
person/object
generalised generalisaon
person/object
Figure 11.3 The cline of (de)contextualization (Cloran 2010)
(5) Mother: There aren’t many passionfruit out there at the moment
Report
Stephen: Why?
Mother: Because passionfruit come when it’s warm
Generalization
In contrast, when the semantic content is picked up on in the Rheme,

there appears to be a more significant switch of angle, as a new rhetorical
purpose is introduced as an expansion upon a topic without a loss of
cohesion, as in example (6):
(6) Mother: It’s too cold for passionfruit now Generalization

They don’t like the cold weather
Do you think we should plant a passionfruit vine at our
new Plan house?
What Cloran’s research showed is that while LAP families stayed on

task with their talk, HAP mothers very often used highly contextual-
ized talk as prompts for discussing less immediate or more abstract
matters, as in example (5) (with example (6), conversely, taking the
conversation back closer to home, though not the immediate task in
hand). Cloran’s results, in terms of both the extent and the nature of
decontextualized talk, therefore supported Bernstein’s claim that
working-class children were less prepared for the decontextualized
language of the classroom and that their relative lack of educational
success was a result of socialization practices rather than a difference
in intelligence.
296 TOM BARTLETT
Hasan and Cloran’s work in this area is a departure from previous

analytical approaches within SFL in a number of significant ways. Firstly,
in relating the variation revealed to extra-linguistic social factors and the
differences in socialization practices according to the parents’ position in
the production of capital, RU analysis posits a causal relationship between
the circumvenient context and the supervenient context. In contrast, where
register theory, for example, accounts for variation, this is largely in terms
of the function of the text as text, and so a linguistic-internal concept of
context is in some way maintained. Secondly, while explorations of GSP
looked for repeated patterns in language-activity pairings where the activity
was strictly defined and the role of incidental surrounding language largely
downplayed, RU analysis looks at all the linguistic behaviour accompanying
non-linguistic activity and, in Cloran’s research at least, highlights the
language that strays from the obvious task at hand. Thirdly, and relatedly,
the structural relations between RUs are not hierarchical, not ‘predeter-
mined’ by the ultimate goal of the activity and combining to realize distinct
stages along the way to this goal; they are, rather, immanent, arising
spontaneously from the ongoing talk and the non-linguistic activity and
not necessarily leading in any particular direction. These are all points I shall
return to below.
11.4 Berry’s Exchange Structure
In the development of both GSP and RU analysis, the importance of elem-

ents of structure above the clause but below the text is clear (leaving open
the question of whether the relationship between the text and such elem-
ents is hierarchical or immanent). Developing such intermediate units was
also the concern of work in Birmingham, where Sinclair and Coulthard
(1975) were working on teacher–pupil interaction in the classroom, and
Nottingham, where Margaret Berry (1981a, 1981b) was developing her ideas
on exchange structure. Focusing on stretches of speech in terms of their
function in interaction, Sinclair and Coulthard adopted a rather more
structured and hierarchical approach than that developed for RUs, with
discrete exchanges between speakers comprising turns from each speaker,
and with each turn comprising one or more moves. Individual moves were
labelled as initiating (I), responding (R), or feedback (F), for example:
(7) 1 A: Have you finished your homework yet? I

2 B: No. What time is it? R;I
3 A: Ten o’clock. R
4 B: Ok, I’d better hurry up. F
This invented dialogue comprises a single exchange which is played out

in four turns, two for each speaker. Turn 1 is a single initiating move,
whereas B’s turn in 2 is made up of two moves, a response to A and an

initiating question. Turn 3 comprises a single move, a response to the
question in Turn 2; and Turn 4 is an acknowledgment of that response,
or feedback, which signals the closure of the exchange.
Writing of her own work in developing exchange structure at this time,
Berry (forthcoming) stresses the need for a complementary approach to
the synoptic mapping of conventional systems networks in SFL, which are
based on norms and tendencies across population sets, and describes
exchange structure as a means of analyzing the dynamic and potentially
unexpected movements of real-time discourse. Taking a metafunctional
perspective, Berry developed interpersonal and experiential descriptions
of exchanges to complement what she considered the textual labelling of
Sinclair and Coulthard.
As Berry (forthcoming) explains, interpersonal exchange structure is
concerned with the roles that the interactants adopt in relation to the
transmission of information or negotiating action (as with mood types),
irrespective of who is initiating the exchange as text. The interpersonal
roles are ‘primary knower’ (k1), the interactant assumed to know the infor-
mation, and ‘secondary knower’ (k2), the interactant assumed not to
know the information (though these roles may not reflect the reality of
who knows what). Using Berry’s illustration below, we can see how the
following examples share the same textual structure but different interper-
sonal structures, as signalled by the intonation patterns across the moves
and the wording of the finishing move. In example (8), the quizmaster
(primary knower) already knows the answer and is testing the knowledge
of the contestant (secondary knower), while in example (9), the son (second-
ary knower) does not know the information and is seeking an answer from
his father (who he assumes is primary knower). In example (8) the quiz-
master defers transmitting the information, so their initial move is labelled
dk1 and their second as k1. In providing the definitive information, the
final move in example (8) contrasts with the final move in (9), which is
merely a follow-up move by the newly enlightened secondary knower and
so marked as k2f:
(8) Quizmaster: In England, which cathedral has the tallest spire dk1
Contestant: Salisbury k2
Quizmaster: Yes k1
(9) Son (doing crossword): Which English cathedral has the k2
tallest spire
Father: Salisbury k1
Son: Oh, right k2f
Experiential exchange structure concerns the establishing and develop-

ment of propositional content, and individual moves include the propos-
itional base (pb), the propositional completion (pc), and the propositional
298 TOM BARTLETT
support (ps). Adding these labels to example (9) above provides the complete
metafunctional analysis of the exchange:
(10) Son (doing crossword): Which English cathedral I k2 pb

has the tallest spire?
Father: Salisbury R k1 pc
Son: Oh, right F k2f ps
Berry (forthcoming) sets out more delicate options within the three
metafunctions, exploring how exchanges can depart from the smooth and
narrow, and discussing the function and variability of side sequences. The
complete framework developed therefore incorporates important insights
from Conversation Analysis and Gricean pragmatics into an SFL frame-
work, though there is not room here to expand on these ideas in depth.
In this way Exchange Structure provides analysts with a more dynamic and
emergent view of discourse as something that happens, and Berry (2016)
provides a systems network that accounts for the different possibilities that
arise at different points in the discourse.
While the formalisms of Exchange Structure all relate to in-text rela-
tions, they provide a useful tool for discourse analysis. As Berry (2016)
explains, labelling for all three metafunctions provides richer analysis of
exchanges, often distinguishing superficially similar structures. Though
the approach was developed within and is particularly suited to educational
settings, it can be applied to a range of contexts to analyze, for example,
how power is played out in various settings in terms of who introduces
the propositional bases that delimit the scope of the conversation and who
act as primary and secondary knowers within these contexts. And while the
approach is based at one level on structure and hierarchy, the localized
range of these hierarchies, which flow from one to the next without
developing into superordinate structures, means the approach can also
be used to analyze less structured genres, including casual conversation.
In Sections 11.5 and 11.7 we will see contrasting approaches to discourse
analysis in these terms: Rhetorical Structure Analysis, which works very
much within the tradition of hierarchicization, combining units of ever-
increasing size to analyze whole texts as single structures, and Phase
Analysis, which emphasizes the flow of texts across the metafunctions
and their emergent properties.
11.5 Rhetorical Structure Theory
While Berry’s approach to Exchange Structure allows for turns of more

than one move, this only occurs in a limited number of cases. As the name
suggests, this framework focuses on the dynamics of interaction rather
than the internal analysis of single-speaker stretches. This complementary
approach is provided by Rhetorical Structure Theory (RST; Mann and
Thompson 1988; Mann and Matthiessen 1991; Mann et al. 1992), a method
for analyzing the logical relations, both syntactic and functional, between
individual messages within what might be classed as a single prolonged
move in Berry’s terms. And while this approach could be extended to cover
dialogic speech, it is therefore most appropriate to monologue. Mann et al.
(1992:43–6) list the basic assumptions underlying RST, which are para-
phrased and condensed here:
a. Texts consist of functionally significant parts, combining as elements

within patterns of text;
b. As with Halliday and Hasan, ‘to be recognised as a text, the writing
must create a sense of overall unity to which every part contributes’
(Mann et al. 1992:43);
c. In contrast to Halliday and Hasan’s emphasis on cohesion in creating
textuality, unity and coherence are a result of the imputed function
of the individual elements and their contribution to ‘a single purpose of
the writer’;
d. Texts are hierarchically organized ‘such that elementary parts are com-
posed into larger parts, which in turn are composed into yet larger parts’
(Mann et al. 1992:43). This contrasts with the more linear though equally
goal-oriented approach within GSP (and Martin, see Section 11.6);
e. The relational structure between elements is the same at every scale;
f. The principal structural pattern is relational, with a small set of highly
recurrent relations linking pairs of elements at all scales;
g. In most cases the structural relationship between pairs of elements is
asymmetrical, comprising a nucleus and a satellite;
h. Relations are functional and can be stated in terms of the effects that
they produce, in other words ‘the purposes of the writer, the writer’s
assumptions about the reader, and certain propositional patterns in the
subject matter of the text’ (Mann et al. 1992:45). In these terms, the
relations are not between the words of the text (in contrast to the early
cohesion approach), but the meanings and intentions behind each
element, of which the wording is the realization;
i. The number of relations is open and additional relations can be identi-
fied through investigation.
Relations commonly used in RST include the following: (i) nucleus-satellite

relations (i.e. where two elements at any scale are in an asymmetrical rela-
tion) such as evidence (for the proposition in the nucleus), concession, elabor-
ation, condition, evaluation, antithesis, purpose, and summary; and (ii) multi-
nuclear relations (i.e. where two or more elements at any scale share nuclear
status) such as sequences and contrasts. Note that elements functioning as
satellites at one scale have their own nucleus with potential satellites at the
scale below. Relations between a nucleus and a satellite are defined in terms
of the constraints on each and the effect ‘that plausibly the writer was trying to
produce in employing the relation’ (Mann et al. 1992:48). For example, for
300 TOM BARTLETT
the evidence relation to be said to hold, the constraints are that the reader
may not believe the proposition in the nucleus; that the reader is likely to
find the satellite credible; and that the reader’s comprehension of the satel-
lite will increase their belief in the nucleus. The effect therefore is that the
reader’s belief of the nucleus is increased. For full examples of analyzed
texts, see Mann and Matthiessen (1991) and Mann et al. (1992).
Given the focus on the recurrent functional relationships between
elements of a text at different scales, Hasan and Fries (1995:xxxiii) suggest
that RST would be better labelled as ‘logical structure theory’. If we take
this conclusion at face value, it could be claimed that RST provides the
missing metafunctional link, adding a description of logical relations across
texts to the experiential (lexical cohesion and cohesive harmony), the text-
ual (RU analysis; Cohesion in English), and the interpersonal (message seman-
tics and exchange structure, though these also consider aspects from the
other metafunctions). In an extension of the original descriptions of
RST that overlays the basic analysis with interpersonal and ideational
features to provide a fuller representation of extended text that is reminis-
cent of Hasan’s conception of register as the accumulation of message
semantics across whole texts, Mann and Thompson (1991) go as far as
to suggest that a correlation between the relations of RST and all the
metafunctions is robust.
In terms of the text/discourse distinction proposed at the beginning
of this chapter, while RST relations are based on predictions of authorial
intention and favoured reading, these are all features that can be read off
from the decontextualized text alone so that, despite the additional analyt-
ical features afforded by RST, it would have to be considered text rather
than discourse analysis according to this definition. Webster et al. (2013) is
an example of RST in analyzing political speeches, while Bateman
(2008) utilizes a somewhat extended version of RST as one of the layers of
description for multimodal static page-based documents.
11.6 Martin’s Discourse Semantics and Genre
Whereas the metafunctional correlates of RST were elaborated almost as

an afterthought, the work of Jim Martin over almost four decades has focused
on developing not just a model but a theory of text that incorporates metafunc-
tional diversity as a defining principle within Martin’s conceptualization of
Genre and Discourse Semantics (see Martin, this volume). As noted above,
Martin (Martin in Anderson et al. 2015:53) critiques Halliday and Hasan’s
approach to textuality and cohesion as being a ‘grammar and glue approach’
in which text is a by-product of cohesive relations between messages. Turning
things around, and focusing on a process-based analysis rather than a product-
based one,9 Martin sees the semantics of the text as a whole as paramount
9
Though Martin sees both perspectives as useful.
(as one dimension of the realization of a genre), with meanings realized within
the text in different ways: periodically and cumulatively across the text, as well
as structurally within the confines of the individual clause or clause complex.
Thus, rather than focusing on realizations of experiential, logical,
interpersonal, and textual meanings at clause rank, Martin (1992) proposes
the following metafunctional categories at the discourse semantic stratum: idea-
tion, conjunction and continuity, negotiation (based largely on Berry’s
exchange structure), and identification, with the further category of texture
‘interleaving’ the features of each together to create texts.
In extended collaborations with Peter White and David Rose, Martin
produced two further landmark books, Evaluation in English (Martin and
White 2005) and Working with Discourse (Martin and Rose 2003, 2007). The
first of these extends the discourse semantic treatment of interpersonal
(including subjective) meaning to comprise three major systems – Attitude,
Engagement, and Graduation – which consider attitudinal language,
interspeaker/intertextual relations, and amplification/moderation respect-
ively. There is no room here to discuss these categories in depth (but see
Martin, this volume). In the second of these books, Martin and Rose update
the analytical framework of Martin’s English Text (1992), re-presenting
the approach from a more text-analytical-up and less theory-down perspec-
tive. The categories presented have now evolved into ideation, conjunction,
appraisal (based on Martin and White’s work on attitude, engagement, and
graduation), negotiation, identification, and periodicity (a revised version of
texture).
An interesting point worth dwelling on here is Martin’s (1992:249–64)
critique of the hierarchical representation foundational to RST, which he
sees as too product-oriented and especially at variance with ‘the dynamics
of text as process, particularly in the spoken mode’ (Martin 1992:258).
In contrast he proposes a linear dynamic which resonates with his
characterization of genre as ‘a staged goal-oriented social process’
(Martin 1992:505). From this perspective, genres unfold as a sequence of
necessary stages, semantically driven and realized through the metafunc-
tionally diversified resources of discourse semantics, lexicogrammar, and
phonology/graphology, to fulfil recognized and recognizable social
activities.
Much of the work of what has come to be termed the Sydney School, after
Martin’s university, has been produced within action research projects
aimed at extending the literacy skills of disadvantaged groups through a
visible pedagogy that focuses on producing highly valued written work
across a range of disciplines at all levels of the curriculum (though Martin
is quick to emphasize this is just one application of the theory, which has a
broader genesis). Within such a framework, a genre can be defined as
situation/language pairings which have become socially recognizable
through repeated association (a point I return to in Section 11.7), and
Martin and Rose (2008) exemplifies this approach through the analysis into
302 TOM BARTLETT
named stages of genres across a range of disciplines. While the goal of this
work is to make visible and hence replicable what is hidden to students and
teachers alike, this apprenticeship approach has regularly invited criticisms
of prescriptivism and acculturation within the oft-heated debate on minor-
ity rights in education.
More recently, the work of Martin and his collaborators has moved into
new fields, particularly restorative justice, and this has led not only to
further socially motivated applications of the approach, but also to a
continuing enrichment of the descriptive and theoretical power of their
work. Bednarek and Martin (2010) is a strong testimony to the achieve-
ments of this prolonged collaborative labour between Martin and his
research students, many of whom are now recognized figures in SFL.
A particularly fruitful collaboration has been with the Bernsteinian soci-
ologist Karl Maton, resulting in the 2014 volume Knowledge and Knowers:
Towards a Realist Sociology of Education, which analyzes a range of discip-
lines across the sciences, social sciences, and humanities in terms of the
kinds of knowledge and epistemic stances favoured and their transpar-
ency and hence their potential for transmission through visible pedago-
gies (see Maton 2014).
One continuous thread in Martin’s work is his insistence on a super-
venient model of context and his occasionally scathing rejection of eth-
nography, or at least what often passes as such in applied linguistic
work, which he labels ‘ethNOgraphy’. Martin consistently takes the
line that, within a social realist ontology such as SFL, culture is realized
in text, via genres, semantics, lexicogrammar, and phonology/graphology
and that non-linguistic information, including observation and the views
of insiders, adds nothing to and potentially distorts our understanding
of ‘what’s going on here’.10 In the following section, I take issue
with Martin’s stance, drawing on the work of Michael Gregory and my
own heavily contextualized approach, before suggesting a potential
reconciliation.
11.7 YESnography: Gregory and Bartlett
The ‘dialect’ of SFL developed by Michael Gregory and colleagues, known

as Communication Linguistics, takes a radically different stand from Martin
with regard to ethnography and the status of the supervenient context in
an overall architecture of discourse. In the opening paragraph of his 1995
paper, Gregory (1995:67) states that the goal of communication linguistics
is to develop a model ‘from the systemic-functional tradition in particular,
10
Martin (personal communication) suggests the following alternative wording: ‘do not provide an answer to
questions addressed in social semiotics analysis informed by SFL (including the multimodal analysis evolving out of
SFL).’ I have decided to keep both my original wording and Martin’s alternative.
ethnographic linguistics in general, and in terms of a dialectical material-

ist theory of knowledge’. The model of discourse presented (Gregory
1995:72) follows other SFL approaches in having a metafunctionally
diverse semantic stratum of register (which would correspond to
Discourse Semantics in Martin’s model) which is realized through special-
ized features of the lexicogrammar. Where it differs most significantly is
that the stratum above the semantics is not the tripartite division of
context into field, tenor, and mode as in Hasan’s and Martin’s models,
which seek to characterize ‘what is going on through language here’ and
therefore tend towards a supervenient approach, but rather a stratum of
knowledge which represents the resources brought to bear in determining
what discourses are possible within a particular environment, so tending
towards a circumvenient approach (though see comments later in this
section). The knowledge brought to bear on the situation is represented as
a conjunctive binary system, with knowledge of the Community Communi-
cating Context informing what generic activities are potentially available in
the present situation, and the knowledge of Language: Dialect Configuration
informing what registerial variables are available as resources in
performing these.
There is thus a crucial difference between Gregory’s model and both
Hasan’s and Martin’s. In Hasan’s model, the contextual configuration of
field, tenor, and mode activates certain meanings as at risk (i.e. potentially
available) within a situation; in Martin’s model, the genre being performed
carries out the same role. For both these models, therefore, description
starts with the activity underway. Gregory’s model goes one stage further
back, however, in stating that there is a range of contextual configurations or
generic situations at risk in any given environment as a function of the know-
ledge the participants bring to bear within that environment (Gregory
1995:71). In order to describe and account for discourses, then, it is neces-
sary to account not only for ‘the linguistic items that occur in them, but
also the relations they enter into with each other and with the knowledge of
users and receivers of the language’ (Gregory 1995:69, emphasis added). Import-
ant to note is that the knowledge brought to bear is not considered in any
mentalist or idealist sense, but ‘as viable knowledge, knowledge as func-
tion, as social fact’ (Gregory 1995:69), a generalized knowledge of recurrent
situations and the potential for action within them that has been internal-
ized in the Vygotskian sense.
I take this to mean that the knowledge brought to bear in a communi-
cating environment can be considered as the distillation of the semiotic
histories of the various participants, histories which act on the text but
cannot be read off it. Discourse analysis therefore demands an ethno-
graphic approach in order to provide an understanding of the rich back-
drop of past meaning-making that is woven into the text as action, and
Gregory’s approach aims to account for texts as instances of discourse
which orient to yet manipulate general tendencies (note the title of
304 TOM BARTLETT
Gregory’s paper), rather than accounting solely for tendencies in terms of

regular language-internal features.11
One important feature which Communication Linguistics does share with
the Martinian approach to genre is that texts are seen as dynamic and linear
rather than hierarchically ordered. However, where in Martin’s model
stages are primarily characterized in terms of their discourse function, for
Gregory a discourse progresses through phases displaying metafunctional
consistency, with a change in meaning in any of the metafunctions resulting
in a new phase. Ultimately, this difference between the two approaches
might prove to be no more than a difference in perspective, given the basic
SFL tenet that form and function are two sides of the same coin, as captured
when Gregory (1995:71) describes textual progression across distinct phases
as reflecting alternatively ‘the dynamic instantiation of microregisters’ (i.e.
real-time shifts in meaning patterns) or the ‘microinstantial situations of the
discourse’ (i.e. real-time shifts in activity).
However, while both Martin’s approach and Gregory’s Communication
Linguistics emphasize the dynamics of discourse, both also focus on generic
situations and their realization as texts, thereby suggesting a closure that is
in some way predetermined. Relatedly, both models rely at some level on a
concept of shared histories within a single overarching context of culture:
Martin (Martin and Rose 2008) describes a culture as the sum of genres
available within it, while Gregory (1995:71) states that ‘[a]ny particular
communicating community context is characterizable in terms of the gen-
eric situations which are potential within it’. My own work (Bartlett 2012;
2017) takes a rather different perspective in not taking generic histories
for granted, deriving as it does from novel contexts of intercultural dis-
course between the Amerindian communities of the North Rupununi Savan-
nahs in Guyana and the international development workers with whom
they negotiate. Adopting Gregory’s notion of phase, I base my analysis on
the different metafunctional configurations within and across phases and
the ways in which these instantiate situation/language pairings that can be
related to the cultural background of the different groups involved and
the distinct semiotic histories and ways of speaking they bring to bear on
the situation. Rather than looking to identify genres, therefore, I was keen
to see the following: the extent to which the different voices of the groups
involved, as represented by these situation/language pairings, were realized
in the intercultural discourse; the degree to which these different voices
were legitimated; and the resultant emergence of new ways of talking – or
hybrid voices. In such novel and evolving instances, including institutional
talk, it would be wrong to talk of genres, a term which assumes social
recognizability through repeated association, and better to focus on recog-
nizable voices which are intertwined in novel ways according to the
11
More recent work in the Martinian tradition has acknowledged the emphasis on the system rather than the
instance in textual analysis and is working to develop better descriptions of individual texts as instances.
dynamics of the situation. If these novel ways of speaking become standard-

ized within this particular social environment, only then will it be possible
to talk of new genres. Given this focus, an emphasis on ethnography is
essential both to relate the different ways of speaking of the different groups
to their distinct semiotic histories and to explain under what conditions a
particular discourse was actualized from the potential afforded by the
environment. However, rather than characterizing such an approach as
circumvenient, with the suggestion of a dualist approach to context and
semiotics, this view is better described as scalar supervenience in that the non-
textual background features that motivate and inflect language in use are
themselves the result of past semiosis which are reactivated as relevant to
specific contexts in different ways for different speakers. My own approach,
therefore, differs from those of Hasan, Martin, and Gregory, in taking
neither a single context of culture nor the idea that all discourse is ‘genred’
for granted, while adhering to the core SFL conception of language as a
stratified, metafunctionally diverse social semiotic system.
11.8 Eggins and Slade on Casual Conversation
If my own work questions whether institutional talk can always be con-

sidered generic, Eggins and Slade’s (1997:6) landmark Analysing Casual Con-
versation appears to come from the opposite direction in suggesting that,
‘despite its sometimes aimless appearance and apparently trivial content,
casual conversation is, in fact, a highly structured, functionally motivated,
semantic activity’. Eggins and Slade (1997:7) argue that such functional
structure is lost in other work on casual conversation, in which ‘analysis
has frequently been fragmentary, dealing only with selected features’ and
‘has not sought to explore the connections between the “social work”
achieved through the micro-interactions of everyday life and the macro-
social world within which conversations take place’. Combining the
approaches of Halliday and Hasan and those developed by Martin and his
colleagues with other traditions such as conversation analysis and narrative
analysis from outside SFL, the authors identify four types of patterning that
occur at different levels of language and that ‘interact to produce the
meanings of casual talk’ (Eggins and Slade 1997:7):
• grammatical patterns at the clause level which indicate power and

subordination within interaction;
• semantic patterns which indicate frequency of contact and familiarity
between interactants;
• conversational structural patterns which indicate affective involvement
and shifting alignments within conversation;
• the use of text types which give some indication of shared worldviews
about normality and predictability.
(Eggins and Slade 1997:18)
306 TOM BARTLETT
These different patterns occur in ‘chunk and chat segments’ in casual

conversation (Eggins and Slade 1997:227), characterized by frequent
exchanges of turn and domination by a single speaker respectively. Both
of these structural types have a social function to play, with the generically
recognizable chunks tending to index shared or opposing group values at a
fairly broad level, while the more fluid chat segments do the hard interper-
sonal labour of signalling and negotiating status and familiarity between
individuals. Taking these two aspects together, Eggins and Slade (1997:22)
conclude that the primary function of casual conversation is to negotiate
the ‘tension between, on the one hand, establishing solidarity through the
confirmation of similarities, and, on the other, asserting autonomy through
the exploration of differences’. Casual conversation thus stands in contrast
to ‘pragmatic conversation’ (Eggins and Slade 1997:19), which is motivated
by a clear pragmatic purpose through which the different interactants
achieve complementary goals (such as buying or selling). Following Hasan’s
distinction, in Section 11.3.4, it would appear that pragmatic conversation
is primarily structured according to the recognizable social roles of the
different interactants at a fairly indelicate degree of differentiation, while
casual conversation is about negotiating the more delicate aspects of the
context in the absence of, or temporary suspension of, such defining roles.
Returning to the comparison made with my own approach at the beginning
of this section, I would suggest that my own research has focused on
contexts in which such defining roles, and the generically structured talk
that accompanies them, are yet to be firmly established, so that much of the
work that goes on in embryonic institutional contexts relies on the pattern-
ings of meaning displayed by casual conversation.
11.9 Matthiessen’s Appliable Linguistic Analysis and

Concluding Remarks
Whereas many of the approaches to discourse analysis outlined so far have

focused on specific areas of language and/or society as meriting particular
attention, Matthiessen (2014) calls for an ‘appliable discourse analysis’ (ADA)
that can serve as a universal resource in responding to social problems and
issues where language in use is a significant element. The use of this label is
an explicit reference to Halliday’s notion of an ‘appliable linguistics’ as a
functionally based theory of language that can be applied across contexts
rather than being applicable in certain cases. For Matthiessen (2014:147),
‘ADA corresponds to Appliable Language Description, the two being comple-
mentary aspects of appliable linguistics’. As such, ADA depends on full
functional descriptions of the language under study in order that the cor-
respondence between specific contexts of situation and the features of
language in use within them can be located within the system as a whole
and hence related in a motivated way to other instances of use. This is of

course a huge undertaking, and one that Matthiessen sees as best solved
through ‘long-term collective discourse analysis designed to address prob-
lems that are beyond the scope of individual research projects’ (Matthiessen
2014:148). There is not the space to discuss the details of this approach in
the present chapter, but the essential properties outlined by Matthiessen
(2014:147–8) provide a flavour of the scope, objectives, and logic of ADA:
• It must be grounded in a holistic theory of language in context;

• It must reference a comprehensive description of the particular lan-
guage in culture that is in focus;
• It must be reasonably explicit so that manual analysis can easily be
related to automated analysis;
• It must be multilingually and multisemiotically oriented;
• It must provide an account of the context in which the analysis was
undertaken in order to reason about the analytical choices made;
• It must be geared towards data sharing and reuse and the goal of long-
term collaborative discourse analysis.
From my own perspective, Matthiessen’s approach is problematical in that,

despite the huge coverage called for, the description of context for each
language event is still linguistically determined (2014:168). It could, how-
ever, be argued that in calling for the analysis of instances to be situated
within a holistic description of the entire discourse system, then the whole
of context as remembered and partially shared semiosis is accounted for. In
this way, ADA shares many of the goals of Communication Linguistics and
can be related to my own perspective of scalar supervenience. However,
I think that the level of description called for in ADA is neither possible in
practice nor representative of what individual speakers bring to bear to
individual instances of real life. As with Hasan’s work on register and
Martin’s work on genre, it is an approach best suited to uncovering tenden-
cies across instances rather than accounting for the specifics of those
instances: not only what is made relevant by text, and what is not, but also
the locally contingent reasons for and effects of such ‘choices’. It is the
purview of ethnography to explore techniques for considering what is
relevant to whom and when, and there is surely room for integrating
ethnographic techniques in a happy compromise with the level of system-
atic description proposed for ADA.
References
Andersen, T. H., M. Boeris, E. Maagerø, and E. S. Tønnessen. 2015. Social

Semiotics: Key Figures, New Directions. London: Routledge.
Bartlett, T. 2012. Hybrid Voices and Collaborative Change: Contextualising Positive
Discourse Analysis. London: Routledge.
308 TOM BARTLETT
Bartlett, T. 2013. ‘I’ll Manage the Context’: Context, Environment and

the Potential for Institutional Change. In L. Fontaine, T. Bartlett, and
Bartlett, T. 2017. Context in Systemic Functional Linguistics: Towards
Scalar Supervenience? In T. Bartlett and G. O’Grady, eds., The Routledge
Handbook of Systemic Functional Linguistics. London: Routledge. 375–90.
Bateman, J. A. 2008. Multimodality and Genre: A Foundation for the Systematic
Analysis of Multimodal Documents. London: Palgrave Macmillan.
Bednarek, M. and J. R. Martin. 2010. New Discourse on Language: Functional
Perspectives on Multimodality, Identity, and Affiliation. London: Continuum.
Berry, M. 1981a. Systemic Linguistics and Discourse Analysis: A Multi-
layered Approach to Exchange Structure. In M. Coulthard and M. Mont-
gomery, eds., Studies in Discourse Analysis. London: Routledge. 120–45.
Berry, M. 1981b. Towards Layers of Exchange Structure for Directive
Exchanges. Network 2: 23–32.
Berry, M. 2016. Dynamism in Exchange Structure. English Text Construction
9(1): 33–55.
2007. Dissociation, Relatedness and ‘Cohesive Harmony’: A Linguistic
Measure of Degrees of ‘Fragmentation’? Linguistics and the Human Sciences
3(3): 263–93.
Relations of Context, Meaning and Grammar. Monographs in Systemic
Linguistics 6. Nottingham: University of Nottingham.
Cloran, C. 2010. Rhetorical Unit Analysis and Bakhtin’s Chronotope. Func-
tions of Language 17(1): 29–70.
Eggins, S. and D. Slade. 1997. Analysing Casual Conversation. London: Cassell.
Gregory, M. 1995. Generic Expectancies and Discoursal Surprises: John
Donne’s The Good Morrow. In P. H. Fries and M. Gregory, eds., Discourse
in Society: Systemic Functional Perspectives. Meaning and Choice in Language:
Studies for Michael Halliday. London: Ablex. 67–84.
Halliday, M. A. K. 1964. Descriptive Linguistics in Literary Studies. In
A. Duthie, ed., English Studies Today: Third Series. Edinburgh: Edinburgh
of Language. London: Edward Arnold.
Language and Meaning. Baltimore: University Park Press.
Halliday, M. A. K. and R. Hasan. 1989. Language, Context and Text. 2nd ed.
Oxford: Oxford University Press.
Hasan, R. 1984. Coherence and Cohesive Harmony. In J. Flood, ed., Under-
standing Reading Comprehension: Cognition, Language and the Structure of
Prose. Newark: International Reading Association. 181–219.
Hasan, R. 1989. Semantic Variation and Sociolinguistics. Australian Journal of

Linguistics 9: 221–75.
Hasan, R. 1995. The Conception of Context in Text. In P. Fries and
M. Gregory, eds., Discourse in Society: Systemic Functional Perspectives. Meaning
and Choice in Language: Studies for Michael Halliday. London: Ablex. 183–284.
Hasan, R. 1996. The Nursery Tale as Genre. In C. Cloran, D. Butt, and
G. Williams, eds., Ways of Saying: Ways of Meaning. London: Cassell. 51–72.
Hasan, R. 2009. Wanted: A Theory for Integrated Sociolinguistics. In J. J.
Webster, ed., Collected Works of Ruqaiya Hasan,Volume 2: Semantic Variation:
Meaning in Society and in Sociolinguistics. Sheffield: Equinox. 5–40.
Metafunctions and Semantics. Functional Linguistics 1(9): 1–54.
Hasan, R. 2016. Collected Works of Ruqaiya Hasan, Volume 2: Context in the
System and Process of Language. Edited by J. J. Webster. Sheffield: Equinox.
Hasan, R. and P. Fries. 1995. Reflections on Subject and Theme: An Intro-
duction. In R. Hasan and P. Fries, eds., On Subject and Theme: A Discourse
Functional Perspective. Amsterdam: John Benjamins. xiii–xlv.
Hasan, R., C. Cloran, G. Williams, and A. Lukin. 2007. Semantic Networks:
The Description of Linguistic Meaning in SFL. In R. Hasan, C. M. I. M.
Lukin, A. 2012. Meanings in Questions: A Case Study of the ABC’s Current
Affairs Coverage of the 2003 Invasion of Iraq. In T. Bartlett and H. Chen,
eds., Special Issue of Journal of Applied Linguistics and Professional Practice 9(1):
424–44.
Lukin, A. 2015. Language and Society, Context and Text: The Contributions
of Ruqaiya Hasan. In W. Bowcher and J. Y. Liang, eds., Essays in Honour of
Ruqaiya Hasan: Society in Language, Language in Society. London: Palgrave.
143–65.
4(2): 187–213.
Mann, W. C. and C. M. I. M. Matthiessen. 1991. Functions of Language in
Two Frameworks. Word 42(3): 231–49.
Mann, W. C. and S. A. Thompson. 1988. Rhetorical Structure Theory:
Towards a Functional Theory of Text Organisation. Text (8)3: 243–81.
Benjamins.
310 TOM BARTLETT
Martin, J. R. 2010. Semantic Variation: Modelling Realisation, Instantiation

and Individuation in Social Semiosis. In M. Bednarek and J. R. Martin,
eds., New Discourse on Language: Functional Perspectives on Multimodality,
Identity and Affiliation. London: Continuum. 1–34.
Martin, J. R. and D. Rose. 2003. Working with Discourse: Meaning beyond the
Clause. London: Bloomsbury.
Clause. 2nd ed. London: Bloomsbury.
Equinox.
Martin, J. R. and P. R. R. White. 2005. The Language of Evaluation: Appraisal in
English. London: Palgrave Macmillan.
Maton, K. 2014. Knowledge and Knowers: Towards a Realist Sociology of Educa-
tion. London: Routledge.
Matthiessen, C. M. I. M. 2014. Appliable Discourse Analysis. In F. Yan and
Moore, A. R. 2017. Register Analysis in Systemic Functional Linguistics. In
T. Bartlett and G. O’Grady, eds., 2017. The Routledge Handbook of Systemic
Functional Linguistics. London: Routledge. 418–37.
Scott, C. 2010. Peace and Cohesive Harmony: A Diachronic Investigation of
Structure and Texture in ‘End of War’ News Reports in The Sydney Morning
Herald. In F. Yan and C. Wu, eds., Challenges to Systemic Functional Linguis-
tics: Theory and Practice. Proceedings of the 36th ISFC. Beijing: ISFC Organising
Committee. 89–96.
Sinclair, J. and R. M. Coulthard. 1975. Towards an Analysis of Discourse:
The English Used by Teachers and Pupils. London: Oxford University Press.
Vygotsky, L. S. 1978. The Mind in Society: The Development of Higher
Psychological Processes. Edited by M. Cole, V. John-Steiner, S. Scribner, and
E. Souberman. Cambridge, MA: Harvard University Press.
Webster, J. J., J. Chan, V. Yan, and K. Wong. 2013. Visualizing the Architec-
ture and Texture of a Text: A Case Study of Selected Speeches of US
President Barack Obama. In F. Shi and G. Peng, eds., Festschrift in Honour
of Prof. William S-Y. Wang’s 80th birthday. Hong Kong: City University of
Hong Kong Press. 301–24.
Widdowson, H. G. 2007. Text, Context, Pretext: Critical Issues in Discourse
Analysis. Malden: Blackwell.
12
Cohesion and
Conjunction
Maite Taboada
12.1 Texture as the Weaving of Cohesion and Conjunction
Most texts we encounter on an everyday basis are coherent. They make sense
in the situation in which they are presented, and their meaning and inten-
tion can usually be understood. Even in cases where the communication
may be more difficult (a ‘different’ accent; spelling mistakes; complex argu-
mentative structure), we tend to accept texts as being coherent, and make an
effort to grasp their meaning. Coherence is such a fundamental property of
texts and of our communication that it is difficult to conceive of completely
incoherent texts. Consider the two invented examples in (1) and (2). In the
first case, we have a set of sentences, each connected to the previous one
through a lexical item. This is an instance of sets of cohesive links, with
items such as last night – at night, which may have a semantic relation of
repetition in most texts, but which do not here. The passage, however, does
not seem to have a common thread; it is not coherent. Conversely, the two
sentences in (2) are coherent in terms of a thread (dark clouds – rained), but
the sentences are not well related, because the conjunction however does not
usually relate two units in this way. It sets up an unfulfilled expectation, or
one contrary to expectation, but rain following dark clouds is actually not
contrary to expectation. Example (2) is coherent, but fails in the way that
coherence is made explicit through the conjunction however.
(1) I went home very late last night. At night, owls come out and hunt.
Harry Potter uses an owl to have his mail delivered. The mail was
very erratic over the Christmas holidays. The holidays were too
short, and short indeed is this paragraph.
(2) There were dark clouds in the sky today. However, it rained.
I would like to thank Geoff Thompson, for inviting me to contribute this chapter, and for providing very insightful
comments shortly before his untimely death.
312 MAITE TABOADA
Cohesive and conjunctive elements constitute the fundamental property

of texts, the property of texture (Halliday and Hasan 1976; Hasan 1985).
Texture is the quality that makes a particular set of words and sentences
a text; what holds them together to give them unity, in the same
way that the weaving of warp and weft create a piece of cloth, a textile
(which has the same root as text and texture). Texture is created through
relationships of choice: the speaker or writer chose to use those words
and sentences over other choices, to make the text meaningful in
context.
Traditionally, and since Halliday and Hasan (1976), texture has been
characterized as the interaction between coherence, how the text relates
to the context outside the text, and cohesion, how the elements in the text
itself contribute to making it a unified whole. Coherence, then, enters into
the realm of intentions (what we want to achieve with the text), and the
representation of the world through propositions and their connections.
Cohesion is more local to the text, and includes links among words (such as
dark clouds – rain in example (2) above). I suggest that coherence and
conjunction form a separate system, distinct from cohesion proper (see
Section 12.5). But before that, let us examine the traditional organization
of cohesion, and examples of cohesive elements.
12.2 Cohesive Devices
A ‘cohesive device’ is an element in the text that requires another element

for its interpretation. The relation between the two is a ‘cohesive tie’. The
first classification of cohesive devices, in Halliday and Hasan (1976), pro-
posed a classification along the following lines:
1. Reference
2. Substitution
3. Ellipsis
4. Conjunction
5. Lexical cohesion
A more abstract classification lists the first three types as instances of

grammatical cohesion, i.e. types which are realized through the grammat-
ical system of the language, as opposed to conjunction and lexical cohesion,
which rely on the lexis to achieve cohesion.
The most recent Introduction to Functional Grammar (Halliday and Matthies-
sen 2014: Chapter 9) lists four systems of cohesion:
1. Conjunction
2. Reference
3. Substitution and ellipsis
4. Lexical organization
Cohesion and Conjunction 313
Conjunction includes conjunction and continuity (Theme-Rheme relations,

in particular, textual Theme, see Berry, this volume). The difference
between conjunction and the other systems is that conjunction links whole
clauses or combinations of clauses, whereas reference creates cohesion by
creating links between elements. Elements may be referents (persons,
things) or facts, including a proposition or a whole passage of text captured
as a fact (Halliday and Matthiessen 2004: Section 9.2, Paragraph 7). Substi-
tution and ellipsis create links between referents by replacing or leaving out
referents that can be easily recovered in the context. Lexical organization
also creates cohesion between referents, but by using exclusively lexical
resources, as opposed to reference, substitution, and ellipsis, which enlist
all of the resources of the lexicogrammar to do so.
The rest of this section briefly examines the types proposed by Halliday
and Matthiessen, with illustrative examples. Unless otherwise indicated,
examples are extracted from informal reviews, posted online on the Inter-
net Movie Database website.1 They are reviews of Spirited Away, a 2001 ani-
mated film by Japanese filmmaker Hayao Miyazaki. The film has been
characterized as one of the top 100 movies of all time, and online reviewers
in IMDB rate it very highly. The reviews are reproduced verbatim, and may
include typos and unorthodox grammatical constructions.
12.2.1 Conjunction
Halliday and Hasan (1976) initially divided cohesion into two types:
grammatical and lexical. Conjunction was listed under the lexical label,
because it makes use of lexical items, i.e. conjunctions such as and,
but, or if. Halliday and Matthiessen (2014) move it out of the lexical
realm, and present it as the first element in the list of cohesive resources.
This is understandable, because, although it deploys lexical items, the
particular items are function words, i.e. not open-class items such as
nouns or verbs. In example (3), we see how the two clauses in the first
sentence are joined by the conjunction but. Conjunction also links sen-
tences, as we can see towards the end of example (3), which uses the
conjunction and to relate the two parallel sentences (I laughed. I cried) to
the last sentence.2
(3) I can rarely say that a movie made me laugh and cry without feeling
like an idiot, but the caliber of this picture is so high that I don’t even
feel embarrassed. I laughed. I cried. And you will too.
1
See www.imdb.com
2
Halliday and Hasan restrict cohesion to links across, not within, sentences (i.e., beyond the clause and clause complex
level). In example (3), then, the connection signalled with and would be cohesive, but not the one indicated by but in
the first clause. Here, I will consider both of these as instances of cohesion, because I believe that cohesion and
conjunction occur across clauses, whether in different sentences or not.
314 MAITE TABOADA
In addition to grammatical conjunctions, the conjunction system deploys a

few other grammatical devices which elaborate, extend, or enhance the
meaning of previous discourse. These are often adverbials (actually, alterna-
tively) or phrasal units (in any case, in the same manner). The class of such
elements is collectively referred to as ‘conjunctive Adjuncts’, and they
partially overlap with what, in other approaches, are referred to as dis-
course markers, filled pauses, or backchannels (Schiffrin 1987; Stenström
1994). We can see one example in (4), where the argument develops around
how different viewers interpret the film differently. This point is inter-
preted and summarized with the help of I mean at the beginning of the last
sentence.
(4) I remember talking to friends after seeing it and we all had interest-
ing points. I felt the film focuses on Chihiro’s innocence as compared
with the other characters she encounters, but her child like views are
so carefree (and naive at times) and her youthful exuberance really
makes it endearing. Another friend said it was a coming of age and
how Chihiro herself progresses throughout the film. I mean, if you
can find so much insight in a film, you know you have a great film.
Conjunction has an uneasy status as a member of the cohesion set of

resources. In fact, many researchers separate it from cohesion, as has been
done in the title of this chapter, presented as a coordinated item with
cohesion, rather than as subordinate to it. This is because conjunction
relates to a separate system, that of the clause complex, since it enables
and signals relations between clauses. Scott and Thompson (2001) charac-
terize the distinction as repetition vs. conjunction. Repetition (i.e. cohesion
in the sense used in this paper) handles continuity. Conjunction, on the
other hand, deals with discontinuity and transitions across units of dis-
course, mostly clauses and clause complexes, but it can extend beyond the
clause complex, relating entire sentences and portions of a text or even
chapters in a book. I discuss this issue in Section 12.5.
12.2.2 Reference
Reference is achieved mostly through relations between a pronoun and an
antecedent, forming a referential chain in the text, which is characterized
as anaphoric reference. Reference links to elements outside the text consti-
tute ‘exophoric reference’, whereas links within the text are ‘endophoric’.
Some reference chains may include both exophoric and anaphoric refer-
ence. For instance, first- and second-person personal pronouns refer to
relationships defined outside the text, but often also create text-internal
relations. In example (5), the personal pronoun I in the first sentence is
exophoric, in that it refers to the writer as somebody outside the text. The
reference chain, however, continues inside the text, with another I in the
second sentence, and the possessive my.
(5) Spirited Away is one of the most perfect movies I have ever seen. The
least I can say about it is that there was not a single moment during
it that my attention wasn’t completely focused.
Reference is often ‘personal’, establishing co-reference between the elem-

ents. The other two main forms of reference are demonstrative and com-
parative. ‘Demonstrative reference’ is probably deictic and exophoric in
origin, pointing to referents outside the text, but can be used endophori-
cally, as text-internal anaphora. In example (6), this world in the second
sentence refers to the world first introduced in the preceding sentence, the
world seen through the main character’s eyes.
(6) We discover the world as Chihiro does and it’s truly amazing to
watch. But Miyazaki doesn’t seem to treat this world as something
amazing.
‘Comparative reference’ is different from personal and demonstrative in

that identity of reference is not established; the link is rather to a compari-
son class. Example (7) is a shortened excerpt where different qualities of the
movie are discussed. The last sentence contains two instances of compara-
tive reference: another great point and the best (part of it). Another establishes a
comparative reference of difference to the other aspects already discussed,
and then goes on to correct that reference to make it particular, stating that
the score is not just another good aspect, but in fact the best.
(7) The story is imaginative and the characters and animations endlessly
unique and strange. . . . What I also loved in this film is that the
animation gives it a real sense of cinematography, . . . Another great
point in fact the best part of it, is the fantastic score.
12.2.3 Substitution and Ellipsis

Substitution and ellipsis are two forms of the same phenomenon, as ellipsis
can be described as substitution by zero. In substitution, a cohesive device
belonging to a closed class in the language is used to replace an open-class
lexical item.
Common substitution devices in English are one, so, or do so. In example
(8), we see one in this one as a substitute term for cartoon movies. The repeti-
tion of the term would have probably made the sentence heavier, with
more lexical items than necessary. The pronoun one helps make it cohesive.
(8) Even if you don’t normally like ‘cartoon movies’, you might give this
one a chance.
Ellipsis makes coherent text possible, and even more so in spoken language.
The simple ‘yes’ and ‘no’ answers that we furnish as answers to questions
would be cumbersome if they were always accompanied by a full answer
which repeats information already present in the question (although
316 MAITE TABOADA
repetition is also possible, and sometimes desired). Example (9) shows an

elliptical verb in the answer to the rhetorical question what does it matter?
The answer is it didn’t [matter].
(9) I can’t imagine seeing a live action foreign language film dubbed into
another language, but hey, this is a kids cartoon, what does it
matter? Up to a point it didn’t, because I loved the film.
Ellipsis seems to display different frequencies across languages, with some

languages allowing ellipsis of Given Subjects, and sometimes Objects, but
ellipsis is common even in English; such is the case in (10), where the object
of eat (presumably, the food in the market) is not mentioned, because it is so
clearly recoverable from the context.
(10) When Chihiro’s parents see the food in the market, they just sit
down and eat but they turn out to turn into pigs.
12.2.4 Lexical Organization

At a very general level, Hasan (1985) classified all cohesive elements (or
cohesive ties; see below) as belonging to one of three types: co-reference, co-
classification, and co-extension. Co-reference is the relation between a
pronoun and its proper name antecedent. In a co-classification relation
between two cohesive elements, A and B, ‘the things, processes, or circum-
stances to which A and B refer belong to an identical class, but each end of
the cohesive tie refers to a distinct member of this class’ (Hasan 1985: 74).
In (11) both Princess Mononoke and Spirited Away belong to the class movies
directed by Miyazaki.
(11) Much like Miyazaki’s previous feature Princess Mononoke, Spirited

Away is an epic fairytale fantasy that deserves no better medium
than the stunning animation work of Studio Ghibli.
Finally, in co-extension, the relation holds between two cohesive elements

that are related to each other by virtue of belonging to the same general
class. Such is the relation between story, animation, and score in example (7)
above, which are all members, i.e. different aspects, of the class film.
This general taxonomy (co-reference, co-classification, co-extension) can
be made more fine-grained by labelling the different types, in particular of
co-classification and co-extension. Hasan (1985) referred to these as sense
relations, and included synonymy, antonymy, hyponymy, and meronymy.
To that, one may also add repetition, which is not a sense relation per se, but
likewise contributes to cohesion.
In Halliday and Hasan (1976), lexical cohesion is classified along slightly
different lines, with lexical cohesion and referential relations as the main
categories (see Table 12.1). Referential relations are included because they
also contribute to creating cohesion across cohesive devices.
Table 12.1 Types of lexical cohesion according to

Halliday and Hasan (1976)
Type of lexical cohesion Referential relation
1. Reiteration
Same word (repetition) Same referent
Synonym Inclusive
Superordinate Exclusive
General word External
2. Collocation
Lexical cohesion may be the hardest to define because different research-

ers have proposed different categorizations of the various phenomena that
can be included as part of lexical cohesion. The superordinate category in
Table 12.1, for instance, seems to call for a subordinate category. And
general nouns (people, thing, stuff) can be considered a class of superordinate
terms. There are multiple reinterpretations and categorizations of the
semantic relations that language, and ultimately, thought, allows (Cruse
2000; Hasan 1985; Martin 1992; Morris and Hirst 1991; Tanskanen 2006).
Perhaps the one that has created the most trouble is the concept of colloca-
tion. Halliday and Hasan (1976:284) define it as ‘the association of lexical
items that regularly co-occur’. This includes pairs such as boy and girl,
which are complementary and not easily defined as either antonyms or
meronyms. It likewise includes words from an ordered series, such as days
of the week, but also a more general category of words that are related to
each other by virtue of a connection, often in the real world: basement . . .
roof; road . . . rail; and box . . . lid are some of the examples suggested by
Halliday and Hasan. More generally, collocation captures the relationship
between lexical items that tend to occur together in certain text types. In a
sense, characterizing collocation involves creating taxonomies of ideas,
concepts, and the world, of the type represented in thesauri and in Word-
Net (Fellbaum 1998).
Eggins (2004) addresses the difficulty of classifying lexical relations by
dividing them into two groups:
• ‘Taxonomic’ relations include the types of relations listed under ‘reiter-

ation’ in Table 12.1, such as synonymy, hyponymy, and meronymy, or
part-whole relations
• ‘Expectancy’ relations hold when there is a predictable relation between
the process and the participants in the process (e.g. mouse – squeak, or
play – a musical instrument).
Eggins defines expectancy relations as those related through transitivity

(which leads to cohesive harmony; see Section 12.4). If we were to make
this category broader, then it could account for collocation, because
318 MAITE TABOADA
Table 12.2 Types of lexical cohesion (Halliday and Matthiessen 2014:644)

Nature of Type of expansion Type of Examples
relation lexical
relation
paradigmatic elaborating identity repetition bear – bear
[lexical set] synonymy sound – noise
sound – silence [antonymy]
attribution hyponymy tree – oak, pine, elm…
oak – pine – elm… [co-hyponyms]
extending meronymy tree – trunk, branch, leaf…
trunk – branch – leaf… [co-
meronyms]
syntagmatic (enhancing) collocation fire – smoke
[collocation]
collocation arises out of expectancy, that in a text we will find words in

certain relations to other words (same semantic field, doer-process,
adjective-noun, etc.).
A different view of the class of lexical items is provided in Halliday and
Matthiessen (2014), where the taxonomic vs. expectancy relation is char-
acterized as paradigmatic vs. syntagmatic. Paradigmatic relations are
those of choice among alternatives. Halliday and Matthiessen describe
them as the types of relations contained in a thesaurus, and they would
be equivalent to Eggins’ taxonomic relations. Syntagmatic relations are
such by virtue of linear relation, otherwise described as collocation. The
classification is further refined in terms of the type of expansion that the
tie provides, as either elaborating, extending, or enhancing. We will see
in Section 12.4 that the same three-way organizational principle applies
to conjunctive relations. The entire classification from Halliday and
Matthiessen (2014) is reproduced in Table 12.2, where we see how the
general list from 1976 which included all types of lexical relations plus
collocation has seen a new level superimposed, to take it to a more
abstract level. This does not mean, however, that collocation has been
made easier to understand. Halliday and Matthiessen note that colloca-
tion relations may or may not be enhancing. Collocation, in their view,
includes the type of expectancy relations mentioned as expectancy above,
but also collocation relations that happen in particular registers, such as
those in technical fields.
12.3 Cohesive Chains
Elements in the text related through cohesion establish a ‘cohesive tie’: the
interpretation of one element in the discourse depends on the interpret-
ation of another, whether preceding (anaphoric relation) or following (cat-
aphoric). The fact that the interpretation is successfully established creates
the cohesive tie between the two elements. As more elements are related to
each other, a ‘cohesive chain’ is created, a series of elements related to one
another. Chains are frequently created through identity of referent (I-I-my
in example (5)), but Hasan (1985:73) clearly states that the relation between
elements in a cohesive tie is semantic, i.e. the elements are tied together
through some meaning relation, even if the relation is expressed through
grammatical resources, such as personal pronouns.
One way of exploring the relationship in a cohesive tie, and ultimately in
cohesive chains, is to measure the distance between components. The
relationship may be immediate (the cohesive element refers to an immedi-
ately preceding one); remote (the referent is more than one clause away); or
mediated (where the ultimate referent is a few clauses earlier in the dis-
course, but has been recaptured in some other element). Although I often
refer to ‘preceding discourse’, this naturally applies only to anaphoric
relations; cataphora works in the opposite direction, by establishing links
that look forward to an element that completes the interpretation. Cata-
phora seems to be extremely rare: in a study of over 11,000 instances of
third-person pronouns, my colleague Radoslava Trnavac and I found only
fifty-seven instances of cataphora (Trnavac and Taboada 2016). Halliday and
Matthiessen (2014:625) also state that cataphora is rare, with the exception
of ‘structural cataphora’, where the reference to a pronoun is resolved
immediately afterwards, in the same clause. Such is the case in (12), where
the referent for those is in the relative clause immediately following (who are
just looking . . .).
(12) Highly recommended to anime fans and those who are just looking
for a film that is unique and interesting.
Cohesive chains run through texts, and provide them with the links to
create texture. Most texts (spoken or written) contain more than one chain.
The short excerpt in (13) contains at least five interrelated chains, as shown
in Table 12.3. The table breaks down the text into units, which are some-
what arbitrary, for ease of presentation, and do not necessarily correspond
to clauses or independent units. We can see that there is a chain relating to
the film under discussion, which includes Spirited Away as the first element
in the chain. The noun group Spirited Away is not technically a cohesive
element yet, as this is the beginning of the text, and the noun group does
not refer to anything preceding, although it naturally establishes links
outside the text proper, on the web page where this review appeared,3
and the web page from which the review is linked.4 The next element in
this chain, the latest, relates to Spirited Away through ellipsis of the Head
noun film. The chain continues with repetition of the noun film. A second
3
See www.imdb.com/title/tt0245429/reviews-25
4
Some of these ties are multimodal, because they relate different modalities. Multimodal cohesion and conjunction
are beyond the scope of this paper, but see Bateman (2008) for an overview of multimodality in discourse.
320 MAITE TABOADA
Table 12.3 Cohesive chains in example (13)
Unit Film Animation Miyazaki Excellence Reference frame
Spirited Away Spirited Away

is the latest in a the latest films animated Hayao
string of incredible Miyazaki
animated films by
Hayao Miyazaki,
the most renowned animator the most Japanese history
animator in Japanese renowned
history
and most say in the the best in world
best in world. (in world)
He takes a couple of this film He best (in the in the world
steps close to the world) title
best in the world
title with this film.
chain, related to the first, establishes that this is an animation film, and the
director then naturally an animator. The third chain refers to the director,
through reference in the form of the personal pronoun he in the second
sentence. Finally, two related chains establish Miyazaki as an excellent
director, first in a narrower frame of reference (Japan), and then more
broadly, as the best in the entire world.
(13) Spirited Away is the latest in a string of incredible animated films by

Hayao Miyazaki, the most renowned animator in Japanese history
and most say in the best in world. He takes a couple steps close to the
best in the world title with this film.
12.4 Cohesive Harmony
Cohesive chains and chain interaction are some of the most interesting
constructs for describing cohesion in text, and how texture is achieved.
Hasan (1984) proposed the idea of ‘cohesive harmony’, a measure of how
well-integrated cohesive chains are (see also Hoey 1991; Khoo 2016; Parsons
1996). Chains do not occur in isolation, but alongside other chains. How-
ever, the mere presence of two or more chains in a text does not guarantee a
cohesive effort. Although chains contribute to cohesion in a text, they need
to be related to each other somehow. This relationship is called ‘chain
interaction’. The relationships are mostly grammatical, as part of the tran-
sitivity structure of the clause, such as the relationship between Processes
and Participants. Hasan establishes a minimum requirement for chain
interaction: at least two members of one chain should stand in the same
relation to two members of another chain. For a better definition of the
interactions, she divides the tokens in a text into three categories:
• Relevant tokens: all tokens that enter into chains, further divided into:
○ Central tokens: relevant tokens that interact.
○ Non-central tokens: relevant tokens that do not interact.
• Peripheral tokens: Tokens that do not enter into any kind of chain.
We are, then, in a position to define cohesive harmony as the function of

three phenomena:
• Low proportion of peripheral tokens to the relevant ones.

• High proportion of central tokens to non-central ones.
• Few breaks in the interaction.
Hasan affirms that coherence is a function of cohesive harmony. Our

perception that a text is coherent, that it somehow makes sense, is depend-
ent on its cohesive harmony. This explains why example (1) is not coherent,
even though it contains cohesive ties: the ties do not form long chains, and
they do not interact with each other very much, outside of the sentence
where they appear. Example (13) above, on the other hand, shows a high
degree of chain interaction: the chain that contains words such as Spirited
Away and films is related to the chain that contains Hayao Miyazaki through
transitivity: by Hayao Miyazaki is the Postmodifier of films in the first sen-
tence, and the Subject in the last sentence, which contains the group this
film also as a Circumstance.
Many scholars have pointed out that cohesion and coherence are not all-
or-nothing categories, but rather a matter of degree. Parsons (1996) stated
that, in any given text type, there is a gradation dependent on the extent to
which a text relies on cohesion to provide coherence. Thus, texts belonging
to different text types and registers will show different degrees of cohesive
harmony. In a study of cohesion in task-oriented dialogue, I found very
little chain interaction (Taboada 2000, 2004). Breaks in cohesive chains, on
the other hand, were indicative of breaks in the text, where one stage of the
genre finished and another one started.
12.5 Conjunction as a Separate System
The system of conjunction has had an uneasy status as a member of the

general cohesion class. All other cohesive relations (reference, substitution,
lexical cohesion, and their cognates) are relations among entities in the
discourse, or propositions presented as entities (such as using that to refer to
a previous sentence). Conjunction, on the other hand, marks relations
between clauses in a clause complex, between text segments realized by
clause complexes, or between longer text segments, named rhetorical para-
graphs by Halliday and Matthiessen (2014:605).
Hasan (1985:81) makes a clear distinction between ‘componential’ and
‘organic’ cohesive devices. Componential devices are items that form part
322 MAITE TABOADA
of a cohesive tie, and are at the same time members of the transitivity
structure of the clause. It is the linking of those devices that creates cohe-
sion across clauses. Organic devices, i.e. conjunctive devices, tie whole
clauses rather than clause components. Hasan points out that other devices
of language organization, such as adjacency pairs, are also organic.
Conjunctive devices (conjunctions, some continuatives, and some
Adjuncts) link clauses in logico-semantic relations such as cause, conces-
sion, or condition. They also serve to indicate temporal and additive rela-
tions. In this sense, conjunction is closely related to the clause complexing
system, and moves cohesion outside of the clause proper. Halliday and
Hasan would always have cohesion act outside of the clause, but the view
taken here is more inclusive, whereby cohesive and conjunctive links occur
both within the clause and across clauses. Conjunction serves to indicate
that a relation exists between clauses (clause complexes, sentences, or
entire text passages), and sometimes it provides an indication of the nature
of the relation. This indication can be quite clear, such as the relation
signalled by because, or it can be underspecified, as is the case with and,
which can indicate a variety of relations.
This general idea, that conjunction relates clauses, or propositions, has
been made specific and instantiated under different theories, and different
taxonomies. Halliday and Hasan (1976) proposed a top-level classification
into additive, adversative, causal, and temporal. The classification is exclu-
sively based on the presence of a conjunctive item (e.g. so, consequently, for this
reason, or it follows for cause). In the first edition of Introduction to Functional
Grammar (Halliday 1985), and in subsequent editions, this semantic classifi-
cation (based on how the conjunction specifies the semantic content of the
linkage) is made more abstract, with a higher-level classification based on
how one clause adds to another, and using three types of connection:
elaboration, extension, or enhancement. This classification is based on the
form of the contribution, rather than the semantic meaning that is contrib-
uted, but typically relies on the presence of a conjunctive item.5
Martin (1992) proposed a slightly different classification of what he
named conjunctive relations, which are outside of cohesion proper. Add-
itionally, he expands on the internal/external distinction, relating
to whether relations refer to external relations, in the real world, or to
the internal organization of the events in the text. The latter are more
‘rhetorical’ in nature, in that they have to do with how arguments are
presented. The distinction is quite clear with temporal relations. An exter-
nal temporal relation describes sequences of activities as they occur in the
world. Internal temporal relations, on the other hand, capture time within
the text, i.e. in relation to what is being said and how the text is organized.
The two following examples are from Martin (1992:182), with the words in
5
Halliday does mention non-finite clauses as examples of conjunctive relations which do not have an explicit marker for
the relation.
bold, from the original text, indicating conjunctive items. Example (14)
presents an external temporal relation, where the events are presented as
they unfolded in the world. In (15), an internal relation, the sequence first-
second could have well been presented in the reverse order, depending on
the effect that the writer or speaker wanted to create.
(14) Ben came in and then had a drink.

(15) Ben wasn’t ready. First he hadn’t studied; and second, he’d been up
all night.
The internal/external distinction is cross-categorized with a semantic classi-

fication, much along the lines of Halliday’s (1985): additive, comparative,
temporal, consequential. What makes Martin’s description different, how-
ever, is the fact that he does not exclusively rely on conjunctions as signals
of a particular relation. For instance, one example given of similarity rela-
tions is that they are signalled by so followed by a Finite verb. That is, it is not
the conjunction alone, but a combination of the conjunction and lexico-
grammatical features that makes it clear which relation the writer/speaker
intends. Martin suggests that there is congruent and metaphorical signal-
ling of relations. Congruent signalling involves conjunctions, but metaphor-
ical signalling takes many other forms. For instance, the verbs enable, cause,
and follow may be signals of relations, equivalent to the conjunctions by,
because, or before. There is, then, a range of lexical and grammatical options
available to signal conjunctive relations. Scott and Thompson (2001) charac-
terize the range of possibilities as explicit signals (conjunctions such as
although or conjuncts such as on the other hand), large-scale signposts (We
can draw three main conclusions from this . . .), or no signal, leaving it to the
reader or hearer to establish the link, as well as the type of link.
Let us examine the issue of signalling through two invented examples.
We can describe the link between the two clauses in example (16) as a
‘classical’ example of cohesive conjunction, given the fact that it contains
the conjunction because.6 The case would not be so clear-cut, however, in
(17). There, the causal connection is presumably still active, but there is no
signal, and the relation occurs across sentence boundaries. There are, of
course, other cohesive devices in the two clauses/sentences that link them,
such as the reference Dominique – he and the lexical connection between job
and long hours (maybe long hours are attributes of some jobs). Regardless of
the other cohesive links, the reader still has to infer that there is a relation,
as well as what specific relation it is.
(16) Dominique quit his job because he was tired of the long hours.
(17) Dominique quit his job. He was tired of the long hours.
6
Not strictly ‘classical’ in the sense of Halliday and Hasan, since for them cohesion only takes place across, not within
sentences. For them, this example would be accounted for within clause complexing, not cohesive conjunction.
324 MAITE TABOADA
In examples (18) and (19) we see instances of explicitly signalled relations,

within and across sentences. In the first example, the concessive relation
between the two clauses is clearly marked by while, and the relation takes
place within the confines of the sentence. In (19), a similar concessive
relation occurs across sentence boundaries. In this case, it is signalled by
but at the beginning of the sentence.
(18) While these two themes are very much current in Japan, they are
also universal themes.
(19) Sometimes in real life the most grim moments contain honest elem-
ents of comedy that do not seem out-of-place. But trying to put that
sort of convoluted emotion into a film creates a very thin line that
too many have fallen off of.
The broadening of the scope of conjunction, beyond the boundaries of the

sentence, and beyond the confines of signalling by conjunctions, is the
realm of the phenomenon explored in theories of discourse, coherence, or
rhetorical relations. The emphasis shifts from the conjunctions themselves
to the relation, whether signalled or not. These relations have received
multiple names: rhetorical predicates (Grimes 1975); combinations of predi-
cations (Longacre 1976); coherence relations (Hobbs 1979; Sanders et al.
1992, 1993); rhetorical relations (Mann and Thompson 1988); or discourse
relations, a label for three different approaches (Polanyi 1988; Renkema
2009; Webber et al. 2003), and including work in Segmented Discourse
Representation Theory (Asher and Lascarides 2003). Despite distinct theor-
etical differences, and diverging treatment of the phenomenon, all the
approaches share an interest in explaining how discourse is coherent
through the combination of ideas or propositions. Among these, one of
the theories better connected to Systemic Functional Linguistics is Rhet-
orical Structure Theory (Mann and Thompson 1988; Mann et al. 1992).
Rhetorical Structure Theory (RST) was developed by Bill Mann, Christian
Matthiessen, and Sandra Thompson in an effort to represent text for compu-
tational purposes. It is closely aligned with SFL because Christian Matthiessen
was one of the developers of the theory, and the overall computational project
had Michael Halliday as a consultant. In RST, the view of conjunctive/rhet-
orical relations is top-down, that is, from the intention of the speaker or
writer, rather than from the lexicogrammatical realization, and the signalling
through conjunctions. Naturally, SFL always takes meaning-making as the
point of departure, but some of the descriptions, including those of cohesion,
are more bound to the lexicogrammar. In RST, on the other hand, intentions
are key, in part because RST evolved in the context of language planning and
Natural Language Generation (Mann 1983a, 1983b).
Rhetorical relations are equivalent to relations among clauses and clause
complexes at the most basic level. The difference is that relations can be
recursive, and apply at all levels of discourse. Thus, units of analysis are no
Figure 12.1 RST representation of example (20)
longer clauses or clause complexes, but ‘spans’. Unlike in the conjunction

system of Halliday and Hasan, or the conjunctive relations of Martin, in RST
there is no overarching taxonomy of relations, and relation lists are quite
flat, i.e. they have no hierarchy, although there is a distinction between
Subject Matter and Presentational relations. Subject Matter relations
(Circumstance, Condition, Elaboration, Cause, or Result) have the intended
effect that the reader recognize the relation. Presentational relations, on
the other hand, attempt to increase some inclination in the reader, such as
the desire to act, or the degree of positive regard for the nucleus of the
relation. Examples are Concession, Enablement, Evidence, Motivation, or
Summary. This is somewhat similar to the external/internal distinction of
Martin, discussed above.
RST is perhaps best explained through an example. In (20), we see a
short excerpt, this time of a review of the book Hot Six by Janet Evanovich,
posted online.7 The excerpt has been divided into clausal units, in the
example marked with square brackets. An RST analysis of this text is
presented in Figure 12.1. In the figure, each of the units of analysis (spans)
has a horizontal line on top. Additionally, they are connected to other units
through either straight or curved lines. For instance, the connection
between Spans 2 and 3 shows that Span 3 has a straight line above it,
marking it as the nucleus, whereas Span 2 has a curved line, an arrow
pointing to Span 3. This indicates that Span 3 is the satellite. The two
clauses are connected in a Condition relation: the action in the nucleus
(do not start the series by reading this book) has a condition attached to it, that
the reader should not have read any of the books in the series.
The terms ‘nucleus’ and ‘satellite’ refer to the relative importance of
spans in the organization of discourse. At the clause level, they are equiva-
lent to the concepts of main and subordinate clause in traditional grammar.
The difference is that this relative importance reaches across clause and
sentence boundaries. We see this in the unit that joins Spans 2–3 to Span 1,
7
www.epinions.com/review/Hot_Six_by_Janet_Evanovich_and_narrated_by_Debi_Mazar/2004218900/963557
326 MAITE TABOADA
which are related through a Background relation: the fact that the book
under review is the sixth in the series is background information necessary
for understanding the Condition conveyed in Spans 2–3. Here, Spans 2–3,
combined, are the nucleus, whereas Span 1 is the satellite. The excerpt has a
second part, composed of a relation which does not contain a nucleus
and a satellite, like the ones we have seen so far, but which instead is made
up of three nuclei, three units of equal importance (Spans 3–6). Such
relations are referred to as ‘multinuclear’ relations. Together, these three
units constitute a Sequence relation, outlining the steps that the author
thinks the reader should take. Finally, these two macro-units or sequences
of spans (1–3 and 3–6) are joined together into another multinuclear rela-
tion, this time of Contrast. The author establishes a contrast between two
possibilities, not reading the book under review and reading the previous
ones in order.
(20) [This is the sixth book in the Stephanie Plum series.] [If you have
never had the fun of reading a book in this series,] [do not start with
this one.] [Go to the library] [and start with One For The Money] [and
work your way up to Hot Six.]
This short example illustrates the principles of nuclearity and recursion

that are fundamental to RST analyses. The important aspect with regard to
cohesion is that these relations are postulated to exist even in the absence of
conjunctive devices to signal them. The Condition relation between Spans
2 and 3 has a nice if to indicate its presence. But the higher-level Contrast
relation does not have any conjunction to guide the reader. It is, in fact, the
presence of other lexicogrammatical items, including other cohesive
devices than conjunction, that gives clues to the relation, like the lexical
chain linking books and libraries, the chain start – work your way up, or the
repeated use of imperatives, first a negated one in Span 3, and then three
consecutive imperative clauses in Spans 4–6.
I return now to the concept of texture. The texture of a text, the way in
which it makes sense, is brought about by this interaction between cohesive
devices proper (reference, substitution and ellipsis, lexical organization)
and a wider view of conjunction in the form of rhetorical relations.
Together, cohesion and conjunction/rhetorical organization create the weft
and warp that weave together a text.
12.6 Computational and Other Applications
The general concept of cohesion has found favour in different areas of

knowledge, because of how elegantly it accounts for why a text may or
may not be seen as coherent, and how adjacent portions of a text are
connected. Thus, it has been used in computational applications to measure
coherence of generated text or to detect breaks where a new topic is being
introduced; in educational contexts to measure coherence of students’

texts; and in multimodal studies to describe the connection among
modalities.
Perhaps the most extensive area of application has been in computational
linguistics. Simplified treatments of the concept of lexical chains, involving
only semantic relations, are frequently used to segment text in chunks
(Purver 2011). This is the idea behind TextTiling, one of the most popular
discourse-oriented computational algorithms (Hearst 1994). In TextTiling,
text is divided into topic units, that is, chunks with a common topic each, by
examining relations of lexical cohesion, and positing a break between
chunks when lexical relations seem to fall below a certain threshold. Morris
and Hirst (1991) showed that it is possible to compute lexical chains, and
that the chains are useful in determining text structure. Cohesive chains are
sometimes computed using only lexical cohesion (in which case they are
referred to as lexical chains), and sometimes include (pronominal) reference.
Cohesive or lexical chains have turned out to be quite useful in multiple
computational applications. For instance, in text summarization Alonso i
Alemany and Fuentes Fort (2003) compute lexical chains and use the char-
acteristics of chains to decide which parts of the text to use in an automatic
summary. Characteristics of chains which are good candidates include their
length, the kinds of cohesive relations in them, and the point of the text
where they start. This lexical information is also enhanced with features
derived from the rhetorical structure of the text. Similar linkages between
lexical cohesive devices and rhetorical or coherence relations are proposed
in other computational work (Cristea et al. 1998; Harabagiu 1999).
Textual coherence is the goal of anyone who desires to master a foreign
language, a new register, or a form of language for specific purposes.
Therefore, an accurate measure of cohesion can show how far from or close
to that goal one is. In educational contexts, cohesion helps establish how
non-native language differs from native-like texts. For instance, Schleppe-
grell (1996) examines and classifies the use of the conjunction because in
native and non-native writing, and finds an overuse by non-native speakers.
Native speakers employ additional constructions, and thus show more
variety in their conjunctions and other connectives. McNamara et al.
(2014) propose Coh-Metrix, a cohesion measure that includes many factors
(such as lexical diversity and syntactic complexity), but also uses the types
of cohesive links established throughout the text as a way to indicate how
coherent the text is. An area where cohesion has been consistently applied
is in the automatic scoring of essays, in particular those produced by
foreign language learners for tests such as TOEFL (Burstein et al. 2010;
Rahimi et al. 2015; Somasundaran et al. 2014). A related application is in
machine translation, where cohesion is used to measure the coherence of
machine-translated text (Wong and Kit 2012).
Outside of the computational arena, interesting treatments of cohesion
include the connection between coherence/rhetorical relations and genre.
328 MAITE TABOADA
Taboada (2000, 2004) showed that breaks in lexical chains correspond to

breaks in the genre structure of task-oriented conversations. This is the
same intuition behind automatic text segmentation methods: a new stage
or a new chunk of the text involves some sort of break in the set of cohesive
chains in the text.
Most of the research described in this chapter has been carried out in English.
Analyses of cohesion in other languages exist, however, and they show subtle
and interesting contrasts with the available descriptions of English. To take
just two examples, Berzlánovich and Redeker (2012) study the interaction
between genre, cohesion, and coherence in an annotated corpus of Dutch
texts, as do Kunz and Lapshinova-Koltunski (2014) for German.8
12.7 Conclusion
Cohesion and conjunction (in the form of rhetorical relations) constitute

the two fundamental properties of text, the properties that give it coher-
ence, or texture. Cohesion establishes links across entities in the text,
whereas conjunction links propositions at all levels (clause, clause complex,
and paragraph). Rhetorical relations capture what we want to achieve with
the text, in terms of the intention of the text creator.
The exact distinction between cohesion, conjunction, coherence, and
texture has sometimes been criticized as unclear (Brown and Yule 1983).
Flowerdew (2013) points out that there is ambiguity in the description,
especially because Halliday and Hasan (1976), in addition to cohesion,
include register and thematic development as sources of texture. The most
likely answer is that all aspects of context contribute to texture, but we tend
to include cohesion and conjunction as the most clear-cut phenomena
to account for the perception of coherence in text. We can summarize
cohesion as links by means of referential, lexical, and logical ties
(Eggins 2004: 53).
8
In addition, for further reading on cohesion, consider the following. The original description of cohesion (in English) is
Halliday and Hasan (1976), and it still remains the most detailed account of the phenomenon, with plenty of examples.
The theory was refined and developed in a book by Halliday and Hasan (1985), and in particular a chapter by Hasan in
that book (Hasan 1985). Several introductions to Systemic Functional Linguistics explain cohesion in very concise
terms, often in a single chapter (Eggins 2004; Thompson 2014), but perhaps the most clear and concise is
Flowerdew’s introduction to SFL (Flowerdew 2013), which emphasizes applications of Systemic Functional Linguistics
to language education. And, of course, Halliday and Matthiessen’s (2014) Halliday’s Introduction to Functional
Grammar remains the ‘official’ version of the theory, with a chapter on cohesion and its place in the functional analysis
of language.
More specialized descriptions focus on cohesion, often with an introduction that then leads to an in-depth study,
typically corpus-based. Tanskanen (2006) does not strictly follow Halliday and Hasan’s classification, but hers is a
thorough corpus-based study. Fox (1987) presents an analysis of both cohesion and coherence in conversational
speech.
References
Alonso i Alemany, L. and M. Fuentes Fort. 2003. Integrating Cohesion and

Coherence for Automatic Summarization. Proceedings of EACL’03 Student
Research Workshop. Budapest, Hungary. 1–8.
Asher, N. and A. Lascarides. 2003. Logics of Conversation. Cambridge:
Berzlánovich, I. and G. Redeker. 2012. Genre-dependent Interaction of
Coherence and Lexical Cohesion in Written Discourse. Corpus Linguistics
and Linguistic Theory 8(1): 183–208.
Brown, G. and G. Yule. 1983. Discourse Analysis. Cambridge: Cambridge
University Press.
Burstein, J., J. R. Tetreault, and S. Andreyev. 2010. Using Entity-based
Features to Model Coherence in Student Essays. Proceedings of
Human Language Technologies: The 11th Annual Conference of the North Ameri-
can Chapter of the Association for Computational Linguistics. Los Angeles.
681–4.
Cristea, D., N. Ide, and L. Romary. 1998. Veins Theory: A Model of Global
Discourse Cohesion and Coherence. Proceedings of the 36th Annual Meeting
of the Association for Computational Linguistics and the 17th International
Conference on Computational Linguistics (ACL-98/COLING-98). Montreal,
Canada. 281–5.
Cruse, D. A. 2000. Meaning in Language: An Introduction to Semantics and
Pragmatics. Oxford: Oxford University Press.
Eggins, S. 2004. Introduction to Systemic Functional Linguistics. 2nd ed. London:
Continuum.
Fellbaum, C., ed. 1998. WordNet: An Electronic Lexical Database. Cambridge,
MA: MIT Press.
Flowerdew, J. 2013. Discourse in English Language Education. New York:
Routledge.
Fox, B. A. 1987. Discourse Structure and Anaphora: Written and Conversational
English. Cambridge: Cambridge University Press.
Grimes, J. E. 1975. The Thread of Discourse. The Hague: Mouton.
Arnold.
Halliday, M. A. K. and R. Hasan. 1985. Language, Context, and Text:
Aspects of Language in a Social-semiotic Perspective. Oxford: Oxford University
Press.
Functional Grammar. 3rd ed. London: Arnold.
Functional Grammar. 4th ed. London: Arnold.
330 MAITE TABOADA
Harabagiu, S. 1999. From Lexical Cohesion to Textual Coherence: A Data

Driven Perspective. International Journal of Pattern Recognition and Artificial
Intelligence 13(2): 247–65.
standing Reading Comprehension. Newark: International Reading Associ-
ation. 181–219.
Hasan, R. 1985. The Texture of a Text. In M. A. K. Halliday and R. Hasan,
Language, Context, and Text: Aspects of Language in a Social-semiotic Perspective.
Oxford: Oxford University Press. 70–96.
Hearst, M. 1994. Multi-Paragraph Segmentation of Expository Text. Proceed-
ings of 32nd Annual Meeting of the Association for Computational Linguistics
(ACL’94). Las Cruces, New Mexico. 9–16.
Hobbs, J. 1979. Coherence and Coreference. Cognitive Science 6: 67–90.
Hoey, M. 1991. Another Perspective on Coherence and Cohesive Harmony.
In E. Ventola, ed., Functional and Systemic Linguistics: Approaches and Uses.
Berlin: Mouton de Gruyter. 385–414.
Khoo, K. M. 2016. ‘Threads of Continuity’ and Interaction: Coherence,
Texture and Cohesive Harmony. In W. L. Bowcher and J. Y. Liang, eds.,
Society in Language, Language in Society: Essays in Honour of Ruqaiya Hasan.
New York: Palgrave Macmillan. 300–30.
Kunz, K. and E. Lapshinova-Koltunski. 2014. Cohesive Conjunctions in
English and German: Systemic Contrasts and Textual Differences. In
L. Vandelanotte, K. Davidse, C. Gentens, and D. Kimps, eds., Recent
Advances in Corpus Linguistics: Developing and Exploiting Corpora. Amster-
dam: Rodopi. 229–62.
Longacre, R. E. 1976. An Anatomy of Speech Notions. Lisse: The Peter de Ridder
Press.
Mann, W. C. 1983a. An Overview of the Nigel Text Generation Grammar: ISI/RR-
83–113. Information Sciences Institute, University of Southern
California.
Mann, W. C. 1983b. An Overview of the Penman Text Generation Grammar: ISI/RR-
83–114. Information Sciences Institute, University of Southern
California.
Mann, W. C. and S. A. Thompson. 1988. Rhetorical Structure
Theory: Toward a Functional Theory of Text Organization. Text 8(3):
243–81.
Benjamins.
McNamara, D. S., A. C. Graesser, P. M. McCarthy, and Z. Cai. 2014. Automatic
Evaluation of Text and Discourse with Coh-Metrix. Cambridge: Cambridge
University Press.
Morris, J. and G. Hirst. 1991. Lexical Cohesion Computed by Thesaural

Relations as an Indicator of the Structure of Text. Computational Linguistics
17(1): 21–48.
Parsons, G. 1996. The Development of the Concept of Cohesive Harmony. In
M. Berry, C. S. Butler, R. Fawcett, and G. Huang, eds., Meaning and Form:
Systemic Functional Interpretations (Meaning and Choice in Language: Studies for
Michael Halliday). Norwood: Ablex. 585–99.
Polanyi, L. 1988. A Formal Model of the Structure of Discourse. Journal of
Pragmatics 12: 601–38.
Purver, M. 2011. Topic Segmentation. In G. Tur and R. De Mori, eds., Spoken
Language Understanding: Systems for Extracting Semantic Information from
Speech. Hoboken: Wiley. 291–317.
Rahimi, Z., D. Litman, E. Wang, and R. Correnti. 2015. Incorporating
Coherence of Topics as a Criterion in Automatic Response-to-Text
Assessment of the Organization of Writing. Proceedings of the Tenth Work-
shop on Innovative Use of NLP for Building Educational Applications. Denver.
20–30.
Renkema, J. 2009. The Texture of Discourse. Amsterdam: John Benjamins.
Sanders, T., W. Spooren, and L. Noordman. 1992. Toward a Taxonomy of
Coherence Relations. Discourse Processes 15(1): 1–35.
Sanders, T., W. Spooren, and L. Noordman. 1993. Coherence Relations in a
Cognitive Theory of Discourse Representation. Cognitive Linguistics 4(2):
93–133.
Schiffrin, D. 1987. Discourse Markers. Cambridge: Cambridge University
Press.
Schleppegrell, M. J. 1996. Strategies for Discourse Cohesion: Because in ESL
Writing. Functions of Language 3(2): 235–54.
Scott, M. and G. Thompson. 2001. Introduction: Why ‘Patterns of Text’? In
M. Scott and G. Thompson, eds., Patterns of Text: In Honour of Michael Hoey.
Somasundaran, S., J. Burstein, and M. Chodorow. 2014. Lexical Chaining for
Measuring Discourse Coherence Quality in Test-taker Essays. Proceedings
of the 25th International Conference on Computational Linguistics (COLING).
Dublin, Ireland.
Stenström, A.-B. 1994. An Introduction to Spoken Interaction. London:
Longman.
Taboada, M. 2000. Cohesion as a Measure in Generic Analysis. In A. Melby
and A. Lommel, eds., The 26th LACUS Forum. Chapel Hill: The Linguistic
Association of Canada and the United States. 35–49.
Taboada, M. 2004. Building Coherence and Cohesion: Task-oriented Dialogue in
English and Spanish. Amsterdam: John Benjamins.
Tanskanen, S.-K. 2006. Collaborating towards Coherence: Lexical Cohesion in
English Discourse. Amsterdam: John Benjamins.
Thompson, G. 2014. Introducing Functional Grammar. New York:
Routledge.
332 MAITE TABOADA
Trnavac, R. and M. Taboada. 2016. Cataphora, Backgrounding and Accessi-

bility in Discourse. Journal of Pragmatics 93: 68–84.
Webber, B., M. Stone, A. K. Joshi, and A. Knott. 2003. Anaphora and
Discourse Structure. Computational Linguistics 29(4): 545–87.
Wong, B. T. M. and C. Kit. 2012. Extending Machine Translation Evaluation
Metrics with Lexical Cohesion to Document Level. Proceedings of the
2012 Joint Conference on Empirical Methods in Natural Language Processing
and Computational Natural Language Learning. Jeju Island, Korea. 1060–8.
13
Semantic Networks
Andy Fung and Francis Robert Low
13.1 Introduction
Discourse studies featuring Systemic Functional Linguistics (SFL) as the

framework of analysis have gained increasing momentum in the past
decades (see Hyland and Paltridge 2011; Gee and Handford 2012; Hyland
2013). It is perhaps not surprising, because this theory of language is
essentially ‘appliable’, attaching fundamental importance to ‘social
accountability’ (Halliday 2006b, 2006a; Mahboob and Knight 2010a,
2010b; Matthiessen 2012, 2014). That is to say, it is a linguistic theory
committed to solving the problems encountered by language users in their
daily social practices (Halliday 2008:189). From a systemic point of view, the
emphasis on ‘appliability’ is particularly relevant to discourse analysis
because, without theory, the analysis of discourse would be ad hoc, incon-
sistent, and ineffective (Halliday 2006b:19), and by the same token, without
the analysis of discourse, there would be no raison d’être of SFL since
scholars working within SFL model language use as their starting point
(Mahboob and Knight 2010b:1). Given the strong orientation to discourse in
SFL, a number of approaches and discourse tools have been proposed and
developed, enabling discourse analysts to interpret and make sense of the
meaning of ‘what people say and write and listen to and read’ (Halliday
1994:xxii).1
Take semantics as an illustration. In SFL, language is conceptualized as a
meaning potential, and semantics is a stratum in the SFL model of lan-
guage. This stratum, together with lexicogrammar, phonology, and phon-
etics, constitute the language-internal strata (for a recent review of
stratification of language, see Halliday and Matthiessen 2014; also see
Hasan 2013, 2014). Above these language–internal strata is the stratum of
1
While recent years have witnessed a growing trend in analyses of multimodal discourse, this chapter, following Hasan
(2014:3), regards language as the central object of enquiry in SFL.
334 AN DY F U NG AN D FR ANCI S ROB ERT LOW
context — the extra-linguistic universe where the text is functioning. Being

a ‘linguistic inter-level to context’ (Matthiessen 1993:227; see also Halliday
2009; Hasan et al. 2007; Hasan 2009c, 2010), semantics serves as the point
of departure in describing and accounting for context and lexicogrammar.
More importantly, it enables analysts to make sense of human life, since
most of our daily social practices are essentially ‘acts of meanings’ (Hasan
2010:267). Given the significance of semantics, various studies have been
undertaken since the late 1960s, each of which has produced a number of
semantic descriptions (for a historical development of semantics in SFL, see
Matthiessen 2007, 2009, 2015a).
So far, there are at least seven orientations in SFL semantic descriptions,
each of which has developed its own set of analytical tools. Though they are
slightly different in their own right, all of them enable discourse analysts to
study the ‘meanings we give language and the actions we carry out when
we use language in specific contexts’ (Gee and Handford 2012:1). Table 13.1
summarizes these research approaches and semantic descriptions.
The semantic description discussed in this chapter is based on Ruqaiya
Hasan’s message semantics system networks. The first publication of mes-
sage semantics networks appeared in 1983 in a mimeo prepared for her
project investigating the different ways of meaning construed in mother–
child talk. Remaining unpublished and circulated only among her team
members, this pioneering work served as the foundation of message
Table 13.1 Semantic descriptions in SFL (building on Cloran et al. 2007)
Orientations Semantic descriptions
1) Text texture – COHESION (e.g. Halliday and Hasan 1976)

– COHESIVE HARMONY ANALYSIS (e.g. Hasan 1984)
2) Sociolinguistics and semantic – SOCIOLOGICAL SEMANTIC NETWORKS
variation (e.g. Halliday 1973; Turner 1973)
– MESSAGE SEMANTICS NETWORKS
(e.g. Hasan 1983, 1996, 2009e; Hasan et al. 2007)
3) Discourse structure in – RHETORICAL UNIT ANALYSIS
constituency terms (e.g. Cloran 1994, 1999)
4) Discourse structure in – RHETORICAL STRUCTURE THEORY
dependency terms (e.g. Matthiessen 1988a, 2004; Halliday and
Matthiessen 1999)
5) Discourse structure in phasal – PHASAL ANALYSIS (e.g. Gregory 1985)
terms
6) Collaborative and interactive – SPEECH FUNCTION NETWORKS
exchange of dialogue (e.g. Halliday 1984; Martin 1992; Matthiessen
1995; Eggins 1990; Eggins and Slade 2004)
7) Discourse semantics – IDEATION,
– CONJUNCTION,
– NEGOTIATION,
– INVOLVEMENT,
– APPRAISAL,
– IDENTIFICATION, and
– PERIODICITY
(see Martin 1992, 2000; Martin and White 2005;
Martin and Rose 2007; Martin 2014)
Semantic Networks 335
semantics, semantic variation, and sociolinguistic research. Despite the fact

that her semantic description was developed specifically for semantic vari-
ation research (e.g. Cloran 1994; Williams 1995; Hasan 2009e), the past
years have witnessed an increasing number of discourse studies featuring
message semantics networks as the tool for the analysis of meaning (e.g.
Hall 2004; Wake 2006; Hasan et al. 2007; Wong 2009; Chu 2011; Lukin et al.
2011; Lukin 2012, 2013). Having benefited from Hasan’s outstanding work
for over three decades, the time is ripe to survey the application of message
semantics.
This chapter is thus organized as follows. Section 13.2 briefly describes
the theoretical foundation of Hasan’s message semantics networks. Elabor-
ating Halliday’s theoretical concepts of sociological semantic networks to a
large extent, Hasan has developed her own position in terms of semantic
description. Such elaborations and subsequent developments constitute the
basis of what is understood as ‘message semantics networks’. Given the
close association between Hasan’s and Halliday’s semantic networks, we
first discuss the concepts postulated by Halliday (see Section 13.2.1),
followed by a discussion on the theoretical constructs which led to Hasan’s
message semantics networks (see Section 13.2.2). Having discussed the
theoretical constructs of message semantics networks, in Section 13.3 we
then move to the primary use of Hasan’s networks, with a particular focus
on the notion of semantic variation. Section 13.4 discusses the extended
uses of semantic networks in discourse studies. Here, we report on two
domains of investigations which employ message semantics networks as
the primary analytical tool, thereby illustrating how the networks are used
and presenting the associated research implications.
13.2 The Theoretical Concepts of Semantic Networks
13.2.1 Halliday’s Sociological Semantic Networks

Early in the 1970s, Halliday published the first paper on semantic networks,
entitled ‘Towards a Sociological Semantics’. As the title suggests, a key
point in Halliday’s work is that the semantic description is grounded in
Bernstein’s theories of socialization and social learning, and attaches fun-
damental importance to the connection between social context and linguis-
tic meanings. In theorizing the meanings accessible to speakers, Halliday
recognizes that verbal behaviour is essentially a phenomenon which can be
described sociologically and linguistically. These descriptions, however,
could not be directly related because the social system is ‘wholly outside
language’ and the grammatical system is ‘wholly within language’ (Halli-
day 1973:88). To relate these descriptions and illustrate how ‘social mean-
ings are organised into linguistic meanings’ (Halliday 1973:72), Halliday
proposed the idea of ‘semantic network’, defining it as a ‘hypothesis about
patterns of meaning’ which forms a bridge between the ‘behavioural
patterns and linguistic forms’ (Halliday 1973:75). The semantic description

is important in that it draws on the system network as representation, that
is, meanings are represented as options within systems, and each option is
systemically related to one another (Halliday 1973:68). These semantic
options, as maintained by Halliday (1973:68), are context-dependent in
the sense that they reflect only ‘what the speaker can do, linguistically,
in a given context’. Such an approach to semantic description is not only
‘context-based’,2 but also ‘strategic’ in nature. As Matthiessen (1990:324–5,
emphasis added) writes:
When we approach semantics from above it is the interface between context

and language that is highlighted. The role of semantics can be stated with
respect to context as follows: semantics is the set of strategies for construing
contextual meanings as linguistic meanings and thus moving into the linguistic
system. Or if we focus on the notion of goal in particular, semantics is the set of
strategies for achieving some goal through symbolic activity. This is a functional
approach to semantics: it interprets semantics in terms of the uses it has
evolved to serve in different communicative contexts.
For example, Halliday (1973) postulates a semantic network of parental

control, illustrating the sets of strategies that a mother could employ in
the regulatory context. Such strategies, or more specially, the goal-oriented
symbolic activities, are semanticized as various semantic options. As shown
in Figure 13.1, in regulating the behaviour of a child, a mother could either
select the option [threat] or [warning] – the former denotes the actions that
are threatened to be undertaken by care-givers, whereas the latter refers to
the possible undesirable consequences of a child doing something that he or
she is being told not to do. Both [threat] and [warning] serve as the point of
entry to further sub-options.
Important in this network representation is that not only are the options
clearly identified and related, but they are also specified in terms of lexico-
grammatical realization statements. That is to say, each semantic option is
viewed from the lexicogrammatical stratum. Take [physical punishment]
as an example. Halliday suggests that [physical punishment] is a sub-
category of [threat], defining it as follows:
The ‘threat’ may be a threat of physical punishment. Here the clause is of the
action type, and, within this, of intentional or voluntary action, not super-
vention (i.e. the verb is of the do type, not the happen type). The process is a
two-participant process, with the verb from a lexical set expressing ‘punish-
ment by physical violence’, roughly that of § 972 (PUNISHMENT) in Roget’s
Thesaurus, or perhaps the intersection of this with § 276 (IMPULSE). The tense
is simple future. The Goal, as already noted, is you; and the clause may be
2
This contrasts with the description of semantics from below, or chooser and inquiry semantics (in Matthiessen’s 1990
terminology), which is typically employed in the model of text generation. Examples include Matthiessen (1988b),
Patten (1988), and Matthiessen and Bateman (1991), to name but a few. See Matthiessen (1990) for details.
Figure 13.1 The semantic network of warning and threat (Halliday 1973:89)
either active, in which case the agency of the punishment is likely to be the
speaker (I as Actor), or passive, which has the purpose of leaving the agency
unspecified.
(Halliday 1973:78, emphases in original)
Given this orientation, the semantic option [physical punishment] is lex-

icogrammatically realized as ‘clause: action: voluntary (do type); effective
(two-participant): Goal = you; future tense; positive; verb from Roget § 972
(or 972, 276)’, as in I will smack you, Daddy will smack you, or You’ll get smacked.
The total set of semantic options, together with their lexicogrammatical
realization statements, constitutes the ‘register-specific semantic potential’
(Hasan 1996:114).
13.2.2 Hasan’s Message Semantics Networks

The semantic network postulated by Halliday in the 1970s was still in a
nascent form, and it is not surprising that there remained much room for
further development. With the subsequent advancement of SFL, the
conceptions of semantic networks have been greatly refined and elaborated.

The major features of Hasan’s message semantics include (1) open context,
(2) unit of analysis, (3) trinocularity, and (4) metafunctional regulation (see
also Williams 1995; Hasan 1996, 2009e, 2013, 2014).
13.2.2.1 Open Context

While Halliday’s network was ‘strategic’ in nature, inviting a description of
the meaning of a given situation (Matthiessen 1990, 2015b), Hasan’s network
has been developed to be contextually open. In Hasan’s view, a situation-
specific semantic network is less desirable, if not impractical, when investi-
gating the meanings at play in her mother–child talk research since the talk
per se entails extensive contexts (Hasan et al. 2007).3 In this sense, rather than
perceiving the context specificity as a ‘categorical one’, she argues that one
should view it as a ‘relative matter’ so that the semantic networks could
serve as a ‘heuristic device for the definition of a specific class of context of
situation’ (Hasan 2009a:151). In so doing, she relocates the networks from
the mid region of the cline of instantiation to the potential end of the cline,
aiming at a description of general semantic systems, or more precisely, an
account of the ‘meaning potential of English’4 (Hasan et al. 2007:712; see also
Fung 2016). One important consequence is that semantic networks are no
longer strategic, but essentially social in the sense that the approach focuses
on the nature of the linguistic meanings in general.
13.2.2.2 Unit of Analysis: Message

In presenting a more general conception of semantic networks, Hasan postu-
lates a hierarchy of units, or more specifically, a four-unit rank scale in
English, moving from the highest to the lowest: text ~ rhetorical unit ~
message ~ seme (Hasan 2013).5 As in other language-internal strata, these
units stand in a relation of constituency, that is, a text is made up of rhetorical
units, a rhetorical unit of messages, and a message of semes. Among the four
units, Hasan regards message as the ‘ultimate descriptum’ in semantics
(Hasan 2014:10), defining it as ‘the smallest semantic unit which is capable
of realising an element of the structure of a text’ (Hasan 1996:117). For Hasan,
it is this descriptum which serves as the object of enquiry in semantics, and is
described exhaustively in system networks in terms of semantic options (for a
recent account, see Hasan 2013, 2014). As pointed out by Hasan (1989:245), a
message can be further categorized in terms of its productivity. A message
3
As illustrated in Hasan’s subsequent work in contextual modelling, mother–child talk is essentially registerially/
contextually inconsistent, entailing frequent reclassifications of con/text as the talk develops (see Cloran 1999; Hasan
1999, 2000 for a detailed discussion on con/textual shift).
4
Though the account of meaning potential is robust, such descriptions, as noted by Hasan, are not yet exhaustive.
Further tests and applications are thus needed.
5
It should be emphasized that the term ‘message’ is also used in another sense in SFL, denoting the textual unit of
meaning (Halliday and Matthiessen 2004:588–9). In this chapter, ‘message’, following Hasan, refers to the semantic
rank scale only.
which is productive and forms structural paradigms is termed a progressive

message. That is, it concerns proposition/proposal exchange and is typically
realized in a major clause. By contrast, a message which is non-productive
and serves to manage the ongoing interaction is termed a punctuative mes-
sage. The default realization of punctuative messages is a minor clause, such
as Hello, Sorry, or Hey. According to Hasan, it is only progressive messages
which are metafunctionally regulated (for detail, see Section 13.2.2.4; see also
Hasan 2013 for a recent review).
13.2.2.3 Trinocularity
Following Halliday, each semantic option in Hasan’s message semantics
network attaches fundamental importance to the ‘concept of trinocularity’
(see Halliday 2009:79–80). That is, Hasan’s semantic networks not only
concern the interrelations among semantic options (i.e. whether the options
postulated are internally duplicate or contradictory (Hasan 1996:110)), but
also emphasize the relations with context (i.e. what contextual features are
construed) and lexicogrammar (i.e. what lexicogrammatical patterns are
activated). In other words, the analysis of meaning through utilizing seman-
tic networks not only illustrates the meanings at risk, but also enables
analysts to explain ‘why and how something is said’ and ‘why these patterns
of wordings appear rather than any other’ (Hasan 2009c:170).
13.2.2.4 Metafunctional Regulation

In Hasan’s message semantics networks, semantic options are ‘multi-focal’
(Hasan 1996:111) in the sense that they relate not only inter-stratally, but
also metafunctionally, i.e. each semantic option pertains to the highly gen-
eralized functions of language, which are known as metafunctions. Early in
the 1970s, Halliday identified three metafunctions: the ideational, the inter-
personal, and the textual, where the ideational metafunction is further
categorized into the experiential and the logical subtypes (for details, see
Halliday and Hasan 1985: Chapter 2). Central to this metafunctional hypoth-
esis is that language is functional in the sense that the functions of a
language are the ‘fundamental principle of language’ and are ‘basic to the
evolution of the semantic systems’. Hasan incorporates Halliday’s metafunc-
tional hypothesis into her semantic descriptions, arguing that a progressive
message entails four simultaneous systems (from Hasan 1989:224):
(a) interpersonal meanings, for example, options in message function

(questioning, informing, commanding . . .), options in personal evalu-
ation, point of view;
(b) experiential meaning, for example, the ascription of actional, evolu-
tional, etc. roles, identification, definition; construction of time;
(c) logical meaning, for example, cause, condition, meta-textual relations;
(d) textual meanings, for example, options in topic maintenance, topic
changes.
Figure 13.2 The overall organization of semantic networks of ‘progressive message’

(Hasan 2013:286)6
Gloss: (1) the arrow indicates the point of entry to the sub-systems; (2) the dots indicate the
name of the sub-systems.
These four systems of meanings are termed relation enactment, con-

tinuation, amplification,7 and classification respectively (Hasan
2013).8 It should be emphasized that these four systems are simultaneous,
that is, they are of equal status and no cluster of meaning is more powerful
or more important than the others. As one moves through the networks
and chooses options, choices are made from most primary (at the left-hand
end of the systems) to most delicate (choices at the right-hand end of the
systems). The increase of degree of delicacy yields a full account of mean-
ings within a single message. Figure 13.2 illustrates the overall organization
of message semantics networks. Hasan’s proposal for metafunctionally
regulated semantic networks, compared with Halliday (1973), constitutes
a significant advance in semantic description because not only can subtle
meaning differences be captured, but it also enables analysts to explore the
calibration of context, semantics, and lexicogrammar (see Section 13.4).
6
Due to space constraints, Figure 13.2 only includes the primary sub-system for a message with the feature
[progressive] under the four metafunctions. For example, the topic indicates the sub-system of CONTINUATION,
selecting between [turn-maintaining] and [turn-changing]. Unfortunately, lack of space precludes a detailed
discussion of each semantic feature. For detail, see Williams (1995); Lukin (2012, 2013).
7
It should be noted that the term AMPLIFICATION was previously used in Martin’s earlier accounts of APPRAISAL (see
Martin 2000). However, it has been re-labelled GRADUATION in Martin and White (2005). Following Hasan, the term
AMPLIFICATION used here refers to the semantic system of logical meanings.
8
Contra Halliday, Hasan separates the experiential metafunction from the logical metafunction, leading to four systems
of meanings.
In summary, these four primary systems constitute a major advance in

semantic description, leading to our current conception of ‘message seman-
tics networks’ (e.g. Hasan 1996, 2009e; Hasan et al. 2007). Such networks, as
shown in the following sections, have a strong descriptive power which
enables a variety of research problems to be tackled.
13.3 Application of Semantic Networks: An Investigation

of Semantic Variation
As noted in our introductory section, the genesis of Hasan’s semantic

networks lies in the need to investigate semantic variation, the central
research agenda of Hasan and her colleagues in the early 1980s (see also
Cloran 2000; Williams 1995, 2005; Hasan et al. 2007). Prior to illustrating
how Hasan’s semantic networks contribute to the understanding of seman-
tic variation, we briefly revisit the conception of semantic variation (see
Section 13.3.1), followed by a review of her research into mother–child
interaction (see Section 13.3.2). Section 13.3.3 then illustrates how analysis
utilizing semantic networks serves as the linguistic evidence of semantic
variation.
13.3.1 The Conception of Semantic Variation

Sociolinguistic studies focusing on linguistic variation are not in themselves
novel; they have a long history, with rich descriptive accounts focusing on
phonological or morphosyntactic variation (e.g. Labov 1972, 1978). While
the Labovian framework of variation has gained widespread acceptance in
sociolinguistic research, Hasan’s work on linguistic variation is unique and
innovative in the sense that her approach is meaning– and sociologically–
oriented, and situated within the Systemic Functional model of language.
Such a pioneering approach, as remarked by Hasan, reflects her dissatis-
faction with sociolinguistic variation studies conducted in the 1970s, in
terms of (i) analytical framework and (ii) variation explanations.
With regard to the former, Hasan recognizes that variation frameworks
which focus almost exclusively on phonology and lexicogrammar are essen-
tially ‘meaning preserving’, thereby giving no place to meaning variations.
For Hasan, neither phonology nor lexicogrammar is the ‘site of socially
significant variation’ (Hasan 2011:xxxvii) – it is the level of semantics which
entails ‘all the necessary characteristics of language varieties’ (Hasan
2009a:144). Thus, rather than perceiving semantics as ‘immune to vari-
ation’ (Hasan 1989:269), Hasan takes semantics as the point of departure
in her study, with a particular focus on the ‘systematic differences in
selection and organisation of linguistic meanings’ (Hasan 2009a:144, see
also Hasan 1989, 2009b, 2009d, 2009f). Hasan postulates that approaches to
linguistic variation which feature no social theory in explaining variation
are undesirable. In so doing, she turns to British sociologist Basil Bernstein,

who postulates a coding orientation concerning the legitimacy and appro-
priateness of meanings.9 According to Bernstein, meanings do not exist
independently from social realities. More specifically, it is social class, the
‘fundamental dominant cultural category’ (Bernstein 1975:175), which
exerts ‘the most formative influence upon the procedures of socialisation’
(Bernstein 1987:37; also see Bernstein 1971, 1990, 2000 for a detailed
discussion).
13.3.2 The Research Design of Hasan’s Mother–Child Research

Bringing these two perspectives together, Hasan argues that if linguistic
meanings vary from one social class to another, it follows that the distinct-
ive meaning patterns implicated by speakers will not be merely ‘expressive,
stylistic matter’ which are ‘totally empty of cognitive content’ (Hasan
2009f:116) but will correlate with the speakers’ social class. To investigate
the extent of correlation, Hasan conducted an investigation focusing on
twenty-four mother–child dyads (children aged between 3.6 to 4.2 years).
These mothers and children were categorized into two contrasting social
class groups termed High Autonomy Professionals (HAP) and Low Auton-
omy Professionals (LAP). According to Hasan, the distinction between HAP
and LAP lies in the degree of professional autonomy of the breadwinners of
the participating families. Breadwinners who exerted a high degree of
autonomy in their workplace were categorized as HAP, while those who
imposed little control over their working life and practices were classified
as LAP. In Phase 1 of her mother–child talk research, Hasan (2009a:144)
asks: ‘Does the selection and organization of linguistic meanings vary in
correlation with variation in social class? If yes, then how can that variation
be interpreted?’
The most pressing issue in answering these questions concerns the way
in which linguistic meanings are conceptualized and analyzed so that viable
claims can be made about the correlation between social class and meaning
patterns. As discussed in Section 13.2.2, the very conceptualization of
meaning adopted in Hasan’s network is a functional one. Central to the
model of language is that it places much emphasis on society and language,
and more importantly, the dialectic relation of realization functioning
across social organization, social context, and language (Hasan 1989:271).
In this sense, one could estimate the meaning orientations based on the
social class of speakers, and by the same token, one could predict the social
class of speakers based on the ways in which meanings are construed.
9
‘Meanings’ refers to all modalities of semiosis in Bernstein’s coding orientation. Hasan, by contrast, takes a restricted
view of meaning, with a particular focus on the modality of language. Such a restricted view on coding orientation, in
Hasan’s word, is termed semantic orientation.
13.3.3 The Selection and Organization of Linguistic Meanings

in Mother–Child Talk
To exemplify Hasan’s approach in analyzing linguistic meanings, we repro-
duce a fragment of the semantic network of relation enactment, viz.,
the network of demanding information, i.e. of questions (see Figure 13.3),
and use a segment of interaction between a mother and her daughter,
Donna, for illustration (see Text 13.1).10 We first describe the semantic
options in the question network, followed by an illustration of how such
a network could enable analysts to illuminate the selection and organiza-
tion of meanings in dialogue.
As shown in Figure 13.3, question is the result of the simultaneous
selection of two semantic options, viz., [demand] and [information]. The
combination of these two options [demand; information] permits an entry
to more delicate systems of question, entailing a selection between two
mutually exclusive options: [confirm] or [apprize].11 According to Hasan,
[confirm] is interpreted as a question which aims to elicit a ‘yes-no
response’, whereas [apprize] aims to elicit ‘some specific element of infor-
mation’ (Hasan 2009d:243).
The system labelled G in the network in Figure 13.3 provides a further
specification of the choice [confirm], selecting either [verify] or [enquire].
Questions selecting the option [verify] constitute ‘tagged questions’ in
Figure 13.3 Options in expressing questions: a simplified fragment (see Hasan 1989:246;
Hasan et al. 2007:713)
Gloss: Each system in the QU ESTION network is labelled. For example, the primary options are
G and H in Figure 13.1, and each of the successive systems is labelled a, b, c . . ., and
finally each of the terms is labelled 1, 2, 3 . . .
10
This short excerpt of interaction is taken from Hasan (2009e).
11
Following Hasan’s recent (2014:17) account, distinctions between the terms (i) option, (ii) choice, and (iii) feature
deserve to be noted. Briefly, the term option is ‘choose-able’ in the sense that it refers to the ‘as-yet-unexplored
property of potential’ in the system; the term choice, by contrast, denotes ‘the option selected for further exploration’;
and the term feature refers to the ‘properties of unit under description’ (see also Hasan 2013 for a distinction
between choice and feature).
Text 13.1 Extract of interaction between mother and Donna
Message
ID Speakers Message
71 Mother: you know who that picture is of?

72 Donna: pardon?
73 Mother: do you know who that picture is of?
74 Donna: my Daddy
75 Mother: No
76 looks like your Dad, doesn’t it?
77 but it’s not him
78 Donna: who is it? [?Daddy]?
79 Mother: it is his Daddy
80 he’d be your grandfather,
81 if he were alive today
82 Donna: pardon?
83 Mother: he would be your grandfather
84 if he were alive today
85 your Daddy’s Daddy
86 Donna: [?his] Daddy Daddy?
87 Mother: your Daddy’s father
88 Donna: my [?grandma]
89 Mother: your grandfather, mm
90 but you don’t know him
91 grandma’s husband, he’s now Martin, isn’t he?
92 but he’s - she’s remarried, I mean
93 she’s remarried
94 your father’s .. father .. died
95 remember we went up to see his grave a little while ago? ..
96 remember? .. where he’s buried in the cemetery?
97 Donna: no ..
98 Mother: you’d remember
99 if we took you back there, I think
English, which entail a further selection of one of two options: [probe] or

[reassure]. According to Hasan, the feature [probe] functions to ‘probe the
veracity of a presented thesis’ whereas the feature [reassure] seeks ‘to be
reassured about its veracity between the interactants’ (Hasan 2009d:246).
While both options imply a need for verification, there is a subtle semantic
difference in the sense that the latter enacts a ‘minimum social distance
between interactants’, whereas the former does not (Hasan 2010:293). The
semantic choice [enquire] permits an entry to both system c and d, selecting
options either [ask] or [check] as well as [assumptive] or [non-assumptive].
Questions with the feature [ask], in Hasan’s view, are the most neutral way
of eliciting a yes/no response in English – what Hasan refers to as ‘non-
attitudinal questions’. Questions selecting the feature [check], in many
cases, are ‘attitudinally marked’, or attempt to ‘draw an attestation from
the addressee’ (Hasan, 2009d:246). These questions will simultaneously
enter system d, or more precisely, the system of assumptiveness, selecting
the option [assumptive] or [non-assumptive]. An [assumptive] question is a
question where an ‘unvoiced assumption’ is made by the speakers (Hasan
2009d:249). Lexicogrammatically, the feature [assumptive] is realized by a
Table 13.2 Realization of options from system G in Figure 13.3
Lexicogrammatical realization
Semantic option Systemic realization Structural realization
G confirm major: indicative S-F

a1 verify major: indicative: declarative: tagged S^F . . . F^S
a2 enquire major: indicative
b1 probe major: indicative: declarative: tagged: constant (i) S^Fneg . . . Fneg^S
(ii) S^pos . . . Fpos^S
b2 reassure major: indicative: declarative: tagged: reserved (i) S^Fneg . . . Fpos^S
(ii) S^Fpos . . . Fneg^S
c1 ask major: indicative: interrogative: polar F^S^P . . .
c2 check major: indicative: declarative: untagged tone 2 S^F^P . . . / Tone 2
d1 assumptive F preselects negative polarity S ^ Fneg . . .
d2 non-assumptive preselects positive polarity S^Fpos . . .
Table 13.3 Selection expressions and examples of some choices in system G
Selection expressions Examples
[confirm: verify: reassure] (i) You love Uncle Matt, don’t you?
(ii) You don’t love Uncle Matt, do you?
[confirm: verify: probe] (i) You love Uncle Matt, do you?
(ii) You don’t love Uncle Matt, don’t you?*
[confirm: enquire: ask: non-assumptive] Do you love Uncle Matt?
[confirm: enquire: ask: assumptive] Don’t you love Uncle Matt?
[confirm: enquire: check: non- You love Uncle Matt?
assumptive]
[confirm: enquire: check: assumptive] You didn’t love Uncle Matt?
* As noted in Hasan (2013), this is normal usage in some varieties of Australian English.
clause with a negative polarity. The lexicogrammatical realizations of

options under system G and their examples are summarized in Table 13.2
and Table 13.3 respectively.
In contrast to system G, system H presents another semantic environ-
ment of questions, which is traditionally known as wh-question in English.
Questions with the option [apprize] permit entry into a more delicate
system, selecting either [precise] or [vague]. According to Hasan, a [vague]
question is ‘vague’ in the sense that the item of information sought has not
been specified clearly and is only interpretable with reference to another
message. A [precise] question, by contrast, states precisely what item of
information is being sought. Depending on the type of information, the
choice [precise] can be further categorized into two options – [explain] and
[specify] – where the former concerns why and how, and the latter focuses
on the identity of participant or the specification of circumstance such as
when and where, etc. (Cloran 2000:164). Like [enquire: confirm], the choice
[explain] permits entry into the system of assumptiveness, selecting the
option [assumptive] or [non-assumptive]. The lexicogrammatical
Table 13.4 Realization of options from system H in Figure 13.3
Lexicogrammatical realization
Semantic
option Systemic realization Structural realization
H Apprize major: indicative: interrogative: non- wh- S^F^P

polar
a1 Precise major: indicative: interrogative: non- wh- S^F^P
polar
a2 Vague major: indicative: non-polar: R- preselects as Adjunct with what about +
interrogative nom group
elliptical: maximum: formal
b1 Explain As a1; wh-conflated with Theme wh-conflated with Residue
And circumstance of cause why or Adjunct or Complement/wh^F^S^P
how ^ F^S^P (or equivalent
expression e.g. what for)
b2 Specify As a1; wh-conflated with Theme
wh- outclassifies why and how
(i)
major: indicative: non-polar: (i) wh-conflated with Residue Adjunct
Residue-interrogative or Complement/wh^F^S^P
(ii) major: indicative: non-polar: (ii) wh-conflated with Subject Subject/
Subject-interrogative wh - ^F^P
d1 Assumptive as b1; Why ^ Fneg^ S ^ P . . .
F preselects negative polarity
d2 non-assumptive as b1; Why ^ Fpos^ S ^ P . . .
F preselects positive polarity
Table 13.5 Selection expressions and examples of some choices in system H
Selection expressions Examples
[apprize: vague] What about swimming?

[apprize: precise: explain: assumptive] Why didn’t you stay?
[apprize: precise: explain: non-assumptive] Why did you leave?
[apprize: precise: specify] (i) When is it now?
(ii) Where did you go?
realization of options under system H and their examples are summarized

in Table 13.4 and Table 13.5 respectively.
Though Figure 13.3 is a highly simplified network, it is sufficient to show
that the construal of English questions, in Hasan’s view, is essentially a
selection and organization of linguistic meaning. That is, the speaker is
making choices in meanings from the semantic network of expressing
questions. In identifying semantic features, shunting across strata is
required. For example, viewing from the co-text, the points of enquiry of
message 71 and message 73 in Text 1 lie in the person in the photograph.
Lexicogrammatically, they are realized in a subject-interrogative where who
conflates with the subject. In this sense, the selection expressions of mes-
sages 71 and 73 are [demand; information: apprize: precise: specify].
Messages 76 and 91, by contrast, select another set of semantic options.
For example, scrutiny of the linguistic context suggests that in message 76,
Donna’s mother is verifying the correctness of her propositional content.
Compared with [probe], the choice [reassure] sounds appropriate in this

situation, considering that the social distance between mother and child is
minimal. Viewed from below, message 76 is realized by a clause preselect-
ing [declarative: tagged: reversed]. Bringing these two perspectives together,
message 76 can thus be analyzed as [demand; information: confirm:
verify: reassure] in the sense that Donna’s mother not only performs a
verification, but also an elicitation of reassurance from Donna. Message
76 can thus be best paraphrased as ‘I think this guy looks like your Dad and
I believe you do, too; is that so?’
Messages 95 and 96, on the other hand, present another semantic aspect
of questioning, with a selection expression [demand; information: con-
firm: enquire: check: non-assumptive]. The use of the feature [check] in
message 95 can be related to Donna’s response in message 86, where she
implies some reservations in conceding that the person in the picture is her
grandfather (i.e. Daddy Daddy?). Recognizing that her daughter’s reserva-
tion is contradictory to the reality, Donna’s mother thus ‘challenges’ such
reservations by asking her daughter if she remembers that they had
recently gone to see her grandfather’s grave. Message 95 is thus a question
loaded with ‘additional information about the attitude of the speaker’
(Hasan 2010:293). This attitudinally marked question, as remarked by
Hasan, is realized both lexicogrammatically (i.e. an elliptical clause prese-
lecting [declarative: untagged]) and phonologically (i.e. a high rising tone).12
Though the transcription precludes a phonological annotation, the clause
realizing message 95 preselects [declarative: untagged], which serves as key
evidence of [check].
It should be emphasized that the semantic features identified typically do
not exist in isolation but relate to others forming identifiable meaning
clusters. As maintained by Hasan, language is not a set of rules but behav-
iour, which can be measured and calculated through a principal compon-
ents technique. Through measuring and calculating the principal
components, one could reveal the patterns in speakers’ ways of meanings.
The statistical calculation of principal components further suggests that
meaning variations exist between HAP and LAP families, in terms of the
mothers’ style of control (Hasan 2009b) and questioning and answering
behaviours (Hasan 2009a, 2009d). Subsequent research adopting Hasan’s
approach in studying semantic variation also yields similar results. For
instance, in the exploration of semantic variation in joint book-reading
between families and schools, Williams (1995) has found that the types of
supplementation and its configuration with speech functions differ signifi-
cantly between HAP and LAP families.
12
Since the transcription is not phonologically annotated, one could equally interpret Message 95 as an elliptical
interrogative, with an ellipted Mood element Do you. In this case, Message 95 would be analyzed as [demand;
information: confirm: enquire: ask].
Presenting only the relation enactment, of course, does not provide

an exhaustive analysis. The perspectives of classification, continu-
ation, and amplification are also needed. However, the relation
enactment analysis, even at this preliminary stage of development, is
already sufficient to suggest that the message semantics framework is
essentially robust and analytical, taking into account a trinocular perspec-
tive. Hasan’s comprehensive ‘mapping’ of semantic features thus serves as
a powerful tool in discriminating meaningful choices in dialogue, and more
specifically, serves as linguistic evidence of social variation.
13.4 The Application of Semantic Networks in Discourse

Studies
The strong descriptive power of message semantics networks has attracted

the attention of discourse analysts who are working on aspects other than
semantic variation. This is perhaps not surprising because the primary
objective of discourse analysis is the ‘study of language in use’ (Gee and
Handford 2012:1). The concern of ‘language in process’ in message seman-
tics networks (Hasan 1996:124) thus fits into this research agenda.
Table 13.6 summarizes the most relevant discourse studies featuring mes-
sage semantics networks as the research tool.
13.4.1 Illustration: Semantic Networks in Pedagogical

and Journalistic Discourses
Due to space constraints, this section reports only two extended uses of
message semantics networks in discourse analysis, and discusses the ways
in which the networks are used and what their research implications are in
discourse studies.
Table 13.6 Applications of message semantics networks
Domain Foci Discourse studies
Legal Court room Maley and Fahey (1991)

Police interview Hall (2004)
Computer-mediated Online chat Wong (2009)
communication
Education Classroom teaching Wake (2006); Chu (2011);
Early childhood education Williams (1995); Torr (2004);
Kim (2014)
Business Service encounter Hasan et al. (2007)
Journalism News interview Lukin (2012, 2013)
Health and medicine Surgical interaction Lukin et al. (2011)
Doctor–patient Fung (2016); Moore (2016)
communication
Hasan’s message semantics networks have been applied in pedagogical

discourse (e.g. Wake 2006; Chu 2011). Wake’s (2006) study, to a certain
extent, shades into Hasan’s (2009d) work, being concerned with how learn-
ing is achieved dialogically. While Hasan is concerned with mode of learn-
ing in mother–child talk, Wake focuses on dialogic learning in a group of
second-language international students, examining its effectiveness in the
context of a university curriculum. Her case study analysis reveals an
interesting phenomenon: students ask more questions than the tutor in
the tutorial talk, with a frequent selection of [explain] in apprize questions
and [ask], [check], and [validate] in confirm questions.13 Central to this
question distribution is that dialogic learning entails a shift of classroom
dynamics. That is, contrary to traditional classroom learning contexts
where tutors enact the majority of the questions, it is students who fre-
quently pose questions to seek explanation and confirmation in the univer-
sity context, thereby unwittingly changing ‘the focus and direction of the
lecturer’s explanation’ (Wake 2006:199).
Similar to Wake (2006), Chu (2011) applies semantic networks to the
classroom context of a New Arrival Programme (NAP), with a particular
focus on the newly arrived students in South Australia. Offered by the
Department of Education and Children’s Services, NAP aims to prepare
the newly arrived students for learning the English needed for living and
studying in South Australia. Chu aims to investigate the exploitation of
meanings of visual and verbal modes in multimodal picture books, as well
as the ways in which teachers engage with students during teacher–student
interaction through picture books. Important in Chu’s (2011) work is the
reconceptualization of the interpersonal functions of questions in the con-
text of picture book reading, drawing on Hasan’s message semantics net-
works. Chu argues that even though teachers pose the same type of
questions during picture book reading, the communicative functions vary
in accordance with student literacy levels. For instance, while [apprize:
precise: specify] questions are widely employed in both higher and lower
literate students, their degree of interaction and points of enquiry differ. In
higher-literate groups, the interaction between teachers and students is less
restrictive, in the sense that teachers aim to invite students to contribute
their ‘personal experience and ideas for interpretation’. In other words,
questions selecting [apprize: precise: specify] in higher-literate groups aim
to ‘probe further into students’ views or opinions’, as in What thoughts do
you have?. By contrast, the degree of interaction between teachers and
students in lower-literate groups is more restrictive, and the questions
posed by teachers aim ‘to retrieve and to verbalise the found information’
of the multimodal texts, as in And what are the pictures we can see? (Chu
2011:228).
13
Wake (2006) uses the feature [validate] to refer to questions which are realized by clauses preselecting
[declarative: Adjunct right?], as in The price is part of the world price, right?
Another illuminating use of message semantics networks in discourse

studies is Lukin (2012, 2013). Lukin’s primary concern lies in journalistic
discourse, or, more specifically, the professional performance of journalists
in current affairs interviews. To investigate and characterize the mode of
interviewing of Kerry O’Brien, the Australian senior political journalist
of the Australian Broadcasting Corporation (ABC), Lukin (2013) adopts
Hasan’s message semantics networks, with a particular focus on the choices
of meaning pertaining to questions. Focusing on Kerry O’Brien as a case
study, Lukin argues that current affairs programmes deserve particular
attention because journalists in news interviews might not be performing
the ‘core democratic functions’ of journalism (Clayman and Heritage
2002:2),14 but working in the service of the interviewees, allowing them
to construe the affairs in accordance with their own purposes.
Lukin’s primary use of message semantics is to discriminate the
meaningful choices enacted by the speaker. She analyzes O’Brien’s ques-
tions from a multidimensional perspective, discriminating among
the choices of meaning in the systems of relation enactment,
continuation, amplification, and classification. She finds
that O’Brien’s questions frequently select the features [confirm], [topic-
changing], and [non-prefaced]. According to Lukin, the feature [topic-
changing] denotes a change of topicality in play, whereas [non-prefaced]
refers to messages which concern ‘what the world is like’ rather than
inquiring about ‘someone’s . . . mental representation of the world’ (Hasan
2009d). The combination of these features suggests that O’Brien
only touches on the issue in a general sense, with fewer follow-up questions
(i.e. [topic-changing]), and his questions fail to invite the mental represen-
tations of interviewees (i.e. [non-prefaced]). In other words, rather than
encouraging the interviewees to account for their views concerning the
Iraqi invasion, Lukin argues that O’Brien’s questioning is essentially
following the interviewee’s ideological direction, and his news interviews
serve as the platform for those ‘military experts’, opening the floor to them
to cast their messages in their own ways. Lukin (2013) demonstrates that
Hasan’s message semantics functions not only as a tool in discriminating
the meanings enacted by speakers, but also as a tool in revealing invisible
ideologies in professional practices, or in Bartlett and Chen’s (2012:10)
words, making ‘visible key features and functions of professional practice
that are, or have become, invisible to the practitioners themselves and so to
those being apprenticed into their practices’.
14
Examples of ‘core democratic functions’, as stated by Clayman and Heritage (2002:2) include ‘soliciting statements of
official policy, holding officials accountable for their actions, and managing the parameters of public debate, all of this
under the immediate scrutiny of the citizenry’.
This chapter has demonstrated one claim, namely, that Hasan’s message
semantics network is essentially a discourse analytical tool enabling dis-
course analysts to study language in use in various contexts. In Section 13.2,
we highlighted the theoretical foundations of semantic networks, and more
specifically, the key advancements that Hasan has made, which contribute
to our current understanding of message semantics. Since it is not our
intention to repeat all the ideas which have been discussed in the previous
literature, we have deliberately kept the discussion short and precise.
However, the issues that we have highlighted are sufficient to demonstrate
that Hasan’s message semantics is a powerful tool for discriminating
among the meaningful choices enacted by interlocutors in a dialogue. Given
this significance, it is therefore not surprising that such a paradigmatic
description of semantics has been extensively applied in various discourse
studies. Hasan’s semantic networks are thus not just a tool for semantic
variation or integrated sociolinguistic research (see Section 13.3), but essen-
tially a discourse analytical tool for meaning analysis. We have also
surveyed the use of semantic networks in discourse studies, presenting an
up-to-date review of the different uses of message semantics (see Table 13.4).
Space precludes a detailed discussion of all the domains of application of
semantic networks; the illustrations of pedagogical and journalistic dis-
courses are, we hope, sufficient to exemplify in what ways message seman-
tic networks are employed by discourse analysts to tackle various research
problems.
For Hasan (2005:56), ‘why and how language works’ and ‘the nature of
the relationship between language and society’ are two sides of the same
coin. While one of the primary tasks of discourse analysts is to interpret
and make sense of the meaning of what people say and write and listen to
and read in context, the ability to calibrate context to the semantics and to
lexicogrammar in Hasan’s contextually open semantic networks thus
serves as a powerful analytical tool in analyzing meanings in dialogue,
enabling discourse analysts to understand ‘why and how language
works’.15
References
Bartlett, T. and H. Chen. 2012. Applying Linguistics in Making Professional

Practice Re-Visible. Journal of Applied Linguistics and Professional Practice 9(1):
1–12.
15
Recent years have witnessed an increase in studies describing context through system networks (e.g. Butt 2004;
Hasan 1999, 2009c, 2014; Bowcher 2007, 2014). Such descriptions enable analysts to integrate network-based
descriptions from context to semantics to lexicogrammar.
Bernstein, B. 1971. Class, Codes and Control: Theoretical Studies towards a

Sociology of Language, Volume 1. London: Routledge.
Bernstein, B. 1975 Class, Codes and Control: Applied Studies towards a Theory of
Educational Transmission, Volume 3. London: Routledge.
Bernstein, B. 1987. Social Class, Codes and Communication. In V. Ammon,
N. Dittmar, and K. J. Matthier, eds., Sociolinguistics: An International Hand-
book of the Science of Society. Berlin: Walter de Gruyter. 536–79.
Bernstein, B. 1990. Class, Codes and Control: The Structure of Pedagogic Discourse,
Volume 4. London: Routledge.
Bernstein, B. 2000. Pedagogy, Symbolic Control and Identity: Theory, Research,
Critique. 2nd ed. Oxford: Rowman and Littlefield.
Bowcher, W. L. 2007. Field and Multimodal Texts. In R. Hasan, C. M. I. M.
Matthiessen, and J. J. Webster, eds., Continuing Discourse on Language,
Bowcher, W. L. 2014. Issues in Developing Unified Systems for Contextual
Field and Mode. Functions of Language 21(2): 176–209.
Butt, D. 2004. Parameters of Context: On Establishing the Similarities and
Differences between Social Processes. Unpublished mimeo, Macquarie
University.
Chu, P. Y. 2011. Picture Book Reading in a New Arrival Context: A Multimodal
Perspective on Teaching Reading. PhD Thesis, University of Adelaide.
Clayman, S. and J. Heritage. 2002. The News Interview. Cambridge: Cambridge
University Press.
Relations of Context, Meaning and Grammar. Monographs in Systemic Lin-
guistics 6. Nottingham: University of Nottingham.
Cloran, C. 1999. Context, Material Situation and Text. In M. Ghadessy, ed.,
Text and Context in Functional Linguistics: Systemic Perspectives. Amsterdam:
Cloran, C. 2000. Socio-semantic Variation: Different Wordings, Different
Meanings. In L. Unsworth, ed., Researching Language in Schools and Commu-
nities: Functional Linguistic Perspectives. London: Cassell. 152–83.
Cloran, C., V. Stuart-Smith, and Y. Young. 2007. Models of Discourse. In
647–70.
Eggins, S. 1990. Keeping the Conversation Going: A Systemic-functional Analysis of
Conversational Structure in Casual Sustained Talk. PhD Thesis, University of
Sydney.
Eggins, S. and D. Slade. 2004. Analyzing Casual Conversation. Sheffield:
Equinox.
Fung, A. 2016. Hasan’s Semantic Networks Revisited: A Cantonese Systemic
Functional Approach. In W. L. Bowcher and J. Y. Liang, eds., Society in
Language, Language in Society: Essays in Honour of Ruqaiya Hasan. London:
Palgrave Macmillan.115–40.
Gee, J. P. and M. Handford, eds. 2012. The Routledge Handbook of Discourse

Analysis. Abingdon: Routledge.
Gregory, M. 1985. Towards ‘Communication’ Linguistics: A Framework. In
J. Benson and W. Greaves, eds., Systemic Perspectives on Discourse, Volume 1:
Selected Theoretical Papers from the Ninth International Systemic Workshop.
Norwood: Ablex. 119–34.
Hall, P. 2004. Prone to Distortion? Undue Reliance on Unreliable Records in
NSW Police Service Formal Interview Model. In J. Gibbons, V. Prakasam,
K. V. Tirumalesh, and H. Nagrajan, eds., Language in the Law. New Delhi:
Orient Longman Private Limited. 44–81.
Halliday, M. A. K. 1973. Towards a Sociological Semantics. In M. A. K.
Halliday, Explorations in the Functions of Language. London: Edward Arnold.
72–102.
Dialogue. In R. P. Fawcett et al., eds., The Semiotics of Culture and Language,
Volume 1: Language as Social Semiotic. London: Pinter. 3–35.
London: Arnold.
Halliday, M. A. K. 2006a. Some Theoretical Considerations Underlying the
Teaching of English in China. The Journal of English Studies (Sichuan Inter-
national Studies University) 4: 7–20.
Halliday, M. A. K. 2006b. Working with Meaning: Towards an Appliable
Linguistics. In J. J. Webster. ed., Meaning in Context: Implementing Intelligent
Applications of Language Studies. London: Continuum. 7–23.
Halliday, M. A. K. 2008. Complementarities in Language. Beijing: The Commer-
cial Press.
Halliday, M. A. K. and R. Hasan. 1985. Language, Context and Text:
Aspects of Language in a Social-semiotic Perspective. Geelong: Deakin Univer-
sity Press.
through Meaning: A Language-based Approach to Cognition. New York:
Continuum.
Functional Grammar. 3rd ed. London: Edward Arnold.
Hasan, R. 1983. A Semantic Network for the Analysis of Messages in Everyday Talk
between Mothers and Their Children. Unpublished mimeo, Macquarie
University.

ation. 181–219.
Hasan, R. 1989. Semantic Variation and Sociolinguistics. Australian Journal of
Linguistics 9(2): 221–76.
Hasan, R. 1996. Semantic Networks: A Tool for the Analysis of Meaning. In
C. Cloran, D. Butt, and G. Williams, eds., Ways of Saying, Ways of Meaning:
Selected Papers of Ruqaiya Hasan. London: Cassell. 104–30.
219–328.
Hasan, R. 2000. The Uses of Talk. In S. Sarangi and M. Coulthard, eds.,
Discourse and Social Life. London: Longman. 30–81.
Hasan, R. 2005. Language in Society in a Systemic Functional Perspective. In
Discourse on Language, Volume 1. Sheffield: Equinox. 55–80.
Hasan, R. 2009a. Language in the Processes of Socialization: Home and
School. In J. J. Webster, ed., Semantic Variation: Meaning in Society and in
Sociolinguistics: The Collected Works of Ruqaiya Hasan, Volume 2. Sheffield:
Equinox. 119–79.
Hasan, R. 2009b. Meaning in Sociolinguistic Theory. In J. J. Webster, ed.,
Semantic Variation: Meaning in Society and in Sociolinguistics: The Collected
Works of Ruqaiya Hasan, Volume 2. Sheffield: Equinox. 271–308.
Hasan, R. 2009c. The Place of Context in a Systemic Functional Model. In
Hasan, R. 2009d. Questions as a Mode of Learning in Everyday Talk. In
J. J. Webster, ed., Semantic Variation: Meaning in Society and in Sociolinguis-
tics: The Collected Works of Ruqaiya Hasan, Volume 2. Sheffield: Equinox.
231–68.
Hasan, R. 2009e. Semantic Variation: Meaning in Society and in Sociolinguistics:
The Collected Works of Ruqaiya Hasan, Volume 2. Edited by J. J. Webster.
Sheffield: Equinox.
Hasan, R. 2009f. A Sociolinguistic Interpretation of Everyday Talk between
Mothers and Children. In J. J. Webster, ed., Semantic Variation: Meaning in
Society and in Sociolinguistics: The Collected Works of Ruqaiya Hasan, Volume 2.
Hasan, R. 2010. The Meaning of ‘Not’ Is Not in ‘Not’. In A. Mahboob and
N. K. Knight, eds., Appliable Linguistics: Texts, Contexts and Meanings.
Hasan, R. 2011. A Timeless Journey: On the Past and Future of Present
Knowledge. In D. N. Cheng, ed., Selected Papers of Ruqaiya Hasan on Applied
Linguistics. Beijing: Foreign Language Teaching and Research Press.
xiv–xliii.
Hasan, R. 2013. Choice, System, Realization: Describing Language as Mean-

ing Potential. In L. Fontaine, T. Bartlett, and G. O’Grady, eds., Systemic
Functional Linguistics: Exploring Choice. Cambridge: Cambridge University
Press. 269–99.
Hasan, R., C. Cloran, G. Williams, and A. Lukin. 2007. Semantic networks:
The Description of Meaning in SFL. In R. Hasan, C. M. I. M. Matthiessen,
and J. J. Webster, eds., Continuing Discourse on Language, Volume 2. Shef-
field: Equinox. 697–738.
Hyland, K., ed. 2013. Discourse Studies Reader: Essential Excerpts. London:
Bloomsbury.
Hyland, K. and B. Paltridge, eds. 2011. Continuum Companion to Discourse
Analysis. London: Continuum.
Kim, J. E. 2014. Parent Child Sharing Reading: The Affordances of Print, Digital
and Hand-held Electronic Storybooks. PhD Thesis, The University of British
Columbia.
Labov, W. 1972. Sociolinguistic Patterns. Oxford: Basil Blackwell.
Labov, W. 1978. Where Does the Sociolinguistic Variable Stop? A response to
Beatriz Lavandera. Working Papers in Sociolinguistics Paper #44. Working
Papers in Sociolinguistics. Austin: Southwest Educational Development
Laboratory. 1–16.
Lukin, A. 2012. Hasan’s Semantic Networks as a Tool in Discourse Analysis.
In J. S. Knox, ed., Papers from the 39th International Systemic Functional
Congress. Sydney: Organising Committee of the 39th International Sys-
temic Functional Congress. 141–6.
Lukin, A. 2013. Evaluating Questions in Journalism: A Case Study of the
Australian Public Broadcaster’s Coverage of the 2003 Invasion of Iraq.
Journal of Applied Linguistics and Professional Practice 9(1): 127–47.
Lukin, A., A. Moore, M. Herke, C. Wu, and R. Wegener. 2011. Halliday’s
4(2): 187–213.
Mahboob, A. and N. Knight. 2010a. Appliable Linguistics. London: Continuum.
Mahboob, A. and N. Knight. 2010b. Appliable Linguistics: An Introduction.
In A. Mahboob and N. Knight, Appliable Linguistics. London: Continuum.
1–12.
Maley, Y. and R. Fahey. 1991. Presenting the Evidence: Constructions of
Reality in the Court. International Journal for the Science of Law 4(1): 3–17.
Benjamins.
Martin, J. R. 2000. Beyond Exchange: Appraisal Systems in English. In
S. Hunston and G. Thompson, eds., Evaluation in Text: Authorial Stance and
the Construction of Discourse. Oxford: Oxford University Press. 142–75.
Clause. Functional Linguistics 1(1): 3.
Clause. 2nd ed. London: Continuum.
English. London: Palgrave.
Matthiessen, C. M. I. M. 1988a. Representational Issues in Systemic Func-
tional Grammar. In J. D. Benson and W. S. Greaves, eds., Systemic Func-
tional Perspectives on Discourse. Norwood: Ablex. 136–75.
Matthiessen, C. M. I. M. 1988b. Semantics for a Systemic Grammar:
The Chooser and Inquiry Framework. In J. D Benson, M. J. Cummings,
and W. S. Greaves, eds., Linguistics in a Systemic Perspective. Amsterdam:
Matthiessen, C. M. I. M. 1990. Two Approaches to Semantic Interfaces in
Text Generation. COLING 90: 322–9.
Matthiessen, C. M. I. M. 1993. Register in the Round: Diversity in a Unified
Theory of Register Analysis. In M. Ghadessy, ed., Register Analysis: Theory
and Practice. London: Painter. 221–92.
Matthiessen, C. M. I. M. 2004. The Semantic System of Relational Expan-
sion: Rhetorical Structure Theory Revised. Unpublished draft, Macquarie
University.
Matthiessen, C. M. I. M. 2007. The Architecture of Language According
to Systemic Functional Theory: Developments since the 1970s. In
505–61.
Matthiessen, C. M. I. M. 2009. Ideas & New Directions. In M. A. K. Halliday
Matthiessen, C. M. I. M. 2012. Systemic Functional Linguistics as Applicable
Linguistics: Social Accountability and Critical Approaches. D.E.L.T.A. 28:
435–71.
Matthiessen, C. M. I. M. 2014. Appliable Discourse Analysis. In Y. Fang and
Matthiessen, C. M. I. M. 2015a. Halliday on Language. In J. J. Webster, ed.,
137–202.
Matthiessen, C. M. I. M. 2015b. Modelling Context and Register: The Long-
term Project of Registerial Cartography. Letras, Santa Maria 25(50): 15–90.
Matthiessen, C. M. I. M. and J. Bateman. 1991. Systemic Linguistics and Text
Generation: Experiences from Japanese and English. London: Pinter.
Moore, A. R. 2016. Can Semantic Networks Capture Intra- and Inter-
Registerial Variation? Palliative Care Discourse Interrogates Hasan’s Mes-
sage Semantics. In W. L. Bowcher and J. Y. Liang, eds., Society in Language,
Language in Society: Essays in Honour of Ruqaiya Hasan. London: Palgrave

Macmillan. 83–114.
Patten, T. 1988. Systemic Text Generation as Problem Solving. Cambridge:
Torr, J. 2004. Talking about Picture Books: The Influence of Maternal
Education on Four-year-old Children’s Talk with Mothers and Pre-school
Teachers. Journal of Early Childhood Literacy 4(2): 181–210.
Turner, G. J. 1973. Social Class and Children’s Language of Control at Age
Five and Age Seven. In B. Bernstein, ed., Class, Codes and Control, Volume 2:
Applied Studies towards a Sociology of Language. London: Routledge. 121–79.
Wake, B. J. 2006. Dialogic Learning in Tutorial Talk: A Case Study of Semiotic
Mediation as a Learning Resource for Second Language International Students.
PhD Thesis, University of Adelaide.
Williams, G. 1995. Joint Book-Reading and Literacy Pedagogy: A Sociosemantic
Examination. PhD Thesis, Macquarie University.
Williams, G. 2005. Semantic Variation. In R. Hasan, C. M. I. M. Matthiessen,
and J. J. Webster. eds., Continuing Discourse on Language, Volume 1.
Wong, Y. T. 2009. The Linguistic Function of Cantonese Discourse Particles in
the English Medium Online Chat of Cantonese Speakers. MA Dissertation,
University of Wollongong.
14
Discourse Semantics
J. R. Martin
14.1 Discourse Semantics
Discourse semantics is the term used by Martin and his colleagues (after
Martin 1992; Martin and Rose 2003) to refer to the stratum of meaning
interfacing lexicogrammar with context (register and genre) in SFL. It
comprises six major systems, organized into three metafunctions: ideation
and conjunction (ideational); negotiation and appraisal (interpersonal); and
identification and periodicity (textual). As reviewed in Martin (2014), this
work reinterprets Halliday and Hasan’s model of cohesion (e.g. Halliday and
Hasan 1976) from the perspective of Gleason’s work on text semantics (e.g.
Gleason 1968) – as a set of text-forming resources realized through gram-
mar, lexis, and intonation. From this perspective, Halliday and Hasan’s
cohesive ties are reinterpreted as discourse structures; and the resources
which Halliday (e.g. 2009:85) positions as non-structural components of the
textual metafunction in grammar are reinterpreted at a deeper level of
abstraction as discourse semantic systems. This reinterpretation fore-
grounds meaning beyond the clause as fundamental to semantic analysis
in SFL and can be usefully compared with the clause semantics research foci
inspired by Halliday and Matthiessen (1999), Hasan (2009), or Fawcett
(2008). Halliday and Hasan (1985:82 in particular) can be read as developing
work on cohesion in a similar direction.
14.2 Discourse Semantic Systems
As noted above, discourse semantic systems are organized by metafunction;

this contrasts with Halliday and Hasan’s treatment of all cohesive devices as
textual. As one would expect in SFL, they are modelled from the comple-
mentary perspectives of system and structure; but the relation of system to
structure is not as tightly bound as in most SFL work on grammar and
Discourse Semantics 359
phonology – where a particular selection of features prescribes a particular

structural configuration, and where by the same token a particular struc-
tural configuration presumes a particular selection of features (see the
discussion of axis in Martin 2013a). Following Bateman’s (2014) deployment
of Peirce, it is more appropriate to think of discourse structures as semantic
relations abduced between parts of a text, and of discourse systems as
constraining the set of possible relations that can be abduced. For example,
the identification relation abduced between discourse semantic systems in the
first sentence of this paragraph and they in the second is presuming anaph-
ora; and the relation abduced between the first sentence and the second in the
previous ranking clause is one of presuming nominal ellipsis. Describing
the discourse semantics of a given language involves establishing the range
of discourse structures which can be abduced from discourse systems.
Since we are abducing discourse relations rather than deriving them, text
analysis necessarily involves some degree of ‘play’ in interpretation. At a
given point in a text there may be more than one analysis available; there
may be a need to revise analysis as the text unfolds; and there may in the
end be multiple analyses available that cannot be resolved. Some analysts
experience this difference between discourse interpretation and grammar
analysis as a source of frustration; others experience it with a sense of
liberation. As usual, coming to appreciate the complementarities involved
in analyzing a clause and interpreting a text is the key to understanding
how language has evolved so we can use it to live.
Discourse semantic systems are briefly summarized below, drawing prin-
cipally on Martin (1992), Martin and Rose (2003), and Martin and White
(2005).
14.2.1 Ideational Systems: Ideation and Conjunction

Two systems are involved here, ideation and conjunction. Their main func-
tion is to construe the register variable field, where field is defined as a set
of activity sequences oriented to some global institutional purpose, includ-
ing the taxonomies of entities involved in these sequences (Martin 1992). In
Halliday and Matthiessen’s (1999) terms, ideation is concerned with the
semantics of figures, and conjunction with the semantics of relations
between figures.
Ideation extends earlier work on lexical cohesion (Halliday and Hasan
1976). Drawing in part on Halliday and Hasan (1985), a more detailed
account of possible relations among semantic units realized through lexical
items is proposed – including repetition, synonym, antonymy, hyponymy,
and meronymy. In addition a model of nuclear relations is proposed for
semantic units typically realized through nominal groups, verbal groups,
and clauses – drawing on Halliday’s notion of logical-semantic relations (i.e.
elaboration, extension, and enhancement; Halliday and Matthiessen 2014);
the purpose of this extension is to capture semantic relations grouped
360 J. R. MARTIN
together under the heading collocation in Halliday and Hasan (1976). The
discourse structures afforded by these ideation relations are termed lexical
strings.
Conjunction integrates earlier work on cohesion (Halliday and Hasan
1976) and expanding clause complexes (Halliday and Matthiessen 2014)
through consideration of semantic relations between figures, whether real-
ized within or between clause complexes – i.e. cohesively between clause
complexes, paratactically or hypotactically between clauses in a clause
complex, or via logical metaphor (Halliday 1998) as a participant, process,
or circumstance within a clause. Four main types of conjunctive relation
are recognized: additive, comparative, temporal, and consequential. Halli-
day and Hasan’s important distinction between external and internal con-
junctive relations is sustained. And the model recognizes the possibility of
implicit conjunctive relations abduced in the interpretation of adjacent
figures (with the constraint that any relation so abduced could be made
explicit). The discourse structures afforded by these conjunctive relations
are modelled as reticula, elaborating on Gleason’s (1968) notion of an
‘event-line’.
14.2.2 Interpersonal Systems: Negotiation and Appraisal

The two systems involved here are negotiation and appraisal. Their main
function is to enact the register variable tenor, where tenor is concerned
with the relations of power and solidarity whereby speakers position them-
selves as interlocutors in discourse. In general terms we can think of
negotiation as focusing attention on the inter-(personal) dimension of
interpersonal meaning and appraisal as focusing attention on the (inter)-
personal.
Negotiation draws on earlier work on exchange structure (Sinclair and
Coulthard 1975) and speech function (e.g. Halliday 1984) and attends to the
relation of one move to another in conversation. Following Berry (1981) and
Ventola (1987), exchanges are modelled as consisting of between one and
five basic moves, including the possibility of expansion through locally
contingent tracking and challenging moves. Exchange rank systems and
structures are realized at a lower rank through move systems, which are in
turn realized inter-stratally through lexicogrammar and phonology. These
inter-stratal relations can be direct or indirect, depending on whether
interpersonal grammatical metaphor is involved. Negotiation analysis pro-
vides a useful scaffolding for consideration of the role in dialogue of
substitution and ellipsis (Halliday and Hasan 1976) and of rhythm and
intonation (Halliday 1967, 1970; Halliday and Greaves 2008).
Appraisal provides a model of resources for evaluation, including types of
attitude (affect, judgement, and appreciation), graduation (the strength and
prototypicality of feelings), and engagement (the range of voices in play).
Affect is concerned with emotion in relation to a trigger; judgement deals
with opinions about targeted behaviour; and appreciation focuses on the

value of targeted semiotic and natural phenomena. All three types of
attitude can be either positive or negative (in the sense of good feelings or
bad ones), and all can be inscribed in a text through explicitly evaluative
lexis or invoked through a choice in ideation which affords a reaction.
Knight (2010, 2013) refers to shared feelings as bonds, inspiring recent
SFL work on identity and affiliation (e.g. Martin et al. 2013).
14.2.3 Textual Systems: Identification and Periodicity

The two systems involved here are identification and periodicity. Their
main function is to compose discourse as waves of information texturing
the register variable mode, where mode deals with the affordances of
various media of communication (speaking, phoning, tweeting, texting,
emailing, posting, writing etc.) in relation to turn-taking, aural/visual feed-
back, abstraction, and inter-modality.
Identification extends the cohesion variable reference to include consid-
eration of both how entities are introduced into a text and how they are
kept track of once there (including consideration of how they are related to
other entities through comparative reference). As explored in Gleason
(1968) and Martin (1983), languages vary in terms of how they introduce
and track entities through nominal deixis, Theme selection, and clause
complex (‘switch-reference’) systems. Phoric entities presume information
which has to be recovered from the co-text or nonverbal context, for
which a range of recovery strategies is proposed (anaphora, cataphora,
exophora, homophora, etc.). The discourse structures afforded by pre-
suming endophoric (i.e. co-textual) identification relations are termed ref-
erence chains.
Periodicity develops SFL work on theme and information structure (Halli-
day 1967, 1970; Halliday and Greaves 2008; Halliday and Matthiessen 2014).
Drawing on work by Fries (1981) in particular, it extends Halliday’s descrip-
tion of information flow in the clause to higher levels of discourse. In this
model Theme is interpreted as a clause resource for focusing our attention
on a region of a field, and New as a complementary resource for developing
what we want to say in that region of meaning. A hyper-Theme can then be
interpreted as predicting the orientation to the field that will be composed
through an ensuing pattern of Theme selections in a phase of discourse
(what Fries refers to as the ‘method of development’ of a phase). And hyper-
New can be interpreted as consolidating, often with some evaluative inter-
pretation, the information composed through a preceding pattern of News
(what Fries refers to the ‘point’ of a phase). Depending on the degree of
planning and editing afforded by the mode, additional layers of thematic
prediction and news consolidation may be found (i.e. macro-Themen and/or
macro-Newn). Higher-level themes may be graphologically foregrounded as
headings and perhaps a table of contents, where such occur.
362 J. R. MARTIN
14.3 Working with Discourse Semantics: Text Synthesis
In this section I will try and illustrate the discourse semantic resources
introduced above and address the practical concerns of this handbook by
engaging in an exercise I will refer to as text synthesis. Text synthesis
proceeds by building up a coherent text and thus contrasts with text
analysis which starts with a finished product and breaks it down. Synthesis
is rhetorical in orientation; it aims to demonstrate how linguistic resources
can be deployed. As such it reflects a long-standing but until recently under-
utilized tradition in language teaching – one which features in my grand-
father’s language textbooks in late nineteenth-century rural Canada, but
had to be re-introduced in the genre-based literacy programmes designed
by the ‘Sydney School’ (Rose and Martin 2012). Synthesis disappeared, one
has to presume, because the knowledge about language available to stu-
dents and teachers across sectors in education became so impoverished that
it could not sustain rhetorically oriented text construction. Now, thanks
particularly to SFL, the knowledge about language we need is readily avail-
able. So let us put it to use.
As our starting point, consider the following string of alphabetically
listed clauses. How might we begin to compose a text from these?
[1] Available finance has not been managed.

Damaged paths and walls have not been repaired.
Frescoes have not been preserved.
In some places ancient lead water pipes have been exposed.
Inexpensive mortar has also been used.
Mangy dogs roam the site.
No conservation and interpretation program has been put in place.
No walkways for viewing platforms have been constructed.
Over thirty different varieties of weed have been identified, includ-
ing ivy, fennel, and fig.
Over time this mortar has cracked.
The roof collapsed.
The roof could not support the weight of the tiles.
The roots grow.
The roots open up further cracks.
The timber roof on the House of Meleager was designed for light-
weight roofing.
This allowed water and vegetation to penetrate.
This allows even more weeds in.
This was to protect ancient stonework.
Tourists enter buildings that are not roped off.
Tourists walk along ancient paths.
Let us begin with ideation. Nuclear relations have been provided for us, in
order to make this exercise fit into a short chapter of this kind (for a
synthesis which builds nuclear relations, see Martin 2013b). So we can

simply accept relations such as timber = roof, poorly x designed, damaged +
paths and so on. As far taxonomic relations among entities are concerned,
let me suggest the following re-organization:
[2] The timber roof on the House of Meleager was designed for light-
weight roofing. The roof could not support the weight of the tiles.
The roof collapsed. Inexpensive mortar has also been used. This was
to protect ancient stonework. Over time this mortar has cracked.
This allowed water and vegetation to penetrate. çè Over thirty
different varieties of weed have been identified, including ivy,
fennel, and fig. The roots grow. The roots open up further cracks.
This allows even more weeds in. çè No walkways for viewing
platforms have been constructed. Tourists walk along ancient paths.
Tourists enter buildings that are not roped off. In some places
ancient lead water pipes have been exposed. çè Damaged paths
and walls have not been repaired. Frescoes have not been preserved.
Mangy dogs roam the site. Available finance has not been managed.
No conservation and interpretation programme has been put in place.
The principal lexical strings underpinning this construal of the field are
outlined in Table 14.1 – with headings highlighting the kinds of relation
involved as mostly based on composition (co/meronymy)1 or mostly based
on classification (co/hyponymy). In general terms then our gaze shifts from
construction to vegetation to vantage points to aspects of the site overall.
Table 14.1 Selected lexical relations in Text 2
‘co/meronymy’ ‘co/hyponymy’ ‘co-meronymy’ ‘co/meronymy’
timber = roof (vegetation) walkways paths

roof varieties = weeds viewing = platforms walls
roof ivy paths frescoes
tiles fennel buildings site
mortar fig lead = water = pipes
stonework roots
mortar weeds
As a next step we can draw on graphology to scaffold the four phases of

discourse as paragraphs.
[3] The timber roof on the House of Meleager was designed for light-
weight roofing. The roof could not support the weight of the tiles.
The roof collapsed. Inexpensive mortar has also been used. This was
1
In Table 14.1 the ‘co/meronymy’ headings indicate that lexical relations involving both meronymy (part/whole
relations) and co-meronymy (part/part relations) are found; the ‘co/hyponymy’ heading indicates that both hyponymy
(class/subclass relations) and co-hyponymy (subclass/subclass relations) are found.
364 J. R. MARTIN
to protect ancient stonework. Over time this mortar has cracked.

ing ivy, fennel, and fig. The roots grow. The roots open up further
cracks. This allows even more weeds in.
No walkways for viewing platforms have been constructed. Tourists
walk along ancient paths. Tourists enter buildings that are not roped
off. In some places ancient lead water pipes have been exposed.
Damaged paths and walls have not been repaired. Frescoes have
not been preserved. Mangy dogs roam the site. Available finance has
not been managed. No conservation and interpretation programme
has been put in place.
At this point we can bring identification into the picture and consider
anaphoric identity chains introducing (technically speaking ‘presenting’)
and tracking (technically speaking ‘presuming’) the entities introduced as
the timber roof on the House of Meleager, poor quality mortar, over thirty different
varieties of weed, and tourists. Each of these entities was fully lexicalized in
Text 3; this redundancy is adjusted as in Text 4 below.
[4] The timber roof on the House of Meleager was designed for light-weight
roofing. It could not support the weight of the tiles. It collapsed.
Inexpensive mortar has also been used. This was to protect ancient
stonework. Over time this mortar has cracked. This allowed water and
vegetation to penetrate.
Over thirty different varieties of weed have been identified, including
ivy, fennel, and fig. The roots grow. They open up further cracks. This
allows even more weeds in.
No walkways for viewing platforms have been constructed. Tourists
walk along ancient paths. They enter buildings that are not roped off.
Damaged paths and walls have not been repaired. Frescoes have
not been preserved. Mangy dogs roam the site. Available finance has
not been managed. No conservation and interpretation programme
has been put in place.
The identity chains at issue here are outlined in Table 14.2. The mortar,
weeds, and tourists chains are initiated non-phorically (via non-specific
Table 14.2 Identity chains in Text 4
‘roof’ ‘mortar’ ‘weeds’ ‘tourists’
the timber roof . . . inexpensive mortar over thirty different varieties . . . tourists
it this mortar the roots they
it they
even more weeds
nominal deixis); the roof chain on the other hand involves both esphora (the
timber roof pointing forward in its nominal group to its Qualifier on the House
of Meleager) and homophora, since the House of Meleager is treated as assumed
knowledge in this field. Thereafter pronouns and specific deixis are used to
track entities (it, it; this mortar; the roots; they); and a ‘sub-entity’, the roots of
the weeds, is introduced via bridging (taking advantage of the part/whole
ideation relation between weeds and roots). The weeds chain also includes a
comparative reference, introducing an additional set of weeds beyond the
weeds initiating the chain.
As we can see, even in a few phases of discourse of this kind, as far as
identification is concerned there is a lot going on. This can be especially
challenging for speakers coming from a language that manages identifica-
tion differently from English. The lack of an obligatory presenting/presum-
ing reference distinction in many languages is especially troubling (as
teachers and supervisors of academic writing well know); and this problem
may be exacerbated by the fact that in such languages the distinction
between specific and generic reference is not explicitly grammaticalized.
We should also note at this point three instances of text reference
(technically ‘extended reference’) in Text 4. With text reference indefinitely
long phases of meaning can be presumed (one or two sentences worth in
the examples below). As the term implies, what is identified is phases of
unfolding discourse rather than specific discourse semantic entities that
are introduced and tracked as part of the construal of a field.
Inexpensive mortar has also been used.

This was to protect ancient stonework.
Over time this mortar has cracked.
The roots grow. They open up further cracks.
This allows even more weeds in.
As an alternative to this text reference we can draw on conjunction to relate

figures to one another. Each instance is reworked as a clause complex below
(with the conjunctive relation connecting the figures specified between
clauses and explicit connectors in italics). Note the contrast with the examples
above, where conjunctive relations were realized inside the clause. There the
text reference incorporated the relevant figure as a participant, which could
then be connected to another figure through Process and Participant transi-
tivity relations (for which see Halliday and Matthiessen 2014).
Inexpensive mortar has also been used

(explicit purpose)
to protect ancient stonework.
Over time this mortar has cracked,
(implicit manner)
allowing water and vegetation to penetrate.
366 J. R. MARTIN
As the roots grow

(explicit simultaneity)
they open up further cracks,
(implicit manner)
allowing even more weeds in.
It is perhaps appropriate at this point to adjust the rest of the external

conjunctive relations at stake here as in Text 5a below.
[5a] The timber roof on the House of Meleager was not designed to
support the weight of the tiles and collapsed. Inexpensive mortar
has also been used to protect ancient stonework. Over time this
mortar has cracked, allowing water and vegetation to penetrate.
ing ivy, fennel, and fig. As the roots grow they open up further
cracks, allowing even more weeds in.
No walkways for viewing platforms have been constructed, so
tourists walk along ancient paths and enter buildings that are not
roped off. In some places ancient lead water pipes have been
exposed.
Damaged paths and walls have not been repaired, frescoes have
not been preserved, and mangy dogs roam the site. Available finance
has not been managed, and no conservation and interpretation pro-
gramme has been put in place.
These relations are specified below; the specification (-) indicates that a
comparative, temporal or consequential conjunctive relation cannot be
made explicit between these two clauses (by convention, implicit additive
relations are left unspecified, by way of lightening the workload for
text analysts). Note that the use of non-finite clauses (e.g. allowing
water and vegetation to penetrate) and branched paratactic clause
complexes (e.g. so tourists walk along ancient paths and enter buildings that
have not been roped off) means that certain entities are ideationally
implicit, and are thus tracked here via ellipsis as far as identity chains are
concerned. This further reduces the ideational redundancy apparent in Text
2 above.
[5b] The timber roof on the House of Meleager was not designed
(explicit purpose)
to support the weight of the tiles
(explicit additive, implicit causal)
and collapsed.
(-)
(explicit purpose)
(implicit succession)
Over time this mortar has cracked,2

(implicit manner)
ing ivy, fennel, and fig.3
(-)
As the roots grow
(explicit simultaneous)
(implicit manner)
No walkways for viewing platforms have been constructed,
(explicit cause)
so tourists walk along ancient paths
(explicit addition)
and enter buildings that are not roped off.
(-)
Damaged paths and walls have not been repaired,
(implicit addition)
frescoes have not been preserved,
(explicit addition)
and mangy dogs roam the site.
(-)
Available finance has not been managed,
(explicit addition)
and no conservation and interpretation program has been put
in place.
At this point we can turn to periodicity and consider how to scaffold the
waves of information in the text. Text 5’s topical Themes are highlighted in
bold below for each of its finite ranking clauses;4 (“) there indicates that the
orientation to the field is being sustained through ellipsis (a more common
pattern in many languages than in English, both within and between clause
complexes).
[5c] The timber roof on the House of Meleager was not designed
2
Over time is interpreted here as a circumstance of Location, ideationally construing the time frame over which the
stonework and mortar problem arose, rather than as a temporal connector realizing conjunction.
3
The dependent non-finite clause, including ivy, fennel, and fig, grammatically elaborates the varieties of weed that have
invaded; accordingly it is not interpreted here as a distinct figure conjunctively related to the identification of the weeds.
4
Theme has not been analyzed in non-finite ranking clauses, with their non-finiteness interpreted as downgrading the
figure as far as textual and interpersonal meaning are concerned.
368 J. R. MARTIN
and (“) collapsed.

Inexpensive mortar has also been used to protect ancient
stonework.
Over thirty different varieties have been identified,
including ivy, fennel, and fig.
As the roots grow they open up further cracks,5
and (“) enter buildings that are not roped off.
and no conservation and interpretation programme has been put
in place.
Theme’s complementary pulse of informational prominence, New, is high-

lighted below – assuming unmarked tonicity for each clause (Halliday and
Greaves 2008). Minimal New is highlighted in bold (i.e. the Process, Partici-
pant, or Circumstance containing the tonic syllable); italics highlights my
reading of the left-ward domain of New as far as including further infor-
mation extending the field of each phase is concerned.6
[5d] The timber roof on the House of Meleager was not designed
and collapsed.
Over thirty different varieties have been identified,
including ivy, fennel, and fig.
As the roots grow
5
Following Martin and Rose (2003) Topical Theme in declarative clauses is analyzed up to and including the Subject.
6
Since intonation does not specify how much of the clause beyond the constituent containing the tonic syllable is
involved, this reading has to be undertaken in relation to the ‘point’ of each phase – by including information relevant to
expanding the field.

and enter buildings that are not roped off.
and no conservation and interpretation programme has been put in place.
At this point we can consider adding a higher layer of periodicity, either

prospective (hyper-Themes) or retrospective (hyper-News). We will lean
toward layers of prediction here,7 adding hyper-Themes, highlighted in
Text 6 below.
[6] Much of the restoration work on Pompeii has been done by local
firms with no knowledge of restoration techniques.
For example, the timber roof on the House of Meleager was not
designed to support the weight of the tiles and collapsed. Inexpensive
mortar has also been used to protect ancient stonework. Over time this
The incursion of uncontrolled weeds has hastened the decay of

the ruins.
Over thirty different varieties have been identified, including ivy,

fennel, and fig. As the roots grow they open up further cracks,
Pompeii’s position as an international tourist attraction brings

half a million visitors each year.
No walkways for viewing platforms have been constructed, so tour-

ists walk along ancient paths and enter buildings that are not roped
off. In some places ancient lead water pipes have been exposed.
There seems to be no overall management plan for the site.
Damaged paths and walls have not been repaired, frescoes have not
been preserved, and mangy dogs roam the site. Available finance has
not been managed, and no conservation and interpretation pro-
7
Writers who proceed from a carefully constructed plan lean towards front-loading of this kind, since they know where
they are going as they write; writers who figure out what they want to say as they work through successive drafts may
find periodic summarizing, via hyper-News, more appropriate.
370 J. R. MARTIN
These scaffolded phases of discourse can be explicitly connected to one

another by drawing on the resources of internal conjunction. In Text 7 below,
the first phase is thus connected to the second via internal succession (to
begin, second), the second to the third via internal addition (in addition), and
the fourth to the rest via internal culminative succession (finally).
[7] To begin, much of the restoration work on Pompeii has been done
by local firms with no knowledge of restoration techniques. For
example, the timber roof on the House of Meleager was not designed
to support the weight of the tiles and collapsed. Inexpensive mortar
Second, the incursion of uncontrolled weeds has hastened the
decay of the ruins. Over thirty different varieties have been identi-
fied, including ivy, fennel, and fig. As the roots grow they open up
further cracks, allowing even more weeds in.
In addition, Pompeii’s position as an international tourist attrac-
tion brings half a million visitors each year. No walkways for viewing
platforms have been constructed, so tourists walk along ancient
paths and enter buildings that are not roped off. In some places
ancient lead water pipes have been exposed.
Finally, there seems to be no overall management plan for the site.
not been managed, and no proper conservation and interpretation
programme has been put in place.
As noted above, identification (text reference) can function as an alternative

to conjunction. So we might have connected up the third and fourth phase
circumstantially in Text 7 above (e.g. in addition to this, on top of all this).
Ideation can also be brought into play if we draw on metadiscourse to name
phases, as factors, for example. This makes it possible to order factors by
bringing comparative reference into the picture (e.g. a second factor, another
factor, a final factor to consider). A range of conjunction (internal), identifica-
tion (text reference and comparison), and ideation (metadiscourse)
resources are used to link phases in Text 8 below.
[8] To begin, much of the restoration work on Pompeii has been done
by local firms with no knowledge of restoration techniques. For
example, the timber roof on the House of Meleager was not designed
to support the weight of the tiles and collapsed. Inexpensive mortar
A second factor is the incursion of uncontrolled weeds which
have hastened the decay of the ruins. Over thirty different varieties
have been identified, including ivy, fennel, and fig. As the roots grow
they open up further cracks, allowing even more weeds in.
In addition to this, Pompeii’s position as an international tourist

attraction brings half a million visitors each year. No walkways for
viewing platforms have been constructed, so tourists walk along
ancient paths and enter buildings that are not roped off. In some
places ancient lead water pipes have been exposed.
not been managed, and no conservation and interpretation pro-
Let us now bring periodicity back into the picture, and cook up a rhetorical
sandwich for the phases as a whole – introducing a macro-Theme to predict
the hyper-Themes of each phase and a macro-New to consolidate their News.
macro-Theme
Since its discovery, Pompeii has been damaged as an archaeological site,
affected by the quality of restoration work, vegetation, tourism, and site
management issues.
[phases 1–4]
Damage to the site remains a key issue for archaeologists and administrators.
There are ongoing concerns arising in relation to restoration work, uncon-
trolled vegetation, tourism, and management which need to be addressed if
the site is to be preserved for future generations of research and public access.
macro-New
To bring out the sense in which lower layers of periodicity reformulate

higher level ones, the symbol ‘=’ (Halliday and Matthiessen’s 2014 elabor-
ation) has been used to annotate the relevant relations in Text 9. In addition
I have used metadiscourse (a number of factors) to name ensuing phases in
the macro-Theme, and I have specified an internal consequential connec-
tion between the macro-New and the rest of the text by drawing on text
reference, metadiscourse, and a circumstantial realization of cause (as a
result of these factors).8 Note in passing that higher levels of periodicity
regularly draw on grammatical metaphor and abstract ideation to ‘general-
ize’ the meaning specified in lower layers (e.g. the quality of restoration work,
vegetation, tourism, site management issues in the Macro-Theme below).
[9] Since its discovery, Pompeii has been damaged as an archaeological

site, affected by a number of factors, including the quality of
restoration work, vegetation, tourism, and site management issues.
=
8
We can compare this selection with some other possibilities: thus (conjunction only); as a result (circumstance of
cause now lexicalized as a cohesive conjunction); as a result of this (conjunction plus text reference); as a result of
these factors (conjunction plus text reference plus metadiscourse).
372 J. R. MARTIN
To begin, much of the restoration work on Pompeii has been done

by local firms with no knowledge of restoration techniques.
= For example, the timber roof on the House of Meleager was
not designed to support the weight of the tiles and collapsed.
Inexpensive mortar has also been used to protect ancient
stonework. Over time this mortar has cracked, allowing
water and vegetation to penetrate.
A second factor is the incursion of uncontrolled weeds which
have hastened the decay of the ruins.
= Over thirty different varieties have been identified, includ-
ing ivy, fennel, and fig. As the roots grow they open up
further cracks, allowing even more weeds in.
attraction brings half a million visitors each year.
= No walkways for viewing platforms have been constructed,
so tourists walk along ancient paths and enter buildings that
are not roped off. In some places ancient lead water pipes
have been exposed.
Finally, there seems to be no overall management plan for the
site.
= Damaged paths and walls have not been repaired, frescoes
have not been preserved, and mangy dogs roam the site.
Available finance has not been managed, and no conserva-
tion and interpretation programme has been put in place.
=
As a result of these factors, damage to the site remains an issue for
archaeologists and administrators. There are decisions which
need to be made in relation to restoration work, uncontrolled
vegetation, tourism, and management which need to be addressed
if the site is to be preserved for future generations of research and
public access.
This bring us to appraisal, and the question of how and how far to make
explicit the evaluative stance appropriate to a history text of this kind. One
fundamental value in historical axiology is the preservation of archaeo-
logical sites, and so propagating a negative prosody is relevant here –
including negative judgements of individuals, agents, and agencies failing
to preserve the site, and negative appreciations of phenomena involved.
Propagation of this prosody is highlighted in Text 10.
[10] To begin, much of the restoration work on Pompeii has been done by
local firms with no specialized knowledge of restoration techniques.
For example, the timber roof on the House of Meleager was so poorly
designed it could not support the weight of the tiles and collapsed.
Poor quality mortar has also been used to protect ancient
stonework. Over time this mortar has cracked, allowing water and
A second factor is the incursion of dangerous weeds which have
hastened the decay of the ruins. Over thirty different varieties have
been identified, including ivy, fennel, and fig. As the roots grow they
open up further cracks, allowing even more threatening weeds in.
attraction brings half a million visitors each year. No special walk-
ways for viewing platforms have been constructed, so tourists walk
along ancient paths and enter buildings that are not roped off. In
some places ancient lead water pipes have been carelessly exposed.
been poorly managed, and no proper conservation and interpret-
ation programme has been put in place.
Here, three negative judgements have been inscribed, the first of which is
intensified (so poorly designed, have been carelessly exposed, has been poorly
managed); and there are six negative appreciations, the fourth quantified (no
specialized9 knowledge of restoration techniques, poor quality mortar, dangerous
weeds, even more threatening weeds in, no special walkways, no proper conser-
vation and interpretation programme). This puts us in a position to predict the
negative prosody of conservation problems in the text’s macro-Theme
(adversely affected by a number of conservation problems) . . .
Since its discovery, Pompeii has been damaged as an archaeological site,

adversely affected by a number of conservation problems, including the
quality of restoration work, vegetation, tourism, and site management
issues.
. . . and to amplify the prosody in its macro-New – reiterating the concerns

raised and appreciating their significance (a key issue, important decisions):
As a result of these problems, damage to the site remains a key issue for
archaeologists and administrators. There are important decisions which
need to be made in relation to poor restoration work, invasive vegetation,
insensitive tourism, and neglectful management which need to be
addressed if the site is to be preserved for future generations of research
and public access.
Through these successive iterations we arrive at the factorial explanation

(Coffin 2006; Martin and Rose 2008) presented as Text 11a, with its generic
structure labelled stage by stage as a list of factors explaining an outcome
9
Strictly speaking there is an interaction of engagement (no realizing contract: deny) and attitude (specialized realizing
appreciation: valuation) enacting the negative evaluation here (as for no special walkways, no proper conservation and
interpretation programme as well).
374 J. R. MARTIN
and culminating in re-appreciation of that outcome in the Reinforcement of

Factors stage.
[11a] Outcome
Since its discovery, Pompeii has been damaged as an archaeological
site, adversely affected by a number of conservation problems,
including the quality of restoration work, vegetation, tourism, and
site management issues.
Factor 1
To begin, much of the restoration work on Pompeii has been done by
local firms with no specialized knowledge of restoration techniques.
For example, the timber roof on the House of Meleager was so poorly
designed it could not support the weight of the tiles and collapsed.
Poor quality mortar has also been used to protect ancient stonework.
Over time this mortar has cracked, allowing water and vegetation to
penetrate.
Factor 2
Factor 3
Factor 4
Reinforcement of factors
As a result of these problems, damage to the site remains a key issue
for archaeologists and administrators. There are important decisions
which need to be made in relation to poor restoration work, invasive
vegetation, insensitive tourism, and neglectful management which
need to be addressed if the site is to be preserved for future gener-
ations of research and public access.
A factorial explanation of this kind can be topologically related to an

exposition, where the point is not to explain historically how something
came to be the case, but rather to argue for a particular interpretation of
events. An appropriate macro-Theme for an argumentative text of this kind
would deploy appraisal to create a debatable issue,10 with implications
across all discourse semantic systems for choices promoting one position
or another in the stages which follow:
Pompeii has been described as a victim of state neglect and indifference and
an archaeological catastrophe of the first order. Its ongoing destruction
since its discovery in the 1590s has arguably resulted in a greater disaster
than its initial destruction by the eruption of Mt Vesuvius one and a half
millennia earlier.
From another perspective, the ideation at risk here could have been used to
construe a classifying report listing the ways in which archaeological sites
are potentially at risk. A suitable macro-Theme might be the following:
There are several types of damage that can affect archaeological sites. These
include the quality of restoration work, vegetation, tourism, and site man-
agement issues.
In a genre of this kind, appraisal has a much quieter role to play since the
focus is on classifying and describing types of phenomena (in this case types
of activities that are nominalized as abstract things). The macro-Theme
classifies the types of damage, and the remainder of the text describes each
type in detail. For relevant discussion of genre typology and topology, see
Martin and Rose (2008).
Returning to our factorial explanation, it is salutary to keep in mind that
most of the scaffolding we have introduced whereby Text 11a in effect
announces its genre is invisible to an untrained eye. In everyday terms,
only the paragraphing is visible (Text 11b below) as a reflection of all that is
going on. Bringing the discourse semantics of genres to consciousness, and
using this knowledge to design and inform teaching practice, has been an
ongoing concern in SFL since the inception of the genre-based literacy
pedagogy of the ‘Sydney School’ (Rose and Martin 2012).
[11b] Since its discovery, Pompeii has been damaged as an archaeo-

logical site, adversely affected by a number of conservation prob-
lems, including the quality of restoration work, vegetation, tourism,
and site management issues.
To begin, much of the restoration work on Pompeii has been done
by local firms with no specialised knowledge of restoration tech-
niques. For example, the timber roof on the House of Meleager was
so poorly designed it could not support the weight of the tiles and
10
Note how expanding engagement resources have been used here to position the inscribed evaluation as contestable:
has been described as, arguably.
376 J. R. MARTIN
collapsed. Poor quality mortar has also been used to protect ancient
stonework. Over time this mortar has cracked, allowing water and
As a result of these problems, damage to the site remains a key
issue for archaeologists and administrators. There are important
decisions which need to be made in relation to poor restoration
work, invasive vegetation, insensitive tourism, and neglectful man-
agement which need to be addressed if the site is to be preserved for
future generations of research and public access.
Because we have chosen to exemplify discourse semantic systems by syn-

thesizing a piece of writing, negotiation resources have not as yet entered
into our discussion here. We might however imagine situating Text 11a as a
response to an exam question, as in the following exchange:
Explain how Pompeii has been affected since its discovery as an

archaeological site.
— Since its discovery, Pompeii has been damaged as an archaeological site,
adversely affected by a number of conservation problems, including the qual-
ity of restoration work, vegetation, tourism, and site management issues . . .
Or we might imagine using Text 11a in a Reading to Learn programme11

and synthesizing a set of detailed reading exchanges that guide students to
recognize features of the genre, such as appraisal, metadiscourse, and genre
staging, in the following:
The first sentence tells us what’s happened to Pompeii since it became an

archaeological site, and previews the factors that have damaged it.
The sentence would then be read aloud, and its elements identified and
elaborated as follows:
11
For access to this pedagogy, see Rose and Martin (2012) and Rose’s R2L website: www.readingtolearn.com.au.
Teacher It says Pompeii has been damaged as an archaeological site,

then there are two words that mean badly affected. Can you
see those two words?
Students – adversely affected
Teacher – Exactly
Let us highlight adversely affected.
Adversely means ‘negatively’, so it’s been affected for
the worse.
Teacher Then it says what’s affected it.
Can you see what it’s been affected by?
Students – a number of conservation problems
Teacher – Right.
Let us just highlight conservation problems.
So these problems are the factors that have damaged it. They
have made it hard to conserve Pompeii, to look after it.
Teacher Then it tells us four factors.
What’s the first factor?
Students – the quality of restoration work
Teacher Yep.
...
Teacher So this paragraph is called the Outcome of the explanation.
It tells us what the outcome is – Pompeii has been damaged –
and then it previews the factors that caused this outcome.
Now each of these factors is explained in more detail.
...
Teacher How do we know we’re up to the first factor?
Students – To begin,
Teacher – Exactly right.
I will leave it to our educational linguists to synthesize possible extensions

to classroom interaction of this kind. The only comment I will add here is to
foreground the ways in which text-focused exchanges of this kind make it
possible for teachers and students to attend explicitly to the discourse
semantic patterns which construe, enact, and compose a genre – the very
patterns I have been exemplifying as Texts 1 to 10 unfold in this brief tour.
14.4 Meaning beyond the Clause
Is this chapter, I have outlined and exemplified discourse semantic

resources, drawing on the description of English resources presented in
Martin (1992), Martin and Rose (2003), and Martin and White (2005). As
noted above, this model reinterprets cohesion as discourse semantic system
and structure – on a deeper stratum realized through lexicogrammar. This
378 J. R. MARTIN
Figure 14.1 An outline of discourse semantic systems (by metafunction)
makes it possible to interpret discourse semantic systems metafunctionally,

as outlined in Figure 14.1 above (in the proportions ideation and conjunc-
tion are to ideational meaning, as negotiation and appraisal are to interper-
sonal meaning, as identification and periodicity are to textual meaning).
This also makes it possible to interpret discourse semantic systems as
realized congruently or incongruently in lexicogrammar and phonology
(i.e. as involving grammatical metaphor or not), as realized congruently
across one or more lexicogrammatical and phonological systems (e.g. affect
realized through Comment Adjuncts, mental processes of reaction and
desire, or attitudinal Epithets), and as connecting phases of discourse of
indefinite extent.
This last point perhaps needs some elaboration. What we are saying here is
that discourse semantic structures involve both ‘local’ and ‘extended’ real-
izations. Conjunction, for example, especially external conjunction, regularly
relates one figure to another (as detailed in Text 5b above); but in writing
internal conjunction regularly relates more than one figure to another (as
illustrated in Text 7 above). Similarly, identification regularly relates one
entity to another, via anaphoric reference (as highlighted in Text 4 above);
but with text reference, indefinitely long passages of discourse can be related
to one another (as exemplified in Texts 9 and 11 above: as a result of these
factors, as a result of these problems respectively). The same holds true for other
discourse semantic systems (see the scope of clause level vs. higher level
periodicity; lexical cohesion vs. metadiscourse; negotiation of the exchange
of goods and services and information vs. regulatory linguistic services;
attitude triggered by or targeted to entities vs. evaluations of discourse).
Discourse semantic relations are abduced, in other words, between stretches
of unfolding discourse of indefinite extent. It is perhaps in this respect that
discourse semantics can be seen as most strongly complementary to
Figure 14.2 Discourse semantics in relation to register and lexicogrammar
lexicogrammatical and phonological resources,12 where units of structure

(e.g. clause, group/phrase, word, and morpheme; tone group, foot, syllable,
and phoneme) bound the scope of the meaning to be made.
Reinterpretation of cohesion as metafunctionally organized discourse
semantic systems also has important implications for analyzing context, in
models which treat context as a higher level of meaning. As outlined in
Figure 14.2 above, discourse semantics is positioned here as realized through
lexicogrammar and as realizing register – where register is a cover term for
field, tenor, and mode systems in a model of context stratifying register and
genre. As outlined in Martin (1999, 2014), this model was developed to
salvage Halliday’s proposal that field is by and large construed through
ideational meaning, tenor by and large enacted through interpersonal mean-
ing, and mode by and large composed through textual meaning. Models in
which the whole of cohesion is interpreted as textual meaning confound this
picture,13 almost as badly, it might be argued, as it is confounded by failing
to position field, tenor, and mode as realizing genre. But this unfortunately
takes us well beyond the remit of this chapter. I will have to stop here.
References
Bateman, J. A. 2014. Text and Image: A Critical Introduction to the Verbal/Visual

Divide. London: Routledge.
12
Lemke (1985:287–8) refers to these semantic relations as covariate relations, contrasting them with the multivariate
and univariate relations constituting lexicogrammatical structure.
13
Equally confounding is the practice in much SFL research of treating field, tenor, and mode as realized directly by
lexicogrammatical systems, bypassing discourse semantic systems altogether in text analysis. The final chapter of
Martin et al. (2010) explores an alternative rite of passage.
380 J. R. MARTIN
Berry, M. 1981. Systemic Linguistics and Discourse Analysis: A Multi-

Coffin, C. 2006. Historical Discourse: The Language of Time, Cause and Evaluation.
London: Continuum.
Fawcett, R. 2008. Invitation to Systemic Functional Linguistics through the Cardiff
Grammar. Sheffield: Equinox.
Fries, P. H. 1981. On the Status of Theme in English: Arguments from
Discourse. Forum Linguisticum 6(1): 1–38.
Gleason, H. A., Jr. 1968. Contrastive Analysis in Discourse Structure. In J. E.
Alatis, ed., Contrastive Linguistics and Its Pedagogical Implications, Report of the
Nineteenth Annual Round Table Meeting on Linguistics and Language Studies.
Georgetown: Georgetown University Press. 39–63.
Halliday, M. A. K. 1967. Intonation and Grammar in British English. The Hague:
Mouton de Gruyter.
Halliday, M. A. K. 1970. A Course in Spoken English: Intonation. London: Oxford
University Press.
Dialogue. In R. Fawcett, M. A. K. Halliday, S. M. Lamb, and A. Makkai,
eds., The Semiotics of Language and Culture, Volume 1: Language as Social
Semiotic. London: Pinter. 3–35.
Halliday, M. A. K. 1998. Things and Relations: Regrammaticising Experience
as Technical Knowledge. In J. R. Martin and R. Veel, eds., Reading Science:
Critical and Functional Perspectives on Discourses of Science. London: Routle-
dge. 185–235.
Halliday, M. A. K. and W. S. Greaves. 2008. Intonation in the Grammar of
English. Sheffield: Equinox.
Halliday, M. A. K. and R. Hasan. 1985. Language, Context, and Text:
sity Press.
through Language: A Language-based Approach to Cognition. London: Cassell.
Hasan, R. 2009. Semantic Variation: Meaning in Society and Sociolinguistics.
Knight, N. 2010 Wrinkling Complexity: Concepts of Identity and Affiliation
in Humour. In M. Bednarek and J. R. Martin, eds., New Discourse on
Language: Functional Perspectives on Multimodality, Identity and Affiliation.
Knight, N. 2013. Evaluating Experience in Funny Ways: How Friends Bond

through Conversational Humour. Text & Talk 33(4–5): 553–74.
Lemke, J. L. 1985. Ideology, Intertextuality and the Notion of Register. In
J. D. Benson and W. S. Greaves, eds., Systemic Perspectives on Discourse,
Volume 1: Selected Theoretical Papers from the 9th International Systemic Work-
shop. Norwood: Ablex. 275–94.
Martin, J. R. 1983. Participant Identification in English, Tagalog and Kâte.
Australian Journal of Linguistics 3(1): 45–74.
Benjamins.
Martin, J. R. 1999. Modelling Context: The Crooked Path of Progress in
Contextual Linguistics. In. M. Ghadessy, ed., Text and Context in Functional
Martin, J. R. 2013a. Systemic Functional Grammar: A Next Step into the Theory –
Axial Relations. Beijing: Higher Education Press.
Martin, J. R. 2013b. Modelling Context: Matter as Meaning. In C. Gouveia
and M. Alexandre, eds., Languages, Metalanguages, Modalities, Cultures: Func-
tional and Socio-discursive Perspectives. Lisbon: BonD & ILTEC. 10–64.
Clause. London: Continuum. 2nd Revised Edition.
Equinox.
Martin, J. R. and D. Rose. 2012. Learning to Write, Reading to Learn: Genre,
Knowledge and Pedagogy in the Sydney School. Sheffield: Equinox.
Martin, J. R., C. M. I. M. Matthiessen, and C. Painter. 2010. Deploying
Functional Grammar. Beijing: Commercial Press.
Martin, J. R., M. Zappavigna, P. Dwyer, and C. Cleírigh. 2013. Users in Uses
of Language: Embodied Identity in Youth Justice Conferencing. Text &
Talk 33(4–5): 467–96.
Sinclair, J. and M. Coulthard. 1975. Towards an Analysis of Discourse: The
English Used by Teachers and Pupils. London: Oxford University Press.
Ventola, E. 1987. The Structure of Social Interaction: A Systemic Approach to the
Semiotics of Service Encounters. London: Pinter.
15
Appraisal
Susan Hood
15.1 Introduction
This chapter is offered in support of educators and scholars interested in

becoming more familiar with the theorization and application of system
choices in appraisal. The chapter moves from theory to practice. First,
appraisal is introduced and located within the broader theoretical frame-
work of Systemic Functional Linguistics (SFL). This is necessarily an abbre-
viated description. For more comprehensive theoretical accounts of
appraisal, readers are referred to sources such as Martin (2000), Martin
and White (2005), and Martin and Rose (2007). Other contributions to the
development of the theory and to its application in research are referenced
throughout the chapter. Descriptions of applied studies address diverse
objects of study and methods of research.
15.2 Appraisal as a System of Interpersonal Meaning
Important to the initial and evolving theorization of appraisal is Martin’s

(1992a) stratification of context as register and genre, and of the content plane
of language as discourse semantics and lexicogrammar. In relation to context,
early work from the 1980s on story genres after Labov (e.g. Rothery 1984; Martin
and Plum 1997) prompted closer attention to the place of evaluation in SFL.
Martin (2014:17) reflects on these contributions, noting ‘the importance of the
type of evaluation used as well as its placement in genre structure’ as a means of
distinguishing among different types of story genres and that the ‘point of a
story depends on the interaction of evaluative language with ideational mean-
ing’. In an anecdote, for example, the point is to share an affective response to an
unexpected event. In an exemplum, it is to share judgement of an incident.
Stratification of the content plane of language as discourse semantics
and lexicogrammar (Martin 1992a; Martin and Rose 2007) provided the
Appraisal 383
foundation for a view of evaluation beyond grammar. From the perspective

of lexicogrammar, evaluative meanings are realized across a range of
systems including, for example, affective mental processes, epithets in
nominal groups, adverbial groups realizing comment adjuncts, and circum-
stances of manner. However, to explore rhetorical expressions at the level
of whole texts, a system was required that could ‘generalise across these
diverse lexicogrammaticalizations, bringing feelings together in relation to
one another’ (Martin 2014:17).
Comprehensive accounts of appraisal as a system of interpersonal mean-
ing in discourse include Martin (1992a), Martin (2000), and Martin and
White (2005). The final publication offers the single most elaborated
account to date. However, the diversity of contexts in which the tools of
appraisal have been applied has generated multiple field-specific accounts,
and in some cases the particular challenges those contexts pose have
continued to push the theory forward.
Notable early contributions include Rothery and Stenglin’s (1997, 2000)
work in educational linguistics. They were motivated to better understand
the linguistic challenges for school students in the effective management of
evaluative discourse in writing literary response texts. Fuller’s (1998) study
of the language of popular science was foundational to the later develop-
ment by White (1998) of the appraisal system of engagement, which
presents options for the negotiation of values in text. White’s work has
predominantly focused on media discourse, and other significant early
publications in that field include Feez et al. (1994) and White (2003). The
analysis of conversational talk in Eggins and Slade (1997) was also influen-
tial. Many others have followed in the footsteps of these early studies,
expanding contexts of exploration and deepening understanding.
15.2.1 Contextualizing Appraisal in SFL Theory

The contextualization of appraisal within the broader theory of SFL is
critical to understanding its potential as a research tool. If appraisal is
amputated from its place in its broader theoretical structure, the interpret-
ations it affords and its value as a framework for empirical study are
necessarily diminished. Lost are the critical relations to other dimensions
of SFL theory, not only within the metafunctional realm of interpersonal
meaning, but also in relation to the other metafunctions. To clarify, I begin
by situating appraisal in relation to some key dimensions of the architec-
ture of an SFL model of language.
Meaning in SFL is theorized as metafunctional, that is, as always and
simultaneously ideational, interpersonal, and textual (Halliday 1985, 1994;
Martin 1992a; Halliday and Matthiessen 2004). Ideationally, language func-
tions to construe kinds of experience: it is about something. Interperson-
ally, language functions to enact roles and share values in the negotiation of
social relations. Textually, language functions to compose the flow of
384 SUSAN HOOD
information into messages that make sense in relation to a mode of inter-

action and co-textual setting. A discussion of appraisal as a model of
evaluative language primarily locates us in the realm of interpersonal
meaning – primarily, as in all instances of text there are interactions across
metafunctions. For example, in any analysis of evaluation in discourse we
need to consider both what is being evaluated, which implicates ideational
meaning, and how evaluation is textured, which implicates textual mean-
ing. This issue is returned to later in the chapter.
SFL models language as tri-stratal, as composed of meaning systems at
the levels of phonology/graphology (Halliday and Greaves 2008), lexico-
grammar (Halliday 1985, 1994; Halliday and Matthiessen 1999), and dis-
course semantics (Martin 1992a; Martin and Rose 2007). The three strata are
in a relationship of realization across relative levels of abstraction. In other
words, patterns in language at the level of text (discourse semantics) are
realized in patterns of language at the level of clause (lexicogrammar),
which are in turn realized as expression in sound or writing systems
(phonology/graphology). We can explore interpersonal meaning across all
strata, but reference to appraisal means we are approaching evaluation
from the stratum of discourse semantics (Martin and White 2005).
A further important note is that appraisal is one of two discourse semantic
systems of interpersonal meaning. Where appraisal models the -personal in
interpersonal meaning, a system of negotiation models the inter- of the
interpersonal. The implication is that in analyzing texts, appraisal is not
loaded with the full responsibility to account for the potential to mean
interpersonally. The differentiated responsibilities of the two systems con-
cern the complementarity of relations of solidarity and power.1 The brief
theoretical contextualization provided here establishes a general frame-
work, which will be extended in descriptions of how appraisal has been
applied in research. Figure 15.1 presents the three sub-systems of appraisal
that comprehensively account for evaluative meaning potential in dis-
course in English.
Attitude models the general categories of values that can be expressed:
affect, appreciation, and judgement. Graduation models options for
scaling attitudinal meanings by degree. Engagement models options for
introducing and managing space for alternate propositions or proposals.
Each dimension of appraisal opens up more delicate system networks of
meaning choices.
15.2.2 The System of Appraisal as Attitude

Figure 15.2 presents the system of attitude as a skeletal network of
choices. Instances of expression select for the category of feeling as affect,
1
For further insights into the complementary system of N EGOTIATION , see Martin and Rose (2007: Chapter 7).
Appraisal 385
Figure 15.1 An outline of the system of appraisal (from Martin and White 2005:38)
affect
category judgement
appreciation
positive
ATTITUDE
‘vibe’
negative
inscribe
mode of
realization
invoke
Figure 15.2 Simultaneous choices in identifying instances of ATTITUDE (adapted from Liu
2017:74)
appreciation, or judgement. Additionally, we select for positive or nega-

tive value or ‘vibe’ (suggested by Martin in person correspondence, to
differentiate from the grammatical system of polarity) and for the mode
of realization.
The mode of realization refers to options for expressing attitude either
explicitly as inscribed attitude or implicitly as invoked attitude (Martin
2000; Martin and White 2005). While this choice is represented in
Figure 15.2 as a simple dichotomy, options for invoking attitude can be
positioned on a cline with some approximating inscription and others more
distanced and relatively more implicit. These more delicate options are
386 SUSAN HOOD
discussed later under graduation. I elaborate first on the categorical

choices in expression in attitude.
15.2.3 Attitude as Affect

An expression of evaluation as affect is one of positive or negative emo-
tion. It may be realis or irrealis phenomena that trigger the emotional
response, or alternatively the emotion may constitute a general state or
disposition. The semantic space of affect is differentiated as follows:
• un/happiness, e.g. miserable, ecstatic

• dis/satisfaction, e.g. disappointed, content
• in/security, e.g. suspicious, assured
• dis/inclination, e.g. fearful, longing for
Affect as dis/inclination expresses an irrealis choice in which ‘feelings

relate to future, as yet unrealized, states rather than present existing ones’
(Martin 2000:150), a distinction reflecting that between desiderative and
emotive mental processes, as in I wanted them to win/I like them winning
(Martin 2017:31).
Each kind of affect opens to more delicate meaning choices. In/security,
for example, is glossed in Martin and White (2005:49) as to do with ‘our
feelings of peace and anxiety in relation to our environs’. Finer categories of
in/security include ‘confidence’ or ‘trust’, and ‘disquiet’ or ‘surprise’. Bed-
narek (2008), approaching a study of attitude from a corpus-based per-
spective, suggests that ‘surprise’ (as triggered insecurity) ought to
constitute a separate category of affect, given that the appearance of the
word ‘surprise’ frequently co-occurs with both positive and negative evalu-
ative expressions, as in It was a lovely/nasty surprise.
Martin (2017) emphasizes the important role of corpus studies in
providing supportive evidence for classification schemes for attitude.
However, he points to some important theoretical features that need to be
accounted for in applying this mode of analysis. First, it is important that
the name of specific categories of emotion should not be confused with
the name of a specific feeling. This can lead to problems in analyses and
claims. The category named ‘surprise’ refers to a realm of emotion as
response to a sudden disruption of expectancy, and hence security. Lexical
realizations, other than surprise itself, include feelings of being ‘disturbed,
shocked, unsettled, stunned, staggered, thrown, taken aback . . . jolted’
(Martin 2017:37), all of which express a sense of negative security. It is
also important to note that the positive/negative value of the affect does
not have to match that of the trigger. So in a lovely surprise, the trigger is
negatively appreciated as disruptive, while the affectual response is posi-
tive. Martin also notes that while available corpora might appear massive,
they are not yet of a size that can generate adequate numbers of instances
Appraisal 387
for other than a few core lexical expressions. For a more detailed linguistic
justification for the categorizations in Martin and White (2005), see Martin
(2017:33–6). In the interest of clarification, Martin (2017) does propose a
renaming of the sub-category of triggered insecurity, from ‘surprise’ to
‘perturbance’.
Martin (2017) explains the development of this typological perspective on
affect as an evolutionary process:
I was parenting a small child at the time and suggested categories based on
my reading of his emotional repertoire in relation to his parents coping (or
not) with his moments of distress – basically asking whether he was
unhappy because he wanted his mother or father (contented sociability), or
because he wanted the comfort of his security blanket (which he called
‘baggy’), or because he wanted the satisfaction of his bottle (‘bopple’). This
gave us the [unhappiness/happiness], [insecurity/security] and [dissatisfaction
/satisfaction] oppositions.
(Martin 2017:31)
He continues:
In retrospect, if work on space grammar had already been available at the

time, I might equally well have drawn on Stenglin’s (e.g. 2009) notions of
bonding (in relation to [un/happiness]) and binding (in relation to [in/secur-
ity]), and McMurtrie’s (e.g. 2013 [2017]) concept of promenade (in relation to
the telos oriented notion of [dis/satisfaction]).
(Martin 2017:31)
This is illustrated in Table 15.1.
15.2.4 Attitude as Judgement

judgement refers to the evaluation of people or their behaviour; it is
appraisal as praise or criticism. Expressions of judgement may relate to
either social esteem or social sanction (Martin and White 2005:52). Social
esteem has to do with judgement as ‘normality’ (how un/usual a person
is); ‘capacity’ (how in/capable they are); and ‘tenacity’ (how ir/resolute they
are). Social sanction has to do with judgement as ‘veracity’ (how un/
truthful a person is) or ‘propriety’ (how un/ethical they are) (Martin and
White 2005:52).
Table 15.1 Additional AFFECT parameters (Martin 2017:31)
‘parenting’ ‘space grammar’
un/happiness Mummy/Daddy bonding

in/security baggy binding
dis/satisfaction bopple promenade
388 SUSAN HOOD
15.2.5 Attitude as Appreciation

Expressions of appreciation evaluate entities and events. At a finer level
of delicacy, instances of appreciation are differentiated as follows:
• reaction, e.g. fascinating, boring, beautiful, ugly

• composition, e.g. complex, simple, harmonious, flawed
• valuation, e.g. exceptional, insignificant
Martin and White (2005:57) offer metafunctional associations for each of

the domains of appreciation, ‘with reaction oriented to interpersonal
significance, composition to textual organization and valuation to idea-
tional worth’. Appreciation expressed as ‘reaction’ describes the degree
to which a reader or audience notices or is made aware of the phenomenon
described. In other words, our evaluations of phenomena are expressed in
terms of how we react to them. Here a further distinction can be made.
Reaction as ‘impact’ is glossed in Martin and White (2005:56) as addressing
the question Did it grab me?, as in ‘a fascinating read’, or ‘an unremarkable
performance’. Reaction as ‘quality’ is glossed as Did I like it?, as in ‘an
interesting artefact’, or ‘an ugly building’. Because we are talking about
appreciation of phenomena in terms of their affectual impact on the
appreciator, we approximate the realm of affect. This approximation
can often prove challenging for learners of English as another language,
where slippage can readily occur between expressions such as It is interesting
and I am interested – resulting in inadvertent exclamations such as I am so
boring!. More delicate options are also noted for appreciation as ‘compos-
ition’, more delicately as ‘balance’ (Did it hang together?), or ‘complexity’
(Was it hard to follow?).
The point has been made, by Macken-Horarik and Isaac (2014) and others,
that attitude as appreciation ‘is the most sensitive to context’, resulting
in a potential need to modify the categorical dimensions in relation to
‘specific discourse contexts, texts and topics’ (Macken-Horarik and Isaac
2014:74). For more comprehensive accounts of categorical distinctions in
attitude, see Martin and White (2005:48–57) and Martin and Rose
(2007:65–71).
The idea of being able to reconfigure typological distinctions as topo-
logical relations was touched on above, and is presented diagrammatically
in Figure 15.3, in which relative proximities of semantic realms of atti-
tude are indicated.
15.3 The System of Appraisal as Graduation
A second semantic domain of appraisal is graduation. This sub-system

concerns the potential available in language for ‘meaning by degree’
(Martin 1992b). Meanings can be adjusted or graduated along dimensions
of force or focus.
Appraisal 389
Figure 15.3 A topological perspective on ATTITUDE resources (from Martin 2000:165)
15.3.1 Graduation as Force

Force refers to the up-scaling or down-scaling that adjusts the ‘intensifica-
tion’ of a value, for example, from important to very important or to not so
important. This kind of adjustment has been analogized as turning up or down
the volume control on a sound system (Martin 1992a:533; Martin and White
2005:37). In the examples above, attitude is realized as a quality, and the
intensity is adjusted up or down. Expressions of attitude can also be realized
in a circumstance of manner around a process, and by such means the process
is adjusted in intensity, as in rigorously investigated. This option is referred to as
‘vigour’ (Hood 2010; Martin and White 2005). A third means of ‘intensifica-
tion’ associates with lexicalized modulation, which adjusts a whole proposal
in terms of degree of obligation, as in It is vital that you follow my instructions.
When a quality is expressed as an entity, as in importance or satisfaction,
degrees of force express +/- ‘quantification’ (rather than ‘intensification’).
This is evident in expressions such as more importance, much happiness, or little
satisfaction. ‘Quantification’ may be in terms of number (amount, volume,
or mass) or extent, with extent as either proximity or distribution in Martin
and White (2005), and as scope or distance in Hood (2010). In both cases the
options apply to time and space. A process may also be quantified as
frequency, as in He was very often anxious.
As an aside, some years ago when observing the teen talk of my stepson
and his friends, I noted their creative play with evaluative language
involved a flipping of the expected association of intensification + quality
and quantification + entity, to arrive at so fun and heaps cool.
15.3.2 Graduation as Focus

graduation as focus refers to the relative sharpening or blurring of
categorical boundaries. As such it can be analogized to the focus function
390 SUSAN HOOD
on a camera. This is exemplified in expressions such as absolute misery, real

research, a vaguely relevant study, a sort of pleasure, pure joy.
In Martin and White (2005) graduation is explored in relation to the
adjustment of options in both attitude and engagement (e.g. Martin and
White 2005:136). In Hood (2010) and Hood and Martin (2007) it is explored
primarily in relation to attitude and particularly as a resource for invok-
ing rather than inscribing attitude. In written academic texts, the play of
inscribed and invoked attitude has been shown to be of strategic import-
ance (Hood 2010). Where writers frequently express inscribed attitude in
reporting on their object of study, more restraint is typically exercised in
reporting on other contributions to knowledge. A preferred strategy is to
adjust objective (experiential) meanings by relativizing or subjectifying
them and by so doing to invoke or flag attitude. Hood (2010) suggests that
an avoidance of dichotomizing inscriptions of attitude enables the aca-
demic writer to maintain a veneer of objectivity while implying stance.
15.3.3 Graduation Invoking Attitude

Force as intensification and quantification are exemplified above with
instances of inscribed attitude. However, force can also function to
grade experiential (non-attitudinal) meanings. For example, the underlined
in the nominal group an action-oriented study is an experiential classifier. If
we adjust this meaning by degree to a more action-oriented study, the process
of ‘intensification’ shifts the function to an epithet and implies an attitu-
dinal position. The grading of experiential meaning in a process may also
invoke an attitudinal interpretation. For example, in choosing to express an
investigative activity as explore rather than look into, we can imply intensi-
fied vigour and again an evaluative potential. The grading of modulation
(e.g. ought to/need to/must) may also invoke an interpretation of a proposal
as relatively more significant. As noted above, this can move to inscription
when lexicalized as in, for example, essential or vital.
Quantifying entities as number (amount, volume, or mass) adjusts experi-
ential meaning by degree, as in many studies, a sizeable volume of literature, the
weight of research. The adjustment of quantity in each case encodes a sub-
jective orientation, and opens a space for it to be read evaluatively. Martin
and White (2005) refer to this as ‘flagging’ an attitudinal interpretation.
There are similar implications in expressions of frequency and extent as
scope or distance, as exemplified in the following:
• graduation as scope: in time – it’s been like this for decades

• graduation as scope: in space – it’s a pervasive phenomenon
• graduation as distance: in time – in a more recent occurrence
• graduation as distance: in space – a further removed example
The study of graduation in the context of academic research papers has

led to some proposed extensions to the system of graduation (Hood 2010).
Appraisal 391
This is particularly so with respect to focus. The model of focus in Martin

and White (2005) concerns the adjustment of boundaries around entities
(e.g. a true father, an apology of sorts).
Hood (2010) suggests an extension to what is considered a category of
ideational meaning beyond that of entity to include process and figure (as a
proposition). The adjustment of focus here is generalized as ‘fulfilment’.
The boundary around a process can be sharpened or softened by adjusting
the degree of ‘completion’ (e.g. tried to show, managed to show). Implicated
here are resources of phase: ir/realis in the verbal group. The boundary
around a figure (as proposition) can be sharpened or softened by adjusting
the degree of ‘actualization’ (e.g. seems to show, definitely shows). Implicated
here are resources of conation and modalization in the verbal group. Again,
the grading of non-attitudinal (experiential) meanings encodes a subjective
positioning on behalf of the speaker/writer, and hence an indication to the
audience to interpret attitudinally (Hood 2010; Hood and Martin 2007).
Graduation options are shown in Figure 15.4.
15.3.4 The Function of Concessive Contractors in Negotiating

Graduated Attitude
In a study of problematic in-bound customer service calls to a call-centre,
Hood and Forey (2008) noted the significant role that concessive contractors
such as just, already, once, yet, and actually played in adjusting and managing
the attitudinal intensity of caller emotions in the flow of interaction. They
Figure 15.4 A network of graduation options (adapted from Hood 2010:105)
392 SUSAN HOOD
Figure 15.5 A network of modes of realization for ATTITUDE (from Hood and Martin
2007:746)
found that rather than mirroring the amplified levels of emotive intensity
on the part of customers, the agents made frequent use of concessive
contractors (and sometimes silence) to rein in or defuse the attitudinal
intensity. For example, an expression such as just a few days on the part of
an agent could defuse an amplification of extended time by the caller.
15.3.5 Degrees of Invocation in Implicit Expressions of Attitude

Graduation is discussed above as a significant resource by which users
imply rather than directly encode an attitudinal position in discourse.
There are also other means by which users can achieve this end.
A network of options for invoking attitude is in Figure 15.5. The options
in the system can also be interpreted as a cline of relative implicitness, or
relative commitment of an attitudinal meaning.
The closest we come to direct inscription of attitude is in the use of
idiom or lexical metaphor. Both are said to provoke an attitudinal inter-
pretation. Chang (2017:324) notes that most idioms are ‘figurative fixed
expressions’, such as cool as a cucumber, too hot to handle, as sharp as a tack. As
idioms are taken up in common usage in a culture, they are essentially
divested of the remnants of an imported field. While not gradable in
themselves, they tend to become overtly positive or negative and can be
readily substituted with an inscription of attitude (Chang 2017). Lexical
metaphor is interpreted as marginally more implicit. The appropriated
reference to the literal field, as in He’s a prisoner of his own volition, brings
an associated value to its metaphoric usage.
Yet more implicit choices are referred to in Figure 15.5 as ‘inviting’ an
attitudinal interpretation. That interpretation becomes more reliant on the
co-text and/or on the field of the discourse. The option of inviting an
evaluative reading as ‘flagging’ refers to the deployment of resources of
graduation to adjust the force or focus of an experiential meaning, as
discussed in some detail above. The option of ‘affording’ an evaluative
interpretation is the most implicit and so the more influenced by the
general field of the text, the broader cultural setting and assumptions of
shared values, and the subjective stance of the reader (Macken-Horarik and
Isaac 2014). So, a description of a classroom as traditional might in one
context be read as part of a constellation of negatively axiologically charged
terms such as old-fashioned, teacher-centred, or oppressive. For another it might
Appraisal 393
assume a positive charge, and for yet another it might be read as purely
categorical, as describing, for example, a room with desks in rows and a
blackboard at the front of the room.
15.4 Prosodic Structuring of Attitudinal Meaning
A final aspect of the theorization of expressions of attitude in texts has to

do with the textual structure associated with interpersonal meanings and
hence evaluation, that of prosody. The term was first used by Firth (see
Palmer 1970) to refer to non-segmental patterns in phonology, and has been
applied more broadly in SFL to refer to the structuring of interpersonal
meanings in lexicogrammar and discourse semantics (Halliday 1985, 1994;
Martin 1992a). Prosodic patterning has been described as a ‘spread, sprawl,
smear or diffusion of interpersonal meanings that accumulate, reinforce, or
resonate with each other to construct an evaluative “key” over an extended
segment of text’ (Hood 2010:141). Other discussions of interpersonal pros-
ody are found in Lemke (1998); Martin (1992a); Martin and Rose (2007);
Macken-Horarik (2003).
Lemke (1998) proposes the term ‘propagate’ to describe the process by
which prosody spreads in text. Hood (2006:38) draws on the metaphor of
colouring to refer to the impact of a prosody radiating out from an inscrip-
tion of attitude. The means by which prosodies propagate across ranks and
strata are illustrated in Hood (2006:37–49). At the clause level, values are
propagated across lexicogrammatical relations. This may occur from the
head of a nominal group (here indicated in bold as it constitutes inscribed
attitude) to an underlined post-modifying phrase, as in example (1).
(1) The pedagogy provided opportunities for scaffolded interaction.
At clause-complex level, Lemke (1998) suggests that explicitly naming an

intertextually valued Sayer (here italicized as it invokes an attitudinal
interpretation) increases the value attributed to the projected proposition
(underlined) in a construction such as example (2).
(2) Vygotsky (1978) stresses that relations of asymmetric expertise are

critical to learning in interaction with others.
Prosodies may also propagate retrospectively. In example (3), showed is

italicized as an instance of graduation: fulfilment. It flags a positive
reading of the projected proposition at the same as it implies a positive
assessment of the Sayer, Milford (both underlined).
(3) Milford (2000) showed that the rate of success fell with age.
Values also propagate across longer phases of text. In example (4), the first
clause constitutes a hyper-Theme (underlined) in which the ideational
focus, ‘methodology’, couples with positive appreciation in ‘refinements’.
394 SUSAN HOOD
As the methodology is progressively elaborated into component features

across the phase (italicized), the value is propagated along that cohesive
line.
(4) His methodology showed certain other refinements. First, he excluded

overseas students. Such students tend to be older than average and
also to fare worse academically, thus influencing any age/perform-
ance relationship. Secondly, he used two measures of performance; the
proportion leaving without obtaining a degree and the degree results
of those taking final examinations. Finally, he weighted the degree
class obtained according to its rarity value in each faculty.
Prosody might also propagate retrospectively across a phase from an

inscription of attitude in a consolidating hyper-New. For further discus-
sion of prosody of evaluation and means for propagation, see Martin
(1992a:22, 553); Lemke (1998); Hood (2006, 2010:141–70).
Three types of prosodic patterns have been identified in texts, namely,
prosodies of ‘saturation’, ‘intensification’, and ‘domination’ (Martin and
White 2005:18–24). The authors describe saturation as an ‘opportunistic’
prosody, one ‘that manifests where it can’, as illustrated, for example, in
the stringing out of expressions of modality in the clause I suppose he might
possibly have, mightn’t he. In a prosody of intensification, a strongly amplified
value ‘reverberates through the surrounding discourse’, as might be
expected, for example, from an appraisal of something as just totally totally
amazing and fantastic. A dominating prosody instantiates attitude at points
of textual prominence, co-opting the textual functions of prediction in
higher–level Themes or consolidation in higher–level News to spread values
prospectively or retrospectively. An understanding of the potential of pros-
odic patterning for interpersonal meaning in discourse is important for
exploring the covert persuasive work that texts do.
15.5 The System of Appraisal as Engagement
The third system of appraisal is that of engagement. It maps options for

negotiating intersubjective stance in discourse. A first cut distinguishes
‘single-voiced’ from ‘multi-voiced’ text, referred to by Bakhtin (1981) as
‘monoglossia’ or ‘heteroglossia’. In monoglossic text, propositional claims
offer no space for negotiation or alternate positions. Martin and White
(2005) refer to ‘bald assertions’, but note that such ‘categorical assertions
within a framework concerned with the resources for dialogic positioning’
should not be interpreted as ‘intersubjectively neutral, objective or even
factual’ (Martin and White 2005:98–9).
Martin and Rose (2007) note key resources by which voices other than the
writer’s can be introduced into text. These include the obvious and explicit
option for mental or verbal ‘projection’ as quoting or reporting positions, as
Appraisal 395
well as options where sources are implied and processes are metaphorized,
as in examples (5) to (10):
(5) Halliday (1993) argues, ‘Science has developed a highly sophisticated

way of representing ideas that makes writing science especially
difficult for students.’
(6) Halliday (1993) argues that science has developed a highly
sophisticated . . .
(7) Halliday (1993) believes that writing science is especially difficult for
students because of the way ideas are represented.
(8) It has been suggested that science has developed a highly sophisti-
cated way of representing ideas that makes writing science especially
difficult for students.
(9) The fact that writing science is especially difficult for students is
widely appreciated.
(10) The suggestion that written science is especially difficult for
students . . .
The graphological resource of scare quotes also signals heteroglossia, as can
be seen in example (11):
(11) . . . little evidence of this in ‘learner-centred’ classrooms
The same is true for the verbalization of scare quotes, as
(12) . . . in so-called teacher-centred classrooms

Less obvious perhaps are resources of ‘modality’ and ‘negation’. The use of
either functions to negotiate with alternative positions. For example, the
claim An alternative method of exploration may be more effective is implicitly
dialogic with It may not be. The negation in The data are not appropriate to
address the questions posed is likewise dialogic with The data are . . . Resources
of counter-expectancy also signal alternative voices in discourse (Martin
and Rose 2007:56–7). These are typically realized as a concessive conjunc-
tion, as in While this reveals positive features, there are some problematic aspects,
or concessive contractors such as already, just, or actually.
In the system network of engagement in Figure 15.6, options for inter-
subjective positioning in heteroglossic text begin with the division of
‘expanding’ or ‘contracting’ space for alternative voices. Each option offers
examples of expression that indicate the kinds of resources that come into
play. In ‘expand’ as ‘entertain’, resources of modality are key (e.g. it might be
the case; this suggests). In ‘attribute’, choices of projection either acknowledge
or distance. Options within ‘contract’ close down negotiation around pos-
itions. As exemplified in Figure 15.6, ‘disclaim’ closes down negotiation
around propositions to be rejected by the writer. ‘Proclaim’ contracts space
for negotiation around a position to be supported.
396 SUSAN HOOD
Figure 15.6 The system of ENGAGEMENT (from Martin and White 2005:134)
15.6 Applying Appraisal in Researching Fields of Practice
Given the broad and rapidly expanding scope of studies drawing on the
system of appraisal, it is impossible to provide anywhere near a comprehen-
sive account of existing literature. Distinct categorizations are also a chal-
lenge. Contributions from applied studies are organized loosely around
dimensions of fields, features, and questions explored.
15.6.1 Ontogenesis of Appraisal

A place to begin is a study that could be described as ‘foundational’ in the
sense that it addresses the ontogenesis of attitude in language develop-
ment. Revisiting her own longitudinal data on first-language development
(e.g. Painter 1984), along with data from Halliday (1984) and Torr (1997),
Painter (2003) focuses on ‘the emergence of evaluative and attitudinal
language’ drawing on appraisal. She is able to show that ‘language itself
should be recognized as founded upon affectual beginnings and that the
earliest “protolanguage” phase can be construed as a system of semioticized
affect’ (Painter 2003:1). A close analysis of the ontogenesis of the language
of two children reveals ‘resources for expressing emotional, moral and
other evaluations’ and ‘interplay of implicit (evoked (or invoked)) and
explicit (inscribed) attitude in mother–child talk’ (Painter 2003:1). The
studies of the ontogenesis of language undertaken by all three scholars
(Halliday, Painter, and Torr) are of profound importance in countering
claims of ‘natural’ language learning.
15.6.2 Appraisal and Educational Discourses

From evaluative language in early childhood there is a logical connection to
studies of appraisal in broader educational contexts, across levels of
schooling and beyond. Important early contributions focused on school
Appraisal 397
subjects that might immediately associate with evaluative texts, such as

subject English and students’ interaction with literary texts and tasks, as in
Rothery and Stenglin (1997) and Macken-Horarik (2003). More recent work
includes Macken-Horarik and Isaac (2014).
Another area in the school curriculum where research drawing on
appraisal has advanced our understanding of the discourses of the subject
is in history, for example, Coffin (2002, 2006) and Matruglio (2010, 2014).
Matruglio (2010) shows how preferences for kinds of attitude (affect,
appreciation, judgement) factor out across a suite of humanities sub-
jects in a senior secondary curriculum reflecting the different targets of
appraisal in each field. The distribution indicates more or less personal or
institutionalized approaches to feelings. One subject, ‘Community and
Family Studies’, favours expressions of affect; another, ‘Society and Cul-
ture’, privileges judgement as capacity. Modern History orients to judge-
ment of human morality, and Ancient History favours appreciation of
artefacts and evidence. The findings highlight challenges for students as
they develop mastery of the resources required to manage the construction
of interpersonal stance with respect to their different subject areas. See also
De Oliveira (2010) on appraisal choices in students’ expository writing with
respect to teachers’ expectations.
The field of tertiary education and academic discourse has generated a
significant body of research that draws on appraisal. Some studies focus on
specific dimensions of appraisal and their role in academic literacies. Hood
(2010), drawing on data from diverse disciplines, reveals the ways in which
inscriptions and invocations of attitude factor out with respect to the
general field that is being appraised, that is, the field of research practices
or the field of the object of study. Lee (2010) compares appraisal choices in L1
and L2 undergraduate writers. Recent research in this educational sector
reveals a rapidly growing interest in disciplinary specific evaluative strat-
egies. See, for example, Hao and Humphrey (2012) on biology, Hao and Hood
(in press) on health science, and Szenes (2017) on business studies. If we
conceive of the apprenticeship into academic disciplines as having to do with
both knowledge and values, then learning to manage the realm of interper-
sonal meaning, including resources of appraisal, is a critical dimension of all
varieties of academic English. It underpins the development of repertoires of
personae required to manage a diversity of roles in diverse interactions.
Some studies explore evaluative strategies by bringing together modes of
interaction with specific disciplinary fields. Lander (2015) studied how
moderator strategies impacted on ‘the linguistic enactment of community’
in online asynchronous discussions in health science. Lander found a ten-
dency on the part of the moderator to invoke rather than inscribe attitude,
and to make frequent use of interpersonal grammatical metaphor and
expressions of heteroglossia. This raised ‘issues of clarity and certainty’
around the building of knowledge, ‘suggesting there may be an inherent
contradiction between community maintenance and the development of
ideational meaning’ (Lander 2015:107). Included in Coffin and Donohue
398 SUSAN HOOD
(2014) are detailed case study accounts of pedagogic practice in academic

literacies in the field of health and social care. Appraisal resources feature
both in the practices themselves and in their analysis.
15.6.3 Appraisal and Media Discourses

As noted earlier, White’s contributions to research on media discourses gave
rise to the development of the engagement system in appraisal (White
2003; Martin and White 2005). Other influential contributions derive from
his work, including studies of media ‘voices’ (White 1998, 2012; Martin and
White 2005). This use of the term ‘voice’ refers to differentiated speaker/
writer roles. The roles are construed semiotically in different profiles of
attitudinal choices. White identifies three conventionalized media roles,
those of ‘reporter voice’, ‘correspondent voice’, and ‘commentator voice’.
Reporter voice is constrained in the use of direct inscribed and unmediated
expressions of attitude. There is no authorial affect, and judgement is
encoded indirectly (invoked) or attributed to others. Correspondent voice is
one in which the inscription of judgement as social esteem is uncon-
strained, while judgement as social sanction is curtailed and if present is
typically attributed. Commentator voice has ‘no co-textual constraints on
judgement’ with ‘free occurrence of unmediated social sanction and social
esteem’. The voices correspond to different media genres, namely, ‘news’,
‘analysis’, and ‘comment/opinion’ (Martin and White 2005:165). For a
greatly elaborated description, see Martin and White (2005:173).
While we may assign institutional labels to ‘voices’ (as in reporter,
commentator, correspondent), it is important to recognize that we are
dealing with semiotic constructs. A more technical term for ‘voice’ is
‘registerial key’. This refers to a typical configuration of co-occurring
appraisal options in a particular situational setting (Martin and White
2005:164). The notion is later applied in Coffin (2002) in identifying a
system of keys in secondary school history discourses. She differentiates
‘recorder voice’, ‘interpreter voice’, and ‘adjudicator voice’. Studies of
registerial key are foundational to more recent developments in SFL theory.
These developments have to do with constructing social semiotic accounts
of notions of identity and community. The concern is about how choices
from multiple systems combine in recurring patterns in kinds of texts, and
how the meaning potential of the system of language as a whole is socially
and individually distributed. Martin refers here to the uses and users of
language (Martin et al. 2013), a focus that is taken up in more detail shortly
with references to studies of affiliation.
The field of media discourse has long attracted the attention of scholars
with an interest in concepts of tacit persuasion and of ideology, and it is not
surprising to find interest in the framework of appraisal in this context.
A useful collection of papers is found in Thompson and White’s (2008)
volume Communicating Conflict: Multilingual Case Studies of the News Media.
Appraisal 399
A study of appraisal and ideology in the French media is Caffarel and

Rechniewski (2009). See also Oteíza (2009), who draws on SFL and CDA in
analyzing patterns of voice in the Chilean National Truth and Reconcili-
ation Commission of 1991, established after the end of the military dicta-
torship (1973–1990).
Other studies have focused on specific media genres, including editorials.
Liu (2017) and Liu and Hood (in press) identify recurring rhetorical strat-
egies in the dynamic flow of couplings of attitudinal and ideational choices
in editorials from the Australian and Chinese media. Almutairi (2015)
explores technologies for visualizing the dynamic flow of evaluative
meaning-making in editorials.
Modes of media dissemination and what they afford with respect to
evaluative meaning-making have also attracted research interest. Bednarek
(2008) explores ‘emotion talk’ in TV sitcoms. Bednarek and Caple (2017)
draw on corpus data from visual and verbal modes to develop a framework
of media values in news discourse with roots in appraisal. Recent rapid
expansions of social media have generated a wave of new studies. Zappa-
vigna (2012) explores the intersubjective functioning of the hashtag in
Twitter interaction and what she calls ‘ambient affiliation’. Inako (2014,
2015) explores the significance of Twitter as a medium of communication
and community building at the time of the tsunami and subsequent
nuclear crisis in in Japan in 2011. One contribution from the study is a
reinterpretation of keigo choices in Japanese beyond conventional descrip-
tions towards their role in negotiating knowledge and values in discourse,
giving rise to an expanded description of the system of engagement.
15.6.4 Appraisal and Legal Discourse

An early study of legal discourse by Korner (2000) analyzed judgements
made in appellate courts. Drawing on engagement and graduation she
mapped variations between everyday, common-sense discourse of the ‘real’
world and the specialized, abstract discourse of the law, with reference to
degrees of heteroglossic diversity and ‘interdiscursivity’. More recently the
system of appraisal has informed a study of Youth Justice Conferencing
(Zappavigna et al. 2008; Zappavigna et al. 2010; Martin et al. 2010; 2013).
This body of work has built upon an earlier study by Knight (2010) to
contribute further to the theorization of processes of affiliation in discourse
(explained further shortly). Other studies have drawn on appraisal to
explore legal discourse in the Chinese system, including the discourse of
judges, for example, Wang and Zhang (2014).
15.6.5 Appraisal and Other Discourses

The field of literature was mentioned earlier in terms of its recontextualiza-
tion into schooling. But other work focuses on the deployment of resources
400 SUSAN HOOD
of appraisal in the study of stylistics. Prominent here is the work of Peng

(2008). See also, for example, the work of Feng and Qi (2014) on prosody and
viewer engagement in film narrative.
In a quite different context, Zhang and Guo (2014) and Zhang et al. (2018)
apply appraisal in neurolinguistic research. Zhang and Guo (2014) explore
the potential of the brain to differentiate types of emotional words.
15.7 Current Directions in Deploying Appraisal

in Studies of Discourse
15.7.1 Appraisal and Studies of Multimodal Discourse

Discourse analysts are now more frequently taking up the challenge of
reinterpreting appraisal in multimodal texts. Painter et al. (2013) undertake
a major study of multimodality in children’s picture books, which includes
an extensive section addressing visual systems of evaluation (Painter et al.
2013:17–52). The publication makes a number of important contributions
to enriching our understanding of the evaluative potential of images. The
authors explore facial expression and bodily posture as expressions of
affect, drawing on earlier contributions by Tian (2011) and Welch (2005).
Welch (2005) concludes that whatever the style of depiction (minimalist,
generic, or naturalistic), we can only interpret faces with certainty as
positive, negative, or neutral affect. Any further delicacy as to kinds of
affect would require an interpretation of contextual or other intermodal
associations. Painter et al. (2013) also explore the evaluative potential of
colour, noting that while this resource can serve each metafunction, from
an interpersonal perspective ‘the significance of colour lies in its emotional
effect of the viewer’ (Painter et al. 2013:35). They present the interpersonal
potential of colour in a detailed system network of ambience (Painter et al.
2013:36).
A further section in Painter et al. (2013) identifies the meaning potential
of graduation (Painter et al. 2013:44–6) in images. With respect to visual
affect, the researchers reinterpret a dimension of Kress and van Leeu-
wen’s (2006) ‘grammar of images’. There Kress and van Leeuwen argue for a
basic distinction between images in which a depicted person gazes out at
the viewer and ones in which there is no such gaze. While in agreement
that direct visual gaze establishes contact and directly addresses the viewer,
Painter et al. reject the claim that this gaze constitutes a ‘demand’ of the
viewer. They argue that ‘facial and bodily postures function primarily to
signify the affect of an actual or depicted person and only in the case of
certain ritualized gestures (e.g. beckoning or raising a hand for halt) place
the viewer in a specific behavioural role’ (Painter et al. 2013:19). They
propose facial and bodily postures ‘as realising meanings equivalent to
the attitudinal resources of verbal language rather than realising the nego-
tiation of dialogic exchange’ (Painter et al. 2013:19), and build a system
Appraisal 401
network of focalization to capture these, as explained and exemplified in

Painter et al. (2013:18–30).
Chen (2010, 2017) applies systems of engagement and graduation to
explore the interactive potential of multimodal features in EFL textbooks
used in schools in China. She identifies features including illustrations and
their labels, dialogue balloons, partially completed texts, and highlighting,
which can all function to establish space for different kinds of intersubjec-
tive negotiation with users.
Economou (2012) explores the evaluative potential of standout features of
large headlines and images in verbal-visual media texts. She argues for the
potential for visual ideational meanings to invoke an attitudinal reading
more powerfully than a verbal description. The function of the images is
likened to that of lexical metaphor in verbal text, and in those terms they
are considered a resource for provoking an attitudinal interpretation
(Martin and White 2005:253). For a discussion of graduation in media
images, see also Economou (2009).
Another body of research on evaluative meaning in multimodal discourse
focuses on body language in face-to-face interaction. These studies build on
seminal work in the social semiotics of body language by Martinec (2001)
and Cléirigh (2011). Hood (2011) analyzes interpersonal functions of body
language in face-to-face teaching. She finds that certain properties of
instantiated gestures invoke interpretations of graduation, and when
coupled with verbal expressions of evaluation, they function to amplify
invoked attitude. Others function to invoke choices of heteroglossic con-
traction or expansion in the system of engagement. This is extended in a
study of the multimodal expression of values in science lectures in Hao and
Hood (in press).
The social semiotic interpretation of interpersonal meaning in body
language informed by Cléirigh (2011) contributed to the study of Youth
Justice Conferencing in Martin et al. (2013) and Zappavigna et al. (2010).
This resulted in the depiction of options in the multimodal construal of
interpersonal personae.
15.7.2 Appraisal in Other Languages

Applications of the system of appraisal in researching evaluation in dis-
course have extended over recent years to studies in a number of languages
other than English. Research in Japanese includes the aforementioned work
of Inako (2014, 2015). Sano (2012) adapts the appraisal system in English to
Japanese on the basis of attitudinal lexis collected from a Japanese diction-
ary. Ngo (2014) and Ngo and Unsworth (2015) explore differences in the
deployment of appraisal resources of attitude and graduation by Viet-
namese students in Australia when participating in Vietnamese and Eng-
lish conversation. Distinctive realizations of attitude in Vietnamese also
suggest a need for adaptation of the appraisal framework. In Spanish,
402 SUSAN HOOD
Moyano (2014) explores features of genre and discourse semantics includ-

ing appraisal choices in written microbiology texts, and Oteíza (2009) and
Oteíza and Pinuer (2016) explore history discourses from a Critical Dis-
course Analytic approach. Vian (2012) discusses the appraisal system in
Brazilian Portuguese. Further studies drawing on appraisal in discourses
of languages other than English are anticipated. In China there is a rapidly
expanding body of work. Peng Xuanwei and colleagues have developed a
Chinese-English Parallel corpus of appraisal meanings as well as software
for visualizing these resources in text. See also Yu et al. (2017).
In the field of language typology, an important volume is that of Munday
(2012). Munday’s comprehensive account of the appraisal system in English
provides a foundation for identifying the many challenges that evaluative
language presents for translators.
15.7.3 Studies of Identity and Affiliation

In the introduction to Martin and White (2005), the authors conceptualize
appraisal as
the subjective presence of writers/speakers in texts as they adopt stances

towards both the material they present and those with whom they commu-
nicate. It is concerned with how writers/speakers approve and disapprove,
enthuse and abhor, applaud and criticise, and with how they position their
readers/listeners to do likewise. It is concerned with the construction by texts
of communities of shared feelings and values, and with the linguistic mech-
anisms for the sharing of emotions, tastes and normative assessments. It is
concerned with how writers/speakers construe for themselves particular
authorial identities or personae, with how they align or disalign themselves
with actual or potential respondents, and with how they construct for their
texts an intended or ideal audience.
(Martin and White 2005:1)
As identified in this chapter, for well over a decade SFL discourse scholars
have applied the tools of appraisal with more or less attention to particular
dimensions of the system, to who is doing the appraising, to the audiences
with whom they are negotiating values, and to the influence of genre, field,
and mode on the potentials for evaluation. Studies have addressed them-
selves to diverse questions in diverse fields, with diverse kinds of data,
including different semiotic modes, and in a growing number of languages.
Analytical perspectives have varied from static to dynamic, and multiple
methods of analysis have been deployed. I conclude the chapter with some
insights into one front of knowledge that is of current interest to those
working to extend SFL theorizations of interpersonal meaning in discourse
semantics. This refers to the modelling of the dynamic complexity of social
interactions and relations as the semiotics of identity and affiliation.
Here, a number of additional dimensions of SFL theory need some intro-
duction. First is the concept of ‘instantiation’ that models how the meaning
Appraisal 403
potential of the system of language is activated in a text (see Webster, this

volume). Any instance of text brings into relationship choices from differ-
ent language systems (potentially in different metafunctions, ranks, or
strata) in a process referred to by Martin (2000) as ‘coupling’. Zappavigna
et al. (2008:169) explain couplings as co-selections of meaning choices.
These might be explored at different levels of generality, ‘as patterns across
texts, text-types and registers, depending on the scope of linguistic data
available’. They can also apply across different semiotic systems, as in the
verbiage and image relations explored in Painter et al. (2013). Relating
meanings across modalities as co-instantiated couplings avoids the problem
that arises when systems of conjunctive relations in language are co-opted
to do this work. Martin (2011) attributes technical descriptors to couplings
that relate metafunctionally across modalities, referring to ‘ideational con-
currence’, ‘interpersonal resonance’, and ‘textual synchrony’ (Martin
2011:255).
The concept of couplings, in particular couplings of ideation and attitude,
underpins Knight’s (2010) important study of affiliation as the formation
and management of communities of shared values. Knight (2010) shows
how recurring instances of kinds of [attitude + ideation] ‘couplings’
constitute the basis for negotiating solidarity and community. To the extent
that couplings are shared between interactants, they constitute ‘bonds’ of
affiliation. Complexes of shared bonds form bond networks, strengthening
the communities of affiliation so engendered. ‘Bond’ as used here is ‘a
technical term (rather than in the general sense often used of ‘social
bonding’) to refer to the social relation generated as we negotiate a particu-
lar shared coupling of ideation and evaluation in language’ (Martin et al.
2013:470). Knight (2010) differentiates among three affiliation strategies for
negotiating bond: ‘communing’, ‘condemning’, and ‘deferring’, with the
latter discussed as ‘laughing off’.
In the context of spoken interaction, for example, in Knight’s conversa-
tional discourse and in the Youth Justice Conferences studied by Zappa-
vigna, Martin, and colleagues (e.g. Zappavigna 2010; Martin et al. 2013),
‘proposing a bond involves a process of discursively sharing a coupling
during an interaction’ (Martin et al. 2013:470). Where interactants are
visible to each other, bonds can also be intermodally negotiated in body
language. Others have since begun to explore affiliation and community
building in recurring couplings of ideation and attitude in written texts.
As evidence of the sharing of couplings and the formation of bonds is not
visible in linguistic or embodied exchanges, researchers have referred to
proposed bonds or ‘putative bonds’, after references in Martin and White
(2005:101) to the putative readers visible where there is an ideological
‘taken-for-grantedness’ in texts. Liu and Hood (in press) draw on this con-
cept in a study of rhetorical strategies in media editorials.
Martin (2007:56) notes that the ‘coupling of knowledge and value is an
important dimension for any field: and negotiating bonds of affiliation is
404 SUSAN HOOD
highly relevant to notions of disciplinary identity’. This view is reflected in

new directions in research into disciplinary discourses that explores field as
intrinsically connected to value and vice versa, as in Hao and Humphrey
(2012) on biology, Hao and Hood (in press) on health science, and Szenes
(2017) on business studies.
The theorization of evaluation in the discourse semantic system of
appraisal has fuelled an extensive body of research across very diverse fields
of social practice. Such studies have in turn pushed the further develop-
ment of theory. This is especially evident in recent efforts to better model
the complexity and dynamism of the enactment of personae, and the
negotiation of alignments and affiliation in language and in multimodal
texts. Martin (2011:254) suggests that what is needed for this project are
‘animated visualisation tools’. While some progress has been made on this
front (see Almutairi 2015), this remains an important and ongoing chal-
lenge for appraisal-based research.
References
Almutairi, B. 2015. Visualizing Evaluative Language in Relation to Constructing

Identity in English Editorials and Op-Eds. PhD Thesis, University of Sydney.
Bakhtin, M. M. 1981. The Dialogic Imagination. Translated by C. Emerson and
M. Holquist. Austin: University of Texas Press.
Bednarek, M. 2008. Emotion Talk Across Corpora. London: Palgrave Macmillan.
Bednarek, M. and H. Caple. 2017. The Discourse of News Values: How News
Organizations Create Newsworthiness. New York: Oxford University Press.
Caffarel, A. and E. Rechniewski. 2009. A Systemic Functional Approach to
Analysing and Interpreting Ideology: An Illustration from French Editor-
ials. Revista Alicantina de Estudios Ingleses 22: 27–43.
Chang, C. 2017. Defining English Idioms in a Bilingual Learner’s Dictionary:
Applications of Systemic Functional Linguistics in Lexicography. In J. J.
Webster and X. Peng, eds., Applying Systemic Functional Linguistics: The State
of the Art in China Today. London: Bloomsbury. 321–36.
Chen, Y. 2010. The Semiotic Construal of Attitudinal Curriculum Goals:
Evidence from EFL Textbooks in China. Linguistics and Education 21: 60–74.
Chen, Y. 2017. Comparing the Semiotic Construction of Attitudinal Mean-
ing in the Multimodal Manuscript, Original Published and Adapted
Versions of ‘Alice’s Adventures in Wonderland’. Semiotica 215(2):
341–64.
Cléirigh, C. 2011. Gestural and Postural Semiosis. Unpublished manuscript.
Coffin, C. 2002. The Voices of History: Theorising the Interpersonal Seman-
tics of Historical Discourses. Text 22(4): 503–28.
Coffin, C. 2006. Historical Discourse. London: Continuum.
Coffin, C. and J. Donohue. 2014. A Language as Social Semiotic-based Approach
to Teaching and Learning in Higher Education. Chichester: Wiley and Sons.
Appraisal 405
De Oliveira, L. C. 2010. Knowing and Writing School History: The Language of

Students’ Expository Writing and Teachers’ Expectations. Charlotte: Informa-
tion Age Publishing.
Economou, D. 2009. Photos in the News: Appraisal of Visual Semiosis and Verbal-
visual Intersemiosis. PhD Thesis, University of Sydney.
Economou, D. 2012. Standing Out on Critical Issues: Evaluation in Large
Verbal-visual Displays in Australian Broadsheets. In W. Bowcher, ed.,
Multimodal Texts from around the World. London: Continuum. 246–69.
Eggins, S. and D. Slade. 1997. Analysing Casual Conversation. London: Cassell.
Feez, S., R. Iedema, and P. White. 1994. Media Literacy. Sydney: NSW Adult
Migrant Education Services.
Feng, D. and Y. Qi. 2014. Emotion Prosody and Viewer Engagement in
Film Narrative: A Social Semiotic Approach. Narrative Inquiry 24(2):
347–67.
Fuller, G. 1998. Cultivating Science: Negotiating Discourse in the Popular
Texts of Stephen Jay Gould. In J. R. Martin and R. Veel, eds., Reading
Science: Critical and Functional Perspectives on Discourses of Science. London:
Routledge. 35–62.
Dialogue. In R. Fawcett, S. M. Lamb, and A. Makkai, eds., The Semiotics of
Culture and Language, Volume 1. London: Pinter. 3–35.
Arnold.
Edward Arnold.
Halliday, M. A. K. and W. Greaves. 2008. Intonation in the Grammar of English.
Functional Grammar. 3rd ed. London: Hodder Arnold.
Hao, J. and S. Hood. in press. Valuing Science: The Cooperation of Language
and Body Language in the Cultivation of Scientific Values in Undergraduate
Lectures. Journal of Pragmatics. https://doi.org/10.1016/j.pragma.2017.12.001.
Hao, J. and S. Humphrey. 2012. The Role of ‘Coupling’ in Biological Experi-
mental Reports. Linguistics and the Human Sciences 5(2): 169–94.
Hood, S. 2006. The Persuasive Power of Prosodies: Radiating Values in
Academic Writing. Journal of English for Academic Purposes 5: 34–79.
Hood, S. 2010. Appraising Research: Evaluation in Academic Writing. London:
Palgrave Macmillan.
Hood, S. 2011. Body Language in Face-to-face Teaching: A Focus on Text-
ual and Interpersonal Meaning. In S. Dreyfus, S. Hood, and M. Stenglin,
eds., Semiotic Margins: Meaning in Multimodalities. London: Continuum.
31–52.
406 SUSAN HOOD
Hood, S. and G. Forey. 2008. The Interpersonal Dynamics of Call-centre

Interactions: Co-constructing the Rise and Fall of Emotion. Discourse and
Communication 2(4): 389–409.
Hood, S. and J. R. Martin. 2007. Invoking Attitude: The Play of Graduation
in Appraising Discourse. In J. J. Webster, C. M. I. M. Matthiessen, and
R. Hasan, eds., Continuing Discourse on Language, Volume 2. Sheffield: Equi-
nox. 739–64.
Inako, A. 2014. The First Month on Plutonium: Physicist’s and Free Journal-
ist’s Reactions to the Fukushima NPP Accident on Twitter. International
Journal of the Oita Text Forum 2: 8–24.
Inako, A. 2015. Affiliating in Crisis: A Linguistic Perspective on Community For-
mation on Twitter after the Nuclear Accident in Japan in 2011. PhD Thesis,
University of Technology, Sydney.
Knight, N. 2010. Wrinkling Complexity: Concepts of Identity and Affiliation
in Humour. In M. Bednarek and J. R. Martin, eds., New Discourse on
Language: Functional Perspectives on Multimodality, Identity, and Affiliation.
35–48. London: Continuum.
Korner, H. 2000. Negotiating Authority: The Logogenesis of Dialogue in Common
Law Judgments. PhD Thesis, University of Sydney.
Kress, G. and T. van Leeuwen. 2006. Reading Images: The Grammar of Visual
Design. London: Routledge.
Lander, J. 2015. Building Community in Online Discussion: A Case Study of
Moderator Strategies. Linguistics and Education 29: 107–20.
Lee S. H. 2010. Differences in the Use of Appraisal Resources between L1
and L2 Writers: Focusing on the Graduation System. Journal of Issues in
Intercultural Communication 3(1): 21–47.
Lemke, J. L. 1998. Resources for Attitudinal Meaning: Evaluative Orienta-
tions in Text Semantics. Functions of Language 5(1): 33–56.
Liu, F. 2017. Strategies for Affiliation in Media Editorials: Persuading and Aligning
Readers. PhD Thesis, University of Technology, Sydney.
Liu, F. and S. Hood. in press. Repositioning Readers in Newspaper Editorials:
The Role of Ir/realis Attitude in Enacting Affiliation. Text & Talk.
Macken-Horarik, M. 2003. Appraisal and the Special Instructiveness of
Narrative. Text 23: 285–312.
Macken-Horarik, M. and A. Isaac. 2014. Appraising Appraisal. In G. Thomp-
son and L. Alba-Juez, eds., Evaluation in Context. Amsterdam: John Benja-
mins. 67–92.
Martin, J. R. 1992a. English Text: System and Structure. Amsterdam: John
Benjamins.
Martin, J. R. 1992b. Macroproposals: Meaning by Degree. In W. A. Mann and
S. A. Thompson, eds., Discourse Description: Diverse Analyses of a Fund-raising
Text. Amsterdam: John Benjamins. 359–95.
S. Hunston and G. Thompson, eds., Evaluation in Text: Authorial Stance
and the Construction of Discourse. Oxford: Oxford University Press. 142–75.
Appraisal 407
Martin, J. R. 2007. Construing Knowledge: A Functional Linguistic Perspec-

tive. In F. Christie and J. R. Martin, eds., Language, Knowledge and Pedagogy:
Functional Linguistic and Sociological Perspectives. London: Continuum. 34–64.
Martin, J. R. 2011. Multimodal Semiotics: Theoretical Challenges. In
S. Dreyfus, S. Hood, and M. Stenglin, eds., Semiotic Margins: Meaning in
Multimodalities. London: Continuum. 125–43.
Martin, J. R. 2017. The Discourse Semantics of Attitudinal Relations: Con-
tinuing the Study of Lexis. Russian Journal of Linguistics 21(1): 22–47.
Martin, J. R. and G. Plum. 1997. Construing Experience: Some Story Genres.
Journal of Narrative and Life History 7: 1–4.
Martin, J. R., M. Zappavigna, and P. Dwyer. 2010. Negotiating Evaluation:
Story Structure and Appraisal in Youth Justice Conferencing. In A. Mah-
boob and N. Knight, eds., Appliable Linguistics: Texts, Contexts, and Meanings.
Martin, J. R., M. Zappavigna, P. Dwyer, and C. Cléirigh. 2013. Users in Uses
of Language: Embodied Identity in Youth Justice Conferencing. Text and
Talk 33(4–5): 467–96.
Martinec, R. 2001. Interpersonal Resources in Action. Semiotica 135
(1/4):117–45.
Matruglio, E. 2010. Evaluative Stance in the Humanities: Expectations and
Performances. In A. Mahboob and N. Knight, eds., Appliable Linguistics:
Texts, Contexts, and Meanings. London: Continuum. 168–84.
Matruglio, E. 2014. Humanities’ Humanity: Construing the Social in HSC Modern
and Ancient History, Society and Culture, and Community and Family Studies.
PhD Thesis, University of Technology, Sydney.
Moyano, E. I. 2014. La Discusión en artículos de microbiología: género,
compromiso y construcción del conocimiento. Onomázein Número Espe-
cial 9 ALSFAL: 161–85.
Munday, J. 2012. Evaluation in Translation: Critical Points of Translator Decision-
making. Abingdon: Routledge.
Ngo, T. 2014. A Comparison of Linguistic Resources for Evaluation in English and
Vietnamese Conversations: An Exploratory Study. PhD Thesis, University of
New England, Australia.
Ngo, T. and L. Unsworth. 2015. Reworking the Appraisal Framework in ESL
Research: Refining Attitude Resources. Functional Linguistics 2(1). DOI
10.1186/s40554-015-0013-x
Oteíza, T. 2009. Evaluative Patterns in the Official Discourse of Human
Rights in Chile: Giving Value to the Past and Building Historical Memor-
ies in Society. DELTA 25: 609–40.
408 SUSAN HOOD
Oteíza, T. and C. Pinuer. 2016. Appraisal Framework and Critical Discourse

Studies: A Joint Approach to the Study of Historical Memories from an
Intermodal Perspective. International Journal of Language Studies 10(2): 5–32.
Painter, C. 1984. Into the Mother Tongue. London: Pinter.
Painter, C. 2003. Developing Attitude: An Ontogenetic Perspective on
Appraisal. Text 23(2): 183–209.
Painter, C., J. R. Martin, and L. Unsworth. 2013. Reading Visual Narratives:
Image Analysis of Children’s Picture Books. Sheffield: Equinox.
Palmer, F. R., ed. 1970. Prosodic Analysis. London: Oxford University Press.
Peng, X. 2008. Evaluative Meanings in Literary Texts: The First Step towards
Appraisal Stylistics. In N. Norgaard, ed., Systemic Functional Linguistics in
Use: Odense Working Papers in Language and Communication, Volume 29.
Odense: University of Southern Denmark. 665–84.
Rothery, J. 1984. The Development of Genres: Primary to Junior Secondary
School. In F. Christie, ed., Children Writing: Study Guide. Geelong: Deakin
Rothery, J. and M. Stenglin. 1997. Entertaining and Instructing: Exploring
Experience Through Story. In F. Christie and J. R. Martin, eds., Genre and
Institutions: Social Processes in the Workplace and School. London: Cassell.
231–63.
Rothery, J. and M. Stenglin. 2000. Interpreting Literature: The Role of
APPRAISAL. In L. Unsworth, ed., Researching Language in Schools and Com-
munities: Functional Linguistic Perspectives. London: Cassell. 222–44.
Sano, M. 2012. Apureizaru riron o kitee to shita hyooka hyoogen no bunrui
to jisho no koochiku [The Classification of Japanese Evaluative Expres-
sions and the Construction of a Dictionary of Attitudinal Lexis: An Inter-
pretation from an Appraisal Perspective]. Kokuritsu Kokugo Kenkyuujo
Ronshuu [NINJAL Research Papers] 3: 53–83.
Szenes, E. 2017. The Linguistic Construction of Business Reasoning: Towards a
Language-based Model of Decision-making in Undergraduate Business. PhD
Thesis, University of Sydney.
Thomson, E. A. and P. R. R. White, eds., 2008. Communicating Conflict:
Multilingual Case Studies of the News Media. London: Continuum Inter-
national Publishing Group.
Tian, P. 2011. Multimodal Evaluation: Sense and Sensibility in Anthony Browne’s
Picture Books. PhD Thesis, University of Sydney.
Torr, J. 1997. From Child Tongue to Mother Tongue: A Case Study of Language
Development in the First Two and a Half Years. Monographs in Systemic
Linguistics 9. Nottingham: University of Nottingham.
Vian, O. Jr. 2012. Avaliatividade, Engajamento e Valoracão (Appraisal,
Engagment and Valuation). D.E.L.T.A. 28(1): 105–28.
Wang, Z. and Q. Zhang. 2014. How Disputes are Reconciled in a Chinese
Courtroom Setting: From an Appraisal Perspective. Semiotica 201: 281–98.
Welch, A. 2005. The Illustration of Facial Affect in Children’s Literature.
Unpublished BA Hons coursework paper, University of Sydney.
Appraisal 409
White, P. R. R. 1998. Telling Media Tales: The News Story as Rhetoric. PhD
Thesis, University of Sydney.
White, P. R. R. 2003. Beyond Modality and Hedging: A Dialogic View of the
Language of Intersubjective Stance. Text 23(2): 259–84.
White, P. R. R. 2012. Exploring the Axiological Workings of ‘Reporter Voice’
News Stories: Attribution and Attitudinal Positioning. Discourse, Context
and Media 1: 57–67.
Yu, L., X. Peng, Z. He, Y. Liu, R. Zhang, X. Tan, and Y. Wang. 2017. Generic
Distributions of English Appraisal Categories Based on Appraisal Corpus.
In J. J. Webster and X. Peng, eds., Applying Systemic Functional Linguistics:
The State of the Art in China Today. London: Bloomsbury. 169–83.
Zappavigna, M. 2012. Discourse of Twitter and Social Media: How to Use Language
to Create Affiliation on the Web. London: Continuum.
Zappavigna, M., D. Cléirigh, P. Dwyer, and J. R. Martin 2010. Visualising
Appraisal Prosody. In A. Mahboob and N. Knight, eds., Appliable Linguistics:
Texts, Contexts, and Meanings. London: Continuum. 150–67.
Zappavigna, M., P. Dwyer, and J. R. Martin. 2008. Syndromes of Meaning:
Exploring Patterned Coupling in a NSW Youth Justice Conference. In
A. Mahboob and N. Knight, eds., Questioning Linguistics. Newcastle upon
Tyne: Cambridge Scholars Publishing. 164–85.
Zhang, M. and T. Guo. 2014. Event-related Brain Potentials Differentiate
Three Types of Emotional Words Categorized from Linguistic Perspective.
Journal of Neurolinguistics 31: 17–27.
Zhang, M., Y. Ge, C. Kang, T. Guo, and D. Peng. 2018. ERP Evidence for
the Contribution of Meaning Complexity Underlying Emotional Word
Processing. Journal of Neurolinguistics 45: 110–18.
16
SFL and Diachronic
Studies
David Banks
16.1 Diachronic Linguistics
Diachronic linguistics is the study of language in terms of its development

over time. According to Crystal (1980, 1999), ‘historical linguistics’ is an
alternative term. Diachronic linguistics is usually opposed to synchronic
linguistics, which is the study of a language at a particular point in time,
without reference to its historical development. Some use the terms ‘his-
torical’ and ‘descriptive’ rather than ‘diachronic’ and ‘synchronic’ (see, for
example, Bynon 1977). This distinction between synchronic and diachronic
linguistics has become a standard ploy in the study of language. When
presented like this, as it usually is, the diachronic/synchronic distinction
sounds like a dichotomy. The true situation, however, is more complex and
probably more like a cline. Lyons points out:
It is important to realize that synchronic description is not restricted in

principle to the analysis of modern spoken language. One can carry out a
synchronic analysis of ‘dead’ languages provided there is sufficient evidence
preserved in the written records that have come down to us.
(Lyons 1968:46)
What Lyons says about dead languages is true of former stages of still living
languages too. However, it might be felt that such studies would involve at
least implicit contrast with the present-day language for a contemporary
reader. For example, the modern Anglophone reader of a purely synchronic
account of, say, Middle English, would inevitably contrast it with his own
language which has developed from Middle English. Of course, such con-
trasts can be made explicit, which would draw the study in the direction of
I would like to thank Geoff Thompson for his insightful and helpful remarks and suggestions on earlier drafts of this
chapter, which he made only shortly before his untimely death in November 2015. It goes without saying that I am
solely responsible for any shortcomings that may remain.
SFL and Diachronic Studies 411
diachrony. Thus, there seem to be at least four points on the scale from
historical synchronic to diachronic: synchronic study of a now dead lan-
guage; synchronic study of the former state of a still living language, with
implicit contrast with its current state; synchronic study of the former state
of a still living language, with explicit contrast with its current state; and
study of the development of a language over time, which is diachronic
study properly so-called; and of course, it is easy to imagine finer gradations
between these four points.
Of course, no one would deny that natural languages are in a constant
state of evolution; they only stop evolving when they are no longer used as a
means of communication, and are thus dead. So, in a sense, since languages
are in a constant state of evolution, a language in its synchronic state is
something which does not exist in the real world. Treating a language
synchronically is a tactic, like taking a photograph or freeze-framing some-
thing in motion, to make it easier to study. This is quite reasonable, but it
must be accepted that a language in its synchronic state is a human
construct, not something which exists in that state. It is the tactic that
has been adopted by Systemic Functional Linguists in general, and, indeed,
by those working within many other frameworks as well, to the extent that
historical, or diachronic, linguistics is often treated, and felt, as a quite
different branch of linguistic studies. As a result, diachronic studies in a
Systemic Functional Linguistics (SFL) framework are relatively rare, and SFL
excursions into the diachronic field are fairly sporadic.
What I shall aim to do in this chapter is to outline these SFL incursions
into the diachronic field. The most important of these, which can be seen as
diachronic studies properly speaking, occur in the work of Michael Halliday
(notably Halliday 1988, 1993), where he studies a number of scientific texts
over a wide period of time, showing such features as the development of
grammatical metaphor, and how this is used in thematic structure to build
up an argument. Michael Cummings (1995, 2010) has used the SFL frame-
work as a basis for his description of Old English (OE). Although dealing
with a single period, in so far as he contrasts OE with present-day English
this might be considered marginally diachronic, or historical, with explicit
contrast to contemporary English. My own work comprises studies which
are both fully diachronic and historical with implicit contrast with present-
day language. In the former category (mainly Banks 2008, 2017), I studied a
corpus of articles from the Philosophical Transactions covering the period
1700–1980. In the latter category, I have more recently considered the
seminal period of 1665–1700, contrasting the first two academic periodic-
als, the Journal des Sçavans in French, and the Philosophical Transactions in
English. To this body of research can be added the work of Martínez-Insua
(2013) on there-constructions, Starc (2010) on advertisements, O’Halloran
(2005) on mathematics, and Urbach (2013) on register variation.
Although diachronic studies using the SFL framework are not numerous,
there is quite a lot of work which, while not strictly speaking SFL, is
412 DAVID BAN KS
nevertheless compatible with an SFL approach. This would include work

such as that of Hopper and Traugott (2003) on grammaticalization, Salager-
Meyer (1999) and Salager-Meyer and Zambrano (2001) on hedging in scien-
tific writing, Valle (1999) on sociolinguistics in the life sciences, and Biber
and Gray (2010, 2011, 2013) on nouns in academic writing. The discussion
will be rounded off with consideration of a recent publication by Halliday
and Webster (2014) in which they argue that English is in the process of
moving from a transitive to an ergative encoding of reality.
16.2 Describing Old English
The work of Michael Cummings on Old English (OE) (1995, 2010) is basically
a description of the language of the OE period. It falls into the class which is
a synchronic study, with explicit contrast to the modern state of the
language. To that extent, this work does not perhaps strictly count as
diachronic linguistics; however, it can be included in this survey, at least
marginally, for two reasons. First, the OE period itself extends over almost
five centuries, from the end of the seventh to the early twelfth century, so it
can be assumed that a description of OE over this period must necessarily
include a certain degree of change over the period, though admittedly this
is not something that Cummings himself explicitly goes into. Secondly, and
more importantly, in addition to describing OE, Cummings sets out to show
the differences between that variety and present-day English: ‘The systemic
approach also highlights areas in which significant changes have taken
place in the transition from Old to modern English’ (Cummings 2010:2).
One of the main areas where present-day English differs from OE is in the
interpersonal metafunction. In OE, the items which function as modal or
perfective operators still retain their original lexical content. This means
that in combination with another lexical verb, they are in the process of
transition between a lexical verb combined with an infinitive or participle
and an operator proper with a lexical predicator. In the case of the peri-
phrastic perfect with the verb habban (to possess), the lexical meaning is
already diminished in OE, as in the example from Beowulf in Table 16.1.
All of the verbs which would evolve into the modal auxiliaries of present-
day English could still be used as full lexical verbs, but also with highly
Table 16.1 An example from Old English (adapted from Cummings

2010:41)
siþðan him scyppen forscrifen hæfde in Caines cynne . . .
after him the creator condemned had amongst the kin of Cain
Subject Predicator Finite
Table 16.2 An example from Old English (adapted from

Cummings 2010:42)
drihten, þæt ic mage geseon
Lord, that I may see

Vocative Subject Finite Predicator
attenuated meaning in combination with other verbs to function as modal

operators, as in the example from Ælfric’s Homilies in Table 16.2.
Moreover, the periphrastic progressive did not yet exist in OE. Overall, OE
had many fewer occasions on which the finite is made explicit, and conse-
quently, Cummings concludes, the distinction between Mood element and
Residue is less sharp than it is in present-day English.
A second area where there are noticeable differences is in the textual
metafunction. For example, OE frequently uses a fused Finite and Predica-
tor in polar questions, where present-day English would use a separate
operator as Finite. This means that the lexical verb in such cases is the first
experiential element in the clause and consequently functions as Theme.
This is the case in the example from the Exeter Book in Table 16.3.
Table 16.3 An example from Old English

(adapted from Cummings 2010:86)
Gehyrest þu Eadwacer?
Hear you Eadwacer?

Finite/Predicator Subject Vocative
Theme Rheme
In both Old and present-day English, it is possible to hive off an element as

Theme at the beginning of the clause and to repeat it, usually as a pronoun,
later in the clause. In present-day English this is usually associated with oral
discourse, but in OE it could occur just as easily in high-register written
discourse, as in the example in Table 16.4 from Wulfstan’s Homilies.
Table 16.4 An example from Old English (adapted from Cummings

2010:102)
Forðan ælc þæra þe ongean þæt to ælc þæra bið Antecrist genamod
swyðe deð oððerne ongean
þæt læreð þe his
cristendome to gebyreð
For each of those who sins too each of those is Antichrist called
greatly against that or who
teaches another contrary to
what belongs to his
Christianity
Theme Rheme
414 DAVID BAN KS
Cummings also notes a number of differences within the noun group. For
example, when a noun group containing a determiner functions as the
Complement of a prepositional phrase, in OE, but not in present-day Eng-
lish, the determiner is frequently placed before the preposition; this is the
case of Him in the example in Table 16.5 from Beowulf.
Table 16.5 An example from Old English

Him On mod
his into mind
Comple… Preposition …ment
determiner preposition noun
There are many points at which Cummings comments on the similarity

between OE and present-day English. One interesting point in this connec-
tion is that in Modern English a high frequency of nominalization is
associated with official documents and scientific discourse. A parallel situ-
ation occurs in OE, where numerous nominalizations are found in sophisti-
cated theological discourse, as in the following extract from Ælfric’s
Homilies, with nominalized forms in bold; the first three are nominalized
processes, the last two, nominalized qualities:
We wyllað to trymminge cowres geleafan eow gereccan þæs hælendes

accnednysse be ðære godspellican endebyrdnysse: hu he on ðisum dæig
þerlicum dæge on soðre menniscnysse acennyd wæs: se ðe æfre buton
angynne of ðam æalmihtigan fæðer acennyd wæs on godcundnesse;
[We intend, for the strengthening of your faith, to tell you about the
Saviour’s birth according to the gospel account: how he, on this very day,
was brought forth in true humanity, he who ever without beginning was
brought forth from the almighty Father in divinity.]
16.3 The Development of Scientific English
It is probably in the area of the development of scientific English that SFL

has made the greatest contribution to diachronic linguistics (see Halliday
and Butt, this volume). It is here too that the work of Michael Halliday is of
prime importance, most notably Halliday (1988) and Halliday and Martin
(1993). Halliday (1988) brings out with force the importance of the resource
of nominalization in scientific writing, and shows how this developed over
time. Halliday starts with what is taken to be the first scientific or technical
document to have been written in English, Chaucer’s Treatise on the Astro-

labe. This was written in 1391, and only two of the originally planned five
parts were actually completed. It was ostensibly written for Chaucer’s son,
then said to be ten years old. Its object is to explain how to use an astrolabe.
Obviously, this is not to be confused with something like a research article,
and is probably closer in genre to a contemporary teenager’s ‘how-it-works’
type of book (Banks 1995). Halliday (1988:165) describes Chaucer’s text as ‘a
kind of technical, perhaps proto-scientific discourse’. The features which
lead him to this conclusion include Chaucer’s use of technical nouns,
extended nominal groups, clause complexes which carry forward the argu-
ment, and clauses expressing events under study, which are mainly Rela-
tional, and the activity of doing science, where the processes are mainly
Material and Mental.
Halliday dates the beginning of scientific English proper from Newton’s
Treatise on Opticks. This work had had a troubled history. Newton worked on
the physics of light from 1664 to 1672, and in that year his article ‘A New
Theory of Light and Colours’ was published in the Philosophical Transactions.
Halliday points out that in the Treatise on Opticks Newton is creating a
discourse of experimentation. It is thus different in genre from Chaucer’s
text: ‘In place of Chaucer’s instructions for use he has descriptions of
action – not “you do this” but “I did that”’ (Halliday 1988:166). The pro-
cesses are mainly Material for the carrying out of the experiment, and
Mental for observation and reasoning. Here, the Mental perception pro-
cesses frequently project. The Material processes are often passive, and this
is motivated by thematic considerations, where an item other than the
Actor is the centre of thematic interest. Halliday finds a difference between
the experimental and mathematical sections of the book. In describing the
experiment, Newton uses intricate clause complexes with expansion and
projection, of the type ‘I observed that, when I did a, x happened’. In the
mathematical sections, however, the clauses are simple but contain long
complex nominal groups. Halliday also finds five basic types of technical
terms: general (e.g. light, colour); field: specific (optical) (e.g. incidence,
refraction); field: general (mathematical) (e.g. proportion, excess); apparatus
(e.g. prism, lens); and methodology (e.g. experiment, trial).
One of the most important points that Halliday makes is that Newton
frequently nominalizes processes. The effect of this is that Newton is
‘packaging a complex phenomenon into a single semiotic entity’ (Halliday
1988:167); in this way its rhetorical function is made fully explicit. Since in
written text the thematic structures of Theme + Rheme are typically
mapped onto the information structures of Given + New, Theme conflates
with Given, and Rheme with New, and since an item which functions as
Given is backgrounded, and one functioning as New is foregrounded, pro-
cesses in nominalized form can be backgrounded or foregrounded by
placing them in thematic or rhematic position. As Halliday (1988:169)
416 DAVID BAN KS
points out, this was to become ‘an essential resource for constructing
scientific discourse. We see it emerging in the language of this period,
when the foundations of an effective register for codifying, transmitting
and extending the “new learning” are rapidly being laid down’. With this
device, instead of clauses dominated by a process, we can have clauses
which express relations between processes, either of an external type, ‘a
causes x to happen’, or of an internal type, ‘b causes me to think y’. Halliday
and Martin (1993) give the following examples from Newton:
(1) The explosion of gunpowder arises therefore from the violent action
whereby . . .
(Halliday and Martin 1993:67)
(2) Now those Colours argue a diverging and separation of the heteroge-
neous Rays . . .
(Halliday and Martin 1993:65)
He points out that in example (1) there are two nominalized processes,
explosion and action, and these are linked by the verb arises, expressing the
fact that one is caused by the other. In example (2), evidence (the colours)
leads the observer to deduce (they ‘argue’) a possible explanation.
Halliday finds that the use of this type of grammatical metaphor (nom-
inalized processes) has been taken even further in Priestley’s The History and
Present State of Electricity, with Original Experiments, published in the 1760s.
The ‘nominal elements in the clause are gradually taking over the whole of
the semantic content, leaving the verb to express the relationship between
these nominalized processes’ (Halliday 1988:171). This process has gone on
developing up to the present day, so that Halliday can set up a schematic
representation of the progress of this phenomenon. This has both external
and internal forms. Externally the schema is as follows:
& a happens; so x happens

& because a happens, x happens
& that a happens causes x to happen
& happening a causes happening x
& happening a is the cause of happening x
Internally it is as follows:
& a happens; so we know x happens

& because a happens, we know x happens
& that a happens proves x to happen
& happening a proves happening x
& happening a is the proof of happening x
Again taking Halliday’s (1988:175) concept into account, in the following,

the last line is an authentic example (slightly adapted) from the Scientific
American; the previous lines are reconstructions (ISOL is an acronym for ‘on-
line isotope-separation system’):
We used an ISOL, and thus could experiment even where it was difficult
By using an ISOL we solved the difficult parts of the experiment
Our using an ISOL resolved the difficulties of the experiment
The experimental difficulties were resolved by the use of an ISOL
The resolution of the experimental difficulties came in the form of
an ISOL
This does not mean that this process is complete and that the final form is
the one we will inevitably encounter in contemporary scientific texts.
Probably all the schema could be found somewhere in texts today. What
Halliday is saying is that over time there is a general trend towards the later
more metaphorical forms, so that there will be a tendency, as time goes on,
to find more of the later forms.
He takes these points up again in Halliday and Martin (1993), where he
uses them to give a detailed analysis of the final section of Darwin’s On the
Origin of Species by Means of Natural Selection, published in 1859. He shows
that grammatical metaphor is not only important at the level of clause, but
also plays a significant role in thematic progression, and hence in the flow
of information and the cohesion of the text as a whole. Items which appear
in non-metaphorical form in rhematic sections, where they are treated as
New, are taken up again in nominalized (grammatical) form as Theme and
Given in a following clause. The argument can thus be built up in this way
from Rheme to Theme. Not only does this give the text a tight cohesion, but
the nominalized process is now backgrounded as Given in thematic pos-
ition. As a noun, it is presented as something fixed and solid, whose
existence cannot be questioned, thus ‘the nominalization picks up
the preceding argument and presents it in this “objectified” form as some-
thing to be taken for granted’ (Halliday and Martin 1993:98). Hence, the
thematic progression moves from the dynamic presentation of the process
as Rheme and New to its presentation in nominal form, and thus as an
objectified item, as Theme and Given. This can be seen in example (3) from
an article in the Philosophical Transactions for the year 1900 (taken from
Banks 2008:153):
(3) I consequently tried to separate it in a pure state by means of plate

cultures, but all my attempts failed owing to a rapid liquefying of
the gelatin solution.
The process of trying is expressed congruently, as a verb, in the Rheme of

the first clause, but then appears in the nominalized form, attempts, as the
Theme of the second clause. As Theme of the second clause it is presented as
something given and established. The following (from Banks 2017:158) is a
very early example from the Philosophical Transactions of 1665 (though the
phenomenon does not appear to have been common at this early date):
(4) But having beheld it with a Telescope, I soon said, that it was joyned
with two small Stars, whereof one was pretty bright, which I had
418 DAVID BAN KS
already seen, on February 28, and 29. And this conjunction gave the
Comet that brightness, as it happens to most of the Stars of the fifth
and sixth magnitude . . .
Here, the verbal form was joyned is taken up by the nominalized form
conjunction as theme of the following clause.
Halliday’s study of the rise of grammatical metaphor, notably in the form
of nominalized processes, in scientific writing and the link between gram-
matical metaphor and thematic progression may well be the most signifi-
cant contribution that SFL has made to date in the area of diachronic
linguistics.
16.4 The Scientific Research Article
In my own work on the development of scientific writing (Banks 2008,

2017), I have concentrated on the scientific research article; I would like
to think that this in general corroborates Halliday’s work, while filling in
some of the detail that Halliday’s broad brush leaves implicit. In the
following, I outline the kinds of analysis that I conducted as an illustration
of how SFL-inspired diachronic studies can be carried out.
Banks (2008) uses a corpus of articles selected from the Philosophical
Transactions. This journal is particularly useful for the study of the scientific
article since it was founded in 1665 by Henry Oldenburg as a bulletin of
scientific news. Since then it has been published continuously, with only
very minor interruptions, and is still in existence today, although its nature
has modified to some extent over the period of three and a half centuries
(Atkinson 1999; Valle 1999; Banks 2008). The corpus I used consists of
thirty articles, two for the year 1700, then two per year at twenty-year
intervals (i.e. 1720, 1740, etc.) up to 1980. For each individual year, one
article is from the physical sciences and one from the biological sciences.
The total number of words is a little over 126,000.
The general picture that emerges is that the physical sciences are experi-
mental from the beginning of this period, and they remain so until the late
nineteenth century, when mathematical modelling becomes a major inter-
est, rivalling experimentation. At the beginning of the period the biological
sciences are basically observational, and it is only in the middle of the
nineteenth century that experimentation begins to have an impact on the
biological sciences; observation and experimentation remain significant
factors up to the end of the period studied.
Use of the passive form is already established in the physical sector at the
beginning of the period, accounting for something of the order of 25 per cent
of the finite verbs in the eighteenth century. Passive use increases to around
30 per cent in the nineteenth century, and this continues to be the case up to
the end of the twentieth century. The verbs which are passivized are mainly
Material processes up to the late nineteenth century, when the introduction

of mathematics leads to a change in the balance in favour of Mental pro-
cesses. In the early years of the period, the biological sciences use passives
rather less than the physical sciences, the difference being of the order of ten
percentage points. The gap narrows in the course of the nineteenth century,
and during the twentieth century the rates in the two fields are more or less
the same. In the biological sciences, passives are mainly Material, but not to
the same extent as in the physical sciences, the difference being made up by
Mental processes. The following examples (5) and (6) show passivized Mater-
ial process and passivized Mental process in early physical and biological
articles respectively:
(5) The ivory was supported horizontally by a stand made of the pre-
pared wood. When the glass was made a little warmer than the
external air, my finger rubbed that side thereof which was furthest
from, and opposite to the ivory.
(Philosophical Transactions 1780, from Banks 2008:102)
(6) This change has been observed in some of the bird tribe, but princi-
pally in the common pheasant.
Use of the first-person pronoun is never particularly common in these texts.

In the physical sciences, those that are used tend to be in association with
Material processes, until the late nineteenth century, when they virtually
disappear, but from which point they are increasingly used with Mental
processes. In the biological sciences, they are used with Mental processes
throughout the eighteenth century and until the mid-nineteenth century.
There is then an increase in use with Material processes until the mid-
twentieth century when there is a return to use with Mental processes. The
following show first-person pronouns with Material processes in a physical
article in example (7), and Mental processes in a biological article in
example (8), both being from the eighteenth century.
(7) In order to do that, I tied a stick of sealing wax to a silk string about a
yard long, and after having excited it very powerfully with flannel,
I plunged it in a tin vessel full of water, and immediately drawing it
out, brought a very accurate electrometer near it, and observed, that
at first it shewed no sign of electricity; but in about half a minute’s
time it manifested a small but very sensible degree of negative
electricity.
(8) This was a Sight I little expected to meet with; and being aware how
much Imagination has frequently had to do with microscopical
Observations, I distrusted my own eyes.
420 DAVID BAN KS
Grammatical metaphor, in the form of nominalized processes, is common

throughout the period considered. In the physical sciences this is of the
order of one per thirty words of running text until 1920, when there is a
sudden increase to about one per fourteen words of running text. The rate
in the biological sector is less stable, but settles at about one per thirty
words in the mid-nineteenth century. There is an increase in the twentieth
century, but it is less marked than that in the physical sector. The majority
of nominalized processes are Material, though the rate is higher in the
physical sector than in the biological until the twentieth century,
when the rates are virtually the same. There is a further development
which takes place in the twentieth century: this is the use of nominalized
processes in modifying position. This feature is absent until 1900, but
becomes relatively common thereafter. There are even some cases where
a complex nominal group made up of a nominalized process plus head is
then reused as a complex modifier in a further complex noun group, as in
example (9), where ionization front functions as one of the modifiers of
structure:
(9) A strong R-type ionization front structure with one shock is the one
which is most likely to occur.
This phenomenon occurs in both sectors throughout the twentieth century,
and is particularly prevalent in the physical sector, rising to 30 per cent of
nominalized processes in one particular article. Although present in the
biological sector, the frequency is much lower, the highest rate in an
individual article being 7 per cent.
The study of topical Themes gives interesting results when these are
analyzed in terms of semantic categories. Topical Themes relating to the
experimental and observational areas, that is, the object of study, the
experiment, equipment, and observation, account for 75 per cent of all
topical themes throughout the period in the biological sciences, and
throughout the eighteenth and nineteenth centuries in the physical
sciences. Example (10) is from an 1860 article in the Philosophical Transactions
(from Banks 2008:168):
(10) By the addition of a little water a small quantity of syncoretin was
then thrown down, in order to carry down the last traces of the less
soluble crystalline compound, in case any were still present.
In the physical sector this rate falls in the twentieth century and is
compensated for by the appearance and rapid increase in Themes of
a mathematical nature. This includes terms relating to mathematics,
but also phrases containing mathematical expressions and equations.
The following examples (11) to (13) are taken from articles which
appeared in the Philosophical Transactions in 1940 and 1960 (from Banks
2008:176):
(11) the minus sign is of no significance

(12) Since z < 1 for all D-types, equation (2.11) shows that v1< c1< a1, so
that all D-type ionization fronts move subsonically relative to the
fluid ahead.
(13) In the present equation, y = a + bx, it might appear at first sight that
a represents the initial size difference between the variates.
This appears to reflect the major change in focus, mentioned above, from
experimentation to mathematical modelling which takes place in the phys-
ical sciences, but not in the biological sciences, at the beginning of the
twentieth century.
In Banks (2017) I have concentrated on the much shorter time scale of
1665–1700. This was a particularly important period for the establishment
of academic writing since the first two academic periodicals both appeared
in 1665. The first was the Journal des Sçavans in Paris, and it was followed
two months later by the Philosophical Transactions in London. The Journal des
Sçavans, founded by Denis de Sallo, had state support, covered the full range
of new knowledge, and was mainly made up of book reviews (Morgan 1928).
The Philosophical Transactions was founded by Henry Oldenburg as a private
venture; it was mainly restricted to science and technology, and was based
on Oldenburg’s voluminous correspondence (Hall 2002). The cut-off point is
1700 because until that point the Académie Royale des Sciences only published
its papers in luxurious limited editions, which were considered the per-
sonal property of the French monarch, Louis XIV. He used them as gifts
for illustrious visitors, but they were not easy to come by otherwise
(Liccope 1994, 1996). At the end of the seventeenth century, the decision
was made to alter this state of affairs and to publish more widely; the
first of these volumes, that for 1699, actually appeared in print in 1702.
Thus, the period from 1665 to 1700 was peculiar, with the Journal des
Sçavans and the Philosophical Transactions as the only real outlets for aca-
demic writing in French and English. From 1700 onwards this situation
changed radically with the appearance of the Mémoires de l’Académie Royale de
sciences.
As might be expected, Subjects function as Theme in the majority of
cases; however, this is more frequently the case in the Journal des Sçavans
(73 per cent) than in the Philosophical Transactions (62 per cent). Corres-
pondingly there are more adjunct Themes in the Philosophical Transactions
(31 per cent) than in the Journal des Sçavans (22 per cent). Thematic
structure functions in virtually the same way in French and English,
and there is no evidence to suggest that differences between the two
languages might account for the differences in these figures. While the
figures for the Journal des Sçavans are fairly stable over the period, in the
Philosophical Transactions the percentage of subject Themes increases and
that of adjunct Themes falls over the years 1665 to 1695. In both journals,
422 DAVID BAN KS
roughly half of the adjunct Themes are clausal in nature. The use of
linear progression rises over time in the Journal des Sçavans (38 per cent
to 45 per cent), but it is stable in the Philosophical Transactions at around 43
per cent. In the French journal the most common semantic category of
Theme is that of humans other than the author (37 per cent), who are
frequently the author of a book under review; the object of study
accounts for a further 22 per cent, and other texts, usually a book under
review, 16 per cent. Use of humans other than the author as Theme is
stable after 1675; that of the object of study shows an increase in 1695;
and that of other texts falls over the period. In the English journal, the
object of study is by far the commonest type of Theme, accounting for 45
per cent overall and rising from 32 per cent to 57 per cent over the period.
Reference to the author as theme accounts for a further 14 per cent, and
humans other than the author, 13 per cent. Hence, it can be seen that the
thematic interest of the Journal des Sçavans is directed primarily towards
humans, and secondarily to objects of study, whereas in the Philosophical
Transactions, this is reversed with objects of study being by far the major
interest, and humans a minor secondary interest. The following examples
show humans other than the author functioning as theme in the Journal
des Sçavans, and the object of study functioning as theme in the Philosoph-
ical Transactions.
(14) L’Auteur de ce livre rejette le premier de ces sentimens, il doute de

la solidité du second, & il approuve fort le troisiéme: mais il en ajoûte
un quatriéme fondé sur l’authorité des Conciles & des Peres . . .
(Journal des Sçavans 1675; from Banks 2017:74) [The author of this
book rejects the first of these proposals; he doubts the soundness of
the second, but he approves the third. He adds a fourth based on the
authority of the Councils and the Fathers . . .]
(15) The following bodys were poured gently into the vessel, and those in
the 12 first Experiments were weighed in scales turning with 2
ounces, but the last 7 were weighed in scales turning with
one ounce.
(Philosophical Transactions 1685; from Banks 2017:89)
The finite verbs were analyzed in terms of process type, using a system of
five processes: Material, Mental, Relational, Verbal, and Existential
(Banks 2005, 2016). In both journals, Relational process accounts for
about 30 per cent of the finite verbs. However, while this is the common-
est category in the Journal des Sçavans, this is not the case in the Philosoph-
ical Transactions, where Material process accounts for 35 per cent, as in
example (16):
(16) I bought a female Rabbet and let it take Buck 3 times in my presence,
(which was quickly done) and then killed it, but did not open the
womb, till a quarter of an hour after; about an Inch from the
beginning of one of the Horns, I found a little fluid matter contain-

ing some few living animals of the Male seed.
In the Journal des Sçavans, Material process accounts for 27 per cent and
Verbal process a further 24 per cent. In the Philosophical Transactions, Verbal
process accounts for only 13 per cent and falls over the period considered.
The sentence in (17) is an example of Verbal process in the Journal des
Sçavans:
(17) Car il pretend qu’il ne l’auoit pas escrit pour estre publié. (Journal des
Sçavans 1665; from Banks 2017:104) [For he claims that he did not
write it in order to be published.]
This shows that while description is a significant feature of both journals,

actions and events are even more important in the Philosophical Transactions,
but not in the Journal des Sçavans, and while communication is relatively
important in the Journal des Sçavans, this is not particularly the case in the
Philosophical Transactions.
Modality was analyzed using the traditional categories of epistemic,
dynamic, and deontic, rather than Halliday and Matthiessen’s (2014)
modalization and modulation, but these are compatible in so far as
epistemic modality corresponds roughly to modalization, and dynamic
and deontic modality to modulation. The analysis shows that dynamic
modality is the commonest type of modality in both journals, but it is even
more prevalent in the Philosophical Transactions (76 per cent) than in the
Journal des Sçavans (58 per cent).
(18) Pour voir si les astres peuvent causer dans les hommes quelques
inclinations, il recherche la cause des differentes humeurs. (Journal
des Sçavans 1675; from Banks 2017:134) [To see whether the heav-
enly bodies can produce certain inclinations in men, he looks for the
cause of different humours.]
One significant feature is that although it is the least common type of

modality, deontic modality accounts for 18 per cent in the Journal des
Sçavans, but is virtually absent (4 per cent) in the Philosophical Transactions.
This can be attributed to the fact that the French journal contains items
from the fields of theology and law, where one might expect to find mater-
ial relating to ethical questions, but these fields are not represented in the
pages of the Philosophical Transactions.
Grammatical metaphor in the form of nominalized processes is slightly
more frequent in the Journal des Sçavans (24 per 100 words of running text)
than in the Philosophical Transactions (20 per 100 words). The majority of
nominalized processes are Material in both journals, but this rate is much
higher in the Philosophical Transactions, where it is 64 per cent, than in the
Journal des Sçavans, where it is 48 per cent.
424 DAVID BAN KS
(19) Les Cometes ne sont qu’vn amas de plusieurs petites estoilles

errantes, qui suiuant la nature des autres planetes qui ont des
mouuemens inegaux, se doiuent necessairement ioindre ensemble
de temps en temps, & se rendre visibles par cette vnion. (Journal des
Sçavans 1665; from Banks 2017:150) [Comets are only a mass of
several small wandering stars, which like other planets which have
irregular motion, must necessarily join together from time to time,
and make themselves visible by this union.]
(20) Nor have I done, as some have fancied of me, who having been able
to observe the Comet, the 27, 28, 29, 30, and 31. of December, and to
see the diminution of its motion, have judged, that I had only
determined that diminution for the time to come, conform to the
augmentation thereof in time passed until the 29. of December.
Mental and Verbal processes account for 24 per cent and 26 per cent of
nominalized processes respectively in the Journal des Sçavans, while the
corresponding figures for the Philosophical Transactions are 18 per cent and
14 per cent. There is no distinct pattern of development over the time
period considered. Once again, the greater leaning of the English journal
towards action and event, and the interest of the Journal des Sçavans in
questions of communication are brought out by these figures.
The linguistic features which are highlighted by this study can be shown
to be the direct result of the editorial decisions made by de Sallo and
Oldenburg, in de Sallo’s case to cover the whole range of disciplines and
to print mainly book reviews, and in Oldenburg’s case to concentrate on
science and technology based on his correspondence. These decisions them-
selves can be understood in the context of the differing historical situations
of France and England in the late seventeenth century. France was at that
time the economic and cultural centre of Europe, with Louis XIV on the
throne, and Colbert as his first minister. Colbert wanted to control every-
thing including new knowledge, which he found potentially dangerous.
Since new knowledge was to be found in books, it was natural that the
French periodical should concentrate on book reviews, and since control
was the object of the exercise, no disciplines could be excluded from its
scope. England on the other hand was basking in the new-found hope of the
Restoration, after decades of chaos. The impoverished crown was unable to
subsidize the Royal Society, and Oldenburg, as one of its secretaries, had to
find supplementary sources of income. Hence, he used his voluminous
correspondence to create a bulletin for his potential readership, the
members and friends of the Royal Society, who were interested in scientific
matters. Thus, it can be seen that the linguistic features found in these two
periodicals derive from the editorial decisions made by de Sallo and Old-
enburg, and these decisions themselves can be seen as being determined by
the historical context in which they were made. This does not mean, of
course, that things could not have been otherwise, but it does mean that the
linguistic features can be seen as consistent with the historical context. This
seems to be an excellent illustration of the SFL contention that there is an
intimate link between language and the context in which it is produced.
16.5 Other Contributions
There are in addition a small number of fairly disparate SFL contributions

on subjects within the field of diachronic linguistics. Without making any
claim to exhaustivity, the following can be given as typical examples.
Martínez-Insua (2013) looks at the use of there-constructions from late
Middle English on. The participant, usually Existent, in such constructions
is normally indefinite, corresponding to its information status of New, since
it occurs in the Rheme of the clause. She nevertheless finds a small but
significant proportion of definite Existents in her corpus, the smallest being
4.3 per cent for the period 1420–1500, and the largest, 13.4 per cent, for the
period 1570–1640. The majority of there-constructions have always had be as
their verb. However, other verbs, both transitive and intransitive, occur,
and while initially fairly numerous, accounting for as much as a third of the
clauses up to 1570, they have dwindled rapidly over time, and account for
less than 1 per cent in present-day English. Similarly, there have always
been a small number of cases where the verb does not agree in number with
the following Existent. Although the number of such cases has fallen in
present-day English, Martínez-Insua (2013:220) hypothesizes that ‘“there +
singular verb” is undergoing a process of grammaticalization and subjecti-
fication as a formula useful for introducing (new) forward-looking referents
into the discourse’.
Starc (2010, 2015) studies the development of Slovene advertisements.
She uses corpora dating from the late nineteenth century onwards, and
distinguishes between commercial advertisements and classified (personal)
advertisements. Initially these were virtually indistinguishable, and were
printed together. The commercial advertisements have changed consider-
ably over time. They have developed a thematic structure which is more
dynamic. Typographical variation was introduced to give salience to
selected items, and this led on to the use of multimodal forms. There is a
gradual tendency to condense the text, so that what frequently appears as
the Theme in later advertisements is what would have been the rhematic
content of earlier versions. In the case of personal advertisements, while
there has been some use of images, and a move towards shorter texts, they
have not changed anywhere near to the same extent as the commercial
advertisements.
O’Halloran’s (2003, 2005) work on mathematical discourse includes con-
sideration of this type of genre in a diachronic perspective. She distin-
guishes between three stages: first, rhetorical algebra, which incorporates
426 DAVID BAN KS
linguistic descriptions; second, syncopated algebra, where quantities and

operations are frequently symbolized; and third, symbolic algebra, where
mathematical symbolism is developed as a semiotic resource in its own
right. The first mathematical texts dealt with commercial arithmetic, such
as the anonymous Treviso Arithmetic of 1478. In the fifteenth and sixteenth
centuries, diagrams in mathematical texts frequently show human figures
involved in some sort of physical or observational activity, such as the firing
of cannons. In the course of the seventeenth and eighteenth centuries, the
body gradually disappears to be replaced by a body part, such as an eye or
hand, and ultimately to a purely symbolic representation. This can be
particularly noted in the work of Descartes. Newton built on the work of
Descartes, and used the possibilities of mathematical symbolism to achieve
a degree of clausal rankshift which would be inconceivable in a natural
language. This allows the ‘unambiguous encoding of deeply embedded
configurations’ (O’Halloran 2003:349).
The interest of Urbach (2013) is in the relationship between text and
context over time. She studies a corpus of newspaper reports from The
Sydney Morning Herald, all dealing with the end of major military conflicts,
from the end of the Boer War in 1902 to the fall of Baghdad in the Iraq War
in 2003. Over this period there have been numerous developments in
technology, notably in sound recording, photography, and printing tech-
niques. Until 1944 it was normal for the main news to appear on page
seven, rather than page one, the preceding pages being devoted to commer-
cial matters. Until the end of World War I, photographs, particularly of
events, were unusual, and even felt to be downgrading for the image of the
newspaper; today images are normal and appear on every page. Urbach
attempts to link these features to the context in terms of Field, Tenor, and
Mode. In terms of Field, she finds that there is a move over time from a
dispersed to a unified temporal perspective. This may be due to the fact that
early telegraphic news bulletins were not necessarily arranged chronologic-
ally, and so getting the time deixis right was important for the readers. By
the time we get to the Iraq War, we have something more like an eyewit-
ness account. Tenor is marked by the increasing individualization of the
writer. Until 1945 the writer remains anonymous; in the period 1950–1975
only the journalist’s location is given; but from the Gulf War of
1990 onwards the individual is named. Mode is conditioned by techno-
logical advances. Early typesetting techniques were laborious, and left little
time for rearranging and editing. Nowadays, computerized techniques
make extensive last-minute editing possible.
16.6 Compatible Approaches
In addition to the above, there are numerous contributions to the field,

which, while not specifically SFL in approach are nevertheless compatible
with it. However, it should be noted that contributions from outside of SFL
tend to concentrate on changes in form, whereas, as the discussion so far
will have made clear, SFL, being a social semiotic approach, sees the
changing form at the discourse level within its social and historical context.
Among these non-SFL contributions one might include the vast literature
devoted to ‘grammaticalization’. Traugott (1988:406) defines grammatica-
lization as ‘the dynamic unidirectional historical process whereby lexical
items in the course of time acquire a new status as grammatical, morpho-
syntactic forms’. Although developed against a generative grammar back-
ground, the evolution of grammaticalization has moved away from that
basis (Hopper and Traugott 2003), and the later findings could easily be re-
expressed in SFL terms. This emerges particularly clearly in the association
that is often noted between grammaticalization and subjectification – in
SFL terms, a move from experiential to interpersonal meaning. Present-day
modal auxiliaries are a prime example since they derive from lexical verbs
in Old English (Hopper and Traugott 2003).
The work of Salager-Meyer on hedging and related matters is basically
within a Swalesian tradition (Swales 1990, 2004), but shares with SFL an
interest in discourse. In Salager-Meyer (1999), she studies references in a
corpus of English medical articles covering the period 1810 to 1995. She
distinguishes between references which are critical and those which are
not. Critical references account for 41.6 per cent of the sample for the
period 1810–1929, but only 17.4 per cent for the period 1930–1995. This,
however, is not due to a fall in the incidence of critical references, but to a
sharp continuous increase in non-critical references from 1930 on. More-
over, although critical references are fairly stable over the period con-
sidered, before 1930 they tend to be pointed and personal, as in the
following:
Mr. Brodie objects to my experiments that they were not exact repetition of
his, and therefore not entitled to much consideration in estimating the
causes of animal heat . . . I cannot conceive, I am afraid, on what Mr. Brodie’s
opinion is founded. (1823)
(from Salager-Meyer 1999:20)
Whereas from 1930 on they are more circumspect and dispassionate, as in

the following:
We have carried out both the test of Akerfeldt and Gibbs and have been
unable to confirm the findings of either investigator. (1960)
(from Salager-Meyer 1999:22)
She sees this as a reflection of a move from private, individual author-

centred medical writing to a tight, professional object-centred style.
Salager-Meyer and Zambrano (2001) compare these results with those of a
parallel French corpus. The two languages give fairly similar results for the
period 1810–1929, but from 1930 onwards there are more critical
428 DAVID BAN KS
references (here termed direct academic conflicts) in the French corpus, and
more non-critical references (indirect academic conflicts) in the English
corpus. In the period before 1930, however, direct criticism in the English
articles is frequently accompanied by gentlemanly polite remarks; these are
absent in the French texts. In the period after 1930, where the English lose
their aggressive tone, the French texts, even if they become a little less
aggressive, remain fairly hard-hitting, until the end of the twentieth cen-
tury, when they start to resemble the English texts on this point. For those
working within an SFL framework, these processes can be seen in terms of
changes in the interpersonal metafunction; and Salager-Meyer’s emphasis
on discourse categories as subject to diachronic change is very much in line
with SFL approaches.
Valle’s (1999) monograph is also in this line of enquiry. It considers the
development of scientific discourse from a sociolinguistic point of view,
focusing on citation practices, and using a corpus of texts dealing with the
life sciences from the Philosophical Transactions for four periods: 1665–1669,
1765–1768, 1865–1869, and 1965–1966. Among her many results, she
notes that the use of short quotations increases over the period, and that
of long quotations falls – to almost zero by the twentieth century. The
majority of quotations in the eighteenth century are for creating debate,
juxtaposition, comparison, and polemic; by the nineteenth century this
category has fallen, and is equalled by those used for knowledge-
embedding, historical background, or filling gaps. By the twentieth century
these have both fallen, and have been overtaken by those used for social
reasons, gaining support or giving credit, and assigning priority. As in SFL
studies, these changes in discourse are seen as reflecting and construing
social and cultural changes across the period.
Biber and Gray in a series of articles (2010, 2011, 2013), again not SFL,
confirm and build on some of Halliday’s findings. They agree with Halli-
day’s contention that in academic writing there has been a move from a
more elaborate style depending on complex clause structures to one which
is more condensed with a relatively simple clause structure but incorpor-
ating complex nominal groups. Their evidence indicates that this is basic-
ally a twentieth-century change, with its origins in the late nineteenth
century. They note that in a corpus of news reporting texts taken
from Time and The New York Times, while direct and indirect quotes and
nominal modifiers have increased over the twentieth century in both
publications, the increase is greater in The New York Times. Similarly,
use of passives and of-genitives has decreased, but the decrease, particularly
of passives, is greater again in The New York Times. In academic writing, the
use of nouns is steady in history texts from the nineteenth century
onwards, increases in the social sciences, and increases dramatically in
the hard sciences. In making these observations, Biber and Gray warn
against the danger of underestimating register (and indeed sub-register)
differences.
16.7 Transitivity and Ergativity
I should like to round off this survey by returning to the work of Michael
Halliday. In a recent work, in collaboration with Jonathan Webster (Halliday
and Webster 2014), the authors hypothesize that over the last thousand years
English has been moving from a transitive to an ergative system. This idea is
mentioned in earlier writing by Halliday (e.g. Halliday 1985:146), but has,
I think, never been expressed with such clarity or force as here. A transitive
system sees things in terms of action and extension, whereas an ergative system
sees them in terms of cause and effect. Thus, in a clause such as Harry opened the
door, the door functions as Goal in a transitive analysis, but as Medium in an
ergative analysis: Goal is the second participant to which the process extends in
a transitive structure, and Medium is the participant which is centrally
involved in, or actualizes, the process. As Medium, door is the participant which
functions as Subject, when there is only one participant (The door opened), thus
showing that it is the participant that is essential for the process to take place.
Traditionally, English verbs are seen as forming a cline from those that are
exclusively transitive, like throw or dig, to those that are exclusively intransitive
like swim or crawl. Halliday and Webster claim that there is a gradual drift, in
modern English, towards the centre of the cline where verbs can function both
transitively and intransitively, like open or ring. What functions as Subject was,
they claim, in earlier periods, selected on the basis of transitivity: that is, the
unmarked choice for Subject was the participant which functioned as Actor in
the transitivity system. But today selection is on the basis of thematic structure:
that is, whatever component the speakers choose as their starting point can be
selected as Subject. The move to an ergative conception of clause structure
makes this easier. They thus reinterpret Modern English as having a cline of
possibilities from verbs which are always non-middle (like throw), to those
which are always middle (like swim). The fact that some verbs function as both
active and passive, such as derive (x derives from y or x is derived from y) is seen as a
natural consequence of the move towards an ergative system. It would seem
that from this point of view English is more ‘mixed’ than many other lan-
guages, and they speculate that a mixture of this kind is unstable. At all events,
this situation, they claim, ‘is the result of extensive change over the last
millennium, the change being away from the transitive type towards the
ergative’ (Halliday and Webster 2014:40), and ‘[t]he ergative interpretation
seems to reflect the direction in which the language is changing’ (Halliday
and Webster 2014:42). It is even possible that we are returning to a state which
obtained in the Indo-European in which English ultimately has its roots.
It is evident from this short survey, that, up to the present moment,

diachronic studies have not been of major interest for those working within
430 DAVID BAN KS
the SFL framework. The main contribution that SFL has made to diachronic
studies is the work in the area of the development of scientific writing. This
is basically the work of Michael Halliday, to which I would like to think my
own work provides a suitable supplement. Systemic Functional Linguistics
contributions to the diachronic study of other text types, while perhaps
significant in their own right, remain fairly piecemeal, and cannot be said
to add up to a sizeable body of work. Nevertheless, these contributions,
which could form the basis of future studies, include the work of Starc on
advertisements, O’Halloran on multimodality in mathematics, and Urbach
on context in journalism. In the area of general English, Cummings has
provided an extensive study of Old English, and to this might be added the
work of Martínez-Insua on there-constructions.
Even though the sum of these contributions is not large, they are
sufficient to show that SFL is capable of providing interesting and signifi-
cant insights into the field of diachronic studies. As has been shown,
there are a number of studies which, although outside of SFL proper, are
nevertheless compatible with it, and which, indeed, could have been carried
out in an SFL framework. Systemic Functional Linguistics, however, does
bring a specific point of view to bear on the questions it treats, and it can
only be hoped that in the years to come, linguists working within the SFL
tradition will turn their attention to questions of language development, so
that the particular insights which SFL is capable of offering will enhance
the field of diachronic studies, as they have so many other fields of
linguistics.
References
Atkinson, D. 1999. Scientific Discourse in Sociohistorical Context: The Philosoph-

ical Transactions of the Royal Society of London, 1675–1975. Mahwah: Law-
rence Erlbaum Associates.
Banks, D. 1995. Your Very First ESP Text (Wherein Chaucer Explaineth the
Astrolabe). ASp 15–18: 451–60.
Banks, D. 2005. Introduction à la linguistique systématique fonctionnelle de l’an-
glais. Paris: L’Harmattan.
Banks, D. 2008. The Development of Scientific Writing, Linguistic Features and
Historical Context. Sheffield: Equinox.
Banks, D. 2016. On the (Non)necessity of the Hybrid Category Behavioural
Process. In D. R. Miller and P. Bayley, eds., Hybridity in Systemic Functional
Linguistics, Grammar, Text and Discourse. Sheffield: Equinox. 21–40.
Banks, D. 2017. The Birth of the Academic Article: Le Journal des Sçavans and the
Philosophical Transactions, 1665–1700. Sheffield: Equinox.
Biber, D. and B. Gray. 2010. Challenging Stereotypes about Academic
Writing: Complexity, Elaboration, Explicitness. Journal of English for Aca-
demic Purposes 9: 2–20.
Biber, D. and B. Gray. 2011. Grammatical Change in the Noun Phrase: The
Influence of Written Language Use. English Language and Linguistics 15(2):
223–50.
Biber, D. and B. Gray. 2013. Being Specific about Historical Change: The
Influence of Sub-register. Journal of English Linguistics 41(2): 104–34.
Bynon, T. 1977. Historical Linguistics. Cambridge: Cambridge University
Press.
Crystal, D. 1980. A First Dictionary of Linguistics and Phonetics. London: André
Deutsch.
Crystal, D. 1999. The Penguin Dictionary of Language. 2nd ed. London:
Penguin.
Cummings, M. 1995. A Systemic Functional Approach to the Thematic
Structure of the Old English Clause. In R. Hasan and P. H. Fries, eds., On
Subject and Theme: A Discourse Functional Perspective. Amsterdam: John
Cummings, M. 2010. An Introduction to the Grammar of Old English: A Systemic
Functional Approach. Sheffield: Equinox.
Hall, M. B. 2002. Henry Oldenburg: Shaping the Royal Society. Oxford: Oxford
University Press.
Halliday, M. A. K. 1985. Introduction to Functional Grammar. London: Arnold.
Halliday, M. A. K. 1988. On the Language of Physical Science. In M. Gha-
dessy, ed., Registers of Written English: Situational Factors and Linguistic
Features. London: Pinter. 162–78.
Halliday, M. A. K. and J. R. Martin. 1993. Writing Science: Literacy and Discur-
sive Power. London: The Falmer Press.
Halliday, M. A. K. and C. M. I M. Matthiessen. 2014. Halliday’s Introduction to
Halliday, M. A. K. and J. J. Webster. 2014. Text Linguistics: The How and Why of
Meaning. Sheffield: Equinox.
Hopper, P. J. and E. C. Traugott. 2003. Grammaticalization. 2nd ed. Cam-
bridge: Cambridge University Press.
Licoppe, C. 1994. The Crystallization of a New Narrative Form in Experi-
mental Reports (1660–1690): The Experimental Evidence as a Transaction
between Philosophical Knowledge and Aristocratic Power. Science in Con-
text 7(2): 205–44.
Licoppe, C. 1996. La formation de la pratique scientifique: Le discours de l’expéri-
ence en France et en Angleterre (1630–1820). Paris: La Découverte.
Lyons, J. 1968. Introduction to Theoretical Linguistics. Cambridge: Cambridge
University Press.
Martínez-Insua, A. E. 2013. There-constructions as a Choice for Coherence in
the Recent History of English. In L. Fontaine, T. Bartlett, and G. O’Grady,
eds., Systemic Functional Linguistics: Exploring Choice. Cambridge: Cambridge
Morgan, B. T. 1928. Histoire du Journal des Sçavans depuis 1665 jusqu’en 1701.
Paris: Presses Universitaires de France.
432 DAVID BAN KS
O’Halloran, K. 2003. Intersemiosis in Mathematics and Science: Grammat-

ical Metaphor and Semiotic Metaphor. In A.-M. Simon-Vandenbergen,
M. Taverniers, and L. Ravelli, eds., Grammatical Metaphor: Views from Sys-
temic Functional Linguistics. Amsterdam: John Benjamins. 337–66.
O’Halloran, K. 2005. Mathematical Discourse: Language, Symbolism and Visual
Images. London: Continuum.
Salager-Meyer, F. 1999. From ‘Mr. Guthrie is Profoundly Mistaken . . .’ to
‘Our Data Do Not Seem to Confirm the Results of a Previous Study on . . .’:
A Diachronic Study of Polemicity in Academic Writing (1810–1995).
Ibérica 1: 5–28.
Salager-Meyer, F. and N. Zambrano. 2001. The Bittersweeet Rhetoric
of Controversiality in Nineteenth- and Twentieth-Century French
and English Medical Literature. Journal of Historical Pragmatics 2(1):
141–73.
Starc, S. 2010. Textual Patterning and Information Flow (Theme^Rheme) in
the Generic Evolution of 19th Century Slovene Newspaper Advertise-
ments. In E. Swain, ed., Thresholds and Potentialities of Systemic Functional
Linguistics: Multilingual, Multimodal and other Specialised Discourses. Trieste:
Edizioni Università di Trieste. 133–57.
Starc, S. 2015. The Difference in Text Structure between Advertisements
and Classified Advertisements from a Diachronic Perspective. In
D. Banks, ed., Aspects linguistiques de la ‘petite annonce’. Paris: L’Harmattan.
25–35.
Swales, J. 1990. Genre Analysis, English in Academic and Research Settings.
Swales, J. 2004. Research Genres: Exploration and Applications. Cambridge:
Traugott, E. C. 1988. Pragmatic Strengthening and Grammaticalization. In
S. Axmaker, A. Jaisser, and H. Singmaster, eds., Proceedings of the Four-
teenth Annual Meeting of the Berkeley Linguistics Society. Berkeley: Berkeley
Linguistics Society. 406–16.
Urbach, C. 2013. ‘Choice’ in Relation to Context: A Diachronic Perspective
on Cultural Valeur. In L. Fontaine, T. Bartlett, and G. O’Grady, eds.,
Systemic Functional Linguistics: Exploring Choice. Cambridge: Cambridge
Valle, E. 1999. A Collective Intelligence: The Life Sciences in the Royal Society as a
Scientific Discourse Community, 1665–1965. Turku: Anglicana Turkuensia.
17
SFL and Multimodal
Discourse Analysis
Kay L. O’Halloran, Sabine Tan, and Peter Wignell
17.1 Introduction
Michael Halliday’s Systemic Functional Theory (SFT), most fully developed

as Systemic Functional Linguistics (SFL) (e.g. Halliday and Matthiessen 2014;
Martin and Rose 2007), provides an unrivalled platform for modelling,
analyzing, and interpreting multimodal texts, interactions, and events
involving language and other resources such as images, scientific symbol-
ism, sound, embodied action, and so forth (see Jewitt 2014; O’Halloran
2011). The resulting approach, systemic functional multimodal discourse analysis
(SF-MDA) (e.g. Jewitt et al. 2015; O’Halloran and Lim-Fei 2014; O’Halloran
2008a; Unsworth 2008) is explored in this chapter as (a) a theoretical
construct and (b) research and analysis in action. Key conceptual ideas in
SFT – such as metafunction, register and genre, realization, stratification,
and constituency – are explicated and extended to other semiotic resources
to provide a vivid account of how meaning arises through combinations of
semiotic choices – that is, from semiotic interactions within and across
different resources rather than from individual system choices – and how
these meanings can be modelled, analyzed, and interpreted.
The SF-MDA approach is demonstrated through analysis of the World
Health Organization (WHO) Ebola webpage as of 20 June 20151
(which has since been revamped) using purpose-built multimodal analysis
software.2 The WHO Ebola webpage was chosen for illustrative purposes
because it contains a variety of discourse types (i.e. reporting, information,
promotionals, and news), offering the opportunity to explore multimodal
semiosis across linguistic text, photographs, scientific graphs, infographics,
hyperlinks, and videos. As part of the discussion, concepts such as
‘intersemiosis’ (e.g. O’Halloran 2008b; Royce 2007) and ‘resemiotization’
1
See web.archive.org/web/20150620204949/http://apps.who.int/ebola. (Last accessed: 07/08/2017.)
2
See multimodal-analysis.com/products/multimodal-analysis-image. (Last accessed 07/08/2017.)
Downloaded from https://www.cambridge.org/core. University of Sussex Library, on 22 Jul 2019 at 01:04:22, subject to the Cambridge Core terms of use, available at
434 K AY L . O ’ H A L LO R A N , S A B I N E T A N , A N D P ET E R W I G N E L L
(e.g. Iedema 2001, 2003), and the major theoretical and analytical chal-
lenges facing multimodal analysts, together with the associated vision for
future research in the field, are explored in detail.
17.2 SF-MDA: Theoretical Constructs
This section presents the theoretical foundations that inform the SF-MDA
approach introduced in this chapter. This approach builds upon Michael
Halliday’s Systemic Functional Theory (SFT) (e.g. Halliday 1978, 1985; Halli-
day and Matthiessen 2014), where semiotic resources are conceptualized in
terms of the functions they serve in society. From an SF-perspective, culture
is defined as a network of semiotic systems, that is, ‘a set of systems of
meaning, all of which interrelate’, and which ‘taken all together, constitute
human culture’ (Halliday and Hasan 1985:4). The practice of modelling the
meaning potential in culture as system networks reflects Halliday’s (1994:
xiv) understanding of systemic theory as ‘a theory of meaning as choice, by
which language, or any other semiotic system, is interpreted as networks of
interlocking options’, whereby the particular choices that are made are not
to be viewed as the result of conscious decisions, but rather as ‘a set of
possible alternatives’ (Halliday 1994:xiv–xxvi).
Although initially applied to language, SFT has since been adapted and
extended to the study of multimodal texts and artefacts to account for
the ways in which linguistic and non-linguistic resources (e.g. spoken
and written language, image, gesture, sound, music, film, page layout,
website design) combine and interact in the communication of meaning
(e.g. Bateman 2014a, 2014b; Bateman and Schmidt 2012; Iedema 2001,
2003; Jewitt 2014; Kress and van Leeuwen 2001, 2006; Machin 2007;
O’Halloran 2004, 2008a, 2008b; O’Toole 2011; Royce 2007, 2015; van
Leeuwen 1999, 2005, 2012).
Because of its adaptability and amenability (e.g. Martin 2002), SFT is
regarded as particularly well suited to provide the theoretical foundation
for SF-MDA (see Jewitt et al. 2015: Chapter 3), as demonstrated, for example,
by Kress and van Leeuwen’s (2006) and O’Toole’s (2011) frameworks for
analyzing images and displayed art (paintings, sculptures, architecture)
respectively. It must be emphasized, however, that while the higher-level
principles of SFT can be applied to the analysis of multimodal texts, the
systems for visual images and other semiotic resources are not the same as
the systems for language. Systemic Functional MDA extends beyond the
simple adaptation of established ‘SF-approaches which were largely
developed for modeling discourse and grammatical systems in language’
(O’Halloran 2008a:446), and requires the development – and integration –
of different, yet complementary, models and approaches for the study of
multimodal semiosis. For this reason, the term SFT is used here rather
than SFL.
SFL and Multimodal Discourse Analysis 435
In what follows, key concepts from SFT that form the guiding principles
for modelling language and other semiotic resources and their interactions
are discussed in detail, followed by practical examples of how aspects of the
SF-MDA model can be applied to the WHO Ebola webpage. The issues which
arise when language and other semiotic resources are considered as inter-
related systems of meaning (i.e. ‘semiotic resources as system’) and as
multimodal texts (i.e. ‘semiotic resources as text’) (see Halliday 2008) are
central to this discussion.
17.2.1 Metafunctions
One of the key tenets in SFT is Halliday’s metafunctional principle, which
posits that language and other semiotic systems are structured to make
three kinds of meanings simultaneously: (a) ideational meaning for constru-
ing our experience and knowledge of the world (i.e. experiential meaning)
and making logical connections in that world (i.e. logical meaning); (b)
interpersonal meaning for enacting social relations and expressing attitudes;
and (c) textual meaning for organizing meanings into coherent messages
(e.g. Halliday 1978; Halliday and Matthiessen 2014).
In language, the three metafunctions are mapped on the structure of
the clause by specifying the grammatical systems through which these
metafunctions are realized. For example, experiential meaning is realized
through the grammatical system of Transitivity which accounts for the differ-
ent types of process that are found in a language, and the structures through
which they are expressed (Halliday and Matthiessen 2014:211–335).
A process potentially has three components: the process itself; participants
in the process; and circumstances associated with the process. The following
process types are found in English: Material Processes of doing and acting;
Mental Processes of thinking, feeling, and perceiving; Relational Processes
of classifying and describing (i.e. having attribute) and identifying; Verbal
Processes of saying and meaning; Behavioural Processes of human physio-
logical behaviour (e.g. smiling); and Existential Processes of existing and
happening; as well as the functional participant roles and circumstances
that are associated with these processes. Logical meaning is concerned
with the relations between happenings at the clause and discourse level,
realized through conjunctions (e.g. if, so, moreover). Logical meaning is
mapped through the systems of logico-semantic relations (i.e. nature of the
semantic relations, which is either to expand meaning or to project what is
said or thought) and taxis (i.e. clause dependency relations). Interpersonal
meaning is realized through the grammatical system of Mood, which – at
the level of the clause – realizes ‘meaning as an exchange’ (Halliday and
Matthiessen 2014:135–9), with choices for giving and demanding informa-
tion (statements, questions), and for giving and demanding goods and ser-
vices (offers, commands); and the system of Modality (expressions of
probability, usuality, obligation, and inclination). Textual meaning is
concerned with the organization of the message, realized through the

thematic structure of the clause which organizes information into clearly
defined points of departure (i.e. Theme) for the remainder of the message
(i.e. Rheme).
The key principle that semiotic resources other than language also have
an underlying organization which permits them to (differentially) fulfil the
three metafunctions has been adopted in SF-MDA. For instance, the notion
of transitivity has been successfully adapted and applied to other semiotic
resources, for example, for analysis of static and dynamic images. Comple-
mented by concepts from social semiotics, Critical Discourse Analysis,
photography, film theory, and visual design, this has resulted in the formu-
lation of visual systems which function to structure our experience of the
world in terms of participants, processes, and circumstances. Similarly,
interpersonal visual systems (e.g. direct and indirect gaze, visual modality)
and compositional systems (e.g. framing and perspective) have also
been formulated and successfully applied for the analysis of visual texts
(e.g. Kress and van Leeuwen 2006; Machin 2007; Machin and Mayr 2012;
O’Toole 2011).
From an SF-MDA perspective, the metafunctional principle plays an
important role in determining the functionalities and underlying organiza-
tion of semiotic resources, and for investigating the ways in which semiotic
choices combine, interact, and integrate in multimodal texts to create
meaning.
17.2.2 Register and Genre

Another key concept from SFT that informs SF-MDA is the notion that
social context is modelled through register and genre (e.g. Eggins 1994;
Martin 1992, 2002; Martin and White 2005). Register theory describes the
impact of three key dimensions on the way language is used in context. The
three key dimensions, or register variables, theorized as field, tenor, and
mode, are directly related to the above-described metafunctions. Field, for
instance, relates to the experiential and logical metafunctions, and
describes what a text is all about, or what is happening, that is, it is
concerned with systems of transitivity, including descriptions of partici-
pants, processes, and circumstances involved. Tenor, in turn, relates to the
interpersonal metafunction, and is concerned with the ways social relations
are enacted through the dimensions of power and solidarity, while mode
relates to the textual metafunction, and is concerned with the role language
plays in discourse, that is, whether it is written or spoken, and the ‘infor-
mation flow’ across different media or channels of communication (speech,
writing, images, webpages, video, etc.) (e.g. Martin 2002:56; Martin and
White 2005:28).
Following Martin (1992), over and above the registerial configurations of
tenor, field, and mode, another contextual level is seen to be operating in
Figure 17.1 Register and genre (reproduced from Martin and White 2005:32)
discourse, referred to as genre. Martin (2002:56) describes genre as a ‘system

of staged goal-oriented social processes through which social subjects in a
given culture live their lives’. From Martin’s (2002:56) perspective, genre is
essentially seen as an integrated system of social processes, ‘where the
principles for relating social processes to each other have to do with tex-
ture, that is, the ways in which field, mode and tenor variables are phased
together in a text’, as illustrated in Martin and White’s (2005:32) schematic
model of register and genre developed for the study of language (repro-
duced in Figure 17.1).
The notions of register and genre, developed for the study of linguistic
texts, are useful concepts to explore how multimodal semiotics opens up
new avenues for theorizing and understanding complex multimodal texts,
such as the WHO Ebola webpage. Such multimodal texts involve a range of
registerial configurations and genres, which, taken together, form recog-
nizable ‘macrogenres’ (e.g. Christie 2002), such as websites. However, as the
discussion of the WHO Ebola webpage will show, registerial configurations
and genres with overlapping boundaries are constantly evolving in inter-
active digital media, providing further challenges for multimodal analysts.
17.2.3 Realization, Stratification, and Constituency

Other key principles adopted from SFT into SF-MDA are the concepts of
stratification and constituency, where semiotic resources are modelled
according to strata and ranks to account for how meaning is organized
and realized through material signs (e.g. words, sounds, and images) (see
O’Halloran and Lim-Fei 2014:138–9). These concepts originate from the idea
that language is a stratified semiotic system ‘involving three cycles of
coding at different levels of abstraction’ (Martin and White 2005:8), as
illustrated in Figure 17.2. Martin and White’s (2005) model sees discourse
Figure 17.2 Language strata (reproduced from Martin and White 2005:9)
semantics realized through lexicogrammar, which in turn is realized

through phonology (for spoken language) and graphology (for written lan-
guage). In this model, which was developed for language, the highest level
of abstraction, discourse semantics, is focused on aspects of discourse
organization, ‘including the question of how people, places and things are
introduced in text and kept track of once there (identification)’ (Martin and
White 2005:9). The lexicogrammar is organized around the clause, the
smallest semantic unit in which the systems for the three metafunctions
operate. These systems are mapped onto the material plane, which is the
actual spoken and written words, with their accompanying phonological
(e.g. intonation, pitch, information focus) and graphological systems (e.g.
font size, colour, style).
The concept of strata has been applied to SF-MDA. For example, in
O’Halloran’s (2008a) framework, developed for printed texts with language
and image components (reproduced in Figure 17.3), the language plane
consists of two strata: (1) the content stratum, realized through discourse
semantics and lexicogrammar; and (2) the expression or display stratum,
realized through typography/graphology for written language. O’Halloran’s
(2008a) framework builds on and expands the concepts of stratification
and constituency by proposing two complementary strata for visual
imagery: (1) the content stratum (comprising visual discourse/grammatical
systems for the whole image and its constituent parts); and (2) the display
stratum (with systems for the material realization of the image) (O’Halloran
2008a:450–1, 2008b). In this framework, different levels in the content
stratum are modelled as constituent parts of higher-level ones, that is,
language is organized according to the constituent ranks of word, word
group/phrase, clause, and clause complex, whilst visual images, following
the work of O’Toole (2011), are organized according to the ranks of Member
(Part), Figure, Episode, Scene, and Work.
Figure 17.3 Adapted from O’Halloran’s (2008a) SF-MDA framework work for language
and visual imagery
The development of an integrative SF-MDA model based on constituency

and ranked scales is a challenging task, particularly as it requires theoriza-
tion of the different systems which operate at each rank for different
semiotic resources. Nonetheless, the approach permits semiotic inter-
actions across system choices at different ranks to be observed, together
with the semantic expansions which occur, as demonstrated in the analysis
of the WHO Ebola webpage below.
A basic principle of SF-MDA is that ‘the whole is other than the sum of its
parts’, following gestalt theory (e.g. Koffka 1935). The approach enhances
our understanding of the intersemiotic relations and mechanisms through
which semiotic interactions construct meaning, both as sets of interrelated
systems and as multimodal semiotic processes and artefacts.
17.2.4 Intersemiosis and Resemiotization

The complexity of multimodal semiosis presents major challenges which
extend beyond those encountered when studying language as an isolated
phenomenon. As we shall see, SFT principles developed for modelling

discourse and grammatical systems in language need to be adapted and
expanded to account for multimodal forms of semiosis, which are different
from language (see Jewitt et al. 2015: Chapter 3). In contrast to linguistic
texts, which tend to unfold ‘syntagmatically as a chain which is sequen-
tially processed, and [in which] meaning culminates progressively as the
text unfolds’ (O’Halloran 2008a:447), the effect of multimodal forms of
semiosis on meaning-making is multiplicative (e.g. Lemke 2005) and multi-
directional. For example, visual texts (including composite multimodal
texts such as websites) structure reality as (multiple) parts which are
immediately perceived in relation to the whole. This is radically different
from language, where reality is structured as sequences of events.
The processes of intersemiosis and resemiotization as the ‘analytical
means for (1) tracing how semiotics are translated from one into the other
as social processes unfold, as well as for (2) asking why these semiotics
(rather than others) are mobilized to do certain things at certain times’
(Iedema 2003:29) are thus key to SF-MDA. Attempts have been made to
theorize the complex relations that exist between text and images, from
both SF as well as wider social semiotic perspectives (e.g. see Bateman 2014b
for a comprehensive discussion of frameworks for text–image relations).
For instance, intersemiotic text–image relations have been modelled in
terms of multimodal cohesion or intermodal complementarity (e.g. Royce
2007, 2015), intersemiotic texture (e.g. Baldry and Thibault 2006; Liu and
O’Halloran 2009; Martinec 1998), and logico-semantic and discursive con-
junctive relations (e.g. Martinec and Salway 2005). In the SF-MDA model
presented here, however, the goal is not to apply or adapt a particular
intersemiotic framework, but rather (1) to closely examine the expansions
of meaning that arise from combinations of semiotic choices as they unfold
within and across metafunctions at different ranks and strata in the WHO
Ebola webpage, and (2) to trace and map out the intersemiotic relations that
account for this multiplicative effect in meaning-making, where the whole
is other than the sum of its component parts.
17.3 SF-MDA: Research and Analysis in Action
In the preceding section, key concepts from SFT that inform SF-MDA have
been outlined. These concepts are now applied for the analysis of texts,
images, and their relations in the WHO Ebola webpage. While one aim is to
discover how the webpage functions to create meaning, a further and wider
aim is to explore the implications of the SF-MDA approach for the analysis
of such complex multi-semiotic texts, and to demonstrate the possibilities,
strengths, and limitations associated with this approach.
The language in the example texts is analyzed using SFL (e.g. Halliday and
Matthiessen 2014; Martin and Rose 2007), while the visual texts are
analyzed using a model based on O’Toole’s (2011) rank-based metafunc-

tional framework for the analysis of paintings, which has been adapted for
the analysis of photographs, together with concepts from Kress and van
Leeuwen (2006). The analysis was undertaken using the interactive software
Multimodal Analysis Image3 which has facilities for entering system net-
works, importing multimodal texts such as image files, annotating the
image files using overlays in the form of geometrical shapes and pins,
attaching a system network choice to each overlay, and exporting the
results to Excel for further data processing. The system networks entered
into Multimodal Analysis Image – which are not exhaustive, but designed for
the purpose of illustrating how SF-MDA can be applied for the analysis of a
complex multimodal text such as the WHO Ebola webpage – are displayed
in Table 17.1.
The analysis begins with an overview of the WHO Ebola webpage, which
is broken down into constituent parts (i.e. sections and ranks) and the
functional elements within those parts. Examples of different image-text
combinations found on the webpage are then discussed in detail.
17.3.1 Constituency-based Analysis of the WHO Ebola Webpage

Previous work on webpage analysis from an SF-perspective (e.g. Baldry and
Thibault 2006; Kok 2004) has shown that concepts like genre, stratification,
and constituency can be adapted and applied to the analysis of hypertext.
Baldry and Thibault (2006:156) describe hypertext discourse as ‘a newly
evolving system of semiotic possibilities’, being composed of ‘a hybrid of
precursor genres such as verbal text, visual images, and multimodal com-
binations of these’. Kok (2004:133) cautions, however, that hypertext pre-
sents ‘different orders of abstraction’ which must not be ‘confused with
ranks or levels which are posited for different semiotic resources’, such as
language, or even those proposed for visual art (e.g. O’Toole 2011). As
shown below, the organization of a webpage is markedly different from
the hierarchical and sequential order that governs most linguistic texts (see
Baldry and Thibault 2006:126).
The constituency of the WHO Ebola webpage’s compositional layout
presented in Figure 17.4 shows, for example, that at the most global level
of organization (Figure 17.4 left), the webpage is organized hierarchically,
comprising the typical hypertextual design elements of a Masthead/Banner
(1), followed by a Content section (2), and a Bottom Sitemap (3).
The Masthead/Banner and Bottom Sitemap function to contextualize the
webpage as belonging to the WHO in general, and the field of Ebola in
particular, and provide the navigational options for the user to traverse the
webpage. In contrast, the levels of organization in the Content section
3
See multimodal-analysis.com/products/multimodal-analysis-image. (Last accessed 07/08/2017.)
Table 17.1 Text and image systems
Semiotic
Resource
Metafunction
Rank System Description
Text
EXPERIENTIAL
Clause Processes; Participant Roles; Happenings, actions, and relations
Circumstance
INTERPERSONAL
Clause Speech Function Exchange of information (e.g.
statements and questions) and
goods and services (e.g. commands
and offers)
TEXTUAL
Clause Information Focus Organization of information, with
Discourse points of departure for what follows
Semantics
Image
EXPERIENTIAL
Work Narrative Theme; Nature of the scene
Representation; Setting
Episode Processes; Participant Roles; Visual happenings, actions, and
Circumstance relations
Figure Posture; Dress Characteristics of the participants
INTERPERSONAL
Work Angle; Shot Distance; Lighting Visual effects
Episode Proportion in Relation to the Happenings, actions, and relations
Whole Image: Focus; with respect to the whole image
Perspective
Figure Gaze-Visual Address Direction of participant’s gaze as
internal to image or external to
viewer
TEXTUAL
Work Compositional Vectors; Framing The organization of the parts as a
whole, with the visual marking (e.g.
framing) of certain parts
Episode Relative Placement of the Position of the happenings, actions,
Episode; Framing and relations in relation to the whole
image, and the visual marking of
certain aspects
Figure Relative Placement of the Position of the figure in relation to
Figure within the Episode; happening, action, or relation,
Arrangement; Framing and the visual marking of certain
aspects of the figure
(Figure 17.4 right) are ordered neither hierarchically nor sequentially, but
comprise the generic mix by which the webpage presents information and
engages the reader. The different discourse types found on the Ebola web-
page are Reporting (4), Information (5), Promotion (6), and News (7), respectively
(see Sharoff 2010 for a discussion of genres on the web).
In terms of the registerial mix deployed, the most global organizational
level (Figure 17.4 left) realizes mostly textual or compositional meaning,
whilst the second layer of organization, which contains different discourse
Figure 17.4 Constituent levels of the Ebola webpage organized in terms of hypertext
components (left): Masthead/Banner (1), Content (2), Bottom Sitemap (3); and discourse
types (right): Reporting (4), Information (5), Promotion (6), News (7)
types (Figure 17.4 right), realizes ideational (i.e. experiential and logical),
interpersonal and textual meaning. In terms of the tenor relations estab-
lished on the webpage, the different sections and subsections are concerned
largely with presenting and reporting information about Ebola, except for
the section entitled ‘Get Involved’, which draws on promotional discourse
(see Section 17.3.5 below).
Each of the webpage’s sections and subsections can be broken down
further into constituent parts and elements (see Figures 17.6 and 17.11
below, for example). In what follows, examples of image-text combinations
found in different sections of the webpage (as displayed in Figure 17.4) are
discussed in detail to demonstrate the possibilities, strengths, and limita-

tions of the SF-MDA approach: (a) a photograph-text complex in the main
visual display in the ‘Reporting’ section (4) at the top of the webpage;
(b) graphs and language in the ‘Information’ section (5); (c) an infographic
in the ‘Information’ section (5); (d) a multi-semiotic text (i.e. a graph
combined with language) in the ‘Promotion’ section (6); and (e) the final
‘News’ section (7) at the bottom of the webpage. In each case, the linguistic
and visual choices are discussed in terms of ideational, interpersonal, and
textual meanings, and the resulting intersemiotic relations within and
across the different registers and genres.
17.3.2 Photograph and Text from the Webpage’s Main

Visual Display
The photograph-text complex in Figure 17.5 is one of seven image-text
complexes which form a moving banner at the top of the Ebola webpage,
and make up the main visual display in the ‘Reporting’ section. The text
accompanying the photograph consists of a headline, a dateline and what
appears to be the lead paragraph (outlined in Figure 17.5) of a longer story,
which can be accessed by clicking on the headline. The bright, colourful
photographs in the display contrast with the rest of the webpage, and are
likely to attract the viewer’s attention when they arrive at the webpage.
This gives the visual display, in part, an interpersonal role in engaging the
viewer with the webpage (see O’Toole 2011:11–16). The following analysis
and discussion focuses principally on how ideational and textual meaning
are realized in the text and the photograph, and the intersemiotic relations
between text and image.
17.3.2.1 Ideational Meaning in the Text

In terms of transitivity, the linguistic text focuses on the family of Bernard
Lansana Soumah, namely, Bernard and his wife, Macire. They engage as
Figure 17.5 Photograph-text complex from the webpage’s main visual display
Participants in a number of different process types: Sensers in Mental

Processes (never expected, realized); Carriers in Relational Attributive Pro-
cesses (are among the lucky ones, are vigilant); and Token in a Relational
Identifying Process (providing a message). Macire also features as Goal in a
Material Process (became infected). The text is located in time and place
through Circumstances (today, in Forecariah). Tense choices separate the text
into two time phases: the text begins in the past and then shifts into the
present. The Processes, Participants, and Circumstances are configured into
clauses, which are logically connected to other clauses, forming clause-
complex structures. In this way, aspects of experience are ordered into
constituent parts which unfold as a series of events which are related to
each other.
17.3.2.2 Textual Meaning in the Text

In language, at clause rank, Theme represents information prominence at
the beginning of a clause (Martin and Rose 2007:189). The other typical
point of information prominence is at the end of a clause, where New
information appears. Martin and Rose (2007:187–218 describe this flow of
information as wave-like. They then extend this analogy to similar, larger-
scale patterns in discourse, describing waves of Theme and New informa-
tion at clause rank as ‘little waves’. ‘Bigger waves’, operating across larger
stretches of discourse, are referred to as hyperThemes and hyperNews,
while ‘tidal waves’, operating across even larger stretches of discourse,
are referred to as macroThemes and macroNews. These resources are
deployed to structure the flow of information at different scales in the
linguistic text. For example, the headline functions as macroTheme for
the story and the dateline serves as hyperTheme, which sets the story in
time. Despite Ebola in the headline functions as Topical Theme for its clause
and sets up an expectation of some kind of success which is announced in
the clause Rheme and then illustrated in the rest of the text. The family and
family members are Topical Themes in most of the rest of the text, apart
from Today, which signals the shift from the past to the present. The
thematic pattern combines with New information, where aspects of the
headline are picked up as New (vigilant, survival, hope) in subsequent clauses
where the family or family members are Theme. This, in conjunction with
the New elements, links back to the headline, creating a unity of Theme and
New information in the text.
Discourse semantic systems of Identification (Martin and Rose 2007) also
work to allow the reader to identify and track participants. First the group
is identified as a family, and then as specific individuals, by name and
through pronouns.
17.3.2.3 Ideational Meaning in the Photograph

In the photograph, ideational meaning is realized very differently. In com-
posing the photograph, the photographer has made semiotic choices based
on a wider experience, and has ordered and arranged those selections.

However, the semiotic ordering of the photograph does not unfold in time
like the linguistic text does. Rather, the scene is presented as a whole,
frozen in time, for the viewer to make sense of. There is no location in
time or passage of time, and there is no information in the photograph that
tells the viewer when or where it was taken.
The photograph focuses on three main participants: a man, a woman,
and a child (the latter, in fact, is a participant not mentioned in the
accompanying text). Although the viewer may perceive them as a family,
this inference is drawn from visual cues, such as their genders and ages,
and the fact that they are sitting close together on the motorbike.
17.3.2.4 Textual Meaning in the Photograph

Similarly, the prominence of information in the photograph is not realized
in the same way as in the text. The written text unfolds in time and space
line by line from top left to bottom right, and the default option for a reader
is to follow this path. Photographs, however, are not necessarily viewed
following a given path. In the photograph, the prominence of the figures
results from the combination of textual, experiential, and interpersonal
choices. That is, textually, or compositionally, the three figures are pos-
itioned centrally and are foregrounded. They occupy about half of the total
space of the photograph. The man’s bright white T-shirt and the woman’s
purple top further help to attract the viewer’s attention. The parallel
alignment of their right arms, along with the angle at which their bodies
are positioned, additionally functions to create a harmony among them,
allowing the viewer to see them collectively as well as individually. This is
augmented by the uniformity of Processes in which they are engaged (i.e.
smiling, looking, sitting).
17.3.2.5 Intersemiotic Relations: Text and Photograph

Although the written text and photograph could each stand alone, there is a
logical connection of implicit similarity between them. In language, con-
junctive relations are typically one-directional: what is said or written next
is usually linked to something that has been said or written earlier in the
text. In contrast, the intersemiotic relations between text and photograph
are multi-directional; that is, the two parts (linguistic and visual) are also
parts of a whole which are perceived in relation to each other. Regardless of
whether we read the text after viewing the photograph or vice versa, the
meanings we make from their combination are different from the mean-
ings we make from each in isolation. This transference of meaning happens
each time we move from image to text or the other way around, with each
pass bringing something new that changes how we perceive both. For
example, we learn the names of the participants in the image from the
text, and we learn from the photograph what the participants in the text
look like. We can see in the photograph that they are happy, and we learn
from the text why they are smiling. The semiotic selections in the photo-
graph and the surrounding contextual factors influence how we perceive
and understand those processes. In this case, the text and image co-
contextualize each other: the main visual participants are a happy smiling
family, and after setting up an expectation of success via the headline
(Despite Ebola, vigilance and hope prevail in Forecariah) the linguistic text pro-
vides the reasons for that happiness.
Although the text and the photograph are drawing on textual, interper-
sonal, and ideational choices, the choices are made from different sets of
systems with different sets of options. Moreover, the selections function
differently in terms of structure: whereas language structures thought and
reality as ordered sequences of events, thought and reality in the photo-
graph are ordered in terms of a part-whole relationship, which makes it
possible to see the family in relation to each other and the immediate
context which is depicted.
Intersemiotic relations are also established within and across ranks (see
O’Toole 2011:11–31). For example, what happens at clause rank in the
transitivity also happens at the rank of Figure in the photograph: i.e.
Participants engage in Processes. Also, what happens at the rank of Episode
in the photograph appears to align with what happens in clause complexes
and in larger pieces of text: multiple participants engage in multiple pro-
cesses. There is, however, a key point of difference: in text, someone (or
something) can be a Participant in only one Process at a time, while in a
photograph the same person or thing can be a Participant in a number of
Processes simultaneously, such as the family members in the photograph,
who are portrayed as Behavers in the Behavioural Processes of smiling,
looking, and sitting.
17.3.3 ‘Current Situation’: A Section of the Webpage

The ‘Current Situation’ in the ‘Information’ section of the webpage consists
of two levels of headings, three graphs, and language (see Figure 17.6). The
section presents information which has been drawn and summarized from
sources which can be found by clicking on the action bars at the end of the
section. The links take the reader to the sources of information for both the
graphs and the text. The section is discussed with reference to textual
meaning, ideational meaning, and intersemiosis. Intersemiotic relations
are discussed as they are found within the section and in relation to other
parts of the webpage. The section (with constituent parts outlined) is shown
in Figure 17.6.
17.3.3.1 Textual Meaning

In its overall textual organization, the section follows the same pattern as
the other sections of the webpage, except for the main visual display. The
section has a large heading ‘Current Situation’ in the upper left corner
Figure 17.6 ‘Current Situation’ section of webpage with constituent parts outlined
which functions as a macroTheme for the section. The two action buttons
in the bottom right corner ‘Ebola data and statistics’ and ‘Ebola Situation
Report – 24 June 2015’ function as a kind of macroNew. The smaller
heading ‘Cases in the most affected countries’ functions as hyperTheme
for the graphs and written text combined. Each of the three graphs has its
own heading structure which identifies the country (i.e. Guinea, Liberia, and
Sierra Leone) and the sources of information for the graphs. Within each
graph, mouse-over pop-ups bring to the foreground information on the
number of cases at a particular time and make that information prominent
by having it stand out from the rest of the graph. The information in these
pop-ups is aggregated and used in the text underneath the graphs, where it
is typically made thematic as part of the Theme (see Figure 17.7). That is,
topical Themes in the text are mainly numbers of cases and countries, and
changes over time and more specific locations within Guinea and Sierra
Leone are found in the New. There is a strong textual link between infor-
mation which is made prominent in the graphs and the Themes in the
linguistic text underneath the graphs.
17.3.3.2 Ideational Meaning

Logically, the whole section is connected through internal implicit similar-
ity (see Martin and Rose 2007:110–44 for a discussion of conjunction), as
each part of the section relates to what precedes it through progressively
more specific exemplification. Within that logical structure the three
graphs could be viewed as a ‘graph complex’ with relations of internal,
Figure 17.7 An example of multiple relational processes in a graph
implicit addition connecting the three graphs into a semantic unit, as each
graph displays the same categories of information but for different coun-
tries, thus providing a visual comparison. This comparison can be made by
looking at the graphs in any order. Each graph shows the name of a
country, number of cases, and dates for that country. The relationship
between the graph complex and the subsequent text could be seen as
implicit similarity, since the graphs and the text present comparatively
similar, but not identical, information.
The x axis of each graph represents time (e.g. x1) and the y axis represents
number of cases (e.g. y1), forming a Relational process with a Token and
Value respectively (see Figure 17.7). The relationship between time and the
number of cases (e.g. (x1, y1) and (x2, y2)) is viewed as a pop-up which
becomes visible through mouse-over (see Figure 17.7). The series of rela-
tions is represented by points on the graph. Each relationship and its point
of intersection can be regarded as a visual Relational Identifying Process,
with the relationship between number and time as the Value/Identifier (x, y)
and the point of intersection as the Token/Identified (the point). However,
this visual relational process is implicit, and is based on knowledge of
mathematics and graph theory, in particular, line graphs. In this case, the
mathematical relation (x, y) is resemiotized as the visual Participant (•),
representing a situation where there is an intersemiotic downranking of a
mathematical identifying process (x corresponds to y) to a visual entity (the
point). Each graph, then, encodes a large number such metaphorical rela-
tions simultaneously (in this case, the number of cases and the number of
dates on which data was entered). While the whole graph encodes these
relationships simultaneously, the mouse-over function allows them to be
viewed one at a time (see Figure 17.7). From the series of identifying
relations which are resemiotized as points, a new visual entity in the form
of a graph is introduced. The visual semiotic (the graph) permits the

patterns of relations to be perceived and understood over time, creating
an abstract form of semiotic ordering in relation to the Ebola outbreak. In
this regard, visual semiosis creates meanings which are not possible using
language or mathematical symbolism.
While the graph displays patterns over time, the text summarizes and
aggregates information in the graphs and provides additional information,
which adds specificity to the situation which is being reported.
17.3.3.3 Intersemiotic Connections: Internal to the Section and to Other

Parts of the Webpage
The intersemiotic relations in this section are complex. Internal to the
section, the headings contextualize both the graph and the text. Without
the headings, the reader would not know what the graphs were about. The
dependency between the headings and the graphs is stronger than it is
between headings and text, or between text and graphs. Without the text
but with headings, the graphs can still be understood, but the meaning is
more specific (e.g. location) when combined with the text. The text is
intelligible without the graphs and the headings, but it structures idea-
tional meaning more concisely when the headings are added and even more
so when the graphs are added. That is, the graphs add more detailed infor-
mation, including the patterns of relations of Ebola cases across different
locations, which is not possible with text. In this regard, the whole section
forms an ‘intersemiotic package’, where the contribution of multiple semi-
otic resources (i.e. language, mathematical symbolism, and mathematical
images in the form of a line graph) allows the reader/viewer to understand
the situation more clearly and precisely than if only one resource had been
deployed.
17.3.4 Ebola Infographic: A Subsection of the Webpage

The Ebola infographic is a subsection of the ‘About Ebola’ section of the
webpage. The overall organization of the section replicates what is found in
other sections of the webpage. Only the infographic (enlarged in Figure 17.8)
is discussed here. In this regard, a number of points raised earlier will be
explored further. These relate principally to how textual and ideational
meaning are realized intersemiotically, and the reciprocal intersemiotic
connections that are established between the infographic and other parts
of the webpage.
17.3.4.1 Textual Meaning in the Infographic

In its composition the infographic could be thought of as an ‘infographic
complex’, where the five separate parts (SYMPTOMS, HOW TO PREVENT, HOW
IT SPREADS, EBOLA IS NOT AIRBORNE, and PEOPLE CAN SURVIVE EBOLA) are
concerned with different aspects of Ebola. Despite being separated from
Figure 17.8 Ebola infographic
each other by bold horizontal lines and by a straight blank space vertically
down the centre, the five parts of the infographic are unified by having
much the same general layout of a bold, large heading in the upper left,
smaller supporting text, and prominent, stylized images which use the
same colours and basic shapes. The position of the headings suggests a
reading/viewing path based on written English, where the headings serve
as Themes for their respective parts, and the accompanying text functions
as New to its respective Theme. The accompanying text ranges from a list of
symptoms in the top left part, two elliptical clauses in the top right part,
and text consisting of full clauses in the other three parts. Each image
resemiotizes some of the information in the written text, reinforcing and
highlighting its function as New information. The combination of linguistic
elements and images in the composition of the individual parts and in all of
the parts together creates a visual cohesion which assists the reader/viewer
in following the flow of information.
17.3.4.2 Ideational Meaning in the Infographic

Some of the features noted in the textual organization of the infographic
are also echoed in its ideational organization. Notably, Participants in the
main Processes in the language of the infographic are also encoded in the
stylized images. For example, in the top left part SYMPTOMS, the text lists
the entities headache, fever, and vomiting as co-hyponyms and as hyponyms of
Figure 17.9 Ebola symptoms resemiotized in images
Figure 17.10 Intersemiotic connections between the infographic and other parts of the
webpage
symptoms of Ebola. These symptoms are also resemiotized as stylized

images, which consist of generic participants who are taking part in a
Process (see Figure 17.9). This has the effect that in the process of
meaning-making we move between the material world of concrete partici-
pants and processes (as depicted in the stylized images) to the abstract
semiotic world constructed by language.
17.3.4.3 Intersemiotic Connections: Internal and to Other

Parts of the Webpage
In addition to the intersemiotic relations established within the info-
graphic, there are also connections between the infographic and other parts
of the webpage. For example, the stylized image displayed at the bottom
right in the infographic is similar ideationally and interpersonally to the
family in the photograph in the main visual display (see Figure 17.10, left).
In both we find three happy figures and the same gender balance, where the
adult male figure is largest. In both, the female figure is on the larger
male’s right and the smaller male figure is on the larger male’s left. In
the photograph all are looking directly at the viewer and smiling. In the
infographic, the figures appear to be jumping for joy, or at least they have
their arms upwards and outwards in symbolic happy gestures. They are also
facing the viewer. Even though they do not have faces, the white space
under their heads indicates that they are facing the viewer.
A further intersemiotic connection can be found where the image of
washing hands in the infographic connects to the opening still frame of
the video ‘Hand hygiene in Ebola care facilities’ in ‘Latest videos’ in the
‘News’ section of the webpage, which shows a woman washing her hands
Figure 17.11 ‘Get Involved’ section, with constituent parts outlined
(see Figure 17.10, right). The video itself contains sequences which illustrate
how to wash hands and how to hand-rub, which are also mentioned in the
text of the infographic. In this instance, the same action is resemiotized in
different media.
17.3.5 ‘Get Involved’: A Section of the Webpage

The screenshot in Figure 17.11 displays the ‘Promotional’ section, titled
‘Get Involved’. Beneath the heading, on the left and on the right, the two
smaller headings, ‘Funding’ and ‘Recruitment for Ebola’, head the two
subsections of the section. In the following discussion only the ‘Funding’
subsection is considered in detail. The main points arising from this section
concern how textual, ideational, and interpersonal resources are deployed
in this multi-semiotic text. Higher-level thematic organization and inter-
personal meaning are discussed in relation to the whole section.
17.3.5.1 Textual Meaning

At the level of the whole section, linguistic resources are used to organize
the flow of information. The heading ‘Get Involved’ functions as macro-
Theme for the section, while ‘Funding’ and ‘Recruitment for Ebola’ func-
tion as hyperThemes for the section and for the two subsections, which are
clearly distinct.
17.3.5.2 Interpersonal Meaning

The ‘Get Involved’ section is the only section of the whole webpage which
uses a speech function other than giving information. Here, the speech
function switches to a demand for goods and services (Halliday and Matthies-
sen 2014:135–9) realized through imperative Mood (‘Get Involved’). This
shift in speech function corresponds with a shift in purpose in this section,
giving rise to a major shift in the registerial selection for tenor. That is,
whereas the other sections of the webpage present information on Ebola,
this section is asking readers to either donate money or volunteer their
Figure 17.12 ‘Funding’ subsection
services. This shift in Mood choice is highlighted as it first occurs as

thematic for the whole section (‘Get Involved’), and it is reinforced in the
action buttons at the end of each subsection (‘View more’).
17.3.5.3 Textual Meaning in the Subsection ‘Funding’

The ‘Funding’ subsection consists of one multi-semiotic text, using lan-
guage, numbers, and a graph (see Figure 17.12). Thematically, the subsec-
tion picks up the hyperTheme of the subsection by having ‘Funding
requirements’ in first position in the nominal group under the heading.
The use of a nominal group (rather than a clause) creates an intersemiotic
construction for a Relational Identifying Process which has a linguistic
Participant (the ‘Funding requirements [for WHO Ebola response plan,
March–December 2015]’) which then connects to a visual Participant (the
graph). The point of departure is the funding requirements (in the text), and
the New information is contained in the graph. The New for the subsection
is the action button ‘View more’.
17.3.5.4 Interpersonal Meaning in the Subsection ‘Funding’

While the heading for the whole section is a demand for action and is in
imperative Mood, there is no explicit Mood choice in the subsection, apart
from imperative Mood in the action button at the end. The rest of the
language in the subsection consists of nominal groups only. The major
information source is the graph, which works to provide objective infor-
mation for the direct command ‘Get Involved’ in the heading.
17.3.5.5 Ideational Meaning in the Subsection ‘Funding’

The Relational Identifying Process construction implicitly relates the lin-
guistic participant (the ‘Funding requirements [for WHO Ebola response
plan, March–December 2015]’) to the graph. The graph shows the relations
between the amount which has been received and the amount which is
required as a dynamic process, as realized by the blurred line which shows
that funding is still being received. The visual impact of the graph, which
Figure 17.13 Screenshot of News section, with constituent parts outlined
objectively depicts the current funding scenario, reduces the need for
further appeals to the viewer, apart from the demands for goods and
services in the heading (‘Get Involved’) and the action button (‘View More’).
17.3.5.6 Summary of Intersemiosis in the Subsection ‘Funding’

The most striking aspect of intersemiosis in this section lies in the idea-
tional meaning realized by the graph, which simultaneously encodes mul-
tiple relationships (i.e. the money received and the money required in the
funding scheme). There also appears to be a semiotic division of labor in
this subsection, where most of the text-organizing and interpersonal work
is done by language and much of the ideational work is done by the graph.
17.3.6 ‘News’: A Section of the Webpage

The ‘News’ section is placed last in the sequence of sections in the webpage,
and contains three subsections: news stories, tweets, and latest videos (see
Figure 17.13, with constituent parts outlined). Each of these contains recent
information, most of which connects to other parts of the webpage, creat-
ing a web of multi-semiotic connections.
The news stories all consist of a link-heading which leads to the full story, a
dateline, and a lead sentence. Three of these news stories connect directly to
other sections of the webpage. The story headed ‘Ebola outbreak: Current
funding requirements’ connects directly to ‘Get Involved’, the immediately
preceding section (Figure 17.14, marked 1). The story headed ‘Despite Ebola
Figure 17.14 Intersemiosis on the Ebola webpage
vigilance and hope prevail in Forecariah’ repeats verbatim the headline in

the photograph-text complex discussed in Section 17.3.3 above (Figure 17.14,
marked 2). The story headed ‘Preparedness of countries to rapidly detect and
respond to Ebola exposure’ connects to the ‘Our Work’ section (Figure 17.14,
marked 3). The featured video ‘Hand hygiene in Ebola care facilities’ con-
nects to information in the infographic, while information in the info-
graphic connects to the photograph-text complex in the main visual
display (Figure 17.14, marked 4 and 5) as discussed in Section 17.3.4 above.
Lastly, the tweet displayed in the tweets-window shows part of the map re-
tweeted from the ‘Ebola data and statistics’ page of the Ebola website, and
functions to re-contextualize the place names in the graph and text in the
‘About Ebola’ section (Figure 17.14, marked 6). There are further connec-
tions to other sections of the Ebola webpage, largely through place names, as
a number of the image-text complexes in the main visual display are located
in places shown on the map and places mentioned in the text of this section.
There are, of course, other connections outside the ‘News’ section, but
this section seems to provide a focal point for intersemiotic connections on
the webpage. Its position as the final section of the webpage suggests that it
functions as macroNew for the webpage as a whole, although this function
is different from that of a macroNew in a written text. Martin and Rose
(2007) describe the function of macroNew as ‘distilling’ information. This
section does link back to other sections, but it also has the additional
function of providing a springboard to other and new (as in both most
recent and previously unseen) sources of information, in which case it
functions as macroTheme for the material to which it links.
17.4 Summary of Main Points
The discussion above shows the significance of SFT for the analysis and
interpretation of complex multi-semiotic texts. As the above analysis has
shown, certain aspects of systemic theory can be generalized across semi-
otic resources, while others are more problematic, as summarized below.
Ideationally, configurations of Participants, Processes, and Circum-
stances realized through language can be realized (albeit, in a different
form and substance) in images, and multi-semiotic configurations. Like-
wise, identifying and attributive relationships realized in language can be
realized in images (again, in a different form and substance) and, even more
efficiently, in mathematical tools such as graphs.
Logically, texts constructed from different semiotic resources can be
connected both to each other and as parts of a larger text. For example,
conjunctive relations like those found in language appear to also apply to
multi-semiotic texts, especially when a whole text is made up of or includes
parts which are constructed from different semiotic resources. The ‘Cur-
rent Situation’ section is a good example of this. Conjunctive relations can
also apply across texts, as is the case in the photograph-text complex
discussed in Section 17.3.2 above, although the application of systems of
conjunction to photographs is more problematic. In language, systems of
conjunction link events logically as they unfold in text. While this staging
of events can also happen visually (e.g. scientific diagrams, visual instruc-
tions), an image (typically) functions by situating happenings, actions, and
relations as parts of a whole, although visual choices make particular
relations more salient or prominent than others.
Ideationally, the combination of different semiotic resources and the
connections between and among them contribute to structuring thought
and reality in new ways, as illustrated in the preceding discussion. For
example, the potential to visually show participants engaged in multiple
processes simultaneously can be advantageous. A graph (or graphs), for
instance, can be used to show patterns of multiple (theoretically infinite)
relationships, which can be compared according to different parameters (in
this case, location, time, and number of cases). Language can then be used
to single out and discuss examples of those relationships considered to be
most important. Accompanying text can isolate which of those configur-
ations is most relevant to the context, to construct further information
beyond what is portrayed visually.
Compositionally, the overall text organization is achieved visually
through spatial layout and framing devices, accompanied by linguistic
headings for each section. Waves or layers of information, realized through
the concepts of Theme and New, are encoded in language. In an image,
while the viewer perceives the whole image at once, there are compos-
itional, ideational, and interpersonal elements which combine to make
some figures or episodes prominent, and which are designed to guide the
viewer through the image to create particular meanings.
In each of the instances discussed above, texts constructed from different
semiotic resources work together to both expand and constrain the range of
possible meanings. As Lemke (2005) explains, multimodal semiosis multi-
plies the semantic possibilities, from which certain selections are made in
order to constrain the possible meanings made within any one instance.
17.5 Conclusions and Future Directions
Following Halliday, multimodal semiosis is conceptualized as sets of inter-

related systems of meaning (i.e. the meaning potential) which are actualized
as semiotic processes and artefacts (i.e. the instance) that together constitute
society and culture. From this perspective, SF-MDA provides an encompass-
ing theoretical platform for conceptualizing multimodal semiosis and for
undertaking multimodal semiotic research. As the above discussion has
shown, multiple semiotic resources are utilized to make meanings which
work together to structure our understanding of the world. For example, the
reader/viewer learns about the Ebola outbreak from the text, photographs,
graphs, infographics, and videos, while hyperlinks provide further informa-
tion as required. Semiotic resources have their own internal organization
with unique systems for constructing ideational, interpersonal, and textual
meaning in different ways, but by working together, they provide an

extremely powerful apparatus for building increasingly complex semiotic
abstractions for structuring thought and reality about the world.
For instance, in the above examples, language is used to identify and
logically relate participants, processes, and circumstances in a sequential
fashion through social interactions; photographs are snapshots where mul-
tiple participants, processes, and parts are seen in relation to the whole, and
order is created through visual system choices which create particular
points of view; mathematical graphs encode perceivable patterns of rela-
tions over time; and infographics distil significant information using lan-
guage and image resources. By working together, further semiotic
abstractions are achieved, building abstraction upon abstraction by utiliz-
ing the semiotic capabilities of each resource. The result is other than the
sum of the parts, because the building of semiotic abstractions upon
abstractions results in expansions of meaning which are not possible using
a single resource. As Halliday claims, language and other semiotic resources
are tools for thinking, and by this account multimodal semiosis is key to
understanding the human condition in terms of experience, logical think-
ing, and interpersonal relations.
However, as the above discussion has shown, multimodal semiosis is
multi-faceted, multi-directional, and thus highly complex, making multi-
modal analysis a challenging and time-consuming task. For this reason,
purpose-built digital tools are required in order to undertake multimodal
semiotic analysis so that the results may be stored and retrieved for model-
ling, visualizing, and mapping of semiotic interactions, patterns, and
trends. These digital tools would need to equal, and if possible, move
beyond existing forms of multimodal semiosis, following the example of
science which managed to rewrite the physical world through the develop-
ment of mathematical symbolism, which integrated with language and
mathematical images to construct new views of the universe. In this
respect, it may be possible to handle the complexity of multimodal semiosis
using existing and possibly new semiotic tools, allowing multimodal ana-
lysts to base their theories, analysis, and interpretations on empirical evi-
dence, rather than surmising how multimodal semiosis takes place
through limited (manual) analyses. Such an approach requires interdiscip-
linary collaboration with the scientific community to address the complex
problem of understanding the abstract semiotic world which humans con-
struct using the range of semiotic tools at their disposal. This is one of the
key challenges in the digitally connected global world of today.
References
Baldry, A. P. and P. J. Thibault. 2006. Multimodal Transcription and Text

Analysis. Sheffield: Equinox.
Bateman, J. A. 2014a. Looking for What Counts in Film Analysis:

A Programme of Empirical Research. In D. Machin, ed., Visual Communi-
cation. Berlin: Mouton de Gruyter. 301–29.
Bateman, J. A. 2014b. Text and Image: A Critical Introduction to the Visual/Verbal
Divide. New York: Routledge.
Bateman, J. A. and K.-H. Schmidt. 2012. Multimodal Film Analysis: How Films
Mean. London: Routledge.
Christie, F. 2002. Classroom Discourse Analysis: A Functional Perspective. London:
Continuum.
Eggins, S. 1994. An Introduction to Systemic Functional Linguistics. London:
Pinter.
Halliday, M. A. K. 1978. Language as Social Semiotic. London: Edward Arnold.
Halliday, M. A. K. 1985. Spoken and Written Language. Waurn Ponds: Deakin
University Press.
London: Arnold.
Halliday, M. A. K. 2008. Complementarities in Language. Beijing: Commercial
Press.
Language in a Social-semiotic Perspective. Waurn Ponds: Deakin University.
Functional Grammar. London: Routledge.
Iedema, R. 2001. Resemiotization. Semiotica 137: 23–39.
Iedema, R. 2003. Multimodality, Resemiotization: Extending the Analysis of
Discourse as a Multisemiotic Practice. Visual Communication 2: 29–57.
Jewitt, C., ed. 2014. The Routledge Handbook of Multimodal Analysis. London:
Routledge.
Jewitt, C., J. Bezemer, and K. L. O’Halloran. 2015. Introducing Multimodality.
London: Routledge.
Koffka, K. 1935. Principles of Gestalt Psychology. London: Lund Humphries.
Kok, A. K. C. 2004. Multisemiotic Mediation in Hypertext. In K. L. O’Hal-
loran, ed., Multimodal Discourse Analysis: Systemic Functional Perspectives.
Kress, G. and T. van Leeuwen. 2001. Multimodal Discourse: The Modes and
Media of Contemporary Communication Discourse. London: Arnold.
Design. 2nd ed. London: Routledge.
Lemke, J. 2005. Multiplying Meaning. In J. R. Martin and R. Veel, eds.,
Reading Science: Critical and Functional Perspectives on Discourses of Science.
New York: Routledge. 87–114.
Liu, Y. and K. L. O’Halloran. 2009. Intersemiotic Texture: Analyzing Cohe-
sive Devices between Language and Images. Social Semiotics 19: 367–88.
Machin, D. 2007. Introduction to Multimodal Analysis. London: Hodder Arnold..
Machin, D. and A. Mayr. 2012. How to Do Critical Discourse Analysis:
A Multimodal Introduction. London: SAGE.

Benjamins.
Martin, J. R. 2002. Meaning beyond the Clause: SFL Perspectives. Annual
Review of Applied Linguistics 22: 52–74.
Martinec, R. 1998. Cohesion in Action. Semiotica 120: 161–80.
Martinec, R. and A. Salway. 2005. A System for Image–Text Relations in
New (and Old) Media. Visual Communication 4: 337–71.
O’Halloran, K. L., ed. 2004. Multimodal Discourse Analysis: Systemic Functional
Perspectives. London: Continuum.
O’Halloran, K. L. 2008a. Systemic Functional-Multimodal Discourse Analy-
sis SF-MDA: Constructing Ideational Meaning Using Language and Visual
Imagery. Visual Communication 7: 443–75.
O’Halloran, K. L. 2008b. Inter-semiotic Expansion of Experiential Meaning:
Hierarchical Scales and Metaphor in Mathematics Discourse. In C. Jones
and E. Ventola, eds., From Language to Multimodality: New Developments in the
Study of Ideational Meaning. Sheffield: Equinox. 231–54.
O’Halloran, K. L. 2011. Multimodal Discourse Analysis. In K. Hyland and
B. Paltridge, eds., Bloomsbury Companion to Discourse Analysis. London:
O’Halloran, K. L. and V. Lim-Fei. 2014. Systemic Functional Multimodal
Discourse Analysis. In S. Norris and C. D. Maier, eds., Texts, Images and
Interactions: A Reader in Multimodality. Berlin: Mouton de Gruyter. 137–53.
O’Toole, M. 2011. The Language of Displayed Art. 2nd ed. London: Routledge.
Royce, T. 2007. Intersemiotic Complementarity: A Framework for Multi-
modal Discourse Analysis. In T. Royce and W. L. Bowcher, eds., New
Directions in the Analysis of Multimodal Discourse. Mahwah: Lawrence
Erlbaum Associates. 63–110.
Royce, T. 2015. Intersemiotic Complementarity in Legal Cartoons: An Idea-
tional Multimodal Analysis. International Journal for the Semiotics of Law –
Revue internationale de Sémiotique juridique 28(4): 719–44.
Sharoff, S. 2010. In the Garden and in the Jungle. In A. Mehler, S. Sharoff,
and M. Santini, eds., Genres on the Web: Computational Models and Empirical
Studies. Dordrecht: Springer. 149–66.
Unsworth, L. 2008. Multiliteracies and Metalanguage: Describing Image/
Text Relations as a Resource for Negotiating Multimodal Texts. In
J. Coiro, M. Knobel, C. Lankshear, and D. J. Leu, eds., Handbook of Research
on New Literacies. Oxford: Taylor and Francis. 377–405.
van Leeuwen, T. 1999. Speech, Music, Sound. Houndsmills: Macmillan Press.
van Leeuwen, T. 2005. Introducing Social Semiotics. London: Routledge.
van Leeuwen, T. 2012. The Critical Analysis of Musical Discourse. Critical
Discourse Studies 9: 319–28.
18
SFL and Critical Discourse
Analysis
Gerard O’Grady
18.1 Introduction
It is no accident that critical linguistics and social semiotics

arose out of SFL or that other work in CDA has drawn upon
it – SFL theorises language in a way which harmonises far more
with the perspective of critical social science than other theories
of language. (Chouliaraki and Fairclough 1999:139)
This chapter looks at the relationship between Systemic Functional Linguis-

tics (SFL) and Critical Discourse Analysis (CDA). Critical Discourse Analysis
is actually a cover term for a transdisciplinary research project which
investigates how power relations in society are established, maintained,
and reinforced by semiotic practice. The approach to CDA I focus on in this
chapter is that presented in Fairclough’s seminal work Language and Power.
I choose to focus on this particular model of CDA as it is the one which most
explicitly illustrates how linguistics can be incorporated into a CDA
framework. By comparison, the dialectal-relational approach found in Fair-
clough (2009) foregrounds the social science aspect of CDA and conse-
quently backgrounds linguistics. Similarly, because their work is further
removed from SFL, I focus less on the work of other significant CDA
scholars, such as Ruth Wodak, Teun van Dijk, and especially Paul Chilton.
Furthermore, Fairclough has recently published the third edition of Lan-
guage and Power, in which the position of language in the approach is once
more made central.
In this chapter, I will briefly summarize and exemplify why SFL emerged
as the linguistic model of choice for many CDA practitioners and their
antecedents in critical linguistics. Then I will specifically focus on the role
I’d like to express my gratitude to the editors, and especially Geoff Thompson, for their thoughts, queries, and corrections.
Downloaded from https://www.cambridge.org/core. Columbia University - Law Library, on 11 Aug 2019 at 09:01:33, subject to the Cambridge Core terms of use,
SFL and Critical Discourse Analysis 463
of SFL in Fairclough’s approach. Next, I will address criticisms of SFL as a

theoretical model from within the CDA community and criticisms of SFL-
flavoured CDA. Finally, I will provide some examples illustrating that SFL
remains a powerful tool in the CDA toolbox.
18.2 Antecedents of CDA and SFL
Critical Discourse Analysis is a multidisciplinary approach to the study of

discourse. It encompasses scholars from a range of disciplines who study
language as a social practice. Critical Discourse Analysis itself emerged in
the late 1970s originally under the title ‘critical linguistics’, which was a
revolutionary attempt to show that linguistic analysis could provide a
rigorous critical account of discourse (Kress and Hodge 1979; Fowler
1991). The then dominant form of linguistics, transformational grammar,
was solely interested in studying language competence in idealized
speakers and in how language is acquired by human infants. Actual
instances of language use were classed as performance and discounted by
the theory. By contrast, SFL, which argues against the study of language as
an autonomous object, proved to be eminently suitable as a theory in which
to ground critical linguistic study. Prior to illustrating how critical linguis-
tic scholars employed SFL in their work, the following three paragraphs
illustrate early SFL and SFL-inspired work which showed its potential for
developing a methodology capable of rigorously analyzing discourse as a
social practice.
Hodge (2017) notes that, even prior to critical linguistics, SFL had already
demonstrated that it had the tools to systematically analyze extended
discourse. For instance, Halliday (1973) is a detailed stylistic investigation
of William Golding’s novel The Inheritors, which shows that a recurring
grammatical feature of the text was that Neanderthals, unlike their
modern human competitors, did not have the option in their grammar of
choosing material processes with Goals. Halliday noted that their language
as represented in the novel had evolved as it had because they had no
purpose that required transitive material processes. While The Inheritors is
a work of fiction, Halliday’s fine-grained text analysis of the ideational
metafunction demonstrated that grammatical analysis had the potential
to reveal how linguistic forms reflected and constrained social action.
At around the same time Bernstein was engaged in his sociological
research into the disparity in educational achievement between working-
and middle-class children. Bernstein used SFL, specifically Halliday (1969),
to consider differences in the types of socialization associated with each
group in terms of the kind of language code each group of speakers had
access to (Bernstein 1971:12). The two codes are the restricted code and the
elaborated code. The restricted code relies on interlocutors sharing assump-
tions and knowledge about society and their place within it. Use of the
464 GERARD O’GRADY
restricted code reflects and facilitates a sense of belonging to a particular

group, but conversely restricts access to out-groups (Bernstein 1971:114).
The elaborated code is more explicit and does not rely on in-group infer-
ences and insider knowledge. Instead, it orients towards context-
independent meanings (Bernstein 1971:136). It is identified by the relative
complexity of syntax and frequency of conjunctions, adverbs, and adjec-
tives (Bernstein 1971:31).1 Working-class speakers were more likely to rely
on the restricted code to express their meanings because of their unfamili-
arity with and lack of access to the elaborated code. Middle-class speakers
had access to both kinds of codes. Like Halliday (1973) the claim is that
linguistic resources construe and constrain social action, in this case
achievement in education: those socialized into speaking situations which
required the use of explicit meanings were advantaged in their learning of
subjects that required explicit reasoning such as maths and consequently
achieved higher grades.
Bernstein’s views have remained highly influential within SFL. For
instance, Hasan’s (2009a) longitudinal study revealed significant differences
in parent–child interactions between families designated high autonomous
professionals (HAP) and ones designated low autonomous professionals
(LAP). Like Bernstein she attributed differences in educational achievement
to differences in socialization and concomitantly communicative orienta-
tion. Hasan’s analysis of her data differed from Bernstein’s earlier work in
that she focused on semantic variation in meaning across all three meta-
functions and did not restrict her study to representational meaning.
A similarly fruitful continuing interaction between Bernstein’s sociological
approach to language and SFL is seen in Legitimation Code Theory (see
Maton 2015).
As noted above, critical linguistic studies focused on the forms of lan-
guage and can be classed as text analysis. In other words, critical linguistics
did not examine the social context in which the discourse was produced nor
try to explain how the discourse influenced the world. But, that said, SFL-
influenced critical linguistics work such as that of Fowler (1991) provided
an excellent analysis of how newspaper language mediated reality. He
conclusively illustrated that the portrayal of events as diverse as the Ameri-
can bombing of Tripoli and the salmonella in eggs scandal were the result of
ideological choices. He focused on a number of linguistic devices, chiefly
transitivity, modality, passivation, and referring expressions. The following
examples (1) to (3) illustrate how these linguistic devices can be used to
mediate news.
1
Bernstein’s claim that code correlates with social class has been considered by some to be controversial. See Jones
(2013) for a discussion of the pros and cons of Bernstein’s approach.
(1) Eight people have been killed in a shooting in a restaurant in

Uhersky Brod, according to a Czech news agency.2
In example (1), the lead sentence of a news article, the text producer has
selected passive voice and so for whatever reason obscured the identity of
the killers. The use of the verb killed carries less negative prosody than
synonymous verbs such as murdered or slaughtered would have. After all,
we speak of the killing of, though not the slaughtering or murdering of,
cells, bacteria, and trees. Finally, the information is presented as hearsay
and not as an evidential fact. The victims are presented as the Goal of the
killing without any Interpersonal evaluation. An alternative representation
is, Named participant murders eight innocent people in a shooting etc.
(2) Syriza’s victory in the Greek elections at the end of January gave
hope to some that the eurozone would change its economic policy.
But a month and many hours of painful diplomatic arm-wrestling
later, those hopes have clearly been dashed.
Even if Friday’s agreement between Greece and its lenders is
approved by eurozone ministers on Tuesday, it decides very little
apart from ensuring that the next four months will be a battle of
attrition. The eurozone will look to keep the new government in
Athens in check, and the Syriza-led coalition will try to eke out fiscal
space for some of the policies it has promised.
Politically, the four-month extension will be a challenge rather
than a blessing for the government. Prime minister Alexis Tsipras
will have to ditch plans to increase low-income pensions and other
similar measures that would affect the country’s fiscal balance.3
In example (2), which is the opening of an opinion piece, the text producer’s
linguistic selections represent the newly elected Greek government as
essentially powerless. Their election is represented as giving [false] hope.
The identity and number of recipients of the hope is left unspecified. The
details of Syriza’s election victory are backgrounded. An alternate repre-
sentation could have been, The greatest number of Greek adults voted for Syriza.
In the second clause, the text producer states that these hopes have clearly been
dashed. The text producer’s selection of the passive voice obscures the
identity of the unwritten Actor of the clause, and more importantly why
and how the Actor dashed those hopes. The use of the modal adjunct clearly,
which represents the author’s summation of the likelihood of the propos-
ition being true, adds to the representation of the Greek government as
powerless. In the second paragraph lexical parallelism of the processes will
look to keep . . . in check and will try to eke out further emphasizes the
2
www.theguardian.com/world/2015/feb/24/czech-republic-restaurant-shooting-multiple-deaths-reported. (Last
accessed 24/02/2015.)
3
www.theguardian.com/commentisfree/2015/feb/24/greece-syriza-victory-euphoria-gone-reforms. (Last accessed
28/02/2015.)
powerlessness of the Greek government, especially in contrast to the euro-

zone – though of course what or who exactly the eurozone refers to is itself
unclear. The second clause of the third paragraph reinforces the powerless-
ness of the government; the Prime minister is presented as an Actor who
will have to ditch plans rather than one who will achieve success. The effect of
his powerlessness is reinforced by the modal will, which construes a high
likelihood. A potential re-writing which would have presented Tsipras as a
more powerful dynamic figure would have been, Prime minister Alexis Tsipras
is reworking plans that would have increased low-income pensions in order not to
negatively affect the country’s fiscal balance.4
(3) Miss Royal showed ‘that she has absolutely no concrete solutions to
respond to the problems of the French people’.5
Fowler (1991: 85) notes that newspapers employ over-lexicalization, which he

defines as ‘the existence of an excess of quasi-synonymous terms for
entities and ideas that are a particular preoccupation or problem in the
culture’s discourse’. In this specific example, originally discussed in
O’Grady (2011), the issue was the Telegraph’s perceived incongruity
between Royal’s political role and her non-marital status and physical
appearance. Royal was at the time in a long-term relationship with a fellow
politician with whom she had four children. O’Grady (2011:2496) found
that the Daily Telegraph and Sunday Telegraph over a period of the year
leading up to the French presidential election in 2007 referred to Royal on
383 occasions. She was referred to as Miss Royal on 164 occasions. However,
on the other occasions, she was referred to by a range of names such as
Segolene Royal, Royal, Ms. Royal, Mrs Royal, Segolene, Sego, and by a series
of other terms such as the Socialist candidate, the local heroine, the well-
groomed media darling, the fifty-two-year-old mother of four, the socialist
in stilettos, and the new pin-up girl of the left. The Telegraph’s use of over-
lexicalization reveals that it had difficulty in accommodating itself to the
fact that a glamorous unmarried mother of four could be a political
heavyweight.
18.3 Critical Discourse Analysis and SFL
Critical linguistics using SFL as a tool to explicate the representation of bias

in media discourse was very successful. But, as noted above, it neither gave
an account of the context in which the media texts were produced, nor was
4
An objection raised against the efficacy of CDA is that it is incapable of describing what was not said or written. But in
this case the text producer’s focus on the negative impact on the country’s fiscal balance rather than on the economic
and physical health of its citizens is in and of itself highly revealing of the author’s ideology.
5
www.telegraph.co.uk/news/worldnews/1543398/Royals-spirits-raised-by-record-for-TV-debate.html. (Last accessed
24/02/2015.)
it able to account for the effect of language upon society. Consequently, the
critical linguistics framework was merged with social and critical theory
and re-emerged as CDA. Slembrouck (2001:35)6 notes the influence on the
formation of CDA of the work of Stuart Hall and his colleagues at the
Birmingham Centre for Contemporary Cultural Studies, who engaged in
influential critical analyses of the social, economic, and political changes
caused by the emergence of the Thatcherite consensus. Hall and his col-
leagues gave a central place to the close study of symbolic practices as a
means for understanding how social relationships were transformed in
everyday social practice. This, as Slembrouck notes, not only accorded well
with the close linguistic analyses practised by the critical linguists, but also
allowed CDA theorists to develop a theoretical approach based on social
theory aimed at explaining the relationship between linguistic practice and
changes in the world.
The publication of Fairclough’s (1989) seminal book Language and Power is
considered to mark the birth of the CDA programme (Blommaert 2005:23).
The following paragraphs will sketch the growth of CDA as an academic
discipline, though I will naturally slant the discussion towards CDA work
which has been more overtly influenced by SFL and illustrate how SFL has
been used to advance the CDA programme. I will postpone discussion of the
various criticisms which have been levelled against SFL-flavoured CDA until
the next section.
Fairclough (1989, 2015) identified three dimensions of discourse. The
first is discourse as text or product. This dimension refers to the formal
linguistic features of the text such as wordings, transitivity choices, modal-
ity, cohesion, and text structure. Readers will have noted that it was this
first dimension of discourse which the critical linguists focused on. The
second dimension is discourse as discursive practice. This dimension refers to
the interactive nature of discourse and describes how discourse is pro-
duced, circulated, and consumed. Analysts examining discourse in this
dimension examine the aspects of the text that link it to its wider social
context, such as speech acts, intertextuality, and coherence. The final
dimension is discourse as social practice. This dimension refers to the ideo-
logical effects and hegemonic practices pre-existing in the context in which
the discourse is produced. For instance, the prevailing ideology may nor-
malize a text as common sense or label it as outside the norm.
Fairclough (1989:26, 2015:58–9) proposed a three-stage methodology to
enable CDA to account for the three dimensions of discourse. Firstly,
analysts must describe discourse in terms of its formal properties. While
engaged in the description of discourse, analysts must not only adopt the
participants’ perspective but also attempt to make their description expli-
cit. Secondly, they must interpret the discourse as interaction in order to
6
It should be noted that Slembrouck’s account of the birth of CDA is slanted heavily towards Fairclough’s approach.
arrive at an understanding of the discourse. Interpretation requires a

degree of distancing between analyst and participants, but the interpret-
ation relies on making explicit the participants’ behaviour. The final stage
is explanation, where analysts draw on social theory in order to explicate the
ideological basis of the interpretative stage. Social theory distances the
analyst from the discourse and in Fairclough’s view transforms discourse
analysis into critical discourse analysis.
The relevance of SFL to the first stage is obvious – see examples (1) to (3) –
but I argue that SFL is equally relevant to the second stage as well, and that
it is theoretically compatible with the third stage. Halliday (1978), drawing
upon earlier work – e.g. that of Michael Gregory, whose views are perhaps
best expressed in Gregory (1988) – used the triad of Field, Tenor, and Mode to
define the context of a text. Field is the nature of the event that is taking
place. Tenor refers to the nature of the participants and their roles and
statuses. Mode is the symbolic organization of language and refers to the
role language is playing in the discourse.
Leckie-Tarry (1995) presents a useful operationalization of the terms. She
breaks down Field into how institutionalized the setting is; who the partici-
pants are based upon – their race, gender, age, appearance, their cultural
knowledge and level of education; and how specialized the semantic
domain is. Tenor refers to how formal the interaction is; how powerful
the social roles occupied by the participants are, and whether the focus of
the interaction is on the interpersonal relationship between the
participants or the ideational content of the message. Mode is operational-
ized in terms of how planned the discourse is; how receptive the speaker is
to listener feedback;7 whether the language is written or spoken;8 and how
embedded the language event is in the context of situation. Leckie-Tarry
(1995) presented the oppositions above in terms of a cline rather than
discrete choices. However, there seems little reason to doubt that a large-
scale investigation using her categories would allow an analyst to draw
networks showing the probabilities of discrete choices in various genres.
Example (4) reprints a longer version of the text used in example (3). I use
it to illustrate how SFL’s contextual parameters can be usefully employed to
interpret the newspaper opinion piece as interactive social practice. Prior to
so doing, however, I will briefly sketch Fairclough’s own interpretative
procedures in order to demonstrate that they are fully congruent with an
analysis couched explicitly in terms of Field, Tenor, and Mode, as presented
by Leckie-Tarry (1995).
7
The potential for feedback is clearly related to the power relations existing between the participants and thus their roles
in the discourse.
8
It is not entirely clear to me whether in an increasingly digital world, where large amounts of the day are spent on social
media, the distinction between written and spoken forms of the language is an entirely accurate way of capturing the
distinction between language intended to have a less permanent inscription from one which was intended to have a
more permanent inscription.
(4) Syriza’s victory in the Greek elections at the end of January gave
hope to some that the eurozone would change its economic policy.
But a month and many hours of painful diplomatic arm-wrestling
later, those hopes have clearly been dashed.
Even if Friday’s agreement between Greece and its lenders is
approved by eurozone ministers on Tuesday, it decides very little
apart from ensuring that the next four months will be a battle of
attrition. The eurozone will look to keep the new government in
Athens in check, and the Syriza-led coalition will try to eke out fiscal
space for some of the policies it has promised.
Politically, the four-month extension will be a challenge rather
than a blessing for the government. Prime minister Alexis Tsipras
will have to ditch plans to increase low-income pensions and other
similar measures that would affect the country’s fiscal balance.
Some on the left wing of the party clearly find this difficult to
swallow. Veteran leftist MEP Manolis Glezos has already likened
the party leadership’s achievements as ‘renaming meat fish’ and
apologised for contributing to the ‘illusion’ that Syriza would change
anything. His comments were politely dismissed by key figures in
the party, and it is clear that there is a damage-limitation exercise in
operation in the wake of the agreement.
Fairclough (1989:140, 2015:154) states that when interpreting text an analyst

‘cannot directly extrapolate from the formal features of a text to these struc-
tural effects upon the constitution of a society’. Text is interpreted by partici-
pants against the set of background assumptions operative in their society. The
formal grammatical features are cues which help to generate meaning poten-
tials in institutional contexts which reflect the hegemonic social order. The
discourses themselves have histories and may have been recontextualized
(Chouliaraki and Fairclough 1999; Fairclough 2006). In example (4), terms such
as fiscal space, low-income pensions, and fiscal balance have been recontextualised
from the technical economics literature. There will be a further brief discus-
sion of intertextuality and presuppositions below. Fairclough (1989:146–52,
2015:159–64) states that, in order to interpret the context in which a discourse
is produced, the analyst should ask the four questions listed below.
• What’s going on in the discourse?

• Who’s involved in the discourse?
• What relations exist between the participants?
• What is the role of language?
The first question probes the Field, the following two the Tenor, and the
final one the Mode.
Example (4): Analyzed for Field

The text is a newspaper comment piece written in the online opinion section
of the Guardian Unlimited website. It is an institutionalized text in that it
conforms to the website’s style sheet. It complies with the expected word
count. The writer, Nick Malkoutzis, is presented as the possessor of specific
cultural knowledge, in this case an account of whether the newly elected
Greek government has honoured its electoral pledges and whether its
actions will benefit or harm the Greek people. The semantic domain is
specialized, and the author presents information from the fields of politics
and economics.
Example (4): Analyzed for Tenor

The writing is formal and the discourse pre-planned. A reader can discover
by clicking on the author’s name that Nick Malkoutzis is the deputy editor of
the English edition of the Greek newspaper Kathimerini English Edition.9 The
focus of the text is ideational. The writer does not overtly develop an inter-
personal relationship with his readers. His use of mostly unmodalized lan-
guage coupled with his use of the ahistorical present tense construes his
thesis as a categorical truth. Mr Malkoutzis is an expert producing a mono-
glossic text. This is somewhat surprising as his piece appears in a section of
the website that invites comments below the line.
Example (4): Analyzed for Mode

The text is written and edited. As noted, while the text is an opinion piece
and as such the author is notionally receptive to feedback, it is clear that the
only feedback welcome or sought is agreement. The text as a newspaper
article constitutes the language event. It construes the situation where the
reader is invited to interact with its construal of the actions of the Greek
government and their likely effects on the Greek people.
Summary
Mr Malkoutzis, while notionally presenting his views as opinion, is in fact
construing a world where his view is the authoritative truth. He is the
expert, and the views he represents are unarguable.
Texts and discourses are not created in isolation. They exist in a dialogic
chain with both previously produced texts and potentially produced future
texts (Voloshinov 1973). In order to interpret the meaning of the text, one
must decide to which discursive chain a text belongs, and therefore what is
presupposed as common ground between the writer and the reader (Fair-
clough 1989:152, 2015:164). Assertions in the example text rely on presup-
positions such as increasing public spending is bad and it is sensible to balance
national budgets. While these project an assumption that readers already
know this, it is impossible for a writer in the mass media to know what his/
her individual reader’s intertextual experiences are. Thus, what is pre-
sented are the presumed intertextual experiences of an ideal reader. This
Fairclough (1989:152, 2015:164) reminds us is a powerful weapon in the
9
The reader is not informed that Kathimerini is a highly partisan Conservative newspaper which keenly supports the
former right-wing governing party New Democracy.
armoury of the mass media. They are free to present contestable or dis-
puted propositions as uncontested or leave them unstated as presupposed
backgrounded common sense which underlie their assertions.
Presuppositions do not of course exist in the text but rather are created by
the readers’ interactions with the intertextual context. As such, prime facie,
SFL might not appear the best tool to unpack pragmatic meaning. However,
Martin and White’s (2005: Chapter 3) Appraisal framework provides a valu-
able social-dialogic framework which can be used to explicate whether or not
alternative positions are acknowledged in texts. As noted in the discussion of
Tenor, Mr Malkoutzis’ discourse is predominantly monoglossic. Compare the
opening clause of the third paragraph Politically, the four-month extension will be
a challenge rather than a blessing for the government with possible dialogic
alternatives Commentators expect/predict/believe/speculate that the four-month exten-
sion will be a challenge rather than a blessing for the government.
However, Mr Malkoutzis intervenes in the text on three occasions, dem-
onstrating his stance through the use of the modal clearly on two occasions
and the modalized metaphor it is clear on one occasion. Halliday and
Matthiessen (2014:190) label clearly as a comment adjunct which occurs in
declarative mood. In their network it is an asseverative subclass of obvious
and functions to assert that it is so. Martin and White’s (2005) framework
labels Mr Malkoutzis’ modality choices as proclaim: pronounce which is a
contracting strategy designed to project a claim as self-evident.
This can be seen particularly distinctly in the following heteroglossic prop-
osition: Some on the left wing of the party clearly find this difficult to swallow. This
example, while notionally up for discussion, is immediately followed by an
illustrative example of a left-wing member who is quoted as apologizing for his
part in the creation of the false hope. Veteran leftist MEP Manolis Glezos10 has
already likened the party leadership’s achievements as ‘renaming meat fish’ and apolo-
gised for contributing to the ‘illusion’ that Syriza would change anything. As such the
text construes a reader who may not necessarily share Mr Malkoutzis’ evalu-
ation of Syriza – one who is perhaps undecided and in need of further evidence.
The effect of the further evidence is to support the contraction of the dialogue
and present Mr Malkoutzis’ view as incontestable common sense.
One further stage remains in Fairclough’s methodology, namely, explan-
ation. The objective of this stage is to portray discourse, itself a social
practice, as part of a social process, illustrating how it is determined by
social structures and explicating how the reproduction of discourses
reinforces or weakens these structures (Fairclough 1989:163, 2015:172).
This stage naturally draws much less from linguistic theory and much
more from social theory and other relevant disciplines – see Fairclough’s
(2009:163) point that an effective CDA must be transdisciplinary. Dis-
courses are examined as part of a social struggle and/or contextualized in
10
Manolis Glezos is a noted World War II resistance fighter famed for his part in removing the swastika from the Acropolis
during the German occupation of Greece.
terms of their effects on social structures. They can be examined as a

situational, institutional, or societal practice. For instance, an analyst could
use example (4) to examine either the ideological view of the Guardian
Unlimited website or the hegemonic growth of neo-liberal discourse and its
effect on society.
Fairclough in his studies has drawn upon a myriad of theories to achieve
his objective of explaining discourse. These have included proponents of
social theory such as Foucault and Habermas; Marxists, for instance,
Gramsci and Harvey; sociologists such as Bourdieu, Bernstein, Giddens,
and Jessup; and political theorists, e.g. Laclau and Mouffe.11 In the next
section, I will sketch SFL’s compatibility with these various theories when
I discuss criticisms of SFL-inspired CDA.
In the meantime, we can note that at an institutional level example (4)
reproduces the neo-liberal economic discourse associated with the Chicago
School of Economics. It treats as obvious the notion that a healthy national
economy is one that balances the books and reduces the role of the state to a
minimum. Alternative economic discourses such as the Keynesian school are
unrepresented in the text. The effect of this is to sustain the prevailing new-
right hegemony that came to power with the election of Margaret Thatcher.
The Eurozone crisis is presumed to have been caused by reckless and feckless
public spending. Alternative explanations such as problems arising from the
spread of toxic financial instruments, the inbuilt disadvantages of currency
union to manufacturers from regions with formerly weak currencies, and
the failure of the Eurozone to incorporate a formal mechanism for recycling
some financial surplus from creditor regions to debtor ones (see Varoufakis
2013) are excluded. An archaeological approach along the lines employed by
Foucault in his exposition of the ontogenesis of the asylum has the potential
to illustrate the role discourse plays in normalizing changes in social prac-
tices. Foucault (1981) observed that the reproduction of discourse is con-
trolled in order to minimize challenges to the existing power structure. He
notes that one of the ways discourse is controlled is through ‘exclusion’ and
specifically the ‘rarefaction’ of speaking subjects (Foucault 1981:61). This
refers to the fact that not all aspects of a discourse are equally transparent
and open to each potential speaking subject.12 Subjects are positioned by
practice, and hence their ability to contract or expand a discourse is deter-
mined by the prior reproduction of the discourse.
11
The observant reader will have noted that all of these scholars’ thinking has to a greater or lesser extent been
influenced by Marxist thought and a view that knowledge and power are formed discursively. I will argue in the
following sections that it is precisely this Marxist orientation that makes SFL congruent with these various approaches.
Fairclough and Graham (2010:340) classify Marx as a discourse theorist based on their reading of his body of work as
one which contains a discursive view of language as an element of social life.
12
A very similar argument could be couched in terms of symbolic capital and its inscribed effect on the habitus
(Bourdieu 1991). But for present purposes the details of the argument are not of relevance. What is of relevance is to
illustrate how critical theory can be employed alongside detailed linguistic description in analyzing a text and in
showing how it fits into a chain of discourse in order to trace its effect on social practice.
To conclude this section, we have seen that explanation rests upon solid
linguistic description and analysis, and that for CDA to be truly effective it
must relate the semiotic to social practices and social practices to societal
structures. In order to do this, an analyst must engage in transdisciplinary
work, fusing linguistics with other relevant approaches. In short, CDA must
not be exclusively a semiotic approach.
18.4 Criticisms of SFL-inspired CDA
In this section, I will focus on addressing criticisms of SFL-flavoured CDA

and not critiques of CDA methodology and theory except where they have
specific resonance for SFL. Widdowson (1995, 1998) criticized CDA for
producing biased readings of discourse. He argued that the bias arose
because CDA conflates semantics and pragmatics and is unable to accom-
modate a range of reading positions. Texts are found to have particular
biases, and these are then assumed to represent the meanings gleaned by
the ordinary consumers of the text. Furthermore, he argued that reliance
on social theory leads to fuzziness and confusion in how discourse is
analyzed. His final objection is that CDA theorists are biased by the choice
of data which supports their preconceived ideas.
I will deal with these criticisms in reverse order. Fairclough (2009:167)
presents a four-step methodology for carrying out CDA as part of a
transdisciplinary project premised upon analysts identifying a social prob-
lem and then selecting relevant texts for analysis. While this approach
could result in bias, it does not necessarily entail it. An analyst who pro-
duces a rigorous hypothesis coupled with careful and extensive corpus-
building would produce relatively unbiased results. It is worth remember-
ing that corpus methods have improved hugely since Widdowson’s cri-
tique. Today an analyst with a solid hypothesis can build an extensive and
encompassing corpus in order to produce replicable and valid findings: see,
for instance, Kriszan (2011), a careful and extensive SFL-framed corpus-
based study of the discursive practices of key politicians in three non-
central members of the EU. In short, bias is a methodological concern but
not necessarily a design flaw in the CDA project.
Widdowson’s second objection is that CDA and especially SFL-flavoured
CDA claims that the function of a text can be deduced from its forms
without reference to the social conditions in which it was produced. How-
ever, Chouliaraki and Fairclough (1999:67) explain that this critique is
misdirected, as interpretation is ‘a multi-layered process’ based upon gram-
matical description and the formal explication of the participants’ under-
standing of the discourse. This is based upon an idealization of the
participants’ presuppositions (members’ resources). The use of the apostro-
phe entails that these are shared resources and there is no implication that
different individuals will share the resources equally. Critical Discourse
Analysis does not, in other words, make predictions about how individuals
will respond to a text. The explanation stage goes further and shows how
different positions emerge historically.
O’Halloran (2003), using insights from Relevance Theory (Sperber and
Wilson 1995), proposes an IR (idealized reader) analysis of news text. This is
potentially a fruitful first step in examining how individual readers interact
with a text by examining differences in terms of reading effort, cognitive
bias, and previous experience with the topic. But, the key point for SFL-
flavoured CDA is that such an interpretation should be grounded in rigor-
ous linguistic description.
The SFL view of context has been robustly criticized by van Dijk (2008).
He argues that it is too linguistic, not cognitive enough, and based on a
limited social theory of language. In the paragraphs that follow, I will
address these issues in turn. As SFL grew out of a sentence grammar
approach, namely, scale and category grammar, it is unsurprising that it
heavily focused on the clause as the unit of analysis. And, indeed, if one
wishes to incorporate a rigorous descriptive linguistic analysis into CDA, a
clause-based grammar, especially for pre-planned formal texts, is a power-
ful device. This of course does not entail that CDA grammatical description
must necessarily restrict itself to the level of the clause.
Systemic Functional Linguistics as a social semiotic theory does not focus
on the individual speaker. Halliday, criticizing the individualist philosophy
underpinning much of linguistics, commented: ‘Creating language and
creating through language, are essentially interactive processes; they can
never take place inside one individual’s skin’ (Halliday 2007:56). This does
not presume that CDA or indeed other forms of language study could not
usefully be informed by insights from the psychological literature. Systemic
Functional Linguistics itself has usefully incorporated Vygotsky into studies
of child language development, e.g. Halliday (1975) and Hasan (2005). In
short, van Dijk seems not to recognize that the issue is not whether SFL
needs to develop a cognitive dimension, but rather whether it, as a semiotic
theory, is compatible with cognitive approaches.
Van Dijk (2008:38) states that, despite what he claims to be the anti-
mentalist view of context inherited by SFL from Firth and Malinowski,
Halliday’s description implies cognitive notions (Halliday 1978). It is indeed
true that Halliday’s view implies cognition, but van Dijk is mistaken in
labelling SFL as anti-mentalist. Thibault (2011) is an SFL description of
semiosis as a dynamic biocultural process distributed across brains, bodies,
and aspects of the social and cultural world. He shows that while semiotic
processes are grounded in the ‘signifying body’, semiosis cannot be reduced
to bodily processes (Thibault 2011:52). A social semiotic approach is entirely
compatible with the theory of distributed cognition (Hutchins 1995).13
13
Fawcett (1980) and the Cardiff Grammar in general is a far more individual speaker/hearer centred version of SFG.
But as I know of no CDA work grounded in the Cardiff Grammar I shall not mention it further.
Distributed cognition is a theory that was developed in the 1980s by Edwin

Hutchins which blends concepts from anthropology with Vygotskian
psychology (Hutchins 1995). Its basic premise is that cognition exists in
real social practices as a socially, temporally, and materially distributed
phenomenon. The resonances with Marx are clear. On page 8 of the German
Ideology, Marx wrote:
The social structure and the State are continually evolving out of the life-
process of definite individuals, but of individuals, not as they may appear in
their own or other people’s imagination, but as they really are; i.e. as they
operate, produce materially, and hence as they work under definite material
limits, presuppositions and conditions independent of their will.
(Marx 1932)
Van Dijk is on slightly firmer ground when he points out inconsistencies

between different SFL accounts of Field, Tenor, and Mode. However, this is
to be expected, as SFL theorists have not claimed to have solved context, and
no theoretical description can ever be complete. For instance, Bowcher
(2017) concludes that the development of descriptions of Field, Tenor, and
Mode within SFL is an ongoing project which includes both more explicit
description and further theorizing. Bartlett (2013)14 is a promising dialect-
ical sketch of the relationship between the environment (context as poten-
tial) and the context of situation (Field, Tenor, and Mode), which serves not
only to develop the SFL theory of context but also to bring SFL closer to
critical work such as Foucault’s genealogical method (Foucault 1977).
In his critique of SFL context, van Dijk omits to mention the influence of
Marxism. Webster (2007:10–11) observes that Marxist thought formatively
influenced the younger Halliday, who was a member of the linguistics
group of the British Communist party. The group was interested in among
other things register variation and describing a grammar in a way that,
while formally explicit, was based on meaning. These two aims remain
central tenets of SFL, as can be seen from the above discussion of Field,
Tenor, and Mode and the centrality of the paradigm to SFL theorizing, e.g.
the chapters collected in Fontaine et al. (2013) and O’Grady et al. (2013).
Halliday (2014:97) himself writes of the continuing influence Marxism has
on his work, specifically, in forming his view of linguistics as an appliable
science (see Mahboob and Knight 2010), and in the dialectical relationship
between theory and practice in his own descriptive work. While one could
argue whether or not van Dijk is correct in saying that SFL as a theory is not
underpinned by strong social theory, SFL is without doubt compatible with
social theory.
14
Bartlett’s approach is prima facie not dissimilar from the Discourse Historic Approach advocated by Ruth Wodak. This is
a problem-oriented approach that examines changes in discursive practice over time. Wodak draws on a range of
resources, such as text linguistics, argumentation theory, and the Frankfurt School of Critical Theory in analyzing
discourse (Reisigl and Wodak 2009).
Cognitive linguistics, which is becoming an increasingly used tool in

CDA,15 is an approach to language which aims for psychological plausibil-
ity. Language according to this view is embodied knowledge grounded in
cognitive processes such as association, schematization, and categorization.
Language emerges through use, and language structure is the product of
repeated interaction between interlocutors grounded in specific and indi-
vidual contexts. Language is viewed as being not only cognitive but also
sociocultural (Butler and Gonzálvez-Garcia 2014). Everett (2013:20) sums
things up nicely: ‘It [language] is a cultural tool as well as a cognitive tool.’
Hart (2015) surveys the relationship between CDA and cognitive linguists,
albeit from a cognitive linguistic viewpoint. He provides numerous
examples of the cognitive linguistic tools which various authors have used
to reveal bias and argues forcefully that, as the ideologies which underpin
social action exist as the set of mental representations shared by a commu-
nity, they are most effectively studied within a sociocultural framework.
Much, though by no means all, of the cognitive linguistic CDA work is
grounded in conceptual metaphor theory (Lakoff and Johnson 1980). For
instance, Hart provides the following extracts from a political speech by
David Cameron and argues that Cameron has construed the man-made
economic downturn as natural threat by construing it in terms of the
natural disaster frame.16
• We are living in perilous economic times . . .

• In keeping Britain safe and building the recovery we face three challenges . . .
• the turbulence coming from the Eurozone . . .
• Despite headwinds from the Eurozone, we are on track . . .17
• As our biggest trading partner, the problems in the Eurozone are affecting
Britain too. As we prepare for the potential storms we should be both resolute
and confident. Resolute because we will do what it takes to shelter the UK from
the worst of the storms.
Government action is required to mitigate the ravaging effects of the

‘weather’ and perhaps to reconstruct the damage. But, in this construal
no government could reasonably be expected to be able to stop ‘the
weather’.18 There is no doubt that analyses such as the above or Lakoff’s
(1991) own construal of the Gulf War in terms of a fairy tale frame are
elegant and rigorous linguistic descriptions which add value to the CDA
15
See, for instance, the dynamic CDA work being produced under the cognitive linguistic influenced CADAAD network
http://cadaad.net/.
16
www.gov.uk/government/speeches/prime-minister-a-speech-on-the-economy. (Last accessed 06/03/2015.)
17
For reasons that are not clear, Hart has altered the order of the speech. In Cameron’s version, the words despite
headwinds . . . were produced after the other extracts.
18
In a world where anthropogenic climate change is changing weather, Cameron could perhaps have opted for an
alternate construal!
literature.19 But, one is troubled by the lack of engagement with social

theory; to his credit, Hart (2015) similarly laments this oversight. It is
difficult though to see how cognitive linguistics, an approach underpinned
by work based on ‘the somatic marker hypothesis’20 (Damasio 1994) and
prototype theory (Rosch 1975), and from a specifically linguistic angle
(Lakoff 1987), can be reconciled with social theory. Tellingly, despite
Lakoff’s well-known political activism, I can find no reference to social
theory in his work. Indeed, Lakoff and Johnson (1999) seemingly reject
the canon of Western philosophy, including critical theory, in favour of
an embodied approach grounded in the findings from cognitive science.
To summarize, cognitive linguistics offers another string to CDA’s bow,
but, contra Hart, it is by no means the ‘missing link’ required to explain
how discursive and social practice connect. Without incorporating or dem-
onstrating compatibility with social theory, it will not be easy to use it to
underpin an explanatory critical analysis. While it offers different insights
than SFL, it has by no means supplanted it.
Butler and Gonzálvez-Garcia (2014) survey the topography of sixteen func-
tional and cognitive approaches to language, including SFL and cognitive
linguistics. Their questionnaire contained fifty-eight items, of which sixteen
are relevant for present purposes, namely, identifying linguistic approaches
compatible with CDA. Out of the sixteen items, SFL received a positive rating
for all but two. None of the other functional approaches scored so highly. The
second-highest scoring approach was emergent grammar, which had a posi-
tive rating for twelve out of the sixteen categories. Excluding SFL, the mean
positive rating for the other fifteen approaches was 7.33, with cognitive
linguistics scoring a positive rating on only seven items. This strongly sug-
gests that SFL, at the very least, remains a linguistic theory with which CDA
approaches can fruitfully engage (see also Butler, this volume).
Blommaert (2005) provides a further critique of the CDA project in
general. He states that much of the work, to date, has focused on first-
world issues such as globalization and the discourses of the new right.
Critical Discourse Analysis is therefore closed to particular kinds of
societies (Blommaert 2005:35). Yet, Blommaert’s point does not have to be
interpreted as anything other than a call for CDA studies among marginal-
ized non-first-world communities. Bartlett (2012), which incorporates SFL,
Blommaert’s concept of voice, ethnography, and some social theory, is a
very promising first step in addressing Blommaert’s concerns.
Blommaert is clearly correct when he points out that a conceptual
weakness of CDA is that it can only explore discourses that are present.21
19
It remains to be seen how applicable conceptual metaphor theory is outside first world and perhaps Indo-European
contexts. See below for Blommaert’s critique of CDA as closed to non-Western societies.
20
This is a mechanism that shows how emotion guides rational thought. In other words, rationality cannot be
disentangled from emotion.
21
It goes without saying that no form of CDA can analyze absent discourses, by which I mean ones that are not available
for recording and hence for analysis.
Yet, adopting a paradigmatic approach allows an analyst to describe both

what was said and what could have been said in the particular genre.
Systemic Functional Linguistics as the paradigmatic approach par excellence
is uniquely situated to shed light on how discursive and social practice
combine to privilege particular constellations of meanings while disprefer-
ring potentially competing meanings in particular genres.
Despite the criticisms made against SFL and the emergence of a competing
(though I would argue not necessarily incompatible) theory of language, it
appears that SFL should remain an integral part of the CDA project. In the next
section I briefly outline how SFL is contributing to the ongoing development of
CDA with a focus on corpus-assisted studies and analyses of semiotic modes
other than language. Critical Discourse Analysis has never been restricted to
language analysis (Fairclough et al. 2011:357). Similarly, Halliday’s work has
informed key studies in the analysis of images, sounds, and moving pictures
(e.g. Kress and van Leeuwen 2006; Baldry and Thibault 2006).
18.5 SFL and the Ongoing Development of the CDA Project
Critical Discourse Analysis aims not only to explain the world but also to
effect a positive change, no matter how limited, in it. An SFL study which
does precisely that is Martin et al.’s (2013) innovative multimodal investi-
gation of identity construction within the genre of youth justice counsel-
ling, a form of restorative justice. Martin et al. note that the genre idealizes
young offenders who construe themselves as both remorseful and rational
about their futures. Such a construal creates the space that permits the
genre to transition from redemption to reintegration. Through detailed
analysis of the meanings created by the coupling of verbal interaction and
gesture, Martin et al. show how bonds are created and maintained over the
interaction. While Martin et al.’s study is not couched in terms of CDA, by
mapping the multimodal generic potential of the youth justice counselling
sessions, they are in a position to advise, i.e. directly influence, the social
practice of participants involved in managing the youth counselling system.
Part of the innovation in the work of Martin and his colleagues is that they
focus not only on what is wrong in discourse but what is right. Their point
is that changes in social practice may come from the spreading of what is
positive in a discourse rather than exclusively from trying to suppress what
is wrong (see Martin and Rose 2003).
In today’s world, ideology is increasingly disseminated by a combination
of hybrid semiotic modes in fields such as advertising and new media
platforms. While there has been a technological revolution since the publi-
cation of Fairclough (1989) which has profoundly changed the way humans
interact, learn, and consume information, CDA has from its inception
recognized that discourse is not formed exclusively from verbal language
(Fairclough 1989:27).
There is a long tradition in SFL of studying semiosis and not solely

language. Within SFL, Kress and Van Leeuwen (2006) examine static images
such as news magazine covers, print advertisements, and textbook illustra-
tions as social practices in terms of Halliday’s three metafunctions, which
they reinterpret as representational, interactional, and compositional
meaning. Jewitt (2009), herself a non-SFL scholar, states that there are
currently three main strands to multimodal research, of which SFL is one.
She defines the other two approaches as ‘social semiotic’ and ‘interactional
analysis’. Taylor (2017), working within SFL, observes that in practice the
borderlines between the three approaches are permeable and that certain
tenets of SFL seem to be omnipresent across all three approaches.
Baldry and Thibault (2006) extend SFL’s scope to moving images by
developing a fine-grained transcription system. Machin (2013), the intro-
ductory chapter of a themed journal, sketches a social semiotic approach
illustrating how the emerging field of critical multimodal discourse analy-
sis can investigate the dissemination of ideologies. Systemic Functional
Linguistics approaches to multimodal discourse are in the unique position
of having rigorous transcription systems, a developing theory of context
compatible with work in social theory, and a firm belief in semiosis as a
social practice. As such we can expect SFL to be a major contributor to
future studies in critical multimodal work.
A further area where we can expect to see fruitful interaction between
SFL and CDA is in the use of corpus studies as a methodological tool for the
analysis of large texts. Thompson and Hunston (2006) is an edited collection
which illustrates how SFL as a theoretical approach to language can use-
fully engage with corpus linguistics as a practice. Systemic Functional
Linguistics and corpus linguistics both prioritize the study of natural lan-
guage in terms of probabilities. While SFL can provide a theoretical basis for
investigating patterns in corpora, the results from corpus linguistics can
inform SFL theory. There are two approaches used in corpus linguistics
research, namely, the corpus-driven approach and the corpus-assisted/based
approach. The former is atheoretical and involves no prior assumptions or
expectations. The latter uses a corpus as a repository which is investigated
to confirm/disconfirm theoretical assumptions. The latter approach is the
more promising approach for an engagement with SFL. Hunston (2013)
studies the ideology of a popular science book using a combination of SFL
and corpus linguistics. She argues that, while there is unresolved tension
between SFL theory and corpus linguistics practice concerning the relation-
ship between paradigmatic primacy and syntagmatic patterning, there was
sufficient accommodation between the approaches to allow her to interpret
the frequency of grammatical choice and lexical patterning in terms of the
ideology of science. For present purposes the key point is that SFL corpus-
based studies are a powerful means of describing the formal linguistic
features of a corpus, and so can provide the textual evidence which CDA
analysts can then use to describe the salient patterns in a particular genre.
For instance, Miller and Johnson (2013) employ an SFL-grounded corpus-

based approach to investigate stance in US Congressional speech concerning
the Iraq war. Utilizing an Appraisal framework, they found that the relative
frequency of evaluation was higher in Congressional debate than in other
text types, though they note the effect of gender, political party allegiance,
and topic on their findings. The most common evaluative pattern judged the
propriety and veracity of behaviour. The patterns of discourse revealed by the
careful combination of SFL and a corpus-based linguistic approach discon-
firmed Lakoff’s (2002) claims that Democratic and Republican discourse can
be respectively categorized in terms of an opposition between ‘a nurturant
parent’ and ‘a strict father’ (Lakoff 2002:452). Once again, it can be seen that
SFL has a large role to play in CDA corpus linguistic projects.
Ruqaiya Hasan (2009b), one of the key thinkers in SFL, has noted the
enormous potential of corpus studies in analyzing regularities in used
language. Yet, Hasan (2009b:350) states that, if corpus studies are restricted
to formal algorithms of lexical patternings, as a method it will prove
incapable of capturing semantic variation and hence in explicating the
ideological effects of linguistic choices. In short, in order for corpus linguis-
tics to be a useful tool for CDA research, corpus linguistics must free itself
of the view that language is an autonomous object of study. Instead, it must
recognize that language use and choice need to be studied in context. This
will no doubt necessitate some rethinking of methodology and one which
could usefully incorporate the SFL variables of Field, Tenor, and Mode.
To conclude, while I have by no means set out to argue that SFL should
provide the entire linguistic toolkit for CDA, I have aimed to show that,
despite the massive technological developments of the past twenty to thirty
years, the quote that opened the chapter remains as valid today as it did
when it was written. Indeed, the fact that much multimodal research has
emerged out of SFL, coupled with SFL’s complementarity with corpus-
assisted studies, ensures that SFL will remain a vibrant and central part of
the CDA programme.
References
Baldry, A. and P. Thibault. 2006. Multimodal Transcription and Text Analysis.

Bartlett, T. 2012. Hybrid Voices and Collaborative Change: Contextualising Positive
Discourse Analysis. Abingdon: Routledge.
Bartlett, T. 2013. ‘I’ll Manage the Context’: Context, Environment
and the Potential for Institutional Change. In L. Fontaine, T. Bartlett,
and G. O’Grady, eds., Systemic Functional Linguistics: Exploring Choice.
Cambridge: Cambridge University Press. 342–64.
Bernstein, B. 1971. Class Codes and Control, Volume 1: Theoretical Studies towards
Sociology of Language. London: Routledge.
Blommaert, J. 2005. Discourse. Cambridge: Cambridge University Press.

Bourdieu, P. 1991. Language and Symbolic Power. Cambridge: Polity.
Bowcher, W. 2017. Field, Tenor and Mode. In T. Bartlett and G. O’Grady,
eds., The Routledge Handbook of Systemic Functional Linguistics. Abingdon:
Butler, C. S. and F. Gonzálvez-Garcia. 2014. Exploring Functional-cognitive
Space. Amsterdam: John Benjamins.
Chouliaraki, L. and N. Fairclough. 1999. Discourse in Late Modernity: Rethink-
ing Critical Discourse Analysis. Edinburgh: Edinburgh University Press.
Damasio, A. 1994. Descartes’ Error. New York: G. P. Putnam.
Everett, D. 2013. Language: The Cultural Tool. London: Profile.
Fairclough, N. 1989. Language and Power. Harlow: Longman.
Fairclough, N. 2006. Language and Globalization. London: Routledge.
Fairclough, N. 2009. A Dialectal-relationship to Critical Discourse Analysis
in Social Science. In R. Wodak and M. Meyer, eds., Methods of Critical
Discourse Analysis. London: SAGE. 162–86.
Fairclough, N. 2015. Language and Power. 3rd ed. Abingdon: Routledge.
Fairclough, N. and P. Graham. 2010. Marx as Critical Discourse Analyst. In
N. Fairclough, ed., Critical Discourse Analysis. 2nd ed. Harlow: Longman.
310–46.
Fairclough, N., J. Mulderrig, and R. Wodak. 2011. Critical Discourse Analy-
sis. In T. A. van Dijk, ed., Discourse Studies: A Multidisciplinary Introduction.
London: SAGE. 357–78.
Fawcett, R. P. 1980. Cognitive Linguistics and Social Interaction: Towards an
Integrated Model of a Systemic Functional Grammar and the Other Components
of an Interacting Mind. Heidelberg: Julian Groos.
Fontaine, L., T. Bartlett, and G. O’Grady, eds. 2013. Systemic Functional
Linguistics: Exploring Choice. Cambridge: Cambridge University Press.
Foucault, M. 1977. Discipline and Punishment. New York: Pantheon.
Foucault, M. 1981. The Order of Discourse. In R. Young, ed. Unifying the Text:
A Post-structuralist Reader. London: Routledge. 48–78.
Fowler, R. 1991. Language in the News: Discourse and Ideology in the Press.
London: Routledge.
Gregory, M. 1988. Generic Situation and Register: A Functional View of
Communication. In J. D. Benson, M. J. Cummings, and W. S. Greaves,
eds., Linguistics in a Systemic Perspective. Amsterdam: John Benjamins.
301–29.
Halliday, M. A. K. 1969. Relevant Models of Language: The State of Lan-
guage. Educational Review, University of Birmingham 22(1): 26–37.
Halliday, M. A. K. 1973. Linguistic Function and Literary Style: An Enquiry
into the Language of William Golding’s ‘The Inheritors’. In M. A. K.
Halliday, Explorations in the Functions of Language. London: Edward Arnold.
103–43.

Halliday, M. A. K. 2007. Some Thoughts on Language and Middle School
Years. Language and Education: Collected Works of M. A. K. Halliday, Volume 9.
Edited by J. J. Webster. London: Continuum. 49–62.
Halliday, M. A. K. 2014. The Continuing Influence of Marxism. In J. J.
Webster, ed., The Bloomsbury Companion to M. A. K. Halliday. London:
Functional Grammar. 4th ed. Abington: Routledge.
Hart, C. 2015. Discourse. In E. Dąbrowska and D. Divjak, eds., Handbook of
Cognitive Linguistics. Berlin: Mouton de Gruyter. 322–45.
Hasan, R. 2005. Semiotic Mediation and Three Exotropic Theories:
Vygotsky, Halliday and Bernstein. In J. J. Webster, ed., Language, Society
and Consciousness: The Collected Works of Ruqaiya Hasan, Volume 1. Sheffield:
Equinox. 130–59.
Hasan, R. 2009a. Everyday Talk between Mothers and Children. In J. J.
Webster, ed., Semantic Variation: Meaning and Society in Sociolinguistics.
Hasan, R. 2009b. Rationality in Everyday Talk: From Process to System. In
J. J. Webster, ed., Semantic Variation: Meaning and Society in Sociolinguistics.
Hodge, B. 2017. Discourse Analysis. In T. Bartlett and G. O’Grady, eds., The
Routledge Handbook of Systemic Functional Linguistics. Abingdon: Routledge.
520–32.
Hunston, S. 2013. Systemic Functional Linguistics, Corpus Linguistics and
the Ideology of Science. Text and Talk 33(4–5): 617–40.
Hutchins, E. 1995. Cognition in the Wild. Cambridge, MA: MIT Press.
Jewitt, C. 2009. Different Approaches to Multimodality. In C. Jewitt, ed., The
Routledge Handbook of Multimodal Analysis. London: Routledge. 28–39.
Jones, P. E. 2013. Bernstein’s Codes and the Linguistics of ‘Deficit’. Language
and Education 27(2): 161–79.
Kress, G. and B. Hodge. 1979. Language as Ideology. London: Routledge.
Krizsan, A. 2011. The EU Is Not Us. Newcastle upon Tyne: Cambridge Scholars
Press.
Lakoff, G. 1987. Women, Fire and Dangerous Things: What Categories Reveal
about the Mind. Chicago: University of Chicago Press.
Lakoff, G. 1991. The Metaphor Theory: The Metaphor System Used to Justify
the War in the Gulf. Journal of Urban and Cultural Studies 2: 59–72.
Lakoff, G. 2002. Moral Politics: How Liberals and Conservatives Think. Chicago:
University of Chicago Press.
Lakoff, G. and M. Johnson. 1980. Metaphors We Live By. Chicago: University of
Chicago Press.
Lakoff, G. and M. Johnson. 1999. Philosophy in the Flesh. New York: Basic Books.
Leckie-Tarry, H. 1995. Language and Context: A Functional Theory of Register.
Edited by D. Birch. London: Pinter.
Machin, D. 2013. What Is Multimodal Critical Discourse Studies? Critical
Discourse Studies 10(4): 347–55.
Mahboob, A. and N. Knight. 2010. Appliable Linguistics. London: Continuum.
Martin, J. R. and P. R. R. White. 2005. The Language of Evaluation: Appraisal of
English. Basingstoke: Palgrave.
of Language: Embodied Identity in Youth Justice Conferencing. Text and
Talk 33(4–5): 467–94.
Marx, K. 1932. A Critique of the German Ideology. Available online at: www
.marxists.org/archive/marx/works/download/Marx_The_German_Ideology.
pdf. (Last accessed 05/03/2015.)
Maton, K. 2015. Knowledge and Knowers: Towards a Realist Sociology of Educa-
tion. Abingdon: Routledge.
Miller, D. R. and J. H. Johnson. 2013. Register-idiosyncratic Evaluative
Choices in Congressional Debate: A Corpus-assisted Comparative Study.
In L. Fontaine, T. Bartlett, and G. O’Grady, eds., Systemic Functional Linguis-
tics: Exploring Choice. Cambridge: Cambridge University Press. 432–52.
O’Grady, G. 2011. The Unfolded Imagining of Segolene Royal. Journal of
O’Grady, G., T. Bartlett, and L. Fontaine, eds. Choice in Language: Applications
in Text Analysis. Sheffield: Equinox.
O’Halloran, K. 2003. Critical Discourse Analysis and Language Cognition. Edin-
burgh: Edinburgh University Press.
Reisigl, M. and R. Wodak. 2009. The Discursive Historical Approach. In
R. Wodak and M. Meyer, eds., Methods of Critical Discourse Analysis. 2nd
ed. London: SAGE. 87–121.
Rosch, E. 1975. Cognitive Reference Points. Cognitive Psychology 7: 532–47.
Slembrouk, S. 2001. Explanation, Interpretation and Critique in the Analy-
sis of Discourse. Critique of Anthropolology 21: 33–57.
Sperber, D. and D. Wilson. 1995. Relevance: Communication and Cognition.
2nd ed. Oxford: Blackwell.
Taylor, C. 2017. Reading Images (Including Moving Ones). In T. Bartlett and
Abingdon: Routledge. 575–90.
Thibault, P. 2011. First-Order Languaging Dynamics and Second-Order
Language: The Distributed Language View. Ecological Psychology 23: 1–36.
Thompson, G. and S. Hunston. 2006. System and Corpus: Exploring Connections.
van Dijk, T. 2008. Discourse and Social Context. Cambridge: Cambridge
University Press.
Varoufakis, Y. 2013. The Global Minotaur: America, Europe and the Future of the
Global Economy. 2nd ed. London: Zed.
Voloshinov, V. N. 1973. Language Speech and Utterance. In V. N. Voloshi-
nov, ed., Marxism and the Philosophy of Language. Cambridge, MA: Harvard
Webster, J. J. 2007. M. A. K. Halliday: The Early Years 1925–1970. In
3–14.
Widdowson, H. 1995. Discourse Analysis: A Critical View. Language and
Literature 4: 157–72.
Widdowson, H. 1998. The Theory and Practice. Applied Linguistics 19:
136–51.
19
Language Development
Geoff Williams
19.1 Introduction
The most influential Systemic Functional Linguistic (SFL) text on language

development is Halliday’s Learning How to Mean (1975). In this volume he
reported an intensive case study of a young child’s exchange of meanings
with his caregivers from approximately 0–2 years. When the book was first
published in 1975, it ‘signalled a radically different orientation from the
prevailing one of language as a set of syntactic structures “acquired”
between the ages of two and four years’ (Painter et al. 2007:563). Instead,
Halliday identified children’s meaning-making in interaction as his theor-
etical and empirical point of departure. This move enabled him not only to
produce a new account of language development but also to show how
findings about the ontogenesis of language could inform a general theory of
language as social semiotic (Halliday 1978; 2004:60–1; also see Hasan 2015a
for an informative account of the early history of social semiotic theory).
However, though children’s development of meaning-making ability
rather than syntax was Halliday’s point of departure, this did not result
in syntax being marginalized. Rather, reconfigured in his theory as
lexicogrammar, it held, and continues to hold, a central role in Halliday’s
account of language development: for example, he refers to grammar as the
‘driving force from primary to higher-order consciousness’ in a later, major
paper (Halliday 2013). Just how and why early meaning-making is develop-
mentally central to children’s later use of the mother tongue, and how it
relates to the development of lexicogrammar, are key questions that his
study explores.
Halliday advanced another major claim about early child language learn-
ing, which in my view deserves far more discussion than it has received. He
suggested that learning language is simultaneously a process of learning
culture, i.e. an account of
488 GEOFF WILLIAMS
how a child is learning all the time through language: how the
microsemiotic exchanges of family and peer group life contain within them-
selves indices of the most pervasive semiotic patterns of the culture.
(Halliday 2004:111, emphasis in original)
While the claim has been very productively explored through case studies,
findings are structurally limited by the fact that they are about individual
children interacting with caregivers in just one socio-semiotic position.
A different perspective on Halliday’s general claim can be achieved by
drawing on samples of interaction between young children, now able to
speak their mother-tongue, and their caregivers in contrasted socio-
semiotic locations. This is a strategy used by Hasan (2009), and subse-
quently by two of her graduate students, Cloran (1994) and Williams
(1995). Neither one perspective on language development is intrinsically
better than the other: they are different ways of investigating the same
phenomenon, provided that phenomenon is construed at a high enough
level of abstraction, i.e. child language development during 0–5 years. Since
there are good, recent introductory overviews of Halliday’s early child
language development theory and research, together with the further
research it has produced (e.g. Painter et al. 2007; Torr 2015), it is possible
to introduce both research approaches here.
There are many advantages to discussing the two perspectives in one
paper, though this does not appear to have been done previously.1 Most
importantly, it is possible to see how some of Halliday’s innovative meth-
odology in Learning How to Mean has been either extended, replicated, or
replaced in subsequent work. Halliday has characterized his research as a
‘diary-based case study’. While this is of course accurate, his methodology
does actually involve much more than is usually understood by that term
because of the new techniques he introduced to ‘map’ meaning-making
development. There is a literal as well as a metaphorical meaning to ‘map’
in play here, which has proved highly significant for SFL explorations of
discourse.
A further important methodological and theoretical feature of Halliday’s
work in his child language development research that has been taken up in
subsequent work is the dialogue he established between linguistics and
other fields, particularly sociology. Even at this early stage of the develop-
ment of SFL theory and research, Halliday adopted a transdisciplinary
approach, not merely by supplementing linguistic research with other
compatible perspectives, but rather by enabling deeper theoretical linguis-
tic questions to be raised through interaction with other disciplines, and
also by developing a sympathetic critique of theory in other disciplines in
which language use plays some role. His dialogue with Bernstein’s theory of
1
One drawback to establishing this scope is that closely related work on infant language development by Painter (1984)
and Torr (1997), though it is significant both as independent support for Halliday’s findings and as sources of new
insights, can only be referred to briefly.
Language Development 489
social transmission, reproduction, and change illustrates both these

features (Bernstein 1990).
I begin where Halliday began, with an account of the emergence of
protolanguage. In the second part of the chapter, I introduce an aspect of
Hasan’s work on child language development,2 focusing particularly on the
questions she asked about language development and family social pos-
itioning, the methodologies she adopted, and, finally, a brief illustration of
findings from her research and their significance for education.
19.2 Protolanguage
What is ‘protolanguage’ in Halliday’s work? Since Halliday advanced a very

specific meaning for the term, it is a crucial question for at least two
reasons: first, because subsequent to his research the word has come to
acquire quite a variety of meanings (see Painter 2005 for an insightful
review of different uses of the term in linguistics), and therefore the specific
focus of Halliday’s work has been somewhat blurred; and second, because,
as researchers in psychology have produced new work on meaning-making
in the first few months of life, protolanguage has to be carefully distin-
guished from other, related phenomena such as ‘prespeech’ and
‘protoconversation’.
Protolanguage comprises a set of content-expression pairs that mean in
contexts of a child’s immediate, intimate circle but not beyond it.
Protolinguistic signs are often pragmatic, in the sense of being used by
the child to satisfy a material need (I want), but they are also interpersonally
oriented in other ways, such as a sign used to greet someone (hello). Proto-
language is thus bi-stratal: it comprises signs with ‘sound/gesture’ and
‘meaning’, but there is no structure (‘syntax’ or lexicogrammar). The signs
do not derive from adult language in the sense of being immature, or
reduced, forms of adult usage, but rather emerge spontaneously from
natural interaction between the child and the most intimate caregivers,
typically the mother and father.
‘Natural’ and ‘interaction’ are particularly significant terms. The child is
not taught to begin to mean, nor is she explicitly taught protolanguage.
Rather, an orientation to mean through interaction with people is con-
sidered to be a natural feature of infancy, beginning from a very
young age. Halliday found this position supported both by his own
very early observations of Nigel and by Colwyn Trevarthen’s observations
of infant interaction with mothers and fathers (Trevarthen 1998). In
2
There is a parallel limitation to the above. Only brief reference can be made to Cloran’s and Williams’ research, though it
supports and complements Hasan’s work in significant ways. Cloran’s work on gender adds the analysis of effects
of this variable operating within the contrasted family social positions.
490 GEOFF WILLIAMS
Halliday’s data, the second diary entry illustrates the point succinctly,
perhaps also rather surprisingly.
When he was two months old, Nigel used to greet me when I came home
from work. He would give a long gurgling account of the day’s events, always
in a cheerful tone. Then I told him how I had got on at the college, and
whether the train had been crowded at Charing Cross.
One day he told me a very sad tale. His face was frowning, his tone was
mournful, and he was barely holding back his tears.
What happened today, I asked his mother.
‘He had his first injections’, she replied.
(Halliday 1984:1)
Psychological research led by Trevarthen (1998, 2009) has similarly shown

infants’ very strong disposition to human interaction, resulting in Tre-
varthen advancing a theory of ‘primary intersubjectivity’. Four propos-
itions are relevant here. The theory of infant subjectivity developed by the
following steps:
1. The foundation was a recognition of the coherent intentionality and active

consciousness of the infant at birth . . .
2. One-month-old babies emitted different patterns of behaviour to persons and
objects . . .
3. In the first six months, communicating with persons develops conversational
proficiency or expressive reciprocity before the infants can perform effective
manipulation of objects . . .
4. The subtle timing and complementary emotional expressions in proto-
conversations by 2- and 3-months-olds were perceived to be preparatory to
linguistic communication . . .
(Trevarthen 1998:17–18, emphases in original)
The theory of primary intersubjectivity was seen as providing important

corroborating evidence for the linguistic significance of the exchange of
intersubjective meanings in the pre-protolanguage phase of development,
and subsequently in protolanguage itself.
However, if protolinguistic signs are idiosyncratic, how do they achieve
meaning, even in the limited environment of the immediate family? They
are able to mean because they form small clusters or, more specifically,
systems within small functions of meaning (‘microfunctions’). Each sign
means through differentiation from others within a system, as in the
classical Saussurean approach to semiosis. For example, between 9 and
10.5 months, Nigel had just two signs for making demands to obtain an
object. A general demand was realized as nā on a mid tone, glossed as ‘give
me that’, while a specific demand was realized by bø on a mid tone, glossed
as ‘give me my bird’. These were the only two protolinguistic signs at this
stage in Nigel’s instrumental function. Similarly, he had just two signs for
regulating the behaviour of others: a general command glossed as ‘do that
(again)’, which was distinguished from an intensified command glossed as

‘do that right now!’
What was Halliday’s basis for the differentiation and descriptions of the
functions? There are two: one intra-linguistic and one extra-linguistic.
From an intra-linguistic perspective, Halliday argues that it is possible to
use the functional organization of the semantic system of language itself
to begin to describe the functions from which a child is likely to start, the
microfunctions as they have subsequently come to be called. He comments:
Somehow, the child moves from the one to the other, from his own system to
that of the adult; and our hypothesis must be such as at least to show that it
would have been possible for him to make the transition.
(Halliday 2004:70)
It is reasonable to think of the microfunctions as prefiguring the metafunc-

tions of language itself, though of course there is a crucial qualitative
difference between the two: language is tri-stratal, with a
lexicogrammatical stratum. But ‘prefiguring’ captures something of the
meaning relevance of the protolinguistic microfunctions to the metafunc-
tions of language, i.e. what makes the move from one to the other possible.
The extra-linguistic basis for the microfunctions is sociological. As Halli-
day commented, the choice of a sociological theory to illuminate these
issues depends on language being allocated at least some prominent role
in social transmission, and on the theory being able to provide a general
characterization of contexts of language use that are crucial in social
transmission. Bernstein’s theory of social transmission and reproduction
met both of these criteria, the first being the specific reason for Halliday’s
and Hasan’s sustained collaboration with his Sociological Research Unit in
the University of London. Bernstein had proposed that there are four such
contexts, which he called critical socializing contexts: regulative, instructional,
interpersonal, and imaginative or innovative contexts.
From these two bases Halliday hypothesized seven protolinguistic
functions, ‘each one having a small range of alternatives, or “meaning
potential”, associated with it’ (Halliday 1975:37). The functions, with Halli-
day’s glosses, are as follows:
Instrumental ‘I want’
Regulatory ‘do as I tell you’
Interactional ‘me and you’
Personal ‘here I come’
Heuristic ‘tell me why’
Imaginative ‘let’s pretend’
Informative ‘I’ve got something to tell you’
It is important to emphasize that these are functions of protolanguage:

they are not a theoretical statement about functions of language.
492 GEOFF WILLIAMS
Figure 19.1 A preliminary illustration of mapping a system in a protolinguistic network
Unfortunately, lists of protolinguistic microfunctions appear in hand-

books for teachers to be used in observing young children’s language
development after they have begun using their first language. This is
clearly counter-theoretical.
In Nigel’s case these small systems of signs began to appear at approxi-
mately nine months. Halliday then represented the state of Nigel’s proto-
linguistic systems at six-weekly intervals, from nine to eighteen months,
when the beginnings of grammar and so entry into language itself made
this mode of representation, organized by microfunctions, too limited.
His method was to represent protolinguistic signs in the form of system
networks. These were, essentially, maps of what the young child could
mean at each six-weekly interval. Each function was the point of origin
from which descriptions of the dependent meaning systems were
developed. A simple example from the instrumental function mentioned
earlier serves to illustrate one of the basic mapping conventions.
The right-facing square bracket is an ‘or’ bracket, which is to say that in
using protolanguage instrumentally at 10.5 months Nigel could either do so
in a general way (give me that) or highly specifically, by making a demand
for his toy bird. This is an exhaustive representation of what he could
demand at this time. This method of presentation provided a way of
representing the systems at the end of each interval that was exhaustive
of the meanings Nigel could make. All of the networks are presented in
Halliday (1975:147–57) and as Appendix 2 in Halliday (2004).
Given the absence of lexicogrammar at this phase of his development,
Nigel could only mean one thing within one function at any one time.
Consequently, ‘or’ brackets occur exclusively in maps in the earliest phase
of protolanguage. However, as his meanings became more complex, more
mapping conventions had to be deployed. In fact, new mapping conventions
have to be introduced to represent qualitative shifts in the complexity of
protolanguage, for example, introduction of braces to show the simultan-
eous selection of different features. They thus form an effective visualiza-
tion device for understanding both the nature and functional location of
these shifts.
Network representations subsequently came to be used widely in SFL,
with greatly expanded representational conventions, especially to represent
both the lexicogrammatical and semantic resources of language itself, with
each stratum obviously requiring very different mapping. To foreshadow,
semantic networks became the key analytic resource for mapping
exchanges of meanings between caregivers and children in different

social positions at a somewhat later phase of their language development
(Hasan 2009).
During the first phase, protolanguage can be extended in two ways:
by adding or deleting options within the small functions, as when during
the next month Nigel simply added a specific demand for powder, and
in the following month deleted the toy bird and added clock; or by adding
more functions, as when between 12 and 13.5 months Nigel began to use
his protolanguage for imaginative play, adding this to the four functions
with which his protolanguage began. The concept of ‘function’, then, is
clearly central to the theory of protolanguage.
Two additional methodological notes are significant for understanding
Halliday’s approach. The first concerns form, the child’s expression in
protolinguistic signs. Halliday presented the expression using IPA conven-
tions, but he notes that this method provided only an approximate descrip-
tion because it is too specific, and that what was actually needed was a
postural notation ‘to represent postures that are taken up by the
articulatory organs’ (Halliday 2004:65–6). The second note concerns mean-
ing. In presenting his observations about Nigel’s protolinguistic systems,
Halliday provided glosses for the meaning of each sign: for example, nana-
nana is glossed as ‘I want that thing now’. But these glosses, by definition,
cannot be ‘translations’ or ‘paraphrases’ into English. So how is the approximate
meaning of each sign determined? Two criteria are used: systematicity and
functionality. Systematicity means that there has to be a constant relationship
between some content and an expression. Systematicity is a bidirectional rela-
tionship, so that when Nigel uttered nananana in some specific context it always
meant ‘I want that thing now’, and when he indicated the meaning ‘I want that
thing now’ he always did so through the expression nananana. Functionality
means that a content could be interpreted by relating it to one of the proposed
microfunctions: the content makes sense in relation to one of the functions, so
to speak. Halliday comments:
there is only content with respect; that is, with respect to the functions that
language serves in the life of the developing child . . . the content of an
utterance is the meaning that it has with respect to a given function, to
one or other of the things that the child is making language do for him. It is a
semiotic act which is interpretable by reference to the total range of semiotic
options, the total meaning potential that the child has accessible to him at
any moment.
(Halliday 2004:67–8, emphasis in original)
The distinctiveness of protolanguage from language itself has been empha-

sized so far. However, this distinction raises an obvious question: if proto-
language and language are so qualitatively different, how does a child begin
to move ‘on’ from her protolinguistic systems and take the enormous steps
‘into’ language itself? In Nigel’s case, he continued to expand the number of
494 GEOFF WILLIAMS
signs within each microfunction up to an overall total of fifty-two, until he

was between 15 and 16.5 months. That is to say, the range of meanings
within each of the microfunctions increased, but the simple, taxonomic
nature of the protolinguistic semiotic system, as illustrated above, stayed
the same. From this point Nigel began to develop a new systematic distinc-
tion between two broad functions, subsequently known as macrofunctions.
To these I will turn shortly, but first it is important to note a very small but
nevertheless significant change in the microfunctional systems that sig-
nalled major changes to come. The change is developmentally important
in itself, but also because Halliday has subsequently cited it as a clear
example of the kind of qualitative change that needs to be taken into
account in a general language-based theory of learning (see, especially,
Halliday 1993:98).
The change involved a new relationship between parallel systems within
the interactional microfunction. These were a system of personalized greet-
ings with three features (Anna ‘an:a’, Mummy ‘ama’, and Daddy ‘dada’),
and a system with two features (seeking [tone = mid-high + high, level] and
finding [tone = mid fall + low level]). Quite suddenly, he was able to combine
elements of the two systems freely by using both articulatory and prosodic
resources to make two meanings simultaneously, i.e. seeking Anna or
finding Anna, seeking or finding Mummy, and so on. In later writing
Halliday has called this a ‘magic gateway’ in the developmental pathway,
resulting in a qualitative change in Nigel’s semantic resources (Halliday
1993) because the features in the systems could now participate in a
multifunctional utterance.
From this perspective we see again the significance of Halliday’s research
into protolanguage for his theoretical account of language. Though proto-
language is distinct from language, there is an important underlying con-
tinuity of functionality. This is one of the key factors supporting the
complex transition into the use of language itself. The restriction is that
in protolanguage there is a comparatively simple relation – function equals
use – whereas all uses of language itself are multifunctional: that is, each
use of language draws on the metafunctionally organized resources of
semantics and lexicogrammar. It is therefore particularly interesting that
the first move on from protolanguage does not come from the acquisition
of referential meaning but rather from the development of a small set of
multifunctional utterances. For an overview of the origins of Halliday’s
proposal of the metafunctions of language, see Hasan (2015a:121–31).
One of the most striking features of the second phase was the emergence
of two generalized functional distinctions, subsequently called
macrofunctions, out of the earlier, more specific microfunctions: the prag-
matic, or ‘language as action’ macrofunction, and the mathetic, or ‘lan-
guage as reflection’ macrofunction (Halliday 1975:87). The significance is
not only that the small functions became more generalized and therefore
could function in a wider range of experience, but also that they were an
important resource for accessing the metafunctional organization of lan-

guage itself. Furthermore, there was a formal basis for confidence in distin-
guishing between macrofunctions. In addition to being distinguished by
use, they were also distinguished phonologically.
Nigel expressed the systematic distinction between the two major modes by
means of the phonological opposition between falling and rising tone: prag-
matic (‘response required’) as rising tone, mathetic (‘response not required’)
as falling tone. The falling tone was a direct continuation of the dominant
intonation pattern of Phase 1, where all tones were falling except in one
interactional system, that of individualized greetings.
(Halliday 1975:87)
Halliday notes that it was the mathetic function that provided entry to
classes of objects, properties of objects, then circumstantial elements (e.g.
toothpaste òn . . . toothbrush; bumblebee on tràin),3 and the pragmatic was the
source of the key interpersonal resource of the mood system. However,
there was no simple, neat distribution of developmental responsibility such
that one macrofunction maps directly onto one metafunction – the process
was more intricate and complex. Halliday provides, for example, interest-
ing observations and suggestions about how the elements noted above tend
to appear first in the mathetic function, but verbs (processes) tend to appear
first in the pragmatic function, which is also possibly the source of agentive
constructions in ergative processes (Halliday 1975:106–8).
In overview, Halliday’s case study demonstrates the value of a sustained
focus and systematic mapping of an individual’s development. Through
such a detailed approach we can see, almost moment by moment, how
Nigel begins to develop a rudimentary understanding of the culture in
which he is located. For example, features of the material environment –
from bananas to pantographs, family roles and food categories – all become
part of everyday experience. In a paper published contemporaneously with
Learning How to Mean, Halliday remarked:
The learning of language and the learning of culture are obviously two
different things. At the same time, they are closely interdependent. This is
true not only in the sense that a child constructs a reality for himself largely
through language, but also in the more fundamental sense that language is
itself a part of this reality. The linguistic system is part of the social system.
Neither can be learnt without the other.
(Halliday 2004:281)
He continues by writing of the need to find some way of conceptualizing the

two systems that would allow them to be ‘brought into some sort of
relevant relationship with one another in the developmental context’. This
he does by considering the two systems as semiotic systems, such that ‘the
3
The downward sloping symbol on the words òn and tràin indicates a falling tone.
496 GEOFF WILLIAMS
social system is realized through (inter alia) the linguistic system’ (Halliday
2004:281–2, emphases in original). Following Hasan, we can interpret this
to mean that co-occurring features of the social system typically activate
features of the linguistic system, and co-occurring features of the linguistic
system construe features of the social system (Hasan 2010).
However, there is a crucial caveat. The social system of a given culture is,
clearly, not a seamless unity. It is fractioned by class, gender, age, and so on.
So, closely following Halliday’s line of argument, is it possible that children
developing language in contrasted social locations within these intersecting
fractions within the same culture might learn to mean in somewhat different
ways? Or do they learn broadly the same ways of meaning across a culture,
with merely non-systematic variation as a result of idiosyncratic features of
their social and personal environments?
19.3 Child Language Development and Semantic Variation
These rather confronting questions preoccupied Ruqaiya Hasan for more

than forty years, culminating in the publication of the second volume of
her Collected Works, Semantic Variation: Meaning in Sociolinguistics (Hasan
2009), but also importantly prefigured by the essays in Ways of Saying, Ways
of Meaning (Hasan 2015b) and a range of other research papers. Variation
between ways of saying and meaning in different cultural and linguistic
groups within the one national or urban setting are well known, but
variation within the same culture associated with family social positioning
is another matter altogether (Bernstein 1990).
Hasan’s focus, presented more specifically, was on the possibility of
semantic variation in ontogenesis in relation to family social positioning.
Fiercely critical of ways in which sociolinguistics had ignored the possibility
of any such variation, or worse, had ridiculed the question, Hasan wished to
open up discussion and research for both linguistic and political reasons:
linguistic, because if systematic semantic variation in families’ everyday
use of language were to be found, there were important consequences for
linguistic theory; and political, because if semantic variation were evident,
there were major practical consequences for the provision of children’s
education by the state. Her critique of sociolinguistics and observations
about the political aspects of her research are well represented in two
chapters written specifically for the third volume of her Collected Works
(Hasan 2009: Chapters 1 and 10).
It is perhaps useful to note that Hasan’s personal biography had pos-
itioned her unusually well as a linguist to investigate semantic variation.
First, she was a deeply reflective bicultural woman who had grown up in
India and Pakistan, where she worked for several years as a teacher of
English literature. Subsequently, she moved to Britain for graduate study
and to do research, and then to Australia where she conducted the research
to which I refer now. She therefore literally lived intricate cultural and
intercultural differences in ways of saying and meaning. Second, while she
was in Britain, she worked in Bernstein’s Sociological Research Unit at the
University of London around the time he was redeveloping his theory of
social transmission and reproduction, including especially his ideas about
how language might function in these processes. The redevelopment
involved maintaining his earlier emphasis on the significance of language
use, but developing a semantic rather than a structural account of its
function (Williams 2005). In fact, as Bernstein warmly acknowledged,
Hasan herself played a key role in that redevelopment soon after her arrival
(see, for example, Bernstein 1973:5, 7). The title paper in her Ways of Saying,
Ways of Meaning, which was first published in 1984, demonstrates her
meticulous use of these two features of her experience (Hasan 2015b).
As the second feature implies, Hasan also adopted a transdisciplinary
approach to her language development and semantic variation research. In
addition to sociology, Hasan looked for a theory of cognitive development
that could be brought into productive dialogue with both linguistic and
sociological theory. This she found in the work of Vygotsky and his student
and collaborator, Luria (e.g. Vygotsky 1978, 1986), though her engagement
with this work was decidedly not uncritical, in large part because she
considered Vygotsky’s proposals in need of extension through linguistic
and sociological theory (Hasan 2005, especially Chapter 3). Nevertheless,
her account was deeply appreciative. In reflecting on her research approach
in an interview with David Butt and Jennifer Yameng Liang towards the end
of her life, she commented:
[I]t seems to me, that to study Basil [Bernstein GW] without Vygotsky, or
Vygotsky without Basil, is to read only half the story: the link between the
social and the psychological in ways of saying and ways of meaning is
provided by these two scholars.
(Butt and Liang 2016:394)
Interaction between the theoretical perspectives of these two scholars is

succinctly presented in a question Bernstein posed: ‘How does the outside
become the inside and how does the inside reveal itself and shape the
outside’ (Bernstein 1990:94). From a methodological perspective, Bern-
stein’s work provided Hasan with a theoretical basis on which to operation-
alize the social positioning of families (a key element of the ‘outside’) and
thus draw systematically contrasted samples from different social loca-
tions. Vygotsky’s theory provided her with insights into the development
of mental functions through linguistic interaction (as semiotic mediation),
and therefore the theoretical and empirical basis on which to be able to
investigate development of different forms of consciousness.
For both Bernstein and Vygotsky it was language use that was claimed to
be central to these processes. However, to probe the general question of the
possible effects of social positioning on language development through
498 GEOFF WILLIAMS
linguistic interaction, two further specifically linguistic resources were

needed: a comprehensive theory of language whose design could allow it
to account for ‘how the outside becomes the inside’, and a set of explicit,
replicable analytic procedures to describe exchanges of meanings. The first
element, Hasan argued, was provided by Halliday’s theory of language as
social semiotic because it proposes a systematic account of relations
between texts and contexts using, particularly, the concept of realization.
The analytic procedures, she argued, had to derive from the theory.
They also had to be both comprehensive and sensitive enough to be able
to detect any systematic variation in patterns of semantic feature selection
associated with the speakers’ social positioning. Making the analytic pro-
cedures comprehensive meant, in practical terms, developing them meta-
functionally, i.e. to describe not only ‘referential’ or experiential meanings
but also interpersonal and logical meanings. Hasan herself developed these
procedures by extending an initial, context-specific proposal offered in
Halliday (1973) into a more general, ‘context-nonspecific’ but more detailed
analytic resource. They will be discussed in further detail later in this
section.
However, there was also a complex methodological problem, which was
the problem of how to recruit participants so as to contrast their social
positions in a theoretically consistent way. Hasan solved this problem by
operationalizing perhaps the key proposal in Bernstein’s theory of social
transmission and reproduction. Bernstein had hypothesized that family
positioning in the social division of labour was the primary, though by no
means the exclusive, determinant of contrasting ways of meaning, or
‘coding orientations’. Language use was central to these processes, he
suggested, though it was not the only modality involved. His position on
what it was about language that was so important developed over time,
involving substantial changes and elaborations of his position as might be
expected in such a complex, general theory of social transmission and
reproduction. Broadly, these changes were from thinking initially in terms
of the role of structural features of language to thinking in terms of the role
of meanings in interaction (for extended discussion, see Bernstein 1990;
Halliday 2007; Williams 2005). Bernstein’s key theoretical proposition
about the originating conditions for the different coding orientations that
informed Hasan’s approach was as follows:
The simpler the social division of labour, and the more specific and local the
relation between an agent and its material base, the more direct the relation
between meanings and a specific material base, and the greater the probabil-
ity of a restricted coding orientation. The more complex the social division of
labour, the less specific and local the relation between an agent and its
material base, the more indirect the relation between meanings and a
specific material base, and the greater the probability of an elaborated
coding orientation.
(Bernstein 1990:20)
Hasan took this complex theoretical proposal and created from it a par-
ticipant selection principle that was specific enough to enable empirical
linguistic research. The alternative, and admittedly much easier, strategy
would have been to opt for ‘socioeconomic status’ as the selection
principle, but this commonly used measure, however defined, would have
prevented her from engaging in a transdisciplinary theoretical as well as
empirical dialogue with Bernstein’s sociology (Hasan 2005: Chapter 2).
The linguistic significance of her findings would also have been markedly
diminished.
The participant selection principle was the degree to which a person’s
position in the social division of labour would typically allow that person to
make decisions that would be brought into effect by others. A higher
autonomy professional (HAP) would thus have considerably more discre-
tion in making workplace decisions than a lower autonomy professional
(LAP), though neither would be either fully autonomous or completely
lacking in autonomy. To exemplify from Williams’ (1995) research, some
occupations of the main breadwinners in the HAP families were engineer,
financial consultant, and barrister, while some in the LAP families were
paint batcher, soldier, and loader driver. It is not the occupations them-
selves that are important – they do not necessarily directly indicate a
family’s social positioning – but they are theoretically well-grounded view-
points from which to explore language in use. Additional to ascertaining
occupational data, it proved necessary to check whether a person had
voluntarily taken a lower autonomy position, such as a person wanting to
have a particular work schedule for domestic reasons or to allow further
study. For further discussion of the selection principle operating in prac-
tice, see Hasan (2009:90–1) and Williams (1995:94–9). Achieving well-
contrasted samples of participants using this principle, while maintaining
respectful recruiting processes, has proved quite feasible, though complex
to manage.
Hasan’s research focus was on features of talk between children and
their mothers. Mothers were selected because they were almost exclu-
sively at this age the primary carers of young children. Audio record-
ings of naturally occurring, everyday interaction between the mothers
and their children, aged approximately 3–5 years were made by the
mothers themselves using small, powerful audio recorders. Mothers
were asked to turn on the recorder while they went about everyday
activities with the focal child that involved talking with him or her, and
only to turn it off when the child had switched to another activity. They
were invited to erase any recording that they did not wish to be heard
outside the family. This approach yielded natural unselfconscious data,
which was crucial to the purpose of the project. It would have been
difficult for the mothers to be self-conscious about their worded mean-
ings while talking with their children and simultaneously completing
domestic tasks. Additionally, Hasan comments that ‘their little children
500 GEOFF WILLIAMS
insisted on the same maternal “face” with which they were familiar’
during the recorded interaction (Hasan 2009:92). Recordings were sub-
sequently transcribed and then explored through detailed linguistic
analysis.
The analysis employed a message semantic approach, developed by
Hasan (2009; see also Low and Fung, this volume). The approach
involved segmenting the interactive discourse into individual messages
and then describing the selection of features for each message using
semantic networks. Messages are defined as the smallest unit at the
semantic stratum, typically realized by a clause at the
lexicogrammatical stratum, with one major exception, one which
proved to be key to the outcomes of the research. To exemplify, Text
19.1a presents a sample of mother–child interaction while reading a
book about a family visit to the beach, and Text 19.1b presents the talk
in this stretch analyzed into messages. Message boundaries are indicated
by bracketed numerals.
Text 19.1a Excerpt of mother–child interaction to

illustrate Hasan’s approach to message analysis
Mother: [READING] When the boys came in, they splashed them too.
Jack swam underwater. He was a shark. He grabbed Rick’s leg
and they were both bowled over by the next wave. It was
glorious.
[INTERACTING] Heavens, look at that lovely picture. Oh look
at that. Doesn’t it look great? Imagine a really hot summer’s
day.
Emily: Mum, when it’s summer I’m going to go to the beach with my
koala and swim right out to the sea and go over the waves
with my koala.
Mother: Well, I don’t know if that’s terribly safe, you know.
Emily: Well, you can come with me.
Text 19.1b Message Analysis of Text 19.1a

Mother: (1) Heavens, look at that lovely picture.
(2) Oh look at that.
(3) Doesn’t it look great?
(4) Imagine a really hot summer’s day.
Emily: (5) Mum, when it’s summer
(6) I’m going to go to the beach with my koala
(7) and swim right out to the sea
(8) and go over the waves with my koala.
Mother: (9) Well, I don’t know if that’s terribly safe, you know.
Emily: (10) Well, you can come with me.
Messages (1–8) and (10) are straightforward examples of the co-occurrence

of ranking clause and message boundaries. However, message (9) clearly
involves two clauses:
||| Well, I don’t know || if that’s terribly safe, you know |||
The basis for this approach lies in the SFL concept of logical metafunction.
I don’t know projects the second clause and is thus considered a
metarepresentation rather than a direct representation. From a semantic
perspective, then, projecting clauses are considered to be part of a message
that also includes the projected clause. It is possible that more than one
projecting clause, and therefore more than one metarepresentation, might
contribute to a message, as in I remembered she had said she was thinking that
she’d go. However, treating these instances as one message does not mean
that the semantic contribution of projecting clauses is ignored: in fact, it
emerges that this contribution is one of a set of features that play a key role
in semantic variation, as will emerge shortly.
There is one further basic descriptive move for messages, which is import-
ant because it makes a substantial difference to interpretation. Hasan dis-
tinguished between ‘progressive’ and ‘punctuative’ messages because it is
only the first type that open the possibility of a full range of metafunctional
meaning. In contrast, punctuative messages are somewhat akin to gram-
matical continuatives. To exemplify, Text 19.2a introduces the concluding
stretch from one of Michael’s joint reading times with his mother, and Text
19.2b presents an analysis of this short stretch into messages.
Text 19.2a Second excerpt of mother–child interaction to

illustrate Hasan’s approach to message analysis
Mother: [READING] They tickled the sea anemones and they made
them close up. Jack put his goggles on and had a clear look
underwater. [LONG PAUSE]
Michael: [INTERACTING] What are those?
Mother: Um . . .
Michael: Those down there?
Mother: I’m not quite sure whether they’re barnacles or something
else. They’re funny little things that are kind of shellfish,
that sort of hang onto the rocks. They’re flat and they’re very
hard and there’s a little animal living inside them.
Michael: Well why do the enemies close up?
Mother: Anemones.
Text 19.2b Message analysis of Text 19.2a

Mother: [READING] They tickled the sea anemones and they made
them close up. Jack put his goggles on and had a clear look
underwater. [LONG PAUSE]
502 GEOFF WILLIAMS
Michael: (1) What are those?

Mother: (2) Um . . .
Michael: (3) Those down there?
Mother: (4) I’m not quite sure whether they’re barnacles or
something else.
Mother: (5) They’re funny little things that are kind of shellfish,
Mother: (6) that sort of hang onto the rocks.
Mother: (7) They’re flat
Mother: (8) and they’re very hard
Mother: (9) and there’s a little animal living inside them.
Michael: (10) Well why do the enemies close up?
Mother: (11) Anemones.
Messages (1) and (3) to (11) are [progressive], but (2) is [punctuative].
Punctuative messages are typically formulaic greetings, hesitations,
addresses, and reactive expressions. Viewed at the lexicogrammatical
stratum, they do not select for Predicator. It is perhaps useful to note that
message (11), though realized by a single word and so perhaps appearing to
be [punctuative], is treated as an elliptical form of They are called anemones
and is therefore analyzed as [progressive].
How then did Hasan analyze her data from a semantic perspective? There
was a rich array of approaches to semantic analysis from which to choose,
but for Hasan’s purposes none provided a sufficiently detailed analysis of
interpersonal, logical, textual, and experiential meanings. What she did
instead was to take up an analytic initiative first introduced by Halliday and
to reconfigure and expand it so it could be used to explore the development
of children’s ways of meaning through language.
Halliday’s initiative was the introduction of semantic networks to map
meaning-making in a specific context type (Halliday 1973). At this point we
can see that the two apparently discrete methodologies in SFL child lan-
guage development research are actually developments of the same basic
techniques. Semantic networks are an extension of Halliday’s approach to
mapping the development of protolanguage, discussed earlier in this paper,
albeit with some major qualitative differences. As in the protolanguage
research, the orientation is to describe what a speaker ‘can mean’ based
on observations of what they ‘do mean’. But a semantic network is also
categorically different from the small, functionally derived protolanguage
networks because it describes language itself, which implicates multifunc-
tionality, i.e. multiple, simultaneous features of meaning-making, together
with multi-stratal interrelationships, because semantic features are related
through realization to lexicogrammar and are themselves realizations of
features of context. The complexity of semantic networks is therefore much
greater.
Halliday’s context-specific semantic network mapped meaning resources
habitually used in maternal control of young children. The relationship
between ‘can do’ and ‘do do’, which was to prove so crucial to subsequent
work in semantic variation, was succinctly summarized by Halliday in the
paper in which semantic networks were first introduced:
A [semantic (GW)] network . . . is a specification of meaning potential. It

shows . . . what the mother is doing when she regulates the behaviour of
the child. Or, rather, it shows what she can do: it states the possibilities that
are open to her, in the specific context of a control situation. It also expresses
the fact that these are linguistic possibilities; they are options in meaning,
realized in the form of grammatical, including lexical, selections.
(Halliday 1973:76)
In contrast, Hasan developed a contextually nonspecific semantic network.

This had the key methodological advantage of enabling her to explore
naturally occurring conversations between mothers and young children
across a wide range of contexts, from caring for children’s physical well-
being to teaching them about life and death. While the points of entry in
Halliday’s semantic network were the primary distinctions between fea-
tures of meanings in control (‘threat’ vs. ‘warning’; ‘condition explicit’ vs.
‘condition implicit’ (see Halliday 1973:89)), the points of entry for Hasan
were the metafunctions themselves, specified in her account as interper-
sonal, logical, textual, and experiential metafunctions.
She provided this comparative comment, contrasting her approach with
those in other theoretical frameworks, and also by implication with Halli-
day’s first SFL-originated proposal:
It transpires that if one is interested in the meanings of the messages, one

could not ignore any aspect; nor could one claim greater centrality for one kind
of meaning in comparison with the others (Hasan 1985a, b, c): a message
packages all kinds of meanings together, and the form of language does not
present any viable justification for the recognition of separate ‘fields’ e.g.,
pragmatics, speech act analysis and so on. The interpersonal functions of the
messages are just as important as the textual functions: it is not much use
knowing that some one is making a statement, without also knowing whether
the statement is in response to something said by some one else or not. Just as it
matters whether the response message is logically related to, say, point of
enquiry in a question. And it is also important to know whether messages are
used to describe a state of affairs centering round a voluntary action or one that
is imposed by some external source. If all metafunctions are seen as equally
important to the meaning potential of a language, then the semantic system
networks must represent all, as do the lexicogrammatical ones.
(Hasan 2009:96)
The semantic networks she devised proved to be productive initially in

caregiver–child interaction in the home, but subsequently in other contexts
such as classroom discourse analysis and courtroom cross-examination.
They cannot be presented in full here for space reasons, but a fragment
for the analysis of ‘demands for information’ can be used to illustrate
504 GEOFF WILLIAMS
reassure 1
b
verify
1 probe 2
a
ask 1
G c
enquire check 2
2
demand d
d assumptive 1 e
d
nonassumptive } simple 1
info
2
alternative 2
explain prompted 1
b c
1
precise
1 specify unprompted 2
a
apprize 2
circumstance 1
H f
tentative event 2
2 g
actant
3
Figure 19.2 Hasan’s semantic network fragment: making choices in demands for
information (reproduced in Williams 1995:158)
(for extended discussion, see especially Hasan et al. 2007). It is important to

note that Hasan continued to revise and refine the networks throughout her
life, never seeing them as a finished product: in fact, and on a personal note,
shortly before her sudden death, she had begun to convene a small group to
work intensively to refine them using data from a new type of context.
Figure 19.2 presents this fragment, which was first published in Hasan
(1989). The framework was used in Hasan’s semantic variation projects and
in Cloran’s (1994) and Williams’ (1995) research. Hasan subsequently pub-
lished a revised version of this fragment (Hasan 2009), but for theoretical
reasons the earlier version is preferred here.
The interpersonal metafunction is the origin of this system, using as
points of departure the basic distinctions proposed by Halliday between a
demand vs. give system, and a system comprising information vs. goods
and services. The entry condition to this network fragment is the simultan-
eous selection of ‘demand’ and ‘information’.
As noted above, a crucial quality of semantic networks is that each
feature is ‘anchored’ by a realization statement. The concept of a realization
statement is familiar from protolanguage research, with the crucial differ-
ence that in Hasan’s work realization is typically stated in
lexicogrammatical terms, which is logically impossible for bi-stratal
protolanguage. Realization statements are, in effect, recognition criteria
for the occurrence of a semantic feature. To illustrate from the fragment in

Figure 19.1, the primary semantic contrast is between the features [con-
firm] and [apprize]. The realization statements for the primary features, as
specified by Hasan, are as follows:
[confirm] major:indicative
[apprize] major:indicative:interrogative:nonpolar
For example, in Text 19.1b message (3), Doesn’t it look great? selects [confirm]
at the primary level, and in Text 19.2b [apprize] is selected by messages (1)
What are those? and (10) Well why do the enemies close up?
Moving toward the right of the networks, each more specific (or delicate)
feature is realized through a set of more delicate lexicogrammatical fea-
tures. For example, in Text 19.1b, message (3) Doesn’t it look great? selects
[confirm], and further selects the features [confirm:enquire:ask:assumptive:
simple]. In Text 19.2b message (1) What are those? selects [apprize:precise:
specify:unprompted:actant:nonspecific].
To emphasize, Figure 19.2 presents just one network fragment. To gain
some sense of the scope and detail of the analysis, it is necessary to envisage
similar network fragments across all four metafunctions, complemented by
interrelations between them. Each message is therefore analyzed on
approximately seventy variables.4
More typically in child language development research, analysis of
discourse data is carried out through a content analysis technique, often
complemented by digital resources such as NVIVO. However, message
semantic analysis has the obvious advantages that, comparatively, it pro-
vides a much more detailed exploration of the interactive language based
on explicit criteria for the recognition of semantic features, which are the
realization statements exemplified above. This is particularly important for
exploring semantic variation since any variation is most likely to be found
in configurations of the more specific semantic features: i.e. not just use of
‘yes/no’ or ‘wh/’ questions, for example, but variation in specific types of
question, the logico-semantic relations between questions and other mes-
sages, patterns of response and responses to responses to different types of
questions, and so on. The equally obvious disadvantage is the time involved
in conducting the analyses, but that is a common feature of any scientific
approach to language.
This brief methodological discussion makes it possible to return to the
initial questions about semantic variation in relation to Halliday’s proposal
concerning ‘learning language, learning culture’, raised at the end of
Section 19.2: is it possible that children developing language in contrasted
social locations within intersecting fractions within the same culture might
learn to mean in somewhat different ways? Or do they learn broadly the
4
Realization statements for all features presented in Figure 19.2 are available in Williams (1995:159), quoting from
earlier work by Hasan that is not readily accessible.
506 GEOFF WILLIAMS
same ways of meaning across a culture, with merely non-systematic vari-

ation as a result of idiosyncratic features of their social and personal
environments?
The first step to address these questions is to assess whether or not there
are systematic differences in exchanges of meaning between caregivers and
children in everyday interaction.
The findings in Hasan’s and Cloran’s work are based on statistical tests of
the linguistic analyses using principal components analyses (PCA), comple-
mented by detailed discussion of transcript examples and linguistic ana-
lyses (for a succinct description of PCA, see Cloran 1989). The statistical
technique enabled them to identify which semantic features loaded onto
the principal components, and then to compare the statistical location of
dyads on those components. If there were differences in habitual ways
of meaning strongly associated with the different social positions, it
was expected that there would be differential clustering of dyads from
the contrasted positions according to the frequencies with which they
selected the semantic features loaded strongly on the primary principal
components.
This has proved to be the case. The range of findings is both extensive and
complex, but it can be briefly illustrated through one specific set of analyses
in Hasan’s study: analyses of features of mothers’ questions and answers
and, reciprocally, of children’s questions and answers. Hasan’s argument
is that a configuration rather than a discrete list of semantic features
is implicated, and that such a configuration realizes a general principle of
social practice. In forming this position Hasan was strongly influenced by
Whorf’s writing, aphoristically conveyed by ‘fashions of speaking’ (Whorf
1956; Hasan 2005:5–8).
Taking PC1 in the maternal data first, there was a strong correlation
between mothers’ scores on PC1 and their social position (HAP
mothers>LAP mothers: p<.0003). HAP mothers were more likely to select
the highest loading features on this principal component, in remarkably
consistent contrast with LAP mothers. The specific semantic features
loading on this PCA would require a chapter-length exposition in them-
selves, together with an exploration of why the major features are selected
together. However, it is possible to illustrate briefly through one of those
features, the feature [prefaced], which was previously introduced in the
methodological discussion earlier in this section. To recall, this feature is
typically realized by projecting clauses such as I don’t know, as in I don’t know
if that’s terribly safe. The feature [prefaced] loads strongly on PC1, but there is
a marked difference in the social distribution: HAP mothers select it far
more frequently than LAP mothers. For extended discussion of these find-
ings, together with discussion of semantic features selected significantly
more frequently by LAP mothers and children, see Hasan (2009: Chapter 6).
Further to Hasan’s findings, Williams (1995) tested specifically for the
frequency of [prefaced] using a different data set and a different statistical
methodology. In this research both the feature [prefaced] and more delicate
features of it were analyzed, e.g. ‘modal’ projections such as I think or Do you
think, and ‘knowledge’ projections such as He didn’t know how they were
hooked on or Do you know what that is? As in Hasan’s research, [prefaced]
was selected differentially to a statistically significant extent within the
contrasted social positions. There were also marked differences in the
frequencies of selection of ‘modal’ and ‘knowledge’ projections (for further
discussion, see Williams 1995:271–5, 2012).
From the perspective of child language development, what is particularly
interesting is that in the children’s talk similar semantic features as in the
mothers’ talk were loaded onto PC1, including [prefaced]. There was again a
strong correlation between children’s scores on PC1 and their social pos-
ition (HAP children>LAP children: p<.009). Again, consistent with Hasan’s
findings, a statistically significant difference was found by Williams
(1995:272). It would appear that by the fourth year of life the children were
taking up different semantic variants of the mother tongue consistent with
features selected differentially by their mothers, and dependent on the
social positioning of their families.
To reiterate, the feature [prefaced] is only one of several which loaded
strongly onto PC1. It was chosen primarily to illustrate the nature of the
findings produced by this SFL approach to child language development
research. However, its selection was not entirely arbitrary. It is now pos-
sible to briefly note a broader research issue, perhaps potentially one of the
most interesting prospects in this theoretical research tradition.
In some of her later work, Hasan advanced some more general pro-
posals based on observations about the semantic functioning of [pref-
aced] and a feature frequently selected in LAP, but not HAP, discourse
which Hasan argues is ‘diametrically opposed’ to [prefaced]. This is the
feature [assumptive], exemplified by such locutions as Didn’t you see me go
out? and Why didn’t you tell me about it? (The location of [assumptive] in
the [demand; information] semantic network can be seen in Figure 19.2.
It is realized lexicogrammatically by an interrogative with negative
polarity.) Selection of [assumptive] implies that the speaker already
knows what an answer should be, and it is ‘diametrically opposed’ to
[prefaced] in the sense that questions with this feature do not carry such
an assumption, i.e. that they are neutral with respect to the addressee’s
knowledge. Thus, if [assumptive] is selected in discourse frequently,
[prefaced] is much less likely to be selected since each implies a different
discourse logic.
Hasan suggested that [assumptive] and [prefaced] function like nuclei in
sets of closely interrelated semantic features, indicated by high-scoring
features on the PCs. These sets of features, she further argued, can be
understood as ‘formative motifs’ in the discourse of caregivers and chil-
dren. She defines a ‘formative motif’ as ‘a cluster of semantic features
which are related to each other by a logic that underlies their configurative
508 GEOFF WILLIAMS
rapport’ (Hasan 2004:173), the latter term being derived from Whorf’s use
of it, here reconfigured to semantics. She comments:
Such clusters are made up of meanings that are held together by a logical
necessity, each cluster being built around one strong node, which bears some
contextual presupposition giving rise to a set of implications. For a cluster to
function as a formative motif, it acts not just as some localized meanings –
e.g. the meaning of items such as did you know, do you think, do you remember –
that are relevant but rather what is implied by their use as [preface]: it is the
implied meanings that appear to be most relevant to a semantic cluster’s
capacity to function as formative motif.
(Hasan 2009:449)
She includes an important methodological and interpretative note for

future research: that the occurrence of the cluster is typically prosodic, by
which she means that
[a]ll relevant features of a cluster cannot be found in the same or even

contiguous message(s); the elements of the cluster are dispersed throughout
the discourse; they have no syntagmatic order . . . The prosodic realization of
the clusters makes them invisible, and their role in the mediation of social
identities and mental dispositions calls for a deep understanding of the
nature of semiotic acts.
(Hasan 2009:449)
The findings so far in semantic variation research into caregiver–child

discourse appear to provide strong support for Halliday’s proposal that in
learning language children are learning culture, or more specifically, a
semantic orientation within a culture. But they also open up many further
questions for research. To me, some of the most interesting questions arise
from the final point above. Are there formative motifs operating in
caregiver–child discourse? If so, what are their semantic characteristics?
Are there formative motifs operating in teacher–child discourse in the first
year of schooling? If so, what are their semantic characteristics? What are
the implications for children’s orientation to learning? Addressing these
questions through linguistic analysis such as through semantic networks
may seem somewhat daunting, but this is because the potential of a corpus-
based semantic approach to understanding the effects of caregiver–child
interaction on child language development, and therefore on learning, is not
yet widely understood.
19.4 Concluding Comments
Both approaches to child language development research in SFL provide

strong support for the value of taking as a point of departure children
learning how to mean. From Halliday’s work we gain insights into both the
ontogenetic and (possible) phylogenetic processes (Williams and Lukin
2004) through which language, beginning as microgenetic, bi-stratal func-

tions develops into a highly complex, tri-stratal semiotic system. The the-
oretical, as well as the empirical, value is very great. Similarly, Hasan’s
approach not only yields very challenging findings for ontogenesis but also
challenging findings for linguistic theory. It seems that most acts of speak-
ing involve not only dialectal and register variation, but also semantic
variation that realizes very different orientations to living and meaning
within the one culture.
Both approaches also dramatize the value of sustained and systematic
attention to language itself. As Halliday commented recently in a series of
lectures devoted to relations between language and learning:
It is hard for any of us to keep language in the focus of attention for very
long: we tend to fly off from it in all directions, to study thought processes,
behaviour patterns, aesthetic values and so on. But I shall try to resist this
tendency and shall ask you to think linguistically, that is, to use your
conscious and unconscious understanding of language as a means of think-
ing about the world, and in particular – since this is my unifying concern in
these lectures – to use language as a tool for exploring how people learn.
(Halliday 2016:1)
Since early childhood is such a remarkable period of learning precisely

because it is a period of intense and rapid language learning, it clearly
merits the sustained attention of linguists interested in children’s develop-
ment so we can better understand ‘how the microsemiotic exchanges of
family and peer group life contain within themselves indices of the most
pervasive semiotic patterns of the culture’ (Halliday 2004:111).
References
Bernstein, B. 1973. Class, Codes and Control, Volume 1: Theoretical Studies

towards a Sociology of Language. London: Routledge.
Bernstein, B. 1990. The Structuring of Pedagogic Discourse: Class, Codes and
Control. London: Routledge.
Butt, D. G. and J. Liang. 2016. In Her Own Words: An Interview with
Ruqaiya Hasan. In W. Bowcher and J. Liang, eds., Society in Language,
Language in Society: Essays in Honour of Ruqaiya Hasan. Basingstoke: Palgrave
Cloran, C. 1989. Learning through Language. In R. Hasan and J. R. Martin,
eds., Language Development: Learning Language, Learning Culture. Meaning and
Choice in Language: Studies in Honour of Michael Halliday. Norwood: Ablex.
111–51.
Cloran, C. 1994. Rhetorical Units and Decontextualization: An Enquiry into Some
Relations of Context, Meaning and Grammar. Nottingham: University of
Nottingham.
510 GEOFF WILLIAMS
Halliday, M. A. K. 1973. Explorations in the Functions of Language. London:

Edward Arnold.
Halliday, M. A. K. 1975. Learning How to Mean. London: Edward Arnold.
Halliday, M. A. K. 1978. Language as Social Semiotic. London: Edward Arnold.
Halliday, M. A. K. 1984. Listening to Nigel: Conversations of a Very Small Child.
Sydney: Department of Linguistics, University of Sydney.
Halliday, M. A. K. 1993. Towards a Language-based Theory of Learning.
Linguistics and Education 5: 93–116.
Halliday, M. A. K. 2004. The Collected Works of M. A. K. Halliday, Volume 4:
The Language of Early Childhood. Edited by J. J. Webster. London:
Continuum.
Halliday, M. A. K. 2007. Language and Socialization: Home and School. In
J. J. Webster, ed., The Collected Works of M. A. K. Halliday, Volume 9: Language
and Education. London: Continuum. 81–96.
Halliday, M. A. K. 2013. On Grammar as the Driving Force from Primary
to Higher-order Consciousness. In J. J. Webster, ed., The Collected Works
of M. A. K. Halliday, Volume 11: Halliday in the 21st Century. London: Blooms-
bury. 159–90.
Halliday, M. A. K. 2016. Aspects of Language and Learning. In J. J. Webster,
ed., The M. A. K. Halliday Library Functional Linguistics Series. Berlin: Springer.
Hasan, R. 2004. The World in Words: Semiotic Mediation, Tenor and Ideol-
ogy. In G. Williams and A. Lukin, eds., The Development of Language:
Functional Perspectives on Species and Individuals. London: Continuum.
158–82.
Hasan, R. 2005. The Collected Works of Ruqaiya Hasan, Volume 1: Language,
Society and Consciousness. Edited by J. J. Webster. Sheffield: Equinox.
Variation: Meaning in Society and in Sociolinguistics. Edited by J. J. Webster.
Hasan, R. 2010. The Meaning of ‘Not’ is Not in ‘Not’. In A. Mahboob
and N. K. Knight, eds., Appliable Linguistics. London: Continuum. 267–306.
Hasan, R. 2015a. Systemic Functional Linguistics: Halliday and the Evolu-
tion of a Social Semiotic. In J. J. Webster, ed., The Bloomsbury Companion to
M. A. K. Halliday. London: Continuum. 101–34.
Hasan, R. 2015b. Ways of Saying, Ways of Meaning: Selected Papers of Ruqaiya
Hasan. Edited by C. Cloran, D. G. Butt, and G. Williams. London:
Bloomsbury.
Hasan, R., C. Cloran, G. Williams, and A. Lukin. 2007. Semantic Networks:
The Description of Linguistic Meaning in SFL. In R. Hasan, C. M. I. M.
Painter, C. 1984. Into the Mother Tongue: A Case Study in Early Language
Development. London: Pinter.
Painter, C. 2005. The Concept of ‘Protolanguage’ in Language Development.
Linguistics and the Human Sciences 1(2): 177–96.
Painter, C., B. Derewianka, and J. Torr. 2007. From Microfunction to Meta-

phor: Learning Language and Learning through Language. In R. Hasan,
C. M. I. M. Matthiessen, and J. J. Webster, eds., Continuing Discourse on
Language: A Functional Perspective, Volume 2. Sheffield: Equinox. 563–88.
Torr, J. 1997. From Child Tongue to Mother Tongue: A Case Study of Language
Development in the First Two and a Half Years. Nottingham: Department of
English Studies, University of Nottingham.
Torr, J. 2015. Language Development in Early Childhood: Learning How to
Mean. In J. J. Webster, ed., The Bloomsbury Companion to M. A. K. Halliday.
Trevarthen, C. 1998. The Concept and Foundations of Infant Intersubjectiv-
ity. In S. Bråten, ed., Intersubjective Communication and Emotion in Early
Ontogeny. Cambridge: Cambridge University Press. 15–46.
Trevarthen, C. 2009. The Intersubjective Psychobiology of Human Meaning:
Learning of Culture Depends on Interest for Co-operative Practical Work,
and Affection for the Joyful Art of Good Company. Psychoanalytic Dialogues
19(5): 507–18.
Vygotsky, L. S. 1978. Mind in Society: The Development of Higher Psychological
Processes. Cambridge, MA: Harvard University Press.
Vygotsky, L. S. 1986. Thought and Language. Edited and translated by
A. Kozulin. Cambridge, MA: MIT Press.
Whorf, B. L. 1956. Language, Thought and Reality: Selected Writings of Benjamin
Lee Whorf. Edited by J. Carroll. Cambridge, MA: MIT Press.
Williams, G. 1995. Joint Book-reading and Literacy Pedagogy: A Socio-semantic
Examination. Volumes 1 and 2. PhD Thesis, Macquarie University.
Williams, G. 2005. Semantic Variation. In J. J. Webster, C. M. I. M. Mat-
thiessen, and R. Hasan, eds., Continuing Discourse on Language, Volume 1.
Williams, G. 2012. Recontextualization and Semantic Variation. Plenary paper
presented at the International Systemic Functional Congress, University
of Technology, Sydney, 11 July 2012.
Williams, G. and A. Lukin, eds. 2004. The Development of Language: Functional
Perspectives on Species and Individuals. London: Continuum.
20
Applying SFL for
Understanding and
Fostering Instructed
Second Language
Development
Heidi Byrnes
20.1 Introduction
This chapter makes two overriding assumptions. First, with regard to

audience, I assume that readers are aware of the basic outlines, interests,
and findings of second-language acquisition (SLA) research. For such
readers the chapter begins with familiar conceptual territory and from
there progresses to SFL-specific constructs. For readers with some know-
ledge of SFL, the intellectual journey is more likely to involve a shift from
an earlier focus on issues in first-language (L1) learning toward the peculi-
arities and particularities of instructed second-language (L2) learning.
Second, with regard to its focus on instructed L2 development, the chap-
ter explicitly foregrounds the fact that, while successful L2 learning is
known to stretch over extended periods of time, conceptualizing and
researching it requires more than a longitudinal optic that is primarily
expressed in terms of elapsed time, the approach taken in the bulk of L2
theorizing and empirical research. The need to go beyond recording
changes over varying time periods (e.g., days, weeks, a semester or aca-
demic year, multiple years) arises from the fact that L2 learning is variable,
dynamic, and nonlinear – with plateaus, U-curves, possibly followed by
noteworthy bursts of development; with macroscopic and microscopic
turning points that require investigation of different grain size; and with
emergences (along with their precursors), their successors, and connected
growers – to name only some of the characteristics of language learning
that have recently captured the attention of SLA researchers (for extended
discussion, see Ortega and Byrnes 2008).
In order to be able to imagine such dynamic characteristics – not to
mention to locate them in instructed L2 learners’ lives, research them
Applying SFL for Understanding and Fostering Instructed L2 Development 513
appropriately, and, ultimately, interpret them for their significance – we

need a theory of language that allows us to envision, in a principled fashion,
what might be the intricate interrelationships between languaging and
meaning-making at different points of the overall trajectory of instruction
and learning. That conclusion has distinct echoes in contemporary SLA
research, as it wrestles with a growing awareness that its hitherto preferred
constructs require thorough reconsideration. For example, while the staple
categories of complexity, accuracy, and fluency (CAF) continue to retain
extraordinary prominence (e.g., the contributions in Housen et al. 2012),
they are simultaneously being critiqued both for their insufficiency as
currently conceptualized (Jackson and Suethanapornkul 2013; Larsen-
Freeman 2009; Ortega 2015) and for their methodological-analytical short-
comings (Norris and Ortega 2000, 2009). Both critiques highlight the need
for much greater sensitivity to developmental issues in our attempts to
understand the nature of learner performance at different proficiency
levels.
A final overarching comment regarding the chapter’s focus on develop-
ment is this: While the previous discussion seemed to imply a clear distinc-
tion between L1 and L2 development, such a dichotomous treatment is
unsustainable in the age of momentous migratory movements, from vol-
untary and opportunistic to forced and oftentimes tragic. The result has
been a remarkable broadening of who the instructed L2 speakers in our
classrooms are and how they will come to differentially use their language
repertoires over a life span as they make sense of and position themselves
within their multilingual worlds (see Douglas Fir Group 2016).
Overall, the chapter seeks to make the case that, from the theoretical, the
research-empirical, and the educational side, SFL, as a comprehensive func-
tional theory of language, presents a welcome framework for capturing
central aspects of instructed L2 development. Its aim is to characterize SFL
as a sufficiently capacious framework that recognizes and can account for
the complexity, contingency, and considerable variability in instructed L2
development as an extended phenomenon. At the same time, as a theory, it
also embraces a certain prototypicality and stability for aspects of language
use, a quality that can support educative action at all stages of the develop-
mental continuum, from beginning levels of instruction to advanced ability
levels.
I begin by locating SFL within dominant approaches to researching L2
development, thereby setting the stage for exploring, in the next section,
this central question: What foci, concepts, and analytical tools does SFL
provide that can support an understanding of L2 development that is usable
in and useful for both research and educational praxis? From among a
number of possible dimensions, I have chosen five for more thorough
discussion and subsequent illustration through research studies.
I conclude with a selection of issues that are likely to occupy societies in
the foreseeable future and to which SFL can contribute substantively.
514 HEIDI BYRNES
Suggested additional readings should enable readers to take their reflection

on the topic of L2 development beyond the confines of these pages.
20.2 Positioning SFL within Approaches to L2 Development

Research and Praxis
Following my initial broad characterization of SFL as well suited for

addressing issues in instructed L2 development, I now contextualize and
support that claim by linking SFL to shifting emphases in contemporary
SLA as a whole.1 These shifts can be traced particularly well through
various terminological dissemblings, whereby previously nearly inter-
changeable usage of core terms is now imbued with particular meanings
that increasingly mark off profound epistemological differences. Three
terms stand out: ‘development’, ‘acquisition’, and ‘learning’.
20.2.1 Relating Development, SFL, and SLA Research

Specifically, the term ‘development’ is now used to indicate a distinct shift
toward what are (still) called ‘alternative’ approaches to SLA (see Atkinson
2011) or, in a less dichotomous and more forward-looking and assured
stance, ‘transdisciplinary’ approaches to the study of language in the age
of globalization. Their essence can be described as foregrounding an under-
standing of language as a socio-semiotic resource for meaning-making in
nested layers of social action, from the micro level of social activity on the
part of the individual, to the meso level of sociocultural institutions and
communities, to the macro level of ideological structures that manifest
belief systems and cultural, political, religious, and economic values, all of
which affect language use in complexly interrelated ways (see Douglas Fir
Group 2016).
For the present discussion, two interrelated aspects are worth noting.
First, this ‘new’ position, which has been evolving for close to three
decades, contrasts with an earlier dominant focus on intra-individual cog-
nitivist qualities expressed in terms of psychological processes, both facili-
tative and constraining, that individual language users draw on in language
use and learning. Second, it is not difficult to detect strong affinities, at least
conceptually if not in the theoretical details, between this stance and
fundamental assumptions in SFL that go back to its beginnings in the
1960s. At that time, Halliday not only took a strong position against the
autonomous cognitivist-conceptual beliefs held by Chomsky, which domin-
ated language theorizing. He also expressed concerns when the rising
‘sociolinguistics’ of the day, associated with Hymes’ communicative
1
Comprehensive discussion that also provides some historical depth is offered in Lafford (2007) and Hulstijn et al.
(2015).
competence and Labov’s approach to capturing variation in language use,

separated out the ‘social’ as though it were not part of any and all language
use. In an interview at the time Halliday had this to say: ‘I would really
prefer to leave out the “socio”, if I had the choice . . . the “socio” is really
unnecessary’ (Martin 2013:1).
With that response he not only rejected the requirement that linguistic
inquiry of value must be pursued as a branch of cognitive psychology; he
also asserted that language is not a self-sufficient system, but always points
outside of itself to the social system. Accordingly, he proposed posing such
questions as ‘What is the role of language in the transmission of culture?
How is it that the ordinary everyday use of the language, in the home, in
the neighborhood and so on, acts as an effective channel for communi-
cating the social system?’ and, ultimately, ‘Why is language as it is? Lan-
guage has evolved in a certain way because of its function in the social
system’ (Martin 2013:3).
It is this fundamentally functional, that is, social-cultural meaning orien-
tation toward the nature of language that profoundly shapes how SFL
imagines and researches language as a semiotic system for culturally
embedded meaning-making: language itself – and that means its very
system of lexicogrammatical resources – is intrinsically or ‘naturally’ func-
tional, that is, oriented toward meaning-making, not just its use (Martin
1991). By implication, SFL seeks constructs that make it possible to describe
any language development, including instructed L2 development, as realiz-
ing the evolving capacity toward such meaning-making in social context,
including the quite real social environment of the language classroom
(Halliday 2007a, 2007b).
20.2.2 Rejecting ‘Acquisition’ – Rejecting What?

As ‘development’ is on the rise in SLA discussions, ‘acquisition’, heretofore
the centre of the second-language ‘acquisition’ enterprise, is increasingly
being rejected (see especially Larsen-Freeman 2015). It is interpreted as
conjuring up inappropriate images of taking possession of ‘language’ as a
fixed object, bounded, rule-governed, and essentially static, and available
‘out there’ as an independent object and commodity to be grasped and
acquired. While variation among individual learners is certainly expected,
in the end, ‘acquisition’ seems to imply the possibility that the process of
language learning is at some point successfully completed (see Larsen-
Freeman 2014, 2015). However, because such a successful completion is
only rarely achieved, there is considerable temptation to locate the cause
for that failure in learners themselves and to express that fact in variously
theorized limitations (e.g., the Critical Period Hypothesis; diverse process-
ing constraints or memory, aptitude, or motivation limitations). Finally, an
acquisition metaphor targets as the goal of L2 learning the idealized per-
formance of the native speaker. The impossibility and intellectually flawed
516 HEIDI BYRNES
nature of that goal have been well exposed by researchers like Cook (2008)
under the construct of ‘multicompetence’. Not only does multicompetence
nullify notions of permanently deficient L2 learning efforts; it reverses
them by asserting that multilingual users must be seen for what they are
in their own right, not under the restrictive lens of monolingualism.
20.2.3 Relating Development to Learning: A Sociocultural View

The final terminological distinction that helps position SFL within contem-
porary SLA deliberations highlights the difference between development
and learning as theorized within Vygotsky-inspired sociocultural theory
(SCT). Importantly, SCT does not claim to be a sociolinguistic theory of
language or language use. Rather, it ‘explains human mental functioning
on the basis of situated sociocultural activity that is mediated in large part
by communicative practices’ (Lantolf 2005:341). Within that interest the
theory differentiates between learning and development as embedded in
one of Vygotsky’s best-known theoretical constructs, the ‘zone of proximal
development’ or ZPD. While Vygotsky himself left the term surprisingly
underspecified (see Chaiklin 2003), it is typically interpreted in terms of the
distance between a learner’s actual developmental level, ascertained
through independent performance, and potential development that bene-
fits from guidance by others. As Lantolf (2005:336) states, ‘Learning, for
Vygotsky, is assisted performance, whereas development is the ability to
regulate mental and social activity as a consequence of having appropri-
ated, or internalized, that assistance.’ Further, Lantolf (2005:336) charac-
terizes the significance of instruction when he refers to it as a kind of
‘projection of a person’s developmental future’. And he adds this claim:
‘When instruction makes it possible for learners to gain a conceptual
understanding of a feature of language, or indeed a feature of any subject
matter, it has a profound influence in shaping how development unfolds,
both with regard to rate and route’ (Lantolf 2005:340). In portraying the
‘rate and route’ of development as being malleable under certain
circumstances, he challenges core psycholinguistically driven assumptions
regarding orders and sequences in L2 development (see the extended dis-
cussion in the contributions to Hulstijn et al. 2015).
Notwithstanding greater or lesser differences regarding development,
acquisition, and learning, I believe this is an opportune time for entering
into fruitful two-way dialogue between the broader language studies field
and SFL. One possible basis for such a dialogue is Hasan’s (1995:194–5)
proposal that ‘[Bernstein’s] sociological theory compels us to explore the
theoretical connections which span the biological, the sociological, and the
linguistic, combining the insights of Vygotsky’s sociogenetic psychology,
Bernstein’s semiotic sociology, and Halliday’s sociological linguistics to
arrive at a theory which will specify the social conditions for the semiotic
mediation of the human mind’ (also see Byrnes 2006a). Among questions to
be posed might be: What are the goals of language development, what counts
as language development in the micro-environment of classroom interaction,
and what would it look like along the extended trajectory of an entire
instructional programme? How might development be ascertained in a prin-
cipled and replicable way that is both attentive to the formal manifestations
of language and unequivocally meaning-oriented? How might it be fostered
and sustained in educational settings, with their relatively proceduralized
assumptions about L2 learning, in light of considerable variation in learning
success among individual learners and, indeed, entire cohorts of learners?
And how might an instructional context prepare learners not only to function
but, ultimately, to thrive in a multilingual environment with its ever-shifting
and complex demands for oral and written literacies?
20.3 Understanding L2 Development: Core Constructs

and Processes in SFL
This section explicates constructs and processes in SFL that are particularly
pertinent to addressing such questions.
20.3.1 Understanding L2 Development as Realizing the Semiotic

Resource Language in Use
As contrasted with previous approaches to treating language as a fixed
‘object’, SFL considers it to be a ‘resource’ for meaning-making whose
potential is realized only in actual acts of meaning-making in diverse
contexts of use in society. It follows that language development – including
L2 development in educational settings – is fostered by expanding oppor-
tunities for different types of language use.
While providing such a variety is a formidable educational challenge that
must be addressed, from the theoretical side, the first challenge is how to
imagine, in the first place, the relation between contexts of language use – a
social phenomenon – and the kind of wording that accomplishes their
realization – a linguistic phenomenon. Here, SFL makes the critical assump-
tion that the entire system of language is metafunctionally organized, a
feature that makes it possible to relate its formal (lexicogrammatical)
resources to social contexts of use.
Specifically, it posits three metafunctions: the interpersonal metafunc-
tion, which recognizes the inherently social-relational nature of language
use as a social semiotic; the ideational metafunction, which recognizes the
fact that it is through language that we turn our experiences with our
physical and our social environments into meaning; and the textual meta-
function, which organizes the information to be conveyed in terms of units
that are recoverable for the listener and reader in a way that itself contrib-
utes to the meanings being conveyed. A shorthand way to refer to the three
518 HEIDI BYRNES
metafunctions is in terms of enactment and exchange, construed represen-

tation, and information structuring. Importantly, all three are realized in
every instance of language use and simultaneously rather than consecu-
tively (also see Berry, this volume).
The metafunctions are linked to context in terms of ‘tenor’, ‘field’, and
‘mode’, a trinocular perspective on how recurrent contexts of situation are
configured in a recognizable way within the larger context of culture.
Specifically, as a theoretical construct within SFL, ‘register’ refers to the
bundling of lexicogrammatical resources into patterned functional var-
ieties. In that sense, registers acknowledge both variety in language use
and systematicity in terms of situation types. Halliday (1978:31) aptly
describes registers as enabling language users to accomplish these central
tasks for meaning construal: ‘First, what is actually taking place; secondly,
who is taking part; and thirdly, what part the language is playing. These
three variables, taken together, determine the range within which mean-
ings are selected and the forms which are used for their expression. In other
words, they determine the “register”.’ Returning once more to issues of
development we might then say that language development can be
described as expanding one’s registerial repertoire.
20.3.2 From the Metafunctional Quality of Language

to Register and Genre
But registers are not directly locatable at any particular place in the lan-
guage system: they are a variety of language. That makes them difficult to
translate into educational practice, one reason for the introduction of the
construct ‘genre’. For Martin (2009:12), genre is a way
to specify just how a given culture organizes this meaning potential

into recurrent configurations of meaning, and phases meaning through
stages in each genre . . . The high-level position of genre in the model
provided a way of talking holistically about the social purposes of texts and
the ways in which different genres marshalled different resources to achieve
their goals.
In other words, although genre is, strictly speaking, not a theoretical

construct within SFL, for educational purposes it can make more transpar-
ent how language functions in social contexts (Martin 1997). By describing
genre as ‘a staged goal-oriented social process’, its defining obligatory
stages, as well as its optional expansions, along with their typical lexico-
grammatical resources, a genre-based approach can help teachers and
learners begin to realize language as a social semiotic resource, even in
classroom settings. Revisiting my previous comment about the role of
registers, we might now amend it to say that L2 development is all about
expanding one’s generic repertoire for confident and competent situated
language use.
20.3.3 Modelling L2 Development as the Dialectic between

System and Instance
The challenge to assure that such resource expansion in fact occurs in an
educational programme is ever-present. SFL seeks to meet it by invoking a
two-way dynamic, that between ‘system’ and ‘instance’. These are not
separate phenomena but ‘two sides of the same coin’: instances of
meaning-making that unfold in a particular communicative situation
(e.g., an oral or written text) are at the same time manifestations of the
meaning potential of the entire system. To translate this somewhat abstract
notion into the reality of learner engagement in an L2 classroom, it is worth
quoting Halliday (2007b:274, emphases in original) at some length:
In all language education, the learner has to build up a resource. It is a

resource of a particular kind: a resource for creating meaning. I call it a
‘meaning potential’. Whether someone is learning the mother tongue, learn-
ing to read and write, learning a second or foreign language, learning the
language of science or mathematics, or learning the styles of written com-
position – all these are forms of meaning potential. What the learner has to
do is to construe (that is, construct in the mind) a linguistic system. That is
what is meant by ‘language as system’: it is a language . . . in the form of a
potential, a resource that you draw on in reading and writing and speaking
and listening – and a resource that you use for learning with. How do you
construe this potential, and how do you use it when you’ve got it? You build
it up, and you act it out, in the form of text. ‘Text’ refers to all the instances of
language that you listen to and read. And that you produce yourself in
speaking and in writing.
The implications of such a stance across the entire spectrum of educational

work in support of L2 development are undeniable.
20.3.4 Linking System and Instance in Language Use as Language

Learning: The Cline of Instantiation and the Hierarchy of
Stratification
Since our concern is with instructed L2 learning and, ultimately, long-
term development, we must, therefore, find a link between the ‘potential’
both of the entire system and of language use and the ‘actual’ meaning-
making in a particular instance of communication. As Matthiessen
(2009:207, emphasis in original) emphasizes, ‘Acts instantiate potential,
and potential emerges from acts. This is where we can locate learning.’ But
along with the learning that educators project into a student’s particular
act of meaning-making (e.g., her ‘realizing’ a particular oral communi-
cative task), they must also consider its developmental potential: Where
does it fit so the student does not ‘merely’ learn, but, in that learning,
continues to traverse the considerable space between the instance and the
meaning potential of the language system that will, gradually, need to be
intuited?
520 HEIDI BYRNES
SFL answers this key question on two levels. On the first and more abstract
level, it proposes the dynamically double-faced construct of the ‘cline of
instantiation’, imagined as a continuum between the instance and the
system. Coming from the system side, a particular instance realizes certain
subpotentials for meaning-making that the full system holds. Coming from
the opposite direction, a particular text is educationally motivated (as con-
trasted with merely being ‘interesting’) to the extent that it well models a
broader instance type, in short, a particular genre. One might visualize this
cline with the construct of genre occupying the middle ground, as it were,
between the instructional instance and the semiotic system (see Byrnes
2015). By ensuring that its instructed language users will, over the duration
of their studies, encounter an increasingly broader range of major genres, a
programme can lay the groundwork for long-term L2 development. Indeed,
how a programme’s educators imagine that trajectory well describes the
central task of curriculum development (Byrnes 2014a, 2017).
On a second, more directly language-based level, SFL invokes the stratal
quality of language through the ‘hierarchy of stratification’, according to
which language shows ‘a split of content into meaning (semantics) and
wording (lexicogrammar) and a split of expression into phonology and
phonetics’ (Matthiessen 2007:519; also see Tarverniers, this volume). That
stratal quality is further characterized in terms of two axes that operate at
each stratum: the syntagmatic axis is concerned with phenomena of
‘chaining’; by contrast, the paradigmatic axis addresses possible and per-
missible contrasts in a network of options within a stratum – in other
words, it addresses ‘choice’ toward meaning-making.
Not surprisingly, SFL favours the paradigmatic axis of choices, in contrast
to the primarily syntagmatically (or rule-)oriented approaches prevalent in
linguistic analysis and SLA research. Relating that preference to L2 develop-
ment, we might then describe the learner’s real task as that of negotiating –
and repeatedly renegotiating in expanding acts of meaning-making – the
realizational space between the paradigmatic and the syntagmatic axes,
between choice and chain.2 Given that different learners are likely to want
to mean different things even in the relatively staged environment of the
language classroom, a kind of bounded variation characterizes the entire
enterprise of instructed L2 learning and development.
20.3.5 Engaging Grammar as the Dynamic Engine of Languages:

Harnessing the Power of Dynamic and Synoptic Forms of
Meaning-Making in Grammatical Metaphor
Discussion thus far has emphasized the functional or meaning orientation of
SFL. Probing more deeply, we now explore the implications for the theory’s
distinctive way of positioning grammar, always understood as lexicogrammar.
Specifically, SFL holds that the semiotic system of a language
2
For a visualization of these constructs as applied to curriculum building, see Byrnes (2015).
means in two phases, having a distinct phase of ‘wording’ serving as the base
for the construction of meaning. In other words, its ‘content plane’ contains
a ‘grammar’ as well as a ‘semantics’ . . . It is the presence of a grammar that
gives such a system its unique potential for creating (as distinct from merely
reflecting) meaning.
(Halliday 1996:5)
Halliday goes further yet, by asserting that grammar is in fact ‘a privileged

part of language . . . the part of language where the work is done. Language
is powered by grammatical energy, so to speak’ (Halliday 1996:4).
With that stance, SFL can embrace the kind of long-term view that
development entails, moving from grammatical features that undergird
spoken interaction to those that characterize written academic language
use, the typical progression in instructed language learning, especially in
higher education. Using a meaning-oriented interpretation of grammar,
Halliday (2002:350) describes the differences like this:
Speech and writing will appear, then, as different ways of meaning: speech as
spun out, flowing, choreographic, oriented towards events (doing, happening,
sensing, saying, being), processlike, intricate, with meanings related serially;
writing as dense, structured, crystalline, oriented towards things (entities,
objectified processes), productlike, tight, with meanings related as components.
(Halliday 2002:350)
He adds that ‘spoken and written language serve as complementary

resources for acquiring and organizing knowledge . . . Educational know-
ledge demands both, the two often relating to different aspects of the same
phenomenon’ (Halliday 2002:350).
The two-phase nature of grammar already addressed under the construct
of the hierarchy of stratification further complexifies the continuum of
development of major text types. It does so by adding perhaps the most
fecund construct for conceptualizing, analyzing, tracing, and fostering L2
development over long periods – the construct of ‘grammatical metaphor’
(GM). Inasmuch as GM has played a central role in the vast majority of SFL
empirical studies, it will receive more in-depth treatment in the next
section. For now, this explication must suffice.
SFL posits the existence of two forms of semiosis, that is, two ways of
relating meaning and grammatical form, referring to them as congruent or
dynamic and non-congruent or synoptic semiosis. From the standpoint of
development, language users begin with congruent ways of realizing mean-
ings, that is, they initially prefer something like a ‘direct line of form to
meaning to experience’ (Halliday 1994:xix). That means, they would select a
verb for the expression of a process, generally use nouns and their proforms
to refer to actors and things in the world, and typically use prepositional
phrases to express the circumstances for action (The teacher starts class at
eight o’clock). However, that initial preference gradually gives way to non-
congruent/synoptic or metaphorical semiosis (Halliday 1993). As a result,
the earlier ‘natural’ relationship between grammatical category and
522 HEIDI BYRNES
meaning becomes skewed in the sense that things might be expressed

metaphorically in relation to the ‘normal’, expected categories. Among
the most prevalent occurrences for grammatical metaphor (GM) is
the nominalization of verbs and adjectives (The start of class is eight o’clock)
in the ideational realm, a transcategorization of lexicogrammatical categor-
ies. But it occurs in all environments, and in fact is perhaps most
readily recognized in speech acts with their strong interpersonal function,
where, for example, the intention of commanding that an open door should
be closed might be expressed grammatically-metaphorically by stating It’s
cold here.
Importantly, there resides in grammatical metaphoricity a tension
between the semantics and grammar that becomes the site for increasing
the meaning potential of an utterance based on the meaning-making power
of the grammatical system itself. Thus, as shown in the example, the
nominalized GM blends both the meanings originally associated with its
verbal base – the process of starting something – and those of the noun
actually chosen, resulting in a new object in conceptual space, the start. As a
profusion of SFL-based studies has shown, the conceptual and linguistic
ability to engage in this kind of objectification of the world, creating its
‘thinginess’, as Halliday calls it, through GM nominalizations turns out to
be the driving trajectory for the development of literacy with its preferred
‘writeable wordings’ as compared with ‘speakable wordings’ (Halliday
2002:345). Because the ability to deploy GM is an extended process with
distinct manifestations in all metafunctional environments, because it is
central to the ability to create textual wholes, and because it is differentially
realized in different genres, from the private and familiar sphere to the
disciplines and language use in public life and institutions, it is, perhaps,
the developmental construct par excellence.3
I conclude this section with the following observations about SFL, all of
which affect our understanding of development: (a) SFL refrains from
distinguishing language from the use of language and from privileging
one or the other dimension of the systems of language and use; (b) it
actively invites ‘both–and’ statements, frequently tying them to the funda-
mentally metaphorical nature of language, particularly through the con-
struct of GM; (c) it upholds foundational complementarities that cannot and
should not be resolved, such as those between a theory of grammar and a
theory of discourse (Halliday 2008); (d) it eschews unsustainable distinc-
tions, such as those between lexicon and grammar or between pragmatic
meaning and purportedly more central conceptual meanings; (e) it
embraces variation and the multiperspectival viewpoints or dimensional-
ities necessary to capture the fluidities of language use; and, finally, (f ) it
3
From the extensive literature on GM use, I highlight Byrnes (2009), Byrnes et al. (2010), and Schleppegrell
(2004a) for L2 learning; Christie and Derewianka (2008) for L1 literacy development; and Bernstein’s sociology of
knowledge as linked to SFL in Martin (2007).
acknowledges relational contingencies such that describing something

adequately calls for relating it, to the extent possible, to everything else in
the system, from constraints to choices (for an excellent treatment of
language developmental issues in L1 education, see Halliday 2007a; for
implications and applications to L2 learning, see Byrnes 2006b).
In sum, SFL celebrates exuberance in how we investigate language and,
by extension, foster its development, including in instructional contexts.
That may appear to be insufficiently ‘scientific’, too much subject to inter-
pretation. It may also appear to make the task of language learning and
teaching more formidable yet than we already know it to be. However – as
I hope has become clear – just such a multidimensional, yet principled
analysis is indispensable if we are to capture the inherently contextual,
adaptively dynamic, and multiply layered ways in which we learn, through
language, to make meaning of and in our world.
20.4 Researching Instructed L2 Development through an

SFL Lens: Foci, Concepts, Analytical Tools
On that basis, we are now ready to explore how core SFL constructs have
guided research and L2 educational practice. Rather than following more
customary divisions by syntax or vocabulary or different modalities of
language use, I have chosen foci derived from SFL constructs in order to
showcase its unique capacity to address long-term development. Also, to
the extent possible, I have chosen studies that involve different languages,
different learner ages, different educational settings, different performance
levels, and different facets of L2 use.
20.4.1 Meaning-Making as Choice: L2 Development as Evolving

Meta-Awareness
The first two studies provide evidence for how conceptualizing meaning-
making as making choices can foster learners’ evolving meta-awareness
about the inner workings of language and, by extension, facilitate L2
learning. Such an approach remains agnostic about the role of explicit
and implicit knowledge and the limitations of processing space, an issue
that continues to dominate cognitivist SLA investigations into how learners
establish form–meaning relationships (see Godfroid 2016). Instead, SFL-
inspired pedagogies focus on enabling learners to construe meaning–form
interfaces at all strata of the system by enlisting the meaning-oriented
analytical constructs the theory makes available – for instance, an aware-
ness of how the kinds of actors and processes chosen in a particular verbal
exchange tend to yield more this vs. another meaning within a probabilistic
space of flexibility and stability. Not only can awareness of available choices
within a networked system and then actually following up on them become
524 HEIDI BYRNES
a powerful form of agency and identity formation; it can engender self-

perceptions on the part of learners as evolving competent knowers that
might, for them, influence the very nature of how they process and use
language.
For example, Schleppegrell (2004b) shows how an immigrant student in a
chemical engineering laboratory course gradually gained awareness of the
way in which the sciences construe meaning through the deployment of GM.
This was made possible by a pedagogy that increased the learner’s awareness
of the valued aspects of the genre laboratory report, among them, the need
to convey technicality through nominalizations, as against the wordiness of
informal language use; the higher information density made possible
through in-clause or within-phrase reasoning (when the temperature is raised
vs. with increasing temperature) vs. subordination; more overt textual structur-
ing through the strategic chaining of theme–rheme structures that involve
the use of passive voice and nominalizations; and the carefully nuanced play
with intersubjective meaning that gradually moves from an implicit subject-
ive way of portraying facts to more implicit objective wordings (these results
must be in error vs. clearly, these results are in error). In other words, this is not
‘mere language learning’; rather, the student gradually becomes a member
of a community of inquirers in a particular area of knowledge construction,
knowledge creation, and knowledge dissemination.
While Schleppegrell’s study addressed the development of text-level abil-
ities in a particular disciplinary area, Teruya’s (2006) focus of inquiry is that
of clause complexing. Within the larger domain of the experiential function,
clause complexing is all about imposing particular logico-semantic relation-
ships on a given figure in semantic space, such as cause, various time
relationships, or conditions. Once more, the issue is not so much learning
how to produce complex clauses as a formal syntactic phenomenon; nor is it
simply about conveying how this particular sentence is part of a larger
narrative episode. Rather, the intention is to show ‘how clause complexing
represents an important gateway between the semantics of text and the
grammar of the clause’ (Teruya 2006:110–11): by using an SFL-grounded
description of how clause complexes function in Japanese, learners, over a
semester-long course, came to develop a kind of ‘grammatical thinking’,
both through analyzing textual passages and through producing them. That
such conscious knowledge is akin to development in the Vygotskian sense is
well demonstrated by the students’ learning journals, which revealed how
they came to understand Japanese clause complexing as realizing logical
relationships that differ decisively from those that are customary in English.
20.4.2 Fostering Development through a Textual Orientation

Numerous studies have shown that awareness of how texts function can
just as successfully be fostered in younger learners and at much earlier
stages of L2 development.
In a study in a 5th grade L2 writing class of minority students in an urban

setting in the United States, Harman (2013) reports how the teacher encour-
aged her students to borrow and play with diverse lexical patterns as they
explored and analyzed other texts, typically narratives, in children’s litera-
ture. Among its highlights are not only the use of literature-based writing
instruction, but the use of a genre-based pedagogy that foregrounded inter-
textuality. Thus, over the five months of the longitudinal study, the teacher
focused on raising students’ awareness of cohesion in literary narratives
through patterns of lexical cohesion based on semantic relations (e.g.,
through class/subclass, part/whole, and similarity/repetition patterns), in
terms of expectancy relations (e.g., predictability between a process and
the object of a process), and through cohesion created through various
conjunctive relations. Noteworthy in the study’s diverse findings is the
considerable expansion of the students’ lexical repertoire, brought on by
the ‘frequent disruption of expected lexical relations and figurative play’
(Harman 2013:137) characteristic of much children’s literature.
A genre-based pedagogy is also in focus in the study by Bunch and Willett
(2013), this time in a middle school with a considerable population of
linguistic minority students. In line with the instructional focus on learn-
ing to write while also learning social studies content, the culminating
writing assignment of the focal unit on the Reformation asked students
to persuade members of their family either to join or to oppose Martin
Luther’s reform movement. A particularly revealing finding, even if ‘nega-
tive’, was this: students’ compositions repeatedly exposed the tension
inherent in the writing prompt, which required an appeal to the emotions
of their family members and adherence to the five-paragraph essay form
that is nearly canonical in US schools for academic-level argumentative
writing. The need to address in a single essay two very different audiences
and purposes virtually guaranteed inauthentic texts because of the mixing
of two different genres.
20.4.3 Exploring the Oral–Literate Continuum from a Semiotic

Perspective
The attainment of literate capacities has long been seen as a marker of L2
development. Investigations have tended to (a) focus on the expansion of
syntactic abilities in terms of CAF (see the earlier discussion); (b) follow the
type of genre analysis prevalent in Language/English for Specific Purposes
(Bhatia 2004; Swales 1990) or a more sociologically oriented New Rhetoric;
or (c) deploy multidimensional register analysis for spoken and written
language, recently enhanced through corpus-based procedures (Biber and
Gray 2016).
Reflecting discussion earlier in this chapter, SFL takes a strongly language-
based approach that emphasizes a continuum of situated semiotic practices.
Under a syntactic optic, development moves from the choreographic
526 HEIDI BYRNES
complexity of spoken language to the crystalline complexity of written

language enabled primarily through GM; under a semiotic optic, a dynamic,
common-sense way of construing the world in terms of function and flow
gradually expands into the ability to see reality in terms of objects; and
under the L2 learning optic, development reflects the considerable cognitive
and linguistic challenge involved in reconstituting both content and
expression.
Among a range of insights, SFL-inspired L2 research well corroborates the
language-based nature of cognitive development. Thus, in Byrnes’ (2009)
longitudinal study into the emergent capacity of L2 college-level learners to
deploy GM, its literate learners who, in their L1, clearly possessed the
semiotic capacity for synoptic semiosis, were not able to simply ‘transfer’
that ability into the L2 German environment. Rather, precisely because the
relationship between meaning and knowing is complexly language-specific,
they needed, once more, to traverse it in the L2. However, while literacy
development in the L1 tends to be years in the making, the study’s early
advanced learners were able to accomplish this considerable feat within one
semester of genre-based instruction that continued the programme’s
efforts to create a sophisticated awareness of the functional nature of
language, here, of the nature of GM.
A similar finding applies to the development of academic registers by
college-level heritage speakers. While their abilities in handling the spoken
discourses of familiarity tend to outpace those of purely instructed L2
learners, they benefit from developing academic-level abilities in writing.
Particularly gratifying is the ‘spill-over effect’, as it were, of the expansion
of resources for academic writing on high-level speaking, something that
allowed them to claim membership in a new aspirational professional
discourse community (Achugar and Colombi 2008).
20.4.4 Attending to Language-Based Knowledge Construction

in the Disciplines
While I have already addressed it indirectly, here I focus on SFL’s consider-
able interest in the link between ways of languaging and ways of knowing,
especially by exploring knowledge construction in different languages, in
different modalities, and in different disciplines (Martin 2007).
A case in point is Caffarel’s (2006) study located in a college-level linguis-
tics class, taught in L2 French, which focused on the resources peculiar to
French for creating cohesion. However, its deeper intention was to lay bare
the nature of the construal of ideology in news media through a particular
literary style. As class participants continuously interwove written and
spoken language and reflexively deconstructed the (French) language of
the media texts through their own use of French in terms of the disciplin-
ary language of linguistics that they were acquiring, they gained first-hand
experience in how the interface of context, register, modality, and
metafunctions contributed to different forms of knowledge construction –

from the printed L1 French media, to the discipline of linguistics, to oral
interactions in this L2 French ‘content class’.
By comparison, the study by Ryshina-Pankova and Byrnes (2013) was
motivated by an interest in establishing how/whether L2 compositions
provided evidence for a collegiate German programme’s foundational
assumption that language and content learning occurred simultaneously
in its integrated genre-oriented and task-based four-year undergraduate
programme (see Byrnes et al. 2010). Its findings revealed that, although
the ability to deploy nominalized GM was important, a crucial additional
step was necessary for such deployment to be legitimately interpreted as
evidence for knowledge creation, that is, ‘the development of thought in
language and, through that thought, the creation of knowledge’ (Ryshina-
Pankova and Byrnes 2013:182). Specifically, confirming earlier research by
Ravelli (2003) on the management of Theme and Information systems in
texts through nominalized GM, the authors concluded that ‘the conditions
for creating and conveying knowledge in writing are created when a writer
reconstrues concrete experiences through conceptual refiguration that
deploys GM; and when these abstract categorizations are related to each
other and to the congruent elements of the text, as, for example, when
strategically deployed in the position of (hyper)-theme and (hyper)-rheme’
(Ravelli 2003:192). This dual capacity of ‘conceptual refiguration’ and ‘text-
ual configuration’ accords well with Halliday’s (2004:193) characterization
of ‘discourse as a progression of steps in reasoning’.
20.4.5 Tapping into the Developmental Quality of SFL Theorizing

for Curricular Thinking
I conclude this section by returning to a claim made at the beginning of
this chapter, namely the opportunity for fruitful dialogue between SFL and
SLA in order to address critical issues in SLA research and L2 education.
Specifically, I refer to the possibility of SFL enabling the creation of inte-
grated coherent curricula as a response to an influential and yet to
be fulfilled call by an MLA committee (2007) for collegiate FL programmes
to thoroughly rethink their offerings in light of the demands of a multilin-
gual globalized world.
I have addressed at some level of detail how a curricular progression
might be imagined and realized by using SFL-inspired considerations
(e.g., Byrnes 2011, 2012, 2014a, 2014b, 2015, 2017; Byrnes et al. 2010). In
this chapter, the following characterization will have to suffice. Systemic
Functional Linguistics provides a sufficiently broad framework for integrat-
ing local conditions and interests and makes sufficiently specific recom-
mendations to enable curricular and pedagogical action of consequence. At
the same time, it offers theoretically grounded, language-based criteria for
assessing L2 development at different performance levels. That facilitates a
528 HEIDI BYRNES
favourable feedback loop into language programmes on the basis of inter-

pretable evidence that, in turn, can lead to further enhancing development.
In many ways, the studies summarized in this chapter provide exemplars
for how educators can use the constructs provided by SFL to envision and,
ultimately, implement curricular progressions that explicitly address con-
tinued L2 development.
Within the space constraints of this chapter, let me exemplify these
points by referring to the SFL-inspired curricular work in my home
department. Using a genre perspective that is translated into genre-based
pedagogical tasks, the curriculum adopts a strong textual focus, with
particular emphasis on reading and writing supported by class discussion
and listening activities, where the chosen texts model how content is
realized through language form. Specifically, in a five-level progression
Level 1 models short functional oral and written texts that highlight the
dialogic nature of meaning construction as that occurs over an entire text.
In Level 2, whose focus is on narrative genres, the prototypical narrative is
the personal story with primary information structuring of the key actors
and their actions through chronology. Level 3 gradually moves from
personal to public narratives, highlighting different positions and rela-
tionships among the actors in terms of expanding logico-semantic rela-
tionships, such as cause and effect, and different facets of expressing time
and place. Although various precursors have already appeared at Level 3,
the pronounced shift from dynamic to synoptic semiosis occurs in
Level 4 by moving from concrete to abstract topics instantiated in genres
of public life with their preference for objectification of the world
and infusion with social and individual values. Level 5 continues this
work in diverse genres that broadly reflect the programme’s literary
cultural content orientation as it seeks to enhance all modalities of
language use.
20.5 Future Work and Outlook
As for future work in SFL on L2 development, the following list extrapolates

from evolving expansions in different contexts and domains of research
and practice.
First, judging from the increasing presence in major ‘mainstream’
research journals (not just monographs where SFL researchers have
tended to create and cater to their own in-house audience) and at profes-
sional conferences, such as the American Association of Applied Linguis-
tics, SFL will increasingly be challenged, in the positive sense of the word,
to take up much more explicitly the assumptions and interests driving
discussions in the larger field of SLA and then to influence them. To
mention just two examples, the proposal for a transdisciplinary frame-
work for the study of language by the Douglas Fir Group (2016), made up
of some of the best-known researchers in SLA, provides an extraordinary

opportunity for entering an expansive discussion at the forefront of the
field. More specifically, a recent collection of research articles focused on
the construct of complexity in writing favourably noted how SFL might
be able to move the writing field, long enamoured with psycholinguisti-
cally framed CAF considerations, forward on account of its meaning-
oriented perspective (Ortega 2015). Such openings can help demystify
SFL, a theory that is often perceived as too complicated, by placing its
constructs into problem spaces that are familiar to the audience. With
such accruing ‘translations’, SFL researchers should be able improve their
ability to present their work and, gradually, to influence discourse in
the field.
Second, there is every indication that SFL researchers engaged in
teacher education (e.g. Schleppegrell and Gebhard in US ESL contexts)
will continue to make a persuasive case that teachers – and, by extension,
their learners – can benefit enormously by incorporating SFL constructs
into their pedagogical practices. Because many teacher education pro-
grammes, particularly at the graduate level, involve students from around
the world, but especially Asian countries where English-language teaching
is now a mainstream effort, SFL might provide a worthy alternative to a
communicative language teaching that is in need of adjustments in vari-
ous cultural contexts. As teaching models not just for EFL/ESL but also for
other languages are proliferating (e.g. dual immersion programmes,
content-based instruction, biliteracy agendas (e.g. see Gentil 2011)), dia-
logues regarding curricular, instructional, and assessment approaches
will also expand.
Third, multimodality has become central to knowledge creation and
transmission in many instructional environments. Systemic Functional
Linguistics has a well-established record in that regard that is likely to
enrich both its internal understanding of this semiotic environment
and also the praxes in diverse educational settings (e.g., Whittaker
et al. 2007; Kress 2009; Bednarek and Martin 2010; O’Halloran 2011;
Painter et al. 2013).
Fourth, while corpus-based work is long-established in SFL, its recent rise
in mainstream SLA is both an opportunity for interchange (see Bednarek’s
2015 investigation of emotion talk from an SFL perspective; or work pub-
lished in Whittaker et al. 2007) and a challenge (see the multidimensional
but, ultimately, limited functional approach by Biber and his influential
research group).
Fifth, the explicit link in SFL between language and content learning can
contribute favourable theoretical underpinnings for content and language
integrated learning (CLIL), one of the most vibrant recent educational
initiatives in Europe (e.g., see Llinares et al. 2012; Nikula et al. 2016).
Sixth, language use in diverse professional settings (e.g., healthcare, law,
business, engineering) is also likely to expand. As demonstrated in
530 HEIDI BYRNES
Matthiessen (2013), SFL offers sophisticated analytical tools for a differenti-

ated understanding of communication in these settings.
Finally, multilingualism and, by extension, the multilingual potential for
meaning-making can be handled with great acuity and insight through a
network-driven approach to meaning-making and choice as it characterizes
SFL (e.g., see Matthiessen 2018; Matthiessen et al. 2008). The contributions
in Kerfoot and Simon-Vandenbergen (2015), which are focused on the
multilingual environment of South Africa, demonstrate well the theory’s
fecundity for enhancing literacy development in diverse educational
settings.
In each of these areas one has every reason to anticipate reaping the
benefits of the kind of appliable research so prominently advocated by
Halliday, including its keen sense of social responsibility for our research
enterprise.
20.6 Recommended Readings
My recommendations for additional readings return to my two imagined

reader groups, SLA researchers and those generally familiar with SFL,
though not necessarily with L2 issues. I have arranged the chosen
monograph-length treatments from the more accessible to the more tech-
nical in SFL; I conclude with some overview treatments of recent work
in SLA.
As their title suggests, Coffin et al. (2009) provide a guide to transitioning
from formal to functional grammars. Intended for language teachers, it
offers a down-to-earth treatment of key SFL positions. Similarly down-to-
earth is Christie’s (2012) treatment of L1 language education, which moves
from early childhood into adulthood in the context of schooling. Byrnes
et al. (2010) is the only treatment of a comprehensive college-level foreign
language programme whose curricular, pedagogical, and assessment prac-
tices were directly inspired by SFL. It occupies a unique position for having
tackled the multiple challenges of language and content learning, including
ab initio learning, in higher education. Moving to publications that refer
more directly to SFL as a functional theory, I recommend in sequence the
edited volume by Martin (2013) of interviews with Halliday that span forty
years of reflection on problems in linguistic theorizing, exemplifying his
remarkably accessible and absorbing manner of presenting complex issues.
Halliday treats some of them extensively in Complementarities in Language
(2008). More challenging but richly rewarding is Halliday and Matthiessen’s
(2006) Construing Experience through Meaning, which provides an excellent,
still mostly non-technical summary of what the theory offers in relation to
language and thinking, including in an L2. Coffin and Donohue (2012, 2014)
present well-argued treatments of how SFL can address issues in higher
education. My two concluding recommendations are intended to give a
sense of the shifting tectonics in linguistic-cognitive SLA research,

Atkinson’s (2011) treatment of alternative approaches to SLA, and the focus
issue of Language Learning, guest edited by Ellis and Larsen-Freeman (2009),
whose contributions on language as a complex adaptive system open up a
host of opportunities for two-way dialogue.
References
Achugar, M. and M. C. Colombi. 2008. Systemic Functional Linguistic

Explorations into the Longitudinal Study of Advanced Capacities: The
Case of Spanish Heritage Language Learners, in L. Ortega and H. Byrnes,
eds., The Longitudinal Study of Advanced L2 Capacities. New York: Routledge.
36–57.
Atkinson, D., ed. 2011. Alternative Approaches to Second Language Acquisition.
London: Routledge.
Bednarek, M. 2015. Emotion Talk across Corpora. Basingstoke: Palgrave
Macmillan.
Bednarek, M. and J. R. Martin, eds. 2010. New Discourse on Language:
Functional Perspectives on Multimodality, Identity, and Affiliation. London:
Continuum.
Bhatia, V. K. 2004. Worlds of Written Discourse: A Genre-based View. London:
Continuum.
Biber, D. and B. Gray. 2016. Grammatical Complexity in Academic English:
Linguistic Change and Writing. Cambridge: Cambridge University Press.
Bunch, G. and K. Willett. 2013. Writing to Mean in Middle School: Under-
standing How Second Language Writers Negotiate Textually-rich
Content-area Instruction. Journal of Second Language Writing 22: 141–78.
Byrnes, H., ed. 2006a. Advanced Language Learning: The Contribution of Halliday
and Vygotsky. London: Continuum.
Byrnes, H. 2006b. What Kind of Resource is Language and Why Does It
Matter for Advanced Language Learning? An Introduction. In H. Byrnes,
ed., Advanced Language Learning: The Contribution of Halliday and Vygotsky.
Byrnes, H. 2009. Emergent L2 German Writing Ability in a Curricular
Context: A Longitudinal Study of Grammatical Metaphor. Linguistics and
Education 20: 50–66.
Byrnes, H. 2011. Beyond Writing as Language Learning or Content Learn-
ing: Construing Foreign Language Writing as Meaning-making. In R. M.
Manchón, ed., Learning-to-write and Writing-to-learn in an Additional
Language. Amsterdam: John Benjamins. 133–57.
Byrnes, H. 2012. Conceptualizing FL Writing Development in Collegiate
Settings: A Genre-based Systemic Functional Linguistic Approach. In
R. M. Manchón, ed., L2 Writing Development: Multiple Perspectives. Berlin:
De Gruyter Mouton. 190–218.
532 HEIDI BYRNES
Byrnes, H. 2014a. Linking Task and Writing for Language Development:

Evidence from a Genre-based Curricular Approach. In H. Byrnes and R. M.
Manchón, eds., Task-based Language Learning: Insights from and for L2
Writing. Amsterdam: John Benjamins. 235–61.
Byrnes, H. 2014b. Theorizing Language Development at the Intersection of
‘Task’ and L2 Writing: Reconsidering Complexity. In H. Byrnes and R. M.
Manchón, eds., Task-based Language Learning: Insights from and for L2
Writing. Amsterdam: John Benjamins. 79–103.
Byrnes, H. 2015. Linking ‘Task’ and Curricular Thinking: An Affirmation of
the TBLT Educational Agenda. In M. Bygate, ed., Domains and Directions in
the Development of TBLT: A Decade of Plenaries from the International Conference.
Byrnes, H. 2017. Meeting the Challenge of Instructed Language Develop-
ment: Reflections on Systemic-Functional Contributions. In S. Neumann,
R. Wegener, J. Fest, P. Niemietz, and N. Hützen, eds., Challenging Boundar-
ies in Linguistics: Systemic Functional Perspectives. Frankfurt am Main: Peter
Lang. 457–91.
Byrnes, H., H. H. Maxim, and J. M. Norris. 2010. Realizing Advanced Foreign
Language Writing Development in Collegiate Education: Curricular
Design, Pedagogy, Assessment. Modern Language Journal 94(Supplement
1): iv–vi, 1–235.
Caffarel, A. 2006. Learning Advanced French through SFL: Learning SFL in
French. In H. Byrnes, ed., Advanced Language Learning: The Contribution of
Halliday and Vygotsky. London: Continuum. 204–24.
Chaiklin, S. 2003. The Zone of Proximal Development in Vygotsky’s Analy-
sis of Learning and Instruction. In A. Kozulin, B. Gindis, V. S. Ageyev,
and S. M. Miller, eds., Vygotsky’s Educational Theory in Cultural Context.
Cambridge: Cambridge University Press. 39–64.
Christie, F. 2012. Language Education throughout the School Years: A Functional
Perspective. Language Learning Monograph Series. Malden: Wiley-Blackwell.
Christie, F. and B. Derewianka. 2008. School Discourse: Learning to Write across
the Years of Schooling. London: Continuum.
Coffin, C. and J. Donohue. 2012. Academic Literacies and Systemic Func-
tional Linguistics: How Do They Relate? Journal of English for Academic
Purposes 11: 64–75.
Coffin, C. and J. Donohue. 2014. A Language as Social Semiotic-based Approach
to Teaching and Learning in Higher Education. Language Learning
Monograph Series. Oxford: Wiley.
Coffin, C., J. Donohue, and S. North. 2009. Exploring English Grammar: From
Formal to Functional. New York: Routledge.
Cook, V. 2008. Multi-competence: Black Hole or Wormhole for Second
Language Acquisition Research? In Z. Han, ed., Understanding Second
Language Process. Clevedon: Multilingual Matters. 16–26.
Douglas Fir Group. 2016. A Transdisciplinary Framework for SLA in a
Multilingual World. Modern Language Journal 100: 19–47.
Ellis, N. C. and D. Larsen-Freeman, eds. 2009. Language as a Complex Adaptive

System. Malden: Blackwell.
Gentil, G. 2011. A Biliteracy Agenda for Genre Research. Journal of Second
Language Writing 20: 6–23.
Godfroid, A. 2016. The Effects of Implicit Instruction on Implicit and Expli-
cit Knowledge Development. Studies in Second Language Acquisition 36:
177–215.
Edward Arnold.
Halliday, M. A. K. 1996. On Grammar and Grammatics. In R. Hasan,
C. Cloran, and D. G. Butt, eds., Functional Descriptions: Theory in Practice.
Halliday, M. A. K. 2002. Spoken and Written Modes of Meaning. In J. J.
Webster, ed., The Collected Works of M. A. K. Halliday, Volume 1: On Grammar.
Halliday, M. A. K. 2004. On the Grammar of Scientific English. In J. J.
Webster, ed., The Collected Works of M. A. K. Halliday, Volume 5: The Language
of Science. London: Continuum. 181–98.
Halliday, M. A. K. 2007a. A Language Development Approach to Education
(1994). In J. J. Webster, ed., The Collected Works of M. A. K. Halliday, Volume 9:
Language and Education. London: Continuum. 368–82.
Halliday, M. A. K. 2007b. The Notion of ‘Context’ in Language Education
(1991). In J. J. Webster, ed., The Collected Works of M. A. K. Halliday, Volume 9:
cial Press.
through Meaning. A Language-based Approach to Cognition. London:
Continuum.
Harman, R. 2013. Literary Intertextuality in Genre-based Pedagogies. Build-
ing Lexical Cohesion in Fifth-grade L2 Writing. Journal of Second Language
Writing 22: 125–40.
Hasan, R. 1995. On Social Conditions for Semiotic Mediation: The Genesis
of Mind in Society. In A. R. Sadovnik, ed., Knowledge and Pedagogy:
The Sociology of Basil Bernstein. Norwood: Ablex. 171–96.
Housen, A., F. Kuiken, and I. Vedder, eds. 2012. Dimensions of L2 Performance
and Proficiency: Complexity, Accuracy and Fluency in SLA. Amsterdam: John
Benjamins.
Hulstijn, J. H., R. Ellis, and S. Eskildsen, eds. 2015. Orders and
Sequences in the Acquisition of L2 Morphosyntax, 40 Years On.
Language Learning 65.
534 HEIDI BYRNES
Jackson, D. O. and S. Suethanapornkul. 2013. The Cognition Hypothesis:

A Synthesis and Meta-analysis of Research on Second Language Task
Complexity. Language Learning 63: 330–67.
Kerfoot, C. and A. Simon-Vandenbergen. 2015. Language in Epistemic
Access: Mobilising Multilingualism and Literacy Development for More
Equitable Education in South Africa. Language and Education 29: 177–85.
Kress, G. 2009. Multimodality: A Social Semiotic Approach to Contemporary
Communication. New York: Routledge.
Lafford, B. A. 2007. Second Language Acquisition Reconceptualized?
The Impact of Firth and Wagner (1997). Modern Language Journal 91:
735–56.
Lantolf, J. P. 2005. Sociocultural and Second Language Learning Research:
An Exegesis. In E. Hinkel, ed., The Handbook of Research in Second Language
Teaching and Learning. Mahwah: Lawrence Erlbaum. 335–53.
Larsen-Freeman, D. 2009. Adjusting Expectations: The Study of Complexity,
Accuracy, and Fluency in Second Language Acquisition. Applied Linguistics
30: 579–89.
Larsen-Freeman, D. 2014. Another Step to Be Taken: Rethinking the
End Point of the Interlanguage Continuum. In Z. Han and E. Tarone,
eds., Interlanguage: Forty Years Later. Amsterdam: John Benjamins. 203–20.
Larsen-Freeman, D. 2015. Saying What We Mean: Making a Case for
‘Language Acquisition’ to Become ‘Language Development’. Language
Teaching 48: 491–505.
Llinares, A., T. Morton, and R. Whittaker, eds. 2012. The Roles of Language in
CLIL. Cambridge: Cambridge University Press.
Martin, J. R. 1991. Intrinsic Functionality: Implications for Contextual
Theory. Social Semiotics 1: 99–162.
Martin, J. R. 1997. Analysing Genre: Functional Parameters. In F. Christie
and J. R. Martin, eds., Genre and Institutions: Social Processes in the Workplace
and School. London: Continuum. 3–39.
Martin, J. R. 2007. Construing Knowledge: A Functional Linguistic Perspec-
tive. In F. Christie and J. R. Martin, eds., Language, Knowledge and Peda-
gogy: Functional Linguistic and Sociological Perspectives. 34–64. London:
Continuum.
Martin, J. R. 2009. Genre and Language Learning: A Social Semiotic Perspec-
tive. Linguistics and Education 20: 10–21.
Martin, J. R., ed. 2013. Interviews with M. A. K. Halliday: Language Turned Back
on Himself. London: Bloomsbury.
Matthiessen, C. M. I. M. 2007. The ‘Architecture’ of Language According
to Systemic Functional Theory: Developments since the 1970s. In
R. Hasan, C. M. I. M. Matthiessen, and J. J. Webster, eds., Continuing Discourse
on Language: A Functional Perspective, Volume 2. Sheffield: Equinox. 505–61.
Emerging from Acts of Meaning. Language Learning 59(Supplement 1):
211–35.
Matthiessen, C. M. I. M. 2013. Applying Systemic Functional Linguistics in

Healthcare Contexts. Text & Talk 33: 437–66.
Matthiessen, C. M. I. M. 2018. The Notion of a Multilingual Meaning
Potential: A Systemic Exploration. In A. Sellami-Baklouti and L. Fontaine,
eds., Perspectives from Systemic Functional Linguistics. London: Routledge.
90–120.
Matthiessen, C. M. I. M., K. Teruya, and C. Wu. 2008. Multilingual Studies as
a Multi-dimensional Space of Interconnected Language Studies. In J. J.
Webster, ed., Meaning in Context: Strategies for Implementing Intelligent Appli-
cations of Language Studies. London: Continuum. 146–220.
MLA Ad Hoc Committee on Foreign Languages. 2007. Foreign Languages
and Higher Education: New Structures for a Changed World. Profession
2007: 234–45.
Nikula, T., E. Dafouz, P. Moore, and U. Smit, eds. 2016. Conceptualising
Integration in CLIL and Multilingual Education. Bristol: Multilingual
Matters.
Norris, J. M. and L. Ortega. 2000. Effectiveness of L2 Instruction:
A Research Synthesis and Quantitative Meta-analysis. Language Learning
50: 417–528.
Norris, J. M. and L. Ortega. 2009. Towards an Organic Approach to Investi-
gating CAF in Instructed SLA: The Case of Complexity. Applied Linguistics
30: 555–78.
O’Halloran, K. L., ed. 2011. Multimodal Studies: Exploring Issues and Domains.
New York: Routledge.
Ortega, L. 2015. Syntactic Complexity in L2 Writing: Progress and Expan-
sion. Journal of Second Language Writing 29: 82–94.
Ortega, L. and H. Byrnes. 2008. Theorizing Advancedness: Setting up
the Longitudinal Research Agenda. In L. Ortega and H. Byrnes, eds.,
The Longitudinal Study of Advanced L2 Capacities. New York: Routledge.
281–99.
Ravelli, L. 2003. Renewal of Connection: Integrating Theory and Practice in
an Understanding of Grammatical Metaphor. In A. Simon-Vandenbergen,
M. Taverniers, and L. Ravelli, eds., Grammatical Metaphor: Views from
Systemic Functional Linguistics. Amsterdam: John Benjamins. 37–64.
Ryshina-Pankova, M. and H. Byrnes. 2013. Writing as Learning to Know:
Tracing Knowledge Construction in L2 German Compositions. Journal of
Second Language Writing 22: 179–97.
Schleppegrell, M. J. 2004a. The Language of Schooling: A Functional Linguistics
Perspective. Mahwah: Lawrence Erlbaum.
Schleppegrell, M. J. 2004b. Technical Writing in a Second Language: The
Role of Grammatical Metaphor. In L. J. Ravelli and R. A. Ellis, eds.,
Analysing Academic Writing: Contextualized Frameworks. London: Con-
tinuum. 172–89.
536 HEIDI BYRNES
Swales, J. M. 1990. Genre Analysis: English in Academic and Research Settings.

Teruya, K. 2006. Grammar as a Resource for the Construction of Language
Logic for Advanced Language Learning in Japanese. In H. Byrnes, ed.,
Advanced Language Learning: The Contribution of Halliday and Vygotsky.
Whittaker, R., A. McCabe, and M. O’Donnell, eds. 2007. Advances in Language
and Education. New York: Continuum.
21
Language and Education
Learning to Mean
Peter Mickan
21.1 Halliday on Language and Education
The distinctive characteristic of human learning is that it is a

process of making meaning – a semiotic process; and the proto-
typical form of human semiotic is language. (Halliday 1993:93)
In this chapter, I review SFL contributions to language education in pri-

mary, secondary, and tertiary sectors. I have included sections on teacher
education and suggestions for continuing the rich exploration of SFL theory
applied in educational practice.
Systemic Functional Linguistics (SFL) continues to be a powerful influ-
ence on education – in the conception of language as social semiotic, in the
analysis of texts in context, and in the understanding of language learning
EDUCATION as a social process of learning to mean. For Halliday (2007b:270) education education is a field of
FOR is a field of activity ‘where we investigate how language functions in activity ‘where
HALLIDAY - we investigate how
various educational contexts, and by doing so, seek to improve our educa-
tional practice’ with our concern ‘always with language in context’. language functions in
Halliday’s early professional interest in the application of linguistic various educational
knowledge in support of effective instruction involved his initiation of contexts,
action research projects working with teachers of English in schools in - and by doing so,
seek to improve our
the UK: Breakthrough to Literacy (Mackay et al. 1978); Language and Communi-
educational
cation (Forsyth and Wood 1977–1980), which was a project aimed at late
practice’ with our
primary and early secondary education; and Language in Use (Doughty et al. concern ‘always with
1971), a project aimed at secondary-school education (see Williams language in context’.‘
2005:284). The projects premised language learning as a social process:
‘At the very least, “being socialised” means entering into and mastering a
large number of discourses’ (Halliday 2007a:90).
The SFL linguistic model is based on naturally occurring texts in contexts,
and on accounting for the relationship between language and the social
context (see Halliday 1975, 1978). The SFL theorization of this relationship
538 PETER MICKAN
places ‘meaning’ at its core. Halliday’s concept of register (or text type)
gives a means of understanding why certain grammatical choices are made
in a text, and relates text type to context, thus providing an explanation for
Register thus provides a language variation across societies (see Bowcher, this volume). Register thus
link between social provides a link between social context and text. Further, the theory recog-
context and text.
nizes and accounts for complementarities in language and different
instances of language use. For instance, proposing a trinocular perspective,
Halliday (2008:141) explains:
When we are observing and investigating language, or any other semiotic

system, our vision is essentially trinocular. We observe the phenomenon we
want to explore – say, the lexicogrammar of language – from three points of
vantage. We observe it from above, in terms of its function in various
contexts. We observe it from below, in terms of its various modes of expres-
sion. And thirdly, we observe it from its own level: from within, or from
round about, according to whether we are focussing on the whole or some of
its parts.
This view of language in relation to social context, or language as social

The process of learning semiotic, underlies Halliday’s educational linguistics. The process of learn-
involves participation in ing involves participation in acts of meaning. Halliday explains language
acts of meaning. learning as the development of a meaning-making resource: ‘In all educa-
language learning
tion, the learner has to build up a resource. It is a resource of a particular
development of a
meaning-making kind: a resource for creating meaning’ (Halliday 2007b:274). As a theory
resource applied in practice, SFL is a core resource for teachers’ understanding of the
role of language in learning. For instance, Hasan’s research has contributed
to an understanding of the influence of the home language and socializa-
tion experience for making sense of school language experiences (Hasan
2002). This can be seen in her work on semantic variation (Hasan 2009c; see
also Low and Fung, this volume).
Halliday has fostered the application of SFL in education across sectors
and levels of education. Systemic Functional Linguistics informs a pedagogy
for building learners’ awareness and management of various types of dis-
course with a focus on texts or genres in social contexts, and equips
teachers with the perspective and access to a system for the analysis
and teaching of language functioning in knowledge-building across social
Functional linguistic contexts, texts, and lexicogrammar.
pedagogy prepare SFL research in education is situated typically in classrooms with the
students with knowledge direct involvement of teachers in action research projects in multiple
about
- language for managing the
educational domains. Functional linguistic pedagogy is designed to prepare
literacy changes students with knowledge about language for managing the literacy changes
- variations across levels of and variations across levels of schooling and in different social domains.
schooling and in different Building students’ knowledge of how language functions across the cur-
social domains. riculum equips them with a resource for cultural participation in schools TEXTS are the
knowledge of how and in society in general. Texts are the units of analysis and the texts which units of analysis
language functions and the texts
comprise the curriculum are examined in contexts of use. The application which comprise
across the curriculum the curriculum
equips students with a are examined in
resource for cultural contexts of use.
participation
Downloaded from in schools
https://www.cambridge.org/core. Columbia University - Law Library, on 11 Aug 2019 at 09:44:17, subject to the Cambridge Core terms of use,
andavailable
in society in general
Language and Education 539
of SFL in education extends from early childhood education through the

years of schooling to adult learning in higher education.
21.2 SFL Applied in Education: Explicit Pedagogy
What is significant about SFL applied in language education is that it

advances our insights into the social function of texts and into the
lexicogrammatical resources we use to realize different meanings
according to context. Researchers since the 1960s have stressed the role
of language in education (Britton 1970; Barnes 1969) but the methodical
documentation of the texts in curricula based on the theoretical frame-
work of SFL has increased our understanding of the centrality of lan-
guage in learning, in teacher education, and in educational research.
SFL depicts with explicitness the reasons for the differences in texts –
between spoken and written language, between formal and casual texts –
and explores the language choices for the realization of different
meanings.
SFL influence also extends across institutions and across the curriculum
with analyses of genres for building students’ knowledge about language
The analysis of
genres depicts (Christie 2012; Christie and Derewianka 2008; Rothery 1996; Painter and
the social Martin 1986). For Martin (2009:159), ‘Genre represents the system of
context staged goal-oriented social processes through which social subjects in a
configured in The analysis of generic
language given culture live their lives.’ The analysis of genres depicts the social
structure and of typical
selections context configured in language selections modelled along the three par- lexicogrammatical
modelled along ameters of Field, Tenor, and Mode, realized in three functions or major selections constituting
the three par different genres underpins
ameters purposes of language in use: Ideational, Interpersonal, and Textual. The
genre pedagogy, designed to
of Field, Tenor, analysis of generic structure and of typical lexicogrammatical selections suport students’ access to
and Mode, powerful language use
constituting different genres underpins genre pedagogy, designed to sup-
realized in three across socioeconomic
functions or port students’ access to powerful language use across socioeconomic con-
conditions
major ditions (Christie 2012; Rose 2015). The pedagogy framed as a teaching and Christie 2012; Rose 2015).
purposes of
learning cycle (see Figure 21.1) applies explicit instruction in the sche-
language in use:
Ideational, matic structure and lexicogrammatical features of genres taught
Interpersonal, in schools.
and Textual.
The teaching–learning cycle has three main phases of activity: learners
deconstruct or analyze a model genre, work together to construct one based
on the model, and then individually construct a genre on their own.
Instruction commences with the teacher leading a discussion on the social
function of a genre and modelling a genre. A genre’s structure and gram-
mar are analyzed from the perspective of the social purpose. The class
composes a text together, based on the model. This is a prelude to students
writing their own genres. The pedagogy asserts the important role of a
teacher in guiding and scaffolding students’ awareness of the social pur-
pose and features of a selected genre, and to build the discourse resources
540 PETER MICKAN
Figure 21.1 Genre teaching and learning cycle (Rothery 1994; also see Rose and
Martin 2012)
for learners’ composing their own texts and managing discourses for suc-
cess in written work in schools.
In broadening a focus from writing instruction to reading comprehension,
Martin and Rose (Martin and Rose 2008; Rose and Martin 2012; Rose 2011a,
2015) have developed a reading to learn pedagogy with a focus on strategies
for reading texts as preparation for students’ comprehension and writing of
texts. In an interaction cycle, teachers scaffold preparation for reading, for
task engagement, and for elaboration. In the preparation phase, teachers
talk with students about the language features of texts for identification and
comprehension of content so that students gain control of patterns of
discourse and of the subject field which they can apply in their own writing.
The approach assists students’ understanding of the grammatical features of
written texts for making sense of the content or subject matter of texts. The
associated teacher professional development programme (Rose 2015) details
strategies to support learners’ independence and success in reading and
writing for meaning across curricula and year levels.
Text-based curriculum
(Mickan 2013, 2017) situates Pedagogy based on authentic texts has been applied to text-based syllabus
texts in communities of and methodology (de Silva Joyce and Feez 2012). Text-based curriculum
practice and assumes people’s
participation in social practices
(Mickan 2013, 2017) situates texts in communities of practice and assumes
with semiotic resources, as people’s participation in social practices with semiotic resources, as illus-
illustrated in Figure 21.2. trated in Figure 21.2. Learning to use
Learning to use the resources of a language system is a process of social- the resources of a
ization which depends on students’ intense and rich engagement with language system
is a process of
texts. By the time children begin schooling, they are familiar with many socialization
different text types and have a register knowledge for working with and which depends
on students’
intense and rich
engagement with
textsof use,
Downloaded from https://www.cambridge.org/core. Columbia University - Law Library, on 11 Aug 2019 at 09:44:17, subject to the Cambridge Core terms
Figure 21.2 People’s participation in society (Mickan 2013:34)
responding to texts. The focus of text-based pedagogy (Mickan 2013, 2017) The focus of text-based
pedagogy (Mickan 2013,
is meaning-making, with language a primary semiotic resource for stu- 2017) is meaning-making,
dents’ development of discourses for engagement in curriculum practices. with language a primary
PREPARING
semiotic resource for students’
STUDENTS Students build on their prior knowledge of language in use with analysis of
development of discourses for
FOR THE context, text types, and lexicogrammar with teachers’ scaffolding support engagement in curriculum
EXPRESSION
(Gibbons 2006). Observing the features of selected text types and respond- practices.
OF MEANING
IN THEIR ing to the content of texts prepares students for the expression of meanings
OWN TEXTS in their own texts.
multisemiotic Text analysis also encompasses multimodality. Ventola and Moya Guijarro
approach (2009:1) write that ‘language rarely stands alone in written and spoken
discourses, that is, mono-modally, and that we urgently need to sharpen our
tools in analysing discourses multisemiotically. We cannot continue analys-
ing language alone, but need an integrated multisemiotic approach.’ SFL
studies reveal the multisemiotic nature of texts and illustrate image–text
interactions in children’s reading and comprehension of texts (Chu 2016;
Painter 2007). The perspective extends to multidimensional analyses of sub-
jects in the curriculum. Examples include mathematics and accounting
(O’Halloran 2004, 2007; Alyousef and Mickan 2016).
A challenge in education is to map and report language learning across
years of schooling to achieve continuity, to verify cumulative learning for
reporting, and to validate measures for assessment (Macken-Horarik et al.
2011). Systemic Functional (SF) linguists have tracked language develop-
ment across levels of schooling. Christie and Derewianka’s (2008) descrip-
tions of writing development across school subjects from familiar informal
and spoken-like language to technical language identify and map grammat-
ical features in different subjects. Christie (2012) gives a broad account
based on a functional linguistic framework of language development across
years of schooling. In early childhood education, children build on spoken
542 PETER MICKAN
and informal discourses, which they adapt and develop for participation in
curriculum genres encompassing class practices increasingly focused on
literacy and the genres of school subjects. Schooling in early adolescence –
upper primary to junior secondary – reveals the importance of literacy
skills with expectations of student management of new genres and regis-
ters. Students develop knowledge, procedures, and practices of subject
specialisms. They encounter the interplay of verbal and non-verbal
resources, for example, in mathematics and science subjects.
In mid-adolescence students are dealing with abstract knowledge: ‘The
language students must read and write becomes dense, its grammatical
organization more noncongruent, increasingly unlike the more familiar
congruent expressions in which much early common-sense experience is
expressed’ (Christie 2012:105–6). Teaching involves genre analysis and
understanding of language choice in scaffolding students’ insights into
the ways language functions in different text types in subject areas such
as history, English, mathematics, and science. In late adolescence to adult-
hood students engage with theoretical knowledge and abstract phenomena
and ideas for the application of language in the ‘distinctive method of
inquiry and knowledge creation’ (Christie 2012:149). In higher education
the technicality of discourses is embedded in disciplinary practices (Mickan
2013), which is a further extension of students’ construction of the
meaning-making resources for operating in different educational domains.
SF linguists have a special interest in literacy research and teaching in
all levels of education (McCabe et al. 2009; Ravelli and Ellis 2005; Uns-
worth 2000; Whittaker et al. 2006). Hasan and Williams (1996), for
instance, documents the literacy demands on school students and the
relevance of understanding and explaining subject-specific linguistic fea-
Hasan (2011a:173) refers to tures of texts for teaching literacy. Hasan (2011a:173) refers to literacy as
literacy as ‘language-based ‘language-based semiosis’, defining literacy as acts of meaning associated
semiosis’, defining literacy as with a semiotic system. Students experience language variation in their
acts of meaning associated
with a semiotic system.
speech communities and come to school with differing discursive abilities
and ways of saying and doing. For some students the different home and
community socialization experiences prepare them for managing the
literacy expectations of schooling tasks. For other students this is not
the case.
‘Recognition literacy Hasan (2011a:179–206) distinguishes three conceptions of literacy: recog-
pedagogy nition literacy, action literacy, and reflection literacy. ‘Recognition literacy’
is teaching words and is teaching words and ‘language as inventory of forms’ (Hasan 2011a:178).
‘language as inventory of
She claims that this is ‘a conception of language far removed from language
forms’ (Hasan 2011a:178).
as social semiotic practice’ (Hasan 2011a:178), even though it remains a
popular idea of what literacy is all about. According to Hasan the teaching
of spelling, punctuation, and grammar based on set rules without explan-
ation of the relationship of lexicogrammar and semantics inculcates con-
formity amongst students: ‘We know that the norms of language children
are taught in school are precisely those which conform to the practices of a
certain segment of the society’ (Hasan 2011a:180). The social implication of

recognition literacy is the maintenance of power relations in society and its
privileges: ‘Recognition literacy would flourish best in societies that are
authoritarian, with a routinization of conduct, or at least where the nature
of educational knowledge is non-negotiable’ (Hasan 2011a:181). Recogni-
tion literacy is not concerned with the processes at the centre of language
use – contexts and language choices relevant to social situations.
‘action literacy Hasan (2011a) defines the development of students’ management of the
pedagogy
language of schooling as ‘action literacy’. Within this category of literacy
she places genre-based pedagogy. As already noted, such pedagogy is
systematic and explicit, with a replicable methodology for teaching edu-
cational genres. It selects for instruction genres valued by educational
systems, such as the genres of specific disciplines, and models the sche-
matic structure of genres to support students producing well-formed texts
The difference
between
with selection of appropriate lexicogrammatical patterning. The differ-
recognition ence between recognition literacy and action literacy as deployed in genre
literacy and teaching is learning genres as acts of meaning relevant to a student’s
action literacy
as deployed in
social life. Hasan (2011a) points out that the teaching of genres as models
genre teaching has a reproductive tendency, valuing existing knowledge structures and
respect for convention: students are assisted to produce societally valued
discourses, thus replicating existing knowledge structures and conven-
tional forms. She thus argues that ‘it is an instrument for the perpetu- CRITICA
ation of the standards of the existing educational system, and one might
go on to argue that if the ideal goal of teaching is to enable pupils to
produce new knowledge, not just replicate the existing knowledge, then
genre-based literacy falls short of this ideal’ (Hasan 2011a:194).
Hasan (2011a:193) believes that literacy is a social process with potential
for powerful action and also for exploitation, and that ‘the endpoint of
education has to be the production of new knowledge’. For this she pro-
poses the concept of ‘reflection literacy’, in which students are taught
‘to analyse, and to challenge the desirability of the prevalent ways of being,
doing and saying’ (Hasan 2011a:193). A reflection literacy pedagogy needs
A reflection to be based on teachers’ explicit knowledge of language as a system – of
literacy texts and their lexicogrammatical composition – and of a deeper under-
pedagogy standing of language as a resource for meaning. Further, ‘the teacher . . .
need[s] to sensitize pupils to not simply the overall schematic structure of
the text, which is just one aspect of discursive ability; . . . [but also] to show
what alternative ways there are of saying the “same thing”’ (Hasan
2011a:198). Thus, in reflection literacy students develop an understanding
of the wording and meaning of utterances and their social implications. She
argues:
Participation in the production of knowledge will call for an ability to use

language to reflect, to enquire, and to analyse, which is the necessary basis
for challenging what are seen as facts. So, if our aim is to enable pupils
544 PETER MICKAN
to produce knowledge, then we would need a view of literacy designed to

develop these faculties. This literacy will necessarily prioritize reflection,
enquiry and analysis.
(Hasan 2011a:196–7)
In the following sections, I review SFL applied in elementary, secondary,

and higher education sectors.
21.3 Early Childhood Education and Primary Schooling
In the transition from early childhood home experiences to schooling,

children encounter an expanding range of organizational and curriculum
practices and characteristic texts. Educational researchers (e.g. Bernstein
2000; Heath 1983; Tough 1977) have identified the critical role of preschool
linguistic experience of children as preparation for schooling and docu-
mented social differences in child–school achievement based on home
HANSAN'S
language experiences. Hasan’s sociolinguistic study of mother–child inter-
actions as ‘habitual orientations to meaning’ (e.g. Hasan 2009a, 2009b;
Hasan and Cloran 1990; also see Bernstein 1971, 1975) examined the role
of everyday talk between mothers and children in relation to the language
of schooling. Mothers audio-recorded their everyday interactions with chil-
dren in the home while eating lunch, bathing, reading books, and doing
household chores. This groundbreaking research into how mother–child
talk prepares children for the language of schooling reveals semantic vari-
ation in the expression of habitual choices of meaning through the gram-
matical selections in their interactions. Hasan (2002) describes the
influence of semiotic meditation on children’s mental disposition for
the development of specialized language resources for schooling. The analy-
sis of cultural transmission in the typical talk between child and mother
provides an explanation for children’s development of orientations to learn-
ing which impact on their success in managing and controlling the dis-
courses of schooling. The semiotic mediation in spoken exchanges
described in the study suggests the development of children’s mental
disposition to the instructional practices and discourses of schooling. The
study of preschoolers gives us insights into the power of language and the
links of text to school achievement, social relationships, and community
practices.
PAINTER AND Working in primary schools in the 1980s, Painter and Martin (1986) and
MARTIN Rothery (1990, 1994) investigated written language tasks to identify stu-
ROTHERY dents’ ‘opportunity for accessing and learning the curriculum’ (Rothery
1996:92). The project collected and analyzed 2000 written texts from a
Sydney primary school. Analysis of the schematic structure of narratives
and the documentation of writing revealed a narrow range of writing
activities and a limited range of genres which were primarily narratives
and unreflective uses of language. Researchers and teachers undertook
action research focusing on literacy requirements across the curriculum,

primarily the identification of genres and their generic structure. The
descriptions of genres in primary schools (Rothery 1994) and subsequently
in secondary schools (Unsworth 2000) informed subsequent genre-based
literacy practices and pedagogy. Genre pedagogy engages students in
looking at the function of genres and their internal stages, recognizing that
different genres serve different purposes. Deliberate selection and explicit
teaching of the features of genres is a preparation for writers to select
appropriate discourse resources for the formulation of their own texts.
Language awareness studies with very young children provide evidence
that children’s work with metalanguage for analysis of texts can go beyond
knowledge transmission to equipping children as creators and composers
of texts (Hasan 2011b; Williams 2016). For instance, a study by O’Hallaron
et al. (2015) draws on Hasan’s concept of reflection literacy. Their study
documents the development of teachers’ and children’s critical language
awareness for reading informational texts. The researchers write:
In analyzing the language resources for interpersonal meaning in science, we

see that writers often: present judgements and evaluation; draw attention to
something to make it salient to readers; represent imagined readers’ experi-
ences or thinking; engage in dialog with readers; direct readers to do things;
and express their own perspectives on how certain or likely something is.
Each of these moves calls on the grammar of the language in varied ways.
(O’Hallaron et al. 2015:57)
The project introduced metalanguage from SFL to focus students’ attention

on authors’ choices for expressing the meaning of texts and found that
‘young children are capable of beginning to engage in such reflection and
inquiry’ (O’Hallaron et al. 2015:67). ‘Participants in the study were 23
second- through fifth-grade teachers and literacy coaches from five schools
in a community where Arabic was widely spoken’ (O’Hallaron et al.
2015:59), and the project involved creating reading and writing units in
science. An inservice programme focused on the analysis of interpersonal
grammatical resources in science texts for presenting author attitudes,
WILLIAMS judgement, evaluation, and perspectives.
Williams (2004:241), too, has investigated the ‘potential for children to
develop abstract resources for thinking about language systematically
through meaning-oriented grammatical study’. Recognizing that a gram-
mar exploring semantics and context requires a complex analytical frame-
work, Williams (2005) asks: What is the potential of working with
grammatical awareness with young children? In 1994 a team of teachers
and researchers ‘commenced a project on children’s development of know-
ledge about language in Sydney with primary school children aged 6 and 11
years’ (Williams 2005:286). The purpose of the project was to explore the
accessibility and efficacy of Systemic Functional Grammar in primary
school literacy education, focusing on the production of written texts and
546 PETER MICKAN
the discussion of literary texts. In doing so, students develop a sense of

grammar awareness, what Halliday calls grammatics (knowledge about
grammar). Williams (2005:282) observes that ‘[a] grammatics definitely
does not develop naturally. It requires systematic and long-term educa-
tional intervention, especially if it is to be part of literacy development
activity.’ He further claims that teaching functional linguistics for teaching
about language is a recontextualization of grammar not as a body of
knowledge but as ‘mapping linguistic meaning-making practices’ (Williams
2005:283). Williams’ (2016) study applies Hasan’s concept of reflection
literacy to elementary school literacy classes. His study suggests that ‘the
grammar appears to be both accessible and efficacious for pedagogic pur-
poses’ (Williams 2016:281).
CHRISTIE Christie (2005) situates language learning across the curriculum in a
detailed and evidence-based review of language education in the primary
years. In a handbook designed for teachers to support students’ successful
writing, exemplars of written genres illustrate language features and stages
of linguistic development typical for learning areas and stages of schooling
(White 2014). The analysis applies genre teaching to composing written
texts across the Australian primary curriculum.
A number of scholars have widened the focus on functional literacy in
primary and junior secondary teaching to incorporate explicit instruction
on meanings conveyed in images, texts, and the visual layout of informa-
tion in electronic environments (e.g. Unsworth 2001; see also Kress 2003;
Kress and van Leeuwen 2006). Various classroom studies examine multi-
literacies and text–image relations in teaching materials and general read-
ing material for young children (e.g. Moya Guijarro 2014; Painter et al.
2013; Rose 2011b). In her analysis of the ideational and interpersonal
metafunctions in over fifty picture books, Painter (2007) explains how
meaning is created visually and verbally in children’s picture books and
that ‘some traditional concerns of narrative theory are realised through the
visual as much as the verbal strand of a picture book’ (Painter 2007:41). Her
analysis points to the need for teachers to facilitate reflection on the
‘interrelations between visual and verbal meaning’, on ‘the relative depend-
ence of the narrative’ on the verbal and visual resources used, and ‘the ways
[a text] might position readers to respond’ (Painter 2007:58). Chu’s (2014;
2017) study of picture book teaching to new arrival students in primary
school and the associated talk around texts illustrates how children exploit
multimodality for understanding picture book texts. In research into a
specific subject area, Hanauer (2008) documents multiliteracy in scientific
discourse in the primary school classroom, depicting the interaction of
texts, pictures, and tasks in the representation of scientific information.
Chan and Unsworth (2011) investigate students’ need to negotiate image–
language interaction in contemporary digital media for understanding
concepts in texts. They describe the challenges of reading materials online
and report that ‘the integration of meanings from complementary semiotic
resources is essential for the effective reading of hyperlinked materials and

web-based screen displays’ (Chan and Unsworth 2011:196). The recognition
of genres or text types and the typical linguistic features of texts which
contribute to knowledge-building demonstrate the significance of linguistic
theory and functional metalanguage for teachers’ and students’ awareness
of the role of language in learning (Schleppegrell 2004).
21.4 SFL Studies in Secondary Education
There are a number of SFL studies focusing on secondary-school education

and the language associated with advanced and new technical texts con-
stituting subject specific registers (Achugar and Schleppegrell 2016; Whit-
taker et al. 2006). In this section, I mention just a few. Schleppegrell
(2004), in a functional linguistic perspective on the language of schooling,
analyzes the grammatical demands of typical written genres in middle
school and college. She describes the grammatical differences between
conversational language and the formal language of schooling with a
focus on the realization of social contexts through grammatical choices,
and identifies the need for students to expand their linguistic resources to
make meaning in the contexts of school subjects. She analyzes the lan-
guage demands of science and history, and argues for a visible pedagogy
with an explicit focus on language in teaching subject-specific knowledge.
Applying Hasan’s concept of reflection literacy in history teaching, Achu-
gar and Schleppegrell (2016:375) propose ‘the idea that students need to
learn to talk about language and its meaning potential as a means of
interrogating the texts they read and develop the resources to challenge
and construct knowledge’. Teachers’ recognition of the literacy experi-
ences of students entails supporting them in interrogating the social
nature of language and in developing their agency for constructing
knowledge.
O’Halloran’s (2005) work on the description of mathematical discourse as
‘multisemiotic’ has clear pedagogical implications. For instance, O’Halloran
(2000) applies a multisemiotic perspective in the analysis of a trigonometry
lesson in a secondary school. The text for analysis ‘arose in a trigonometry
lesson in Perth, Western Australia in a Year 10 class with 15-year-old stu-
dents’ (O’Halloran 2000:368). The analysis focuses on the oral discourse and
board texts of the lesson, which displayed the semiotic resources of math-
ematical symbolism, visual display, and language, as well as the ‘unique
lexicogrammatical systems for encoding meaning’ (O’Halloran 2000:359).
More recently, O’Halloran (2007) applies a multisemiotic frame of
reference in her examination of the literacy demands of secondary-school
mathematics. She conceptualizes mathematics as a discourse which is
constructed through the use of three semiotic resources – language,
visual imagery, and mathematical symbolism. She demonstrates the
548 PETER MICKAN
interplay in selected texts, revealing the multimodal nature of mathematics

and the intricacy of mathematical, pedagogic discourse.
The technicality of scientific discourse challenges students. Halliday and
Martin (1993) provides an historical perspective on the grammatical con-
strual of knowledge in science, including grammatical metaphor and nom-
inalization, and describes the nature of scientific language and some of the
linguistic demands in the development of scientific literacy. Martin and
Veel (1998) builds on this work with a collection of essays describing
different features of the discourse and pedagogy of science, including
technicality, abstraction, and the multisemiotic nature of scientific dis-
course. Mickan (2007) has analyzed science practices and associated oral
and written texts in an upper secondary class. He bases his analysis on
transcripts of lessons of a class working with a science textbook and talking
together while conducting an experiment. The teacher’s explanation of a
science experiment scaffolded students’ comprehension of scientific dis-
course as preparation for carrying out an experiment in the science labora-
tory. Veel (1997) describes the process of learning science in the secondary
school as an apprenticeship into scientific discourses in the conduct
of enquiries, for recording observations in taxonomies, for classification
of phenomena, and for experimental procedures.
Other subject-related work includes Macken-Horarik’s (e.g. 2013) work
on school English, and Derewianka’s (2009) and Coffin’s (2006) work on
history. Derewianka (2009) analyzes history texts, focusing on the interper-
sonal dimension of genres. Her work spans essay writing in classes from
early secondary to university. Coffin (2006) applies SFL in the examination
of the writing requirements in history classes. From a corpus of 1,000 texts
from students across years 7–12 in seventeen schools, she maps texts in the
history curriculum. Her analysis identifies the main text types and their
purposes in the context of the discourses of history – arguing, recording,
and explaining. She analyzes the structure and lexicogrammar of the
genres, integrates the linguistic description into teaching activities,
and uses the resources for intensive professional development of teachers.
21.5 SFL Applied in Higher Education
Functional analyses of the language of disciplines and the discourses of

academic practices is a focus of SFL research in higher education. Coffin
and Donohue (2014) frame their study of language in higher education in
terms of language as social semiotic (LASS). The approach views language
as a resource for meaning-making, with meaning inseparable from texts.
They make the case that language is central to teaching and learning at
the university level, stating that ‘language is seen as the powerhouse of
meaning-making, and language, teaching and learning are therefore
inseparable’ (Coffin and Donohue 2014:2). Described as a content- and
language-integrated approach, their account is based primarily on written

data from higher education programmes which include students’
responses to viewing a film in Film Studies, recording the range of genres
students write in Health and Social Care, and online discussion forums.
They propose procedures for the implementation of a LASS approach in
higher education by applying evidence from ethnographic research
linking language, teaching, and learning.
In response to colleagues’ reports of the difficulties postgraduate stu-
dents face with written assignments, Mickan (2013) initiated investiga-
tions into the texts and practices constituting the epistemologies of
disciplinary subjects taught at the postgraduate level. Included in the
studies were courses in Environmental Impact Assessment, Nursing,
Interior Architecture (Yee Wong 2016), and Curatorial Studies. These
qualitative studies situate the analysis of literacies in the context of the
social practices and discourses of disciplines and subdisciplines. Docu-
mentation of selected subjects included examining institutional represen-
tations of subjects in curriculum statements and assignments, and
incorporated recording of lectures and interviews with students. The
studies revealed a wide variety of oral, written, and multimodal texts
characteristic of academic coursework, and suggested the need for multi-
semiotic frameworks for the analysis of the epistemology of disciplinary
subjects in the context of institutional requirements, instructional events,
and utilization of technologies.
Analyses of texts which constitute academic practices such as writing
literature reviews, research proposals, conference presentations, and jour-
nal articles depict the diverse literacy demands in building disciplinary
knowledge (Hood 2011; Ravelli and Ellis 2005). Alyousef and Mickan
(2016) have investigated the academic practices of international students
in a postgraduate Management Accounting module in a Master of Com-
merce Accounting programme. They identified the need for students to
interpret tables and graphs in technical texts and manage multimodal
texts in computing applications for their assignments. SFL analyses of
specific features of disciplinary writing include undergraduate students’
introductions to research reports (Hood 2005), hyper-themes in history
essays (Ravelli 2005), and written features of geography essays (Hewings
2005).
Universities have introduced programmes designed to prepare students
for the demands of disciplinary assignments. Jones (2005) reports the appli-
cation of SFL in a university writing centre to make explicit for students the
writing practices in disciplines. Jones describes working with postgraduate
research students writing a literature review, where they analyze the gen-
eric structure of a review, and the linguistic resources related to textual
and interpersonal metafunctions. SFL theory and metalanguage informs
students’ awareness of the social context and purpose for the linguistic
features of academic literacies.
550 PETER MICKAN
The validity of high-stakes tests designed to assess students’ language

proficiency for entry to tertiary study is highlighted in academic literacies
research. Schleppegrell (2006:144) observes that what an assessor can look
for in terms of advanced language development is vague and underspecified
in the American Council on the Teaching of Foreign Languages (ACTFL)
Proficiency Guidelines. For teaching and assessment, Schleppegrell proposes
a focus on more specific information about the linguistic resources students
need in order for them to work on complex texts and tasks. In Australia, the
International English Language Testing System (IELTS) has been selected by
universities as a gateway to tertiary entrance for international students.
Mickan has described the complexity of evaluating IELTS candidates’ texts
written in response to prompts, has proposed explicit instruction in the
discourse of text types in preparation programmes, and has recommended
describing text-defining criteria for valid assessment (see Mickan 2003; Mic-
kan and Motteram 2009). Taking a text-focus into account in test prepar-
ation, Moghaddam and Mickan (2013) use a genre framework for the
analysis of spoken and written texts in an IELTS preparation programme in
Iran. Hedeboe (2009) proposes explicit teaching of features of genre for
students taking tests and for informing examiners, and Martinez-Lirola
(2015) reports that an explicit linguistic modelling of genres improves stu-
dents’ written texts. These studies suggest attending to the semantic stratum
of language tasks with a focus on the social purpose of the assessment tasks
(Mickan 2013) and the adoption of a social semiotic perspective on the
design, the validation, and the marking of proficiency of literacy tests.
SFL analyses have also highlighted the importance of situating texts in
the contexts of disciplines and disciplinarity (Christie and Maton 2011),
thus generating insights for transdisciplinary application, as demonstrated
by Coffin and Donohue (2014), who refer to language as the powerhouse for
learning to mean in higher education. Language as social semiotic frames
language learning across the spectrum of educational disciplines and pre-
sents a challenge to tertiary educators and researchers.
21.6 SFL Research and Teacher Education
SFL linguists have established the crucial role of teachers’ knowledge of

language for analyzing their own teaching, and for working with students.
The metalanguage of SFL is a formidable linguistic resource for interro-
gating the role of language in education. Macken-Horarik et al. (2011:13)
argue for a grammatics for teaching English: ‘Teachers need “ways in” to
working with the relationship between systems that describe language in
all its potential, and specific choices from these systems that co-pattern in
particular ways with particular texts’. They point out that ‘[c]ontinuity of
learning for students in years four, six, eight and 10, however, can only be
provided if teachers have access to a shared metalanguage for developing

such understandings’ (Macken-Horarik 2011:17; see also Love et al. 2014;
Macken-Horarik et al. 2015). Achugar et al. (2007) incorporate functional
metalanguage related to knowledge construed with language into profes-
sional development programmes. They claim that ‘[t]eachers need know-
ledge about language and tools to analyze language to understand the
demands their subject matter poses to students, to support their students’
literacy development and to critically approach the texts they use’ (Achu-
gar et al. 2007:8). O’Halloran et al. (2015) in a longitudinal inservice
programme combine fieldwork and action research with the study of
functional linguistic theory. The programme includes the development
of units of work for teaching science with the application of functional
grammar from second through to fifth grade. As an application of critical
language awareness in science reading, the analysis of information texts
made overt authors’ attitudes and perspectives in their language choices
in scientific writing.
Extensive experience of genre-based teaching informs Martin and Rose’s
(2008) and Rose and Martin’s (2012) systematic training programme in
Reading to Learn, a pedagogy which scaffolds students’ reading of genres. In
the programme, readers’ attention is focused on the language choices for
the expression of different meanings. The critical and selective reading of
genres provides insight into the composition and ideology of texts, and
prepares students for the formulation of their own texts. Other training
programs include an action research inservice program, ESL in the Main-
stream, established by the Department of Education and Children’s Services
in South Australia (Polias and Dare 2004). Inservice sessions for primary and
secondary teachers from across the curriculum introduced functional lin-
guistics for teachers to investigate and explore the language of teaching. The
programme has been influential in developing teachers’ awareness of stu-
dents’ language needs across the curriculum, demonstrating the idea that
every teacher is a language teacher.
SFL theory is a powerful vehicle for the collaboration of teachers with
researchers in school-based action research, enacting the concept of teacher
as researcher in partnership in applying theory in practice (Burton and
Mickan 1993; Unsworth 2001). Typically, teachers and researchers working
with children and students record and analyze language use in different
contexts. The principled recording of teaching practices and linguistic analy-
sis of genres across the curriculum show the value of sustained fieldwork,
of observation of natural language use, and of the systematic analysis of
language in action. The studies in many cases have transformed pedagogy,
practices, and teacher education with applications of the theory of language
as social semiotic and the understanding that knowledge about language is
essential for teachers’ and students’ building of the meaning-making
resources for dynamic cultural engagement (Christie 2012; Mickan 2017;
Mickan and Lopez 2017; Rose 2015).
552 PETER MICKAN
21.7 Continuing Exploration of Language in Education
Exploration of SFL applied in language in education is ongoing (Martin

2014), with SFL theory applied to understanding texts in our lives. The
implications of Halliday’s (2004) view of the child as a semiotic being
continue to challenge teachers and researchers, as do the implications of
Hasan’s (2002) work on semiotic mediation and mental dispositions
developed in intimate relationships in childhood, and on language variation
in context (Hasan 2014). In pedagogy Williams has raised the bar and
challenged the extent to which young children are able to utilize knowledge
about language in their literacy work. Based on classroom research with
six-year-olds and eleven-year-olds, Williams (2000) observes children using
concepts of ‘Actor’ and ‘Goal’. He notes that ‘children’s ability to reason
abstractly and enthusiastically about language has been grossly underesti-
mated’ (Williams 2000:127).
Williams (2016) emphasizes the need for systematic exploration of the
potential for children’s development of metalanguage for writing success-
fully and understanding language choices for texts serving different pur-
poses. First and additional language learners’ experiences with genre- and
text-based teaching are evidence of this.
Extensive experience with genre- and text-based teaching has created a genre based
need to fine-tune pedagogies from a trinocular perspective in deciding what text based
to focus on for instruction within the complex domain of individual texts
and across text types in relation to context. Halliday’s (Halliday and Mat-
thiessen 2014) grammar of texts and SFL analysis of meaning beyond the
clause (Martin and Rose 2007) and of language as a system for the expres-
sion of meaning potential raises the question of the selection of language
features of texts or genres selected for instruction. The works by Christie
(2012), Martin and Rose (2008), and Rose and Martin (2012) are sources of
experience for tackling what is to be taught as a priority. The issue suggests
need for research on children's
engagement at the semantic the need for research on children’s engagement at the semantic level of
level of texts texts in order to begin to capture what is being semiotically mediated in
explicit genre- or text-based teaching. Matthiessen’s (2009, 2014) research
into context and register in his ‘registerial cartography’ project involves
mapping registers and register variation according to language use. The
focus on language as a system of variation and on the organization of texts
in context is an orientation relevant to curriculum design and pedagogy.
SFL studies SFL studies pose robust alternatives to the limitations of national curric-
ula, the shortcomings of national and international assessment regimes,
and the reductionism of local meaning-making in the commercialization of
educational programmes and materials. Taking the case of international
tests, rather than reducing parameters for the purpose of standardized
measurement of children’s language development and academic achieve-
ment, assessment based on language corpora which conceptualize language
as genres or texts in social contexts across stages of development (Christie
2012; Christie and Derewianka 2008) offers a foundation for valid testing.
Given the immense complexity of measuring the multiple linguistic fea-
tures of texts (Mickan 2003) and the need for multidimensional frameworks
to capture learners’ meaning-making in multimodal texts (Alyousef and
Mickan 2016; Macken-Horarik et al. 2017; O’Halloran 2007; Unsworth
2001), the use of concordancing and computing programmes for analyzing
learners’ achievements is open to further study, along with ICT networks
with options for students to engage with language and other modes of
communication in acts of meaning (e.g. Jewitt 2006).
Teachers and linguists working within the SFL model have both raised
awareness and provided a robust theory of language learning as a life-long
experience for entry and participation in new sociocultural domains, with
language as a meaning-making resource with the capacity to explore new
spheres of human experience which come with social changes in societies
(Hasan 2011b; Mickan and Lopez 2017). Halliday (1996:363–4) comments
that ‘[l]iteracy today includes many contexts . . . To be literate is to operate
in such complex, multiple contexts, to write with many voices, still ending
up with a text, and to read such texts with kaleidoscopic eyes’. The strength
of applying SFL in educational contexts can help us to make sense of human
relationships, of cultural change, of the material world, and of the quality
of peoples’ lives. SFL analyses and applications can increase students’ and
teachers’ awareness of the power of language in daily life and of the
potential for acting to change cultural conditions.
References
Achugar, M. and M. Schleppegrell. 2016. Reflective Literacy and the Teach-

ing of History. In W. L. Bowcher and J. Y. Liang, eds., Society in Language,
Language in Society: Essays in Honour of Ruqaiya Hasan. New York: Palgrave
Achugar, M., M. Schleppegrell, and T. Oteiza. 2007. Engaging Teachers in
Language Analysis: A Functional Linguistic Approach to Reflective Liter-
acy. Teaching English: Practice and Critique 6(2): 8–24.
Alyousef, H. and P. Mickan. 2016. Literacy and Numeracy Practices in
Postgraduate Management Accounting. In A. Archer and E. Breuer, eds.,
Multimodality in Higher Education. Studies in Writing, Volume 33. Leiden:
Brill. 216–40.
Barnes, D. 1969. Language, the Learner and the School. London: Penguin.
Bernstein, B. 1971. Class Codes and Control, Volume 1: Theoretical Studies toward
a Sociology of Language. London: Routledge.
Bernstein, B. 1975. Class Codes and Control, Volume 3: Toward a Theory of
Educational Transmission. London: Routledge.
Critique. Oxford: Rowman and Littlefield.
554 PETER MICKAN
Britton, J. 1970. Language and Learning. London: Allen Lane.

Burton, J. and P. Mickan. 1993. Teachers’ Classroom Research: Rhetoric and
Reality. In J. Edge and K. Richards, eds., Teachers Develop Teachers Research:
Papers on Classroom Research and Teacher Development. Oxford: Heinemann.
113–21.
Chan, E. and L. Unsworth. 2011. Image–Language Interaction in Online
Reading Environments: Challenges for Students’ Reading Comprehen-
sion. Australian Education Researcher 38: 181–202.
Christie, F. 2005. Language Education in the Primary Years. Sydney: University
of New South Wales Press.
Christie, F. 2012. Language Education throughout the School Years: A Functional
Perspective. Malden: Wiley-Blackwell.
Christie, F. and B. Derewianka. 2008. School Discourse: Learning to Write across
the Years of Schooling. London: Continuum.
Christie, F. and K. Maton, eds. 2011. Disciplinarity: Functional Linguistic and
Sociological Perspectives. London: Continuum.
Chu, C. 2014. Supporting New Arrival Students’ Engagement with Picture
Books: Analysis of Teacher Talk Using the Appraisal Theory. Functional
Linguistics 1(12): 1–17.
Chu, C. 2017. Student Questions in Talk around Children’s Picture Books. In
P. Mickan and E. Lopez, eds., Text-based Research and Teaching: A Social
Semiotic Perspective on Language in Use. Basingstoke: Palgrave Macmillan.
139–56.
Coffin, C. 2006. Learning the Language of School History: The Role of
Linguistics in Mapping the Writing Demands of the Secondary School
Curriculum. Journal of Curriculum Studies 38(4): 413–29.
Coffin, C. and J. Donohue. 2014. A Language as Social Semiotic-based
Approach to Teaching and Learning in Higher Education. Malden: John Wiley
and Sons.
Derewianka, B. 2009. Using Appraisal Theory to Track Interpersonal Devel-
opment in Adolescent Academic Writing. In A. McCabe, M. O’Donnell,
and R. Whittaker, eds., Advances in Language and Education. London:
de Silva Joyce, H. and S. Feez. 2012. Text-based Language and Literacy Educa-
tion: Programming and Methodology. Sydney: Phoenix Education.
Doughty, P., J. Pearce, and G. Thornton. 1971. Language in Use. London:
Edward Arnold.
Forsyth, I. and K. Wood. 1977–1980. Language and Communication. London:
Longman.
Gibbons, P. 2006. Bridging Discourses in the ESL Classroom: Students, Teachers
and Researchers. London: Continuum.

Halliday, M. A. K. 1996. Literacy and Linguistics: A Functional Perspective.
In R. Hasan and G. Williams, eds., Literacy in Society. London: Longman.
339–76.
Halliday, M. A. K. 2004. Representing the Child as a Semiotic Being (One
Who Means). In J. Foley, ed., Language, Education and Discourse: Functional
Approaches. London: Continuum. 19–42.
Halliday, M. A. K. 2007a. Language and Socialization: Home and School. In
Halliday, M. A. K. 2007b. The Notion of ‘Context’ in Language Education. In
Halliday, M. A. K. 2007c. On the Concept of Educational Linguistics. In J. J.
Webster, ed., The Collected Works of M. A. K. Halliday, Volume 9: Language and
Education. London: Continuum. 354–67.
cial Press.
Halliday, M. A. K. and J. R. Martin, eds. 1993. Writing Science: Literacy and
Discursive Power. London: Falmer Press.
Hanauer, D. 2008. Scientific Discourse: Multiliteracy in the Classroom. London:
Continuum.
Hasan, R. 2002. Semiotic Mediation and Mental Development in Pluralistic
Societies: Some Implications for Tomorrow’s Schooling. In G. Wells and
G. Claxton, eds., Learning for Life in the 21st Century. Oxford: Blackwell.
112–26.
Hasan, R. 2009a. Language in the Process of Socialisation: Home and School.
In J. J. Webster, ed., The Collected Works of Ruqaiya Hasan, Volume 2: Semantic
119–79.
Hasan, R. 2009b. Semantic Variation and Sociolinguistics. In J. J. Webster,
ed., The Collected Works of Ruqaiya Hasan, Volume 2: Semantic Variation:
Meaning in Society and in Sociolinguistics. Sheffield: Equinox. 180–230.
Hasan, R. 2009c. The Collected Works of Ruqaiya Hasan, Volume 2: Semantic
Variation: Meaning in Society and in Sociolinguistics. Edited by J. J. Webster.
Hasan, R. 2011a. Literacy, Everyday Talk and Society. In J. J. Webster, ed.,
The Collected Works of Ruqaiya Hasan, Volume 3: Language and Education:
Learning and Teaching in Society. Sheffield: Equinox. 169–206.
Hasan, R. 2011b. The Collected Works of Ruqaiya Hasan, Volume 3: Language and
Education: Learning and Teaching in Society. Edited by J. J. Webster. Sheffield:
Equinox.
556 PETER MICKAN

Metafunctions and Semantics. Functional Linguistics 1(9): 1–54.
Hasan, R. and C. Cloran. 1990. A Sociolinguistic Interpretation of Mother–
Child Talk. In M. A. K. Halliday, J. Gibbons, and H. Nicolas, eds., Learning,
Keeping and Using Language: Selected Papers from the 8th World Congress of
Applied Linguistics. Amsterdam: John Benjamins. 67–99.
Hasan, R. and G. Williams. 1996. Literacy in Society. London: Longman.
Heath, S. 1983. Ways with Words: Language, Life, and Work in Communities and
Classrooms. New York: Cambridge University Press.
Hedeboe, B. 2009. On the ‘Internal Dialogue’ between the Examination Task
and Pre-university Students’ Responses. In A. McCabe, M. O’Donnell,
and R. Whittaker, eds., Advances in Language and Education. London:
Hewings, A. 2005. Developing Discipline-specific Writing: An Analysis of
Undergraduate Geography Essays. In L. Ravelli and R. Ellis, eds., Analysing
Academic Writing: Contextual Frameworks. London: Continuum. 131–52.
Hood, S. 2005. Managing Attitude in Undergraduate Academic Writing:
A Focus on the Introduction to Research Reports. In L. Ravelli and
R. Ellis, eds., Analysing Academic Writing: Contextual Frameworks. London:
Continuum. 24–44.
Hood, S. 2011. Writing Discipline: Constructing Inscriptions of Knowledge
and Knowers in Academic Writing. In F. Christie and K. Maton, eds.,
Disciplinarity: Functional Linguistic and Sociological Perspectives. London:
Jewitt, C. 2006. Technology, Literacy and Learning: A Multimodal Approach.
London: Routledge.
Kress, G. 2003. Literacy in the New Media Age. London: Routledge.
Design. London: Routledge.
Jones, J. 2005. Learning to Write in the Disciplines: The Application of
Systemic Functional Linguistic Theory to the Teaching and Research of
Student Writing. In L. Ravelli and R. Ellis, eds., Analysing Academic Writing:
Contextual Frameworks. London: Continuum. 254–73.
Love, K., C. Sandiford, M. Macken-Horarik, and L. Unsworth. 2014. From
‘Bored Witless’ to ‘Rhetorical Nous’: Teacher Orientation to Knowledge
about Language and Strengthening Student Persuasive Writing. English in
Australia 49(3): 43–56.
Mackay, D., B. Thompson, and P. Schaub. 1978. Breakthrough to Literacy:
Teacher’s Manual: The Theory and Practice of Teaching Initial Reading and
Writing. London: Longman.
Macken-Horarik, M. 2013. English in the Tempest: The Value of Metaphor
and Re-imagining Grammar in English. English in Australia 48(3): 46–53.
Macken-Horarik, M., K. Love, and L. Unsworth. 2011. A Grammatics Good
Enough for School English in the 21st Century: Four Challenges in Real-
izing Potential. Australian Journal of Language and Literacy 34(1): 9–23.
Macken-Horarik, M., C. Sandiford, K. Love, and L. Unsworth. 2015. New

Ways of Working ‘with Grammar in Mind’ in School English: Insights
from Systemic Functional Grammatics. Linguistics and Education 31:
145–58.
Macken-Horarik, M., K. Love, C. Sandiford, and L. Unsworth. 2017. Func-
tional Grammatics: Reconceptualising Knowledge about Language and Image for
School English. London: Routledge.
Martin, J. R. 2009. Discourse Studies. In M. A. K. Halliday and J. J. Webster,
eds., Continuum Companion to Systemic Functional Linguistics. London:
Equinox.
Martin, J. R. and R. Veel, eds. 1998. Reading Science: Critical and Functional
Perspectives on Discourses of Science. London: Routledge.
Martínez-Lirola, M. 2015. Using Genre Theory for Teaching Writing at
Tertiary Level: A Practical Example of Teaching to Write Effective
Recounts. Revista Científica Guillermo de Ockham 13(1): 59–66.
Emerging from Acts of Meaning. Language Learning 59(Supplement 1):
206–29.
Matthiessen, C. M. I. M. 2014. Registerial Cartography: Context-based Map-
ping of Text Types and Their Rhetorical-relational Organization. Proceed-
ings of the 28th Pacific Asia Conference on Language, Information and
Computation. Available online at: www.aclweb.org/anthology/Y14-1003.
(Last accessed 24/10/2017.)
McCabe, A., M. O’Donnell, and R. Whittaker, eds. 2009. Advances in Language
and Education. London: Continuum.
Mickan, P. 2003. ‘What’s Your Score?’ An Investigation into Language
Descriptors for Rating Written Performance. In L. Hyam, ed., International
English Language Testing System Research Reports, Volume 5. Canberra: IELTS
Australia Pty Ltd. 125–57.
Mickan, P. 2007. Doing Science and Home Economics: Curriculum Social-
isation of New Arrivals to Australia. Language and Education 21(1): 1–17.
Mickan, P. 2013. Language Curriculum Design and Socialisation. Bristol: Multi-
lingual Matters.
Mickan, P. 2017. Text-based Research and Teaching from a Social Semiotic
Perspective: Transformative Research and Pedagogy. In P. Mickan and
E. Lopez, eds., Text-based Research and Teaching: A Social Semiotic Perspective
on Language in Use. Basingstoke: Palgrave Macmillan. 15–35.
Mickan, P. and E. Lopez, eds. 2017. Text-based Research and Teaching: A Social
Semiotic Perspective on Language in Use. Basingstoke: Palgrave Macmillan.
558 PETER MICKAN
Mickan, P. and J. Motteram. 2009. The Preparation Practices of IELTS

Candidates: Case Studies. In J. Osborne, ed., IELTS Research Reports, Volume
10. Canberra: IELTS Australia Ltd. 223–62.
Moghaddam, S. and P. Mickan. 2014. An Investigation of IELTS Preparation
Courses for University Study: Development of Written and Verbal Argumentative
Texts. Saarbrucken: Scholars’ Press.
Moya Guijarro, A. J. 2014. A Multimodal Analysis of Picture Books for Children.
O’Halloran, K. 2000. Classroom Discourse in Mathematics: A Multisemiotic
Analysis. Linguistics and Education 10(3): 359–88.
O’Halloran, K., ed. 2004. Multimodal Discourse Analysis: Systemic Functional
Perspectives. London: Continuum.
O’Halloran, K. 2005. Mathematical Discourse: Language, Symbolism and Visual
Images. London: Continuum.
O’Halloran, K. 2007. Systemic Functional Multimodal Discourse
Analysis (SF-MDA) Approach to Mathematics, Grammar and Literacy. In
A. McCabe, M. O’Donnell, and R. Whittaker, eds., Advances in Language and
O’Hallaron, C., A. Palincsar, and M. Schleppegrell. 2015. Reading Science:
Using Systemic Functional Linguistics to Support Critical Language
Awareness. Linguistics and Education 32: 55–67.
Painter, C. 2007. Children’s Picture Book Narratives: Reading Sequences of
Images. In A. McCabe, M. O’Donnell, and R. Whittaker, eds., Advances in
Painter, C. and J. R. Martin, eds. 1986. Writing to Mean: Teaching Genres
across the Curriculum. Occasional Paper no. 9. Series: Occasional Papers
Applied Linguistics Association of Australia. Brisbane: Applied Linguistics
Association of Australia.
Polias, J. and R. Dare. 2004. Language and Literacy: Classroom Applications of
Functional Grammar: Teacher Development Course. Hindmarsh: Department
of Education and Children’s Services.
Ravelli, L. 2005. Signalling the Organization of Written Essays: Hyper-
themes in Management and History Essays. In L. Ravelli and R. Ellis,
eds., Analysing Academic Writing: Contextualized Frameworks. London: Con-
tinuum. 104–30.
Ravelli, L. and R. Ellis, eds. 2005. Analysing Academic Writing: Contextualized
Frameworks. London: Continuum.
Rothery, J. 1990. ‘Story’ Writing in Primary School: Assessing Narrative Type
Genres. PhD Thesis, University of Sydney.
Rothery, J. 1994. Exploring Literacy in School English (Write it Right Resources for
Literacy and Learning). Sydney: Metropolitan East Disadvantaged Schools
Program.
Rothery, J. 1996. Making the Changes: Developing an Educational Linguis-

tics. In R. Hasan and G. Williams, eds., Literacy in Society. London:
Longman. 86–123.
Rose, D. 2011a. Beyond Literacy: Building an Integrated Pedagogic Genre.
Australian Journal of Language and Literacy 34(1): 81–97.
Rose, D. 2011b. Meaning beyond the Margins: Learning to Interact with
Books. In S. Dreyfus, S. Hood, and M. Stenglin, eds., Semiotic Margins:
Meaning in Multimodalities. London: Continuum. 177–208.
Rose, D. 2015. New Developments in Genre-based Literacy Pedagogy. In
C. MacArthur, S. Graham, and J. Fitzgerald, eds., Handbook of Writing
Research. 2nd ed. New York: Guilford.
Rose, D. and J. R. Martin. 2012. Learning to Write/Reading to Learn: Genre,
Knowledge and Pedagogy in the Sydney School. Mahwah: Lawrence Erlbaum
Associates.
Schleppegrell, M. 2004. The Language of Schooling: A Functional Linguistics
Perspective. London: Lawrence Erlbaum Associates.
Schleppegrell, M. 2006. The Linguistic Features of Advanced Language Use:
The Grammar of Exposition. In H. Byrnes, ed., Advanced Language Learning:
The Contribution of Halliday and Vygotsky. London: Continuum. 134–46.
Tough, J. 1977. Development of Meaning: A Study of Children’s Use of Language.
London: Allen & Unwin.
Unsworth, L., ed. 2000. Researching Language in Schools and Communities:
Functional Linguistic Perspectives. London: Cassell.
Unsworth, L. 2001. Teaching Multiliteracies across the Curriculum: Changing
Contexts of Text and Image in Classroom Practice. Buckingham: Open Univer-
sity Press.
Veel, R. 1997. Learning How to Mean – Scientifically Speaking: Apprentice-
ship into Scientific Discourse in the Secondary School. In F. Christie and
J. R. Martin, eds., Genre and Institutions: Social Processes in the Workplace and
School. London: Pinter. 161–95.
Ventola, E. and A. J. Moya Guijarro. 2009. Introduction. In E. Ventola and
A. J. Moya Guijarro, eds., The World Told and the World Shown: Multisemiotic
Issues. Basingstoke: Palgrave Macmillan. 1–8.
White, B. 2014. Genres across the Curriculum. South Australia: Catholic
Education Office.
Whittaker, R., M. O’Donnell, and A. McCabe. 2006. Language and Literacy:
Functional Approaches. London: Continuum.
Williams, G. 2000. Children’s Literature, Children and Uses of Language
Description. In L. Unsworth, ed., Researching Language in Schools and
Communities: Functional Linguistic Perspectives. London: Cassell. 111–29.
Williams, G. 2004. Ontogenesis and Grammatics: Functions of Metalan-
guage in Pedagogical Discourse. In G. Williams and A. Lukin, eds.,
The Development of Language: Functional Perspectives on Species and Individuals.
560 PETER MICKAN
Williams, G. 2005. Grammatics in Schools. In J. J. Webster, C. M. I. M.

Matthiessen, and R. Hasan, eds., Continuing Discourse on Language, Volume
1. Sheffield: Equinox. 281–310.
Williams, G. 2016. Reflection Literacy in the First Years of Schooling:
Questions of Theory and Practice. In W. L. Bowcher and J. Y. Liang,
eds., Society in Language, Language in Society: Essays in Honour of Ruqaiya
Hasan. New York: Palgrave Macmillan. 333–56.
Yee Wong, T. S. 2016. Academic Literacies in the Field of Interior Architec-
ture: A Multimodal Analysis. In P. Mickan and E. Lopez, eds., Text-based
Research and Teaching: A Social Semiotic Perspective on Language in Use.
Basingstoke: Palgrave Macmillan. 341–61.
22
Systemic Functional
Linguistics and
Computation
New Directions, New Challenges
John Bateman, Daniel McDonald, Tuomo Hiippala, Daniel

Couto-Vale, and Eugeniu Costetchi
22.1 Introduction
Computational linguistics (CL), natural language processing (NLP), and

language technologies (LT) are closely intertwined fields that employ com-
putational techniques for various tasks related to treatments of language.
Key tasks include both the development of software tools (for information/
document retrieval, document summarization, sentiment analysis, named
entity extraction, corpus analysis, and machine translation) and the devel-
opment and refinement of linguistic theories via algorithmic models. The
importance of algorithmic models became clear perhaps most significantly
with Peters and Ritchie’s (1973) formal proof that Chomsky’s emerging
transformational grammar was unlearnable in the form then under discus-
sion. Results of this kind led to a new awareness that algorithmic properties
could be deeply revealing of properties relevant for linguistic theorizing as
well. More recently, work exploring the conditions under which language
and language behaviour can ‘emerge’ from situated interaction has also
made considerable progress – progress that simply would not have been
possible without the computational models necessary to conduct experi-
mentation (Steels 2005).
Since the advent of computational approaches to language in the 1950s,
computational models have in fact always been at the forefront of linguistic
theorizing, serving to push the degree of explicitness of such models fur-
ther and revealing both gaps in existing treatments and new capabilities.
Most grammatical theories have now received computational instantiation,
and there is extensive experience in designing algorithms for their process-
ing, as well as substantial computationally accessible resources such as
562 J O H N B AT E M A N E T A L .
corpora, treebanks (i.e. grammatically annotated corpora), lexicons, and

more. The models employed usually attempt to ensure that their algorith-
mic properties are reasonable so that they do not suffer the same fate as
early transformational grammar and fall foul of the unlearnability trap. It
is this attention to computational properties that has allowed broad-
coverage components that can be applied to real (large-scale) data to become
the norm.
Given this, it should not then be surprising that Systemic Functional
Linguistics (SFL) has a long history of interaction with computational
linguistics (a detailed review of the historical engagement between the
fields is given in O’Donnell and Bateman 2005). Bateman and O’Donnell
(2015) trace the deep involvement of SFL’s primary founder, Michael A. K.
Halliday, in this process from its beginnings. Halliday participated in
some of the earliest attempts to achieve automatic translation systems
in the 1950s, bringing together linguistic theoretical considerations and
practicalities of computational processing. He was also instrumental in
shaping the linguistic foundations of some of the most well-known lan-
guage-oriented systems to emerge in computational linguistics and Artifi-
cial Intelligence in the 1970s and 1980s. This included both Terry
Winograd’s SHRDLU (Winograd 1972), a landmark natural language dia-
logue system that demonstrated that natural dialogic interaction with
computers was an achievable goal, and William C. Mann’s Penman system
for large-scale automatic natural language generation (Mann 1983; Mann
and Matthiessen 1985).
From the 1980s and up until the mid-1990s, interaction between SFL and
computation was consequently well established, with other significant
initiatives bringing the fields together. Robin Fawcett’s COMMUNAL, for
example, also approached automatic language generation, applying a dif-
ferent variant of SFL (Fawcett 1988), while Michael O’Donnell attempted to
extend the capabilities on offer by developing automatic analysis compon-
ents similarly based on SFL (O’Donnell 1994); many further systems are
described in O’Donnell and Bateman (2005). There was, however, a marked
difference in the relative successes and acceptance of these efforts. Systemic
Functional Linguistics, as a broadly functional theory of language focusing
on language use as ‘motivated choice’, appeared well suited to address
natural language generation: here the abstract task is often characterized
as precisely one of ‘making the right choices’ given a description of a
language as a resource (see McDonald 1980). Most approaches to automatic
text generation of that time, regardless of theoretical orientation, thus
turned to questions concerned with finding the functional conditions under
which particular (primarily) grammatical choices would be appropriate for
the communicative goals being pursued. This characterization echoed dir-
ectly that offered by Halliday concerning the main descriptive apparatus
used within SFL, the system network (see Halliday 1996: 10).
SFL and Computation 563
The situation with analysis, i.e. the construction of computational com-

ponents capable of moving from provided strings of words or sounds to
more abstract, grammatical or semantic representations, was very differ-
ent. Here there was very little success compared to the rapid growth of
general purpose analysis systems in computational linguistics more
broadly. The reasons for this asymmetry are themselves of considerable
theoretical note, revealing deeper issues in the formalizations of linguistic
resources offered by the theory. The most significant of these are discussed
in Bateman (2008b) and so will not be repeated here – essentially, however,
questions of how to manage computational complexity of the kind men-
tioned above with respect to transformational grammar played a
substantial role.
For the purposes of this chapter, what followed from the lack of success
in the area of computational analysis using SFL specifications is more
central. Whereas other approaches and systems began to achieve consider-
able success in automatic analysis, SFL’s lack of success in this critical task
led to it not being considered a viable approach for computation. This
brought many consequences of its own, some of which were particularly
important for the subsequent development of theory. For example, while
SFL has always been oriented toward corpus-based work, pursuing such
large-scale empirical research necessitates the availability or development
of corresponding computational tools: one cannot examine large bodies of
data by hand. Several other linguistic approaches have been more support-
ive of automated analysis of large-scale corpora, and so it was natural that
research – even research based on naturally occurring examples – would
come to orient more to the kinds of theories for which analyzed corpora
were, or could be made, available.
The success of these latter approaches had significant consequences
for computational linguistics in general. As statistical methods, machine
learning, automatic grammar construction, and the like became ever
more central to computational approaches to language throughout the
1990s, it became essential to create richly annotated data sets from which
computational models could be derived. Machine learning, for example,
works by taking a set of ‘correct’ examples and automatically deriving
decision procedures capable of classifying previously unseen examples in
the same way. To be effective, the quantity of ‘correct’ examples required
to bootstrap the process can be quite large, and SFL has simply not had
resources of this kind – again largely due to a lack of automated analysis
capabilities to get the entire process going. By the end of the 1990s,
therefore, the position of SFL within computational linguistics had
become relatively marginal.
Some legacy systems relying on SFL resources, such as the general natural
language generation system KPML (Bateman 1997), descended from the
Penman system, continued to be used and extended because of the
considerable linguistic information they had come to include. For example,

the development of the English grammar available with the KPML system
stretches back to work by Matthiessen and Halliday in the 1980s in the
Penman project (Mann and Matthiessen 1985); since then the grammar has
come to include additions made by many further contributors, giving
rise to a grammar with very broad coverage – even by today’s standards.
The system as a whole thus came to occupy a particular niche among ‘high-
quality, high-effort’ computational systems. More recently, less flexible but
largely automatically produced generation systems relying on a variety of
statistical methods have become common; an introductory overview of the
field of natural language generation, its development over time, and cur-
rent methods and challenges can be found in Bateman and Zock (2017).
Internally to SFL, the rapid developments in computational linguistics
also began to have a significant, if largely indirect, impact. Corpus work, for
example, frequently demands that bodies of data be ‘marked up’, or anno-
tated, with particular categories that can subsequently be examined for
meaningful patterns. This can only sensibly be done using computational
tools that support the annotation process and manage the corpus. Here
Michael O’Donnell’s UAM Corpus Tool (O’Donnell 2008) has been of enor-
mous benefit to many researchers, operating both within and outside of
SFL. This freely available programme allows researchers to define their own
classification systems and then supports application of those systems to
bodies of text. In contrast to many such tools, UAM adopts the system
network as its basic resource for defining classifications, thereby allowing
deeply nested classifications of a kind particularly supportive of functional
linguistic work. Preconfigured networks for commonly used areas of
Systemic Functional Grammar are also provided.
Although the UAM Tool is intended to support manual annotation, i.e.
annotation where the human researcher makes the choices of classification
according to the options available, in its more recent instantiations it also
provides access to some of the now standard computational components
capable of producing structural analyses of unrestricted text, such as the
freely available Stanford Parser (Manning et al. 2014). This now supports
automatic structure analysis for several languages (including English,
Arabic, French, German, and Chinese). Because the Stanford Parser provides
Phrase Structure and Universal Dependency (UD) parses, the relationship
between its analyses and the categories of SFL is often far from straightfor-
ward. Currently, however, no comparable capabilities exist for systemic
functional grammars. This critical task is still unresolved from the perspec-
tive of SFL.
The general availability and wide-scale take-up of the UAM Tool demon-
strates in addition how computational tools now form a normal part of the
linguist’s world, and this is sure to increase as such tools gain even further
in capabilities. Kay O’Halloran and team have, for example, produced a
series of tools extending capabilities for corpus analysis both with respect to
the depth of analysis, including semantics and discourse organizations, and

to the breadth of analysis, moving into considerations of image, video,
and text–image combinations (O’Halloran 2014) as well. We return to this
line of development below.
Ultimately, the current situation involving interactions between SFL and
computation is complex. Whereas the lack of contact between computation
and SFL by the early 2000s had led to a hiatus in new theoretical and
practical engagements of SFL with computational techniques, the growing
capabilities and sophistication of computational approaches to language
have made that work increasingly relevant and difficult to ignore. As a
consequence, there are now signs that a new revival in interaction is in
progress. The availability of a far broader range of computational tech-
niques, together with more accessible, robust, and extensible infrastruc-
tures for developing and combining computational components, has made
the development of new generations of computational SFL tools both pos-
sible and beneficial. This then forms the focus of the remainder of this
article. We pick out several core areas in this newly emerging state of the
art, describing current activities and identifying some key areas for future
developments.
22.2 Parsing
We begin with the core task of providing SFL analysis capabilities, or

‘parsing’. As mentioned above, the lack of such capabilities was one
of the main reasons why interactions between SFL and computation
faltered. Building a natural language parser can be seen as a task of
creating an artificial text reader which understands the meaning
expressed in some text. The depth and the kind of text understanding
varies according to the tasks addressed. Different levels of abstraction are
required when enabling tasks such as summarizing documents,
answering questions about them and their content, deriving new know-
ledge, interacting with a human user using natural language, and so on.
These and many other tasks are currently moving out of the field of
artificial intelligence research and more and more into everyday life and
practical applications.
Broad coverage natural language processing modules now exist for sev-
eral levels of linguistic abstraction, ranging from the least abstract tasks of
‘stemming’ – i.e. removing inflectional information to reveal the basic
lexical forms employed – and part of speech tagging, through intermediate
tasks such as syntactic analyses, to highly complex tasks such as semantic
analysis or argument extraction. But there are two caveats: in general, the
higher the degree of abstraction, the less accurate the coverage becomes;
moreover, the richer the linguistic description, the slower the parsing
process. This is then particularly problematic for SFL because its grammars
and other levels of description are rich and multi-layered in ways that differ
from many other theoretical accounts.
More specifically, the descriptive power of a Systemic Functional Gram-
mar (SFG) lies to a considerable extent in its separation of descriptive work
across ‘structure’ (i.e. syntagmatic organizations) and ‘choice’ (i.e. paradig-
matic organizations). This comes at the cost of high computational com-
plexity, which still presents today the biggest challenge in parsing broad
coverage texts with full SFGs. O’Donnell and Bateman (2005) discuss how
each successive attempt to construct parsing components using SFL then
necessarily led to the acceptance of limitations either in grammar size or in
language coverage in order to proceed.
A parsing process for full SFGs needs then to derive both syntagmatic (e.g.
constituency structure) and paradigmatic (i.e. selections from the system
networks) descriptions. Providing a syntagmatic description is crucial for
parsing, as it is this organizational frame that serves as an anchor for
structured paradigmatic details – that is, it is not sufficient to know that
some feature has been selected; we also need to know precisely which
grammatical unit that feature constitutes. Moreover, we need to be able
to derive constraints on structure that are given by compatible feature
selections and ruled out by incompatible feature selections. This latter task
is a major source of computational complexity and, as Bateman (2008b)
explains, brings with it significant theoretical implications for the construc-
tion of SFL theory as well as of computational systems. Today, it is common
for parsers to rely on simpler syntactic trees (or other non-SFL grammars) as
starting points for the parsing process. First, a syntagmatic organization, or
‘structural backbone’, would be defined, followed by an enrichment by
paradigmatic selections. This technique was subsequently adopted as a
beneficial heuristic for reducing complexity by most attempts to parse with
SFGs (Kasper 1988; O’Donnell 2005; Costetchi 2013).
The first attempt to achieve larger-scale parsing capabilities for SFG was
that of Robert Kasper (Kasper 1988). The structural backbone employed
was provided by a context-free Phrase Structure Grammar (PSG), similar to
Chomsky’s use of a PSG to generate kernel sentences that would subse-
quently be subject to transformations (Chomsky 1957). In Kasper’s case,
each phrase-structure rule was given additional information for mapping
the phrase structure onto a parallel systemic tree. After all possible sys-
temic trees had been created, they were further enriched using informa-
tion from the Nigel Grammar (the SFG generation grammar developed
within the Penman system – see Matthiessen 1985). This process was
extremely slow and worked only on a limited-size grammar because it
involved ‘multiplying out’ all of the combinatorial possibilities inherent
in the grammar. O’Donnell (1993) and Weerasinghe (1994) subsequently
wrote parsers that attempted to build systemic functional constituency
trees directly, avoiding the need to construct full phrase structure gram-
mars. Again, however, the production of phrase-structure rules for
systemic clause structure was only feasible when limited grammatical

possibilities were considered.
More recent approaches to syntactic parsing in computational linguistics
more broadly have now advanced the field and offer far more powerful
capabilities than the more experimental analysis components of the 1990s.
It has therefore become logical to consider to what extent these may now
better bootstrap the systemic parsing process as well. Costetchi (2013)
consequently has designed and implemented a parser that still uses a
syntactic backbone, but this time employing not a context-free phrase
structure grammar but the Universal Dependency approach (UD: see
Marneffe et al. 2014; Nivre 2015) used within the Stanford Dependency
Parser (Marneffe et al. 2006; Socher et al. 2013). The approach remains
broadly familiar: a structural backbone is derived using the broad coverage
of the Stanford parser, and this is then ‘converted’ into a form compatible
with the systemic-functional syntagmatic organization.
Although the partially ‘functional’ nature of dependency relations pro-
vided by a UD parse can be used to support a more intuitive mapping to the
functional elements defined in SFG, the conversion still raises challenges.
UD is a single-layered grammar, oriented toward cross-linguistic validity
and minimal redundancy, and as a consequence collapses features which in
SFG would be assigned to different ranks (e.g. word class and roles within a
clause, as in the ‘nsubj’ relation) and metafunctions (e.g. experiential
Agents and interpersonal Subjects). The limited number of functional labels
UD defines also means that UD descriptions distinguish far fewer cases than
is commonplace within SFG. The ‘nsubj’ role provided by UD may corres-
pond in an SFG to the Actor, Behaver, Sayer, Senser, Token, or Existent,
among others. While the less delicate UD has obvious computational bene-
fits (less data required for training, faster manual annotation, etc.), it comes
at the expense of descriptive breadth, especially considering the delicate
orientation of many investigations carried out within SFL.
Costetchi’s system approaches these difficulties as follows. First, the
parser transforms the dependency graph into an SFG constituency struc-
ture, specifically following the Mood structure, and afterwards enriches
this with a series of features from Mood, Determination, and Person system
networks as described for SFG by Halliday and Matthiessen (2013). Next, the
parser assigns process types and participant roles to constituents as defined
in the Transitivity system network defined for Fawcett’s (2008) variant of
SFG. The assignment of these semantic features is not unique, however:
the parser assigns all possible semantic configurations as opposed to the
most probable one for the given clause, drawing on a database of process-
type structures (PTDB: see Neale 2002) and an auxiliary process that detects
and creates placeholders for syntactically empty elements as described in
Government and Binding theory (Haegeman 1991). Finally, for multiple-
clause sentences, the parser provides possible assignments of inter-clause
tactic relations as described in Halliday and Matthiessen’s (2013) system
network for clause combinations. Transforming the dependency graph into

a systemic Mood constituency tree and enriching it with features is per-
formed by an implementation of a generic graph matching framework
(Costetchi 2013). This allows descriptions of graph patterns with rich fea-
ture structures and embedded operations that are executed when their
patterns match a target. The same mechanisms based on graph patterns
allow both enrichment with systemic features and creation of placeholders
for syntactically null elements.
The works by Kasper (1988), O’Donnell (2005), and Costetchi (2013) high-
light that approaches combining aspects of different grammars as well as
cross-theoretic transformations are now practical and feasible options,
especially when such computationally simpler grammars yield good per-
formance and high levels of accuracy. This direction of research is then
worth exploring further with other dependency grammars, such as the
Link Grammar (Sleator and Temperley 1993), constituency grammars such
as HPSG (Collins 2003; Oepen et al. 2000) or XTAG (XTAG Research Group
2001), or Combinatorial Categorial Grammars (CCG – see Steedman 1993).
All of these accounts are currently producing high-quality results with very
broad coverage grammars. Other NLP tasks currently advancing in compu-
tational linguistics, such as semantic role labelling, temporal annotations,
spatial annotations, named entity or concept identification, and many
others, might now also be incorporated by employing similar mapping
mechanisms (Costetchi 2013).
The task of providing full paradigmatic analyses remains challenging,
however. This task can be formulated as a reasoning problem by treating
existing systemic-functional grammars (such as the Nigel Grammar) as a set
of logical constraints constituting a combinatoric possibility space (of con-
stituency structure, functions, and descriptive features). The input of parses
with other grammars can then be seen as ‘factual’ evidence to be con-
sidered when searching for solutions that resolve all the constraints given
in the problem space. Although this abstract task still exhibits very high
(computational) complexity, there is a long history in computer science of
considering solutions precisely to this problem (Kotthoff 2014). Particularly
promising heuristics are offered by the problem reduction principle defined
in computational complexity theory (Arora and Barak 2009) and the decom-
position principle in probability theory (Grinstead and Snell 2012). Given a
large high-dimensional distribution θ representing the domain knowledge,
the task is to decompose it into a set of smaller lower-dimensional distribu-
tions {θ1,θ2, . . . θn} from which the original distribution θ can be recon-
structed with no errors. With such a decomposition one could draw any
conclusions from {θ1,θ2, . . . θn} that could be inferred from θ without actu-
ally reconstructing it.
This procedure already has implementations in terms of probabilistic
graphical models (Airoldi 2007) using Bayesian Belief Networks and Markov
Random Fields, which form the foundation for probabilistic logics such as
Bayesian (Kersting and De Raedt 2007) and Markov Logic (Richardson and
Domingos 2006; Domingos et al. 2010). They might therefore also offer
good candidates for expressing system networks together with those net-
works’ constraints on syntagmatic structure, especially given that highly
efficient (polynomial and even log linear) learning and inference algorithms
already exist for them (Guo and Hsu 2002).
Another direction worth exploring is building the syntactic backbone
as native SFG constituency structures. For this task, an SFG corpus might
be built either from scratch, such as the one in Fawcett’s (1993) COMMU-
NAL project, or by transformation of existing corpora (Honnibal 2004;
Honnibal and Curran 2007). The latter option is more feasible and in line
with the idea of cross-theoretic transformations across distinct grammat-
ical accounts defended above. Such resources could then be employed
in well-established practice from computational linguistics as training
data sets for machine learning algorithms, which we will return to
briefly below.
Automated parsing with functional categories promises a number of
important applications, both within and outside of research settings. Parsed
data can help in determining the register dimensions of a text, assisting in
document classification or analysis of diachronic change. These approaches
can also help develop linguistic theory – in the case of SFL, automated
frequency counting is perhaps the only feasible way of accomplishing the
‘grammarian’s dream’ (Hasan 1987) whereby grammatical distinctions and
lexical alternatives become one unified resource. Such counting would
enable accounts to systematically extend characterizations from system to
instance while respecting statistically derived probabilities for given
contexts.
Functional linguists approach corpora both from above (i.e. looking at
collections of texts as assemblages of registers) and from below (i.e. by
building profiles of lexicogrammatical frequencies). Teich et al. (2016), for
instance, use register theory and selected elements of SFG to analyze a
large, metadata-rich corpus of scientific writing. Automated tagging and
manual annotation are used in tandem to extract frequency counts for
various lexicogrammatical features. Statistical modelling is then used to
model phylogenetic change, as well as disciplinary specialization. Findings
show that disciplines differentiate themselves not only through experien-
tial choices but through differing probabilities within tenor and mode as
well. Teich et al. point out that tenor and mode are often neglected in tool
and method development within NLP, and SFL approaches could then help
achieve more inclusive accounts.
Using similar methods, McDonald and Woodward-Kron (2016) investigate
language change over the course of membership in an on-line support
group for bipolar disorder. Over time, members’ talk increasingly comes
to align with a biomedical ideology: members’ Mood choices shift in order
to allow advice and foreground declaratives to provide hedged advice that
foregrounds lay experience – experientially, users come to construe health

problems as Possessions, rather than Identities. Zinn and McDonald (2015)
apply a similar methodology in order to track shifting lexicogrammar and
semantics of risk in print news journalism over the past thirty years.
Focusing on experiential and group-level features, they found that risk is
increasingly nominal, negatively appraised, and construed as possible,
rather than calculated.
In each of these contemporary projects, dedicated systemic parsing
would radically increase the potential features chosen for analyses, and
allow more precise division of meaning-making along metafunctional lines.
22.3 Dialogue Systems: Situated Language Use
Although the prospect of computational systems that converse with

humans has always been upheld as one of the primary goals of artificial
intelligence, progress has been relatively slow. The early system mentioned
above, SHRDLU from Winograd (1972), was a landmark system that in fact
proved a difficult act to follow. Over the past decade, an increasing number
of computational systems with impressive dialogic capabilities have been
produced, however, and the area is now coming back into the mainstream:
consumer electronics increasingly ship with computational ‘assistants’ of
one kind or another, with each supporting at least spoken interaction
between information systems and their human users.
When dialogue systems are constructed by incorporating insights from
linguistic theory, the interactional behaviour that results offers a powerful
source of additional knowledge and evaluation of the adequacy of the
theories employed. It is generally immediately apparent when interaction
does not run smoothly, and so this can be considered in terms of whether
the theories employed were adequate or not in the guidance they offered.
This applies to all levels of abstraction within dialogue systems: for
example, even when the speech recognition component is of insufficient
quality to reliably resolve what was said acoustically, one might still expect
the dialogue management component to respond more or less gracefully by
politely inquiring again as to what was said, rather than simply failing with
an internal system error.
The usual components of a computational dialogue system therefore
span a considerable breadth of linguistic knowledge as well, ranging from
spoken language recognition, parsing, semantic analysis, contextualization,
recognition of speech acts, designing responses appropriate both to the
context and to the addressee, to producing grammatical forms for those
responses and converting them into intelligible spoken output. Working on
computational dialogue systems is then of considerable value for refining
our linguistic theories in each of these areas and in combination. This
applies equally to Systemic Functional Linguistics, and particularly to all
of the varied components of the theory spread over levels of linguistic

abstraction from phonetics to context of situation.
In general, developing a dialogue system can be seen as the task of
creating an artificial ‘persona’, a non-organic being who will speak to
humans. There are many reasons for developing such beings: some devel-
opers aim at alleviating human loneliness by building great listeners and
companions while others focus on automating labour, work, and entertain-
ment by implementing virtual call-centre attendants, autonomous vehicles,
vending machines, intelligent speakers, intelligent TVs, and intelligent
home devices. To a greater or lesser extent, dialogues in all these contexts
exhibit fundamental properties of ‘situatedness’. Here, a host of well-
known linguistic phenomena, such as deixis, i.e. referring to the speech
situation, and recipient design, i.e. making sure that what is said is appro-
priate both to the context and to the state of knowledge of the addressee
(see Fischer 2016), come to play central roles. Ensuring that aspects of
the linguistic account are able both to access and to influence situation
models appropriately to control these phenomena is then an important
requirement.
These properties raise some particular theoretical challenges for systemic-
functional theory that are currently unresolved: this concerns the entire
area of establishing relations between the linguistic system and context.
Although SFL has always placed considerable weight on the notion that
language use, on the one hand, depends crucially on context and, on the
other, plays a constitutive role in constructing such contexts, the mechan-
isms available within SFL for modelling this are still schematic. When
building computational dialogue systems, however, precisely this property
must be specified and implemented in detail. A major difficulty here is the
highly ‘dynamic’ nature of the linkage between language and context. Each
utterance is dependent on the context, while also changing that context for
subsequent utterances.
Approaches to dialogue systems outside of SFL most often deal with this
flexibility by combining the so-called ‘information state’ approach of
Traum and Larsson (2003), whereby semantic representations of the cur-
rent ‘question under discussion’ and the states of knowledge of the respect-
ive interactional participants are used to trigger actions, and statistically
derived probabilities concerned with which speech acts have occurred and
which are most likely to follow. Within SFL-based systems, an early model
of this interaction was set out by O’Donnell (1990), while Fawcett (1989)
adopted more directly a procedural account based on flow charts. More
recent models interacting with systemic-functional descriptions also
include levels of description concerned with transitions between dialogue
states, modelled in a variety of ways (Teich et al. 1997; Shi et al. 2011),
but these approaches still exhibit drawbacks in comparison to the non-SFL-
based dialogue systems available. Non-SFL-based dialogue systems typically
include far stronger formalizations of processes of reasoning with
presuppositions, of the influence of the knowledge of addressees, of the

communicative goals of the interactants, of the discourse history, of impli-
cations following from the semantics, and more besides. They can, there-
fore, by no means still be seen as impoverished or simplistic with respect to
their purely linguistic competitors.
Even at the level of fine-grained lexicogrammatical choices, context-
sensitivity of many linguistic phenomena must be catered for and theoret-
ically characterized. Consider the case of a single proper name. Depending
on the states of knowledge of the participants, uttering that name may be a
simple mention, an invitation to participate, a direct address or call for
action, and so on. The same participant may also be picked out by partici-
pant ‘status’ – i.e. according to whether the participant is a speaker
(at some specific time), an addressee, or a third-person participant or over-
hearer. Thus, choices in the grammar need direct access to various organ-
izational features of the situation, and, moreover, those features change
with each utterance and with time. Furthermore, depending on what
precisely is being done with an utterance in a dialogue, the interpretation
required may be quite different.
As a concrete example, consider the case of an autonomous wheelchair
capable of spoken language interaction. Here, even uttered terms for
objects, such as ‘sofa’ or ‘TV’, may function as the referenced object of a
relative location playing the appropriate role in a command to move some-
where. In contrast, other uttered terms for objects, such as ‘wheelchair’,
may serve as a way to direct a command at the intended addressee, as in
‘Wheelchair, go to the kitchen’. Such variation demonstrates that
anchoring linguistic analysis in the agency and affordances of things pre-
sent in the situation is essential to determine the speech function of an
utterance.
Situated dialogues therefore differ from monologues and from non-
situated dialogues in substantial ways. On the one hand, a monologue such
as an argumentative essay consists of a series of uncontested statements.
As a result, each statement corresponds to a process that the author
assumes took, is taking, or will take place. On the other hand, in a dialogue
each statement must be acknowledged by the addressee(s) before interact-
ants agree that something took, is taking, or will take place. Computational
dialogue systems must then also incorporate explicit models of such
‘grounding’ as well in order to support natural interaction. Of course,
interactants may also disagree about what can be accepted as having taken
place, which can lead to interruptions and further subdialogues at any
point. In other words, while truth is proposed in monologues, in dialogues
it is negotiated. In addition, information is not only offered by interactants,
it is also demanded in the form of questions: in short, interactants
exchange goods and services, and this must be explicitly modelled.
A computational model addressing the forms of flexibility described
above at various levels has been developed by Couto-Vale (2017), building
on SFL principles. The application domain for this system is that of the
above mentioned ‘intelligent wheelchair’. Such wheelchairs primarily play
a role in interactive exchanges of goods and services. In particular, they
offer a set of services, including going or taking someone to specified
destinations and recharging themselves at an appropriate docking station.
When interacting with such a wheelchair, a human normally gives the
wheelchair a sequence of tasks for it to perform. These tasks are executions
of the wheelchair’s services. The context of a wheelchair offering services
consequently motivates particular interpretations and descriptions of the
utterances occurring. When humans and wheelchairs exchange services
linguistically, the human takes the role of a service client and the wheel-
chair takes the role of a service provider. The human tells the wheelchair
what to do and so is the source of the request; the wheelchair is the
request’s ‘destination’.
This can then be captured in terms of functional configurations in the
linguistic analyses that a dialogue system needs to perform when participat-
ing in natural dialogue in this scenario. Functional linguistic analyses appro-
priate for the context and constructed automatically on the basis of
situation-specific resources are illustrated in Tables 22.1 and 22.2; in this
scenario, the wheelchair goes by the name ‘Rolland’. In Table 22.1, what is
transferred from the service client (Agent) to the service provider (Medium)
is a request, whereas in Table 22.2 what is described is a service of the service
provider. Selecting the appropriate analysis for these grammatically very
similar cases demands that a language analyst knows about the respective
capabilities of the represented participants: only the wheelchair is capable of
moving the client, whereas only the client is capable of calling for services.
Following this line of discussion further: not every process represented in
human–wheelchair interaction is a service, and so the system, and the
underlying theoretical description, must be able to tease these distinct
Table 22.1 Description of request (by human)

I called Rolland to my bed
I brought Rolland to my bed
Client Process Provider Route
Agent – Medium Update
Table 22.2 Description of service (by human)

Rolland took me to my bed
Rolland brought me to my bed
Provider Process Client Route
Agent – Medium Update
interpretations apart. For example, humans also use descriptions of their

own future actions in order to communicate to the wheelchair that services
need to be performed that enable users’ actions to take place. For instance,
let us assume a gait-impaired woman is interacting with an intelligent
wheelchair and wishes to wash her hands. In this situation, only the
woman can wash her hands; the wheelchair cannot. However, the woman
can only wash her hands if she is at a given position in relation to a sink.
Resolving the intention of an utterance such as ‘I must wash my hands’ is
therefore complex and again demands that the system can access just who
can perform what actions in order to assign functional roles in the gram-
matical analysis appropriately.
Specifically, because the wheelchair offers services, it needs to construct
interpretations even of representations of human actions (such as ‘I must
wash my hands’) in terms of possible actions that would support those actions.
It is this that makes it possible for the wheelchair to respond in the present
case: ‘Ok, then I’ll take you to the sink.’ In other words, it is the fact that
interactants can perform some actions and cannot perform other actions, and
that interactants offer services that enable each other’s actions, that enables
such clauses to be construed as commands for the wheelchair to do some-
thing. Such knowledge must then be accessible if situationally appropriate
interpretation of utterances is going to be possible (Couto-Vale 2017).
Several practical approaches to implementing computational dialogue
systems of this kind make simplifying assumptions that provide basic func-
tionality, but with little scope for extension to cover more natural or complex
interactions of the kind illustrated here. For example, spotting keywords is a
straightforward way of ‘guessing’ what an utterance might have meant.
When the word ‘kitchen’ is spotted somewhere in an utterance, the wheel-
chair or other assistant may just assume that the intended service is to take
the user to the kitchen. This strategy becomes increasingly unwieldy when
extended to interpret recognition of a keyword such as ‘wash’ as an instruc-
tion to go somewhere. What happens in practice with such straightforward
approaches is that multiple keywords such as ‘wash’ + ‘hands’, ‘wash’ + ‘hair’,
‘wash’ + ‘dishes’, and ‘wash’ + ‘clothes’ are necessary for guessing the com-
mand. These sets of keywords become the conditions of interpretation rules,
which get progressively more complex as the number of potential clauses
increases. For this reason, although such an implementation strategy works
for simple scenarios, it does not result in reusable linguistic resources nor in
resources that are easy to maintain as the domain of application expands.
Being able to describe features of utterances as systemic options with
reference to the situation as suggested here and following SFL principles
then promises an economical way of automating understanding and
generation in general. It is by recognizing the features of the situation that
enable and disable people to mean something in particular that we can
produce resources for one situation and reuse them in another. Developing
more complex dialogue systems by building on fine-grained and dynamic
linkages between language and situation is in fact not only of interest for
applications; it is also a highly effective strategy for pushing theory devel-
opment. When a computational system is brought to the level of explicit-
ness and completeness that it can actually produce behaviour (at any levels
of linguistic abstraction), weaknesses or gaps in the theories implemented
can become glaringly evident in a way that is simply not accessible when
considering the theories ‘on paper’. When the produced behaviour does not
meet expectations, this is a good indicator of places where theoretical
frameworks may need refinement.
22.4 Multimodality
Another development gaining momentum in several areas of computa-

tional linguistics is the focus beyond language to include other modalities
or forms of expression. With this move, the range of computational work
relevant for non-computational theory building is also extended consider-
ably – particularly for the area of multimodality. Multimodality is the study
of how multiple modes of expression interact with each other in diverse
communicative situations (Bateman et al. 2017). To draw an example: a
multimodal approach to the dialogue systems discussed above might pay
attention to gestures alongside spoken language, thus extending the scope
of relevant work from natural language processing to gesture recognition, a
subfield of human–computer interaction (Rautaray and Agrawal 2015).
Alternatively, a multimodal perspective on page-based documents might
in turn involve document analysis, a subfield of computer vision, to auto-
matically infer the structure of the document and identify its contents, and
apply appropriate processing, depending on whether they consist of photo-
graphs, diagrams, other graphic elements, or written language (Doermann
and Tombre 2014) and to derive text–image relations (Bateman 2014).
While this illustrates how issues of multimodality and computation can
often be intertwined, a discussion capturing their breadth is well outside of
the scope of this chapter. Therefore, to limit and structure the discussion on
computational methods in SFL-inspired work on multimodality, we begin
with a body of work strongly rooted in SFL, namely, that directed by Kay
O’Halloran, before considering how recent advances in computational
methods may benefit and inform future research on multimodality.
In work undertaken at the Multimodal Analysis Lab at the National
University of Singapore between 2008 and 2013, O’Halloran and her team
aimed to
develop and use interactive digital technologies for multimodal analysis of
different media and to develop computational, mathematical, and visualiza-
tion techniques for interpreting semantic patterns in the resulting multi-
modal data. The research program also aimed to develop automated
computational techniques for analysis of large cultural data sets, and to
develop digital technologies that promote a systematic approach to teaching

and learning multimodal literacy and communication skills for the twenty-
first century.
(O’Halloran 2015:390)
Much has been written elsewhere about the tools for supporting multi-
modal analysis (see e.g. O’Halloran et al. 2014b); in the following, therefore,
we focus on the application of computational techniques to analyzing
multimodal data.
Despite the computational emphasis, the underlying theoretical frame-
work of O’Halloran’s projects draws heavily on what O’Halloran (2008)
conceptualizes as Systemic Functional-Multimodal Discourse Analysis
(SF-MDA). Following the social semiotic and systemic-functional approaches
to multimodality (e.g. Kress and van Leeuwen 2006; O’Toole 2011), SF-MDA
considers language and image as resources for meaning-making, building
on the rich theoretical framework provided by systemic-functional theory.
This involves, for instance, applying the concepts of metafunctions and
rank to visual images, so as to provide an integrative framework for
describing multimodal data. With this kind of framework at hand, another
question quickly emerges from the computational perspective: namely,
how to move beyond hand-picked examples and bring the powerful theor-
etical apparatus of SFL to bear on large volumes of multimodal data. Indeed,
this question is well known and long-discussed in non-SFL computational
corpus analysis.
Similar questions are posed in the emerging field of digital humanities,
which studies how computational methods and techniques can help to
answer research questions raised in the humanities (see e.g. Berry 2012;
Schreibman et al. 2016). Multimodality, however, has not been theorized
extensively in the digital humanities, except in discussions of developing
new ways of representing and disseminating research results (Svensson
2010). O’Halloran et al. (2014a:565) consequently characterize their work
as a further extension into what they define as ‘multimodal digital human-
ities’, which involves collecting large volumes of linguistic and visual data,
which are then aggregated for analysis and converted into interactive
visualizations for exploration.
O’Halloran et al. (2014a) provide a good example of this approach at
work, exploring the dynamics of urban life in Singapore. To study large
volumes of data collected from various social media services, O’Halloran
et al. (2014a) examined interpersonal meanings across two different semi-
otic resources by using computational techniques. For written language,
O’Halloran et al. (2014a:572) evaluated the sentiment of 98,733 Twitter
messages sent from specific locations by calculating a lexicon-based emo-
tion vector for each message, which captured the basic emotions of happi-
ness, sadness, fear, anger, disgust, and surprise. For photographs, they
applied automatic face detection to 301,865 images retrieved from
Instagram (O’Halloran et al. 2014a:573). By aggregating this information

into a grid defined over the map of Singapore, in which the locations
are enriched with venue information combined from FourSquare and
Wikipedia, O’Halloran et al. (2014a) investigated differences between resi-
dential and tourist areas.
In another study, Cao and O’Halloran (2015) explored differences in
photo-shooting patterns between different groups of users on Flickr, focus-
ing on differences in shot scale (close-up vs. long-distance). Trained using
texture patterns extracted by computer vision algorithms from the photo-
graphs, a machine learning algorithm – a Support Vector Machine (SVM) –
learned to distinguish between shot scales with 91.3 per cent accuracy. Cao
and O’Halloran (2015) then applied this classifier to examine photographs
taken by different groups. This revealed a strong correlation between the
user’s geographic location and shot scale: the users were more likely to take
close-up shots while in their home country, while taking more distant shots
abroad. Cao and O’Halloran (2015) suggest that this may result from indi-
viduals taking pictures of more mundane objects at home, while capturing
sights and sceneries abroad, reflecting photographic practices associated
with tourism.
Although the results discussed above show the potential of computa-
tional methods, their level of detail remains far from those commonly
found in manual analyses of multimodality within SF-MDA and other
approaches. O’Halloran et al. set out to bridge this gap in subsequent work,
identifying the following challenges:
First, it is not possible to model and predict discourse patterns extrapolating

from a limited number of detailed analyses. Second, the modeling of multi-
modal data using dimensionality reduction and clustering techniques results
in visual patterns that require a human analyst to make sense of them,
rather than delivering explicit, computable accounts of the semantic pat-
terns which have been derived.
(O’Halloran et al. 2016a:10)
To this end, they propose a ‘multimodal mixed methods research frame-

work’ that uses qualitative methods to identify key semiotic resources in
the collected data set, which are then targeted using quantitative methods,
such as mining the data with the help of machine learning algorithms.
Finally, these analyses are synthesized into visualizations for exploring the
results.
As the work of O’Halloran and her colleagues shows, rapidly developing
fields of study such as computer vision, natural language processing, and
machine learning will undoubtedly make a significant contribution to the
study of multimodality in the coming years. The work of Bateman et al.
(2016) exemplifies emerging work in this area, combining manually and
automatically generated annotation layers in a corpus describing the multi-
modality of film. Whereas various visual and aural events in film, such as
shot boundaries and background music, are captured automatically by

algorithms, filmic cohesion is described manually using the framework
set out in Tseng (2013).
Bateman et al. (2016) show that automatically generated annotation not
only reduces the time and resources spent on compiling multimodal cor-
pora, but also extends their scope by introducing layers of description
which could be otherwise considered too demanding for manual annota-
tion. Moreover, these benefits are not limited to the description of complex
dynamic multimodal phenomena in film, but apply to page-based artefacts
as well, whose annotation has proven equally time- and resource-intensive.
To draw on an example, Hiippala (2016) presents a tool which uses open
source computer vision, natural language processing, and optical character
recognition libraries to generate XML annotation from document images,
which is designed to support the manual application of the annotation
framework presented in Bateman (2008a).
Developments of this kind are now being driven forward by advances in
several interrelated areas of computer science, including machine learning,
computer vision, and NLP. In particular, results in a specific subfield of
machine learning known as ‘deep learning’ are now bringing about signifi-
cant developments in all of the aforementioned fields. This subfield focuses
on the design and use of artificial neural networks for a broad variety of
tasks (LeCun et al. 2015). Artificial neural networks follow principles of
operation inspired by the structure and activation of neurons in the human
brain and have become increasingly popular in recent years due to
increases in computing power and the volume of data available for training
the networks.
In contrast to many ‘traditional’ machine learning algorithms, in which
the features necessary for the task at hand – such as classifying an image
based on its contents or finding an object in the image – are crafted
manually by humans, neural networks learn these representations auto-
matically by adjusting their parameters during training. The epithet ‘deep’,
in turn, refers to how these parameters are organized into successive layers,
which learn increasingly abstract representations of the data. Whereas the
first layers may contain representations of changes in illumination or
texture, the subsequent ones may construe combinations of these features
into representations of particular objects. These developments are highly
relevant to multimodal research, as exemplified by now common computer
vision tasks such as image captioning and object detection, recognition, and
classification (see Bateman et al. 2017:162–6).
As a concrete example, the work of Kembhavi et al. (2016) focuses on
understanding the content and structure of diagrams and shows how far
deep learning has pushed the joint processing of language and images.
Kembhavi et al. (2016) train multiple neural networks to parse the diagrams
for constituents and their relationships. This representation of the diagram
is then fed into another neural network for ‘diagram question answering’,
which involves predicting the correct answer to a multiple choice question.

How diagrams make meanings has also been explored from an SFL-inspired,
multimodal perspective by Guo (2004), but as in many other cases, these
analyses have been limited to very few examples. In contrast, the parser
presented in Kembhavi et al. (2016) could be used to radically increase the
size of multimodal corpora for diagram description. In fact, their data set,
which includes 5,000 human-annotated diagrams, warrants a multimodal
investigation in its own right.
Finally, it is useful to consider the potential contribution of SFL and
multimodal research to what may be broadly described as the field of
artificial intelligence. Although O’Halloran et al. (2016b) envisage that
multimodal analyses conducted by experts could be reused as training data
for machine learning algorithms, state-of-the-art techniques such as deep
learning are notoriously hungry for data. This data is needed both for
training models for various tasks and to measure their performance. Typic-
ally, this data is crowd-sourced through services such as Amazon Mechan-
ical Turk, which allows individuals to bid on and undertake small tasks,
such as labelling objects or drawing their outlines in images. At this stage,
given their experience in developing rich and systematic annotation
schemes, researchers working on multimodality could develop annotation
frameworks for multimodal data and investigate how best to instruct the
non-expert annotators in their task, thus improving the quality of data.
22.5 Toward Further Integration
Better engagement with CL and NLP has a number of benefits for SFL. First,
computational methods can facilitate automatable and reproducible work.
The large amounts of time taken for manual annotation of data mean that
many SFL projects face time and cost constraints. Heavily automated work-
flows, on the other hand, can be deployed on new data at little to no cost.
This seems to be a practical path to Matthiessen’s notion of language as ‘an
assemblage of registers’ (Matthiessen 2015b:44): the same set of routines,
automatically applied to corpora of different domains, could provide an
insightful account of how Field, Tenor, and Mode influence the probabil-
ities for content-stratum phenomena.
At the same time, computational approaches make it possible to empiric-
ally test key components of an SFG. Automated text processing, for
example, may be able to shed light on the oft-noted difficulty of process-
type identification: if a model trained on large, well-annotated collections of
process types cannot accurately predict process-type labels in unseen data
from a similar text type, we may have an indication that our current
understanding of experiential semantics is incomplete. Moreover, as Halli-
day discovered, the number of words needed to collect a quantitatively
useful sample grows exponentially with the delicacy of the phenomenon
of interest. While only 2000 main clauses are needed to create a profile of
Mood or Process Type, hundreds of thousands of words (or a smaller, highly
specific sample) may be needed in order to develop frequency profiles for
lexical alternatives (Matthiessen 2015a). Realizing the statistical component
of SFG at the grammatical pole of lexicogrammar is therefore dependent on
computational methods.
Another important benefit of combining SFL and NLP is that the high-
quality data produced by human annotators with detailed training in SFG
could be effectively repurposed as training data for machine learning algo-
rithms. As noted above, a consistent, well-annotated data set is the main
prerequisite for the development of a high-quality statistical parser.
Finally, for NLP practitioners, SFL provides several well-articulated
hypotheses concerning the relationship between critical linguistic ques-
tions, such as the relationship between text and context, lexis and gram-
mar, and words and meaning. In clearly differentiating between form,
function, words, and meaning, SFG may be able to avoid the pitfalls that
limit the utility of more popular computational grammars for functional-
semantic tasks. It is certainly possible that current limitations in NLP are
not the result of limitations in statistical methods, but in the grammars
accepted by algorithms as input and output. Put another way, even per-
fectly accurate automatic annotation may have limited usefulness if what is
being annotated does not correspond to meaningful distinctions within the
grammar of a language. Goals that are still distant in NLP, such as semantic
parsing and discourse-level annotation, could foreseeably stand to benefit
from the relatively holistic account provided by SFL.
22.6 Conclusion
SFL is a complex and rich theoretical account in which several distinct

representational resources are regularly used. Many of these are currently
at the limit of or beyond what can be modelled computationally. Computa-
tion has, however, made enormous strides over the past ten years, and
many of the phenomena at the core of SFL theorizing at the outset –
particularly, for example, reliance on data and corpora – are now coming
within reach. In fact, such approaches to language are only achievable with
computational methods, and so it is both necessary and logical that inter-
action between SFL and computation should continue and, in some areas,
intensify. In previous rounds of interaction, the efforts of particular indi-
viduals have been central. For example, beginning with work on machine
translation and then, later, natural language generation, the willingness of
Michael Halliday to engage with emerging technologies played a crucial
role (see Bateman and O’Donnell 2015). There is now considerably more
need for researchers who are trained both in computation and SFL as both
of these disciplines continue to evolve and grow.
References
Airoldi, E. M. 2007. Getting Started in Probabilistic Graphical Models.

PLoS Computational Biology 3(12): e252.
Arora, S. and B. Barak. 2009. Computational Complexity: A Modern Approach.
Bateman, J. A. 1997. Enabling Technology for Multilingual Natural
Language Generation: The KPML Development Environment. Natural Lan-
guage Engineering 3(1): 15–55.
Bateman, J. A. 2008a. Multimodality and Genre: A Foundation for the Systematic
Analysis of Multimodal Documents. Basingstoke: Palgrave Macmillan.
Bateman, J. A. 2008b. Systemic Functional Linguistics and the Notion of
Linguistic Structure: Unanswered Questions, New Possibilities. In J. J.
Webster, ed., Meaning in Context: Strategies for Implementing Intelligent Appli-
cations of Language Studies. Sheffield: Equinox. 24–58.
Bateman, J. A. 2014. Text and Image: A Critical Introduction to the Visual/Verbal
Divide. London: Routledge.
Bateman, J. A. and M. O’Donnell. 2015. Computational Linguistics: The
Halliday Connection. In J. J. Webster, ed., The Bloomsbury Companion to
Bateman, J. A. and M. Zock. 2017. Natural Language Generation. In
R. Mitkov, ed., Oxford Handbook of Computational Linguistics. 2nd ed. Oxford:
Oxford University Press. 284–304.
Bateman, J. A., C. Tseng, O. Seizov, A. Jacobs, A. Lüdtke, M. G. Müller, and
O. Herzog. 2016. Towards Next-generation Visual Archives: Image, Film
and Discourse. Visual Studies 31(2): 131–54.
Bateman, J. A., J. Wildfeuer, and T. Hiippala. 2017. Multimodality: Founda-
tions, Research and Analysis: A Problem-oriented Introduction. Berlin: Mouton
de Gruyter.
Berry, D. M., ed. 2012. Understanding Digital Humanities. Basingstoke:
Palgrave Macmillan.
Cao, Y. and K. L. O’Halloran. 2015. Learning Human Photo Shooting
Patterns from Large-Scale Community Photo Collections. Multimedia Tools
and Applications 74(24): 11499–516.
Chomsky, N. 1957. Syntactic Structures. The Hague: Mouton and Co.
Collins, M. 2003. Head-driven Statistical Models for Natural Language
Parsing. Computational Linguistics 29(4): 589–637.
Costetchi, E. 2013. A Method to Generate Simplified Systemic Functional
Parses from Dependency Parses. In Proceedings of the Second International
Conference on Dependency Linguistics. Prague, August 27–30, 2013. Charles
University in Prague. Prague: Matfyzpress. 68–77.
Couto-Vale, D. 2017. How to Make a Wheelchair Understand Spoken Commands.
PhD Thesis, Bremen University.
Doermann, D. and K. Tombre, eds. 2014. Handbook of Document Image
Processing and Recognition. London: Springer-Verlag.
Domingos, P., S. Kok, D. Lowd, and H. Poon. 2010. Markov Logic. Journal
of Computational Biology: A Journal of Computational Molecular Cell Biology
17(11): 1491–508.
Fawcett, R. P. 1988. Language Generation as Choice in Social Interaction. In
M. Zoch and G. Sabah, eds., Advances in Natural Language Generation.
Fawcett, R. P. 1989. Towards a Systemic Flowchart Model for Discourse
Analysis. In R. P. Fawcett and D. Young, eds., New Developments in Systemic
Linguistics: Theory and Application. London: Pinter. 116–43.
Fawcett, R. P. 1993. The Architecture of the COMMUNAL Project in NLG
(and NLU). In The Fourth European Workshop on Natural Language Generation.
Pisa.
Fawcett, R. P. 2008. Invitation to Systemic Functional Linguistics through the
Cardiff Grammar. Sheffield: Equinox.
Fischer, K. 2016. Designing Speech for a Recipient: The Roles of Partner Modeling,
Alignment and Feedback in So-called ‘Simplified Registers’. Amsterdam: John
Benjamins.
Grinstead, C. M. and J. L. Snell. 2012. Introduction to Probability. Providence:
American Mathematical Society.
Guo, H. and W. Hsu. 2002. A Survey of Algorithms for Real-time Bayesian
Network Inference. In The Joint AAAI/KDD/UAI02 Workshop on Real-time
Decision Support and Diagnosis Systems. Edmonton, Canada.
Guo, L. 2004. Multimodality in a Biology Textbook. In K. L. O’Halloran, ed.,
Multimodal Discourse Analysis: Systemic Functional Perspectives. London:
Haegeman, L. 1991. Introduction to Government and Binding Theory, Volume 2.
London: Blackwell.
Halliday, M. A. K. 1996. On Grammar and Grammatics. In R. Hasan,
C. Cloran, and D. Butt, eds., Functional Descriptions: Theory in Practice.
Halliday, M. A. K. and C. M. I. M. Matthiessen. 2014. Halliday’s Introduction
to Functional Grammar. 4th ed. Oxford: Routledge.
Hasan, R. 1987. The Grammarian’s Dream: Lexis as Most Delicate
Grammar. In M. A. K. Halliday and R. P. Fawcett, eds., New Developments
in Systemic Linguistics, Volume 1. London: Frances Pinter. 184–211.
Hiippala, T. 2016. Semi-automated Annotation of Page-based Documents
within the Genre and Multimodality Framework. In Proceedings of the
10th SIGHUM Workshop on Language Technology for Cultural Heritage, Social
Sciences, and Humanities. Berlin: Association for Computational Linguistics.
84–9.
Honnibal, M. 2004. Converting the Penn Treebank to Systemic Functional
Grammar. In Proceedings of the Australasian Language Technology Workshop
(ALTW04). Available online at: www.alta.asn.au/events/altw2004/publica
tion/04-27.pdf. (Last accessed 12/09/2017.)
Honnibal, M. and J. R. Curran. 2007. Creating a Systemic Functional Gram-

mar Corpus from the Penn Treebank. In Proceedings of the ACL 2007 Work-
shop on Deep Linguistic Processing. Prague: Association for Computational
Linguistics. 89–96.
Kasper, R. 1988. An Experimental Parser for Systemic Grammars. In Pro-
ceedings of the 12th International Conference on Computational Linguistics.
Budapest: John von Neumann Society for Computing. 309–12.
Kembhavi, A., M. Salvato, E. Kolve, M. Seo, H. Hajishirzi, and A. Farhadi.
2016. A Diagram Is Worth a Dozen Images. In Proceedings of the 14th
European Conference in Computer Vision. Amsterdam, The Netherlands, October
11–14, 2016. Cham: Springer. 235–51.
Kersting, K. and L. De Raedt. 2007. Bayesian Logic Programming: Theory
and Tool. In L. Getoor and B. Taskar, eds., Introduction to Statistical Rela-
tional Learning. Cambridge, MA: MIT Press. 291–321.
Kotthoff, L. 2014. Algorithm Selection for Combinatorial Search Problems:
A Survey. AI Magazine 35(3): 48–60.
LeCun, Y., Y. Bengio, and G. Hinton. 2015. Deep Learning. Nature 521:
436–44.
Mann, W. C. 1983. An Overview of the PENMAN Text Generation System. In
Proceedings of the Third National Conference on Artificial Intelligence. Menlo
Park: The AAAI Press. 261–5.
Mann, W. C. and C. M. I. M. Matthiessen. 1985. Demonstration of the
Nigel Text Generation Computer Program. In J. D. Benson and W. S.
Greaves, eds., Systemic Perspectives on Discourse, Volume 1. Norwood: Ablex.
50–83.
Manning, C. D., M. Surdeanu, J. Bauer, J. Finkel, S. J. Bethard, and
D. McClosky. 2014. The Stanford CoreNLP Natural Language Processing
Toolkit. In Proceedings of the 52nd Annual Meeting of the Association for
Computational Linguistics. Baltimore: Association for Computational
Linguistics. 55–60.
Marneffe, M., T. Dozat, N. Silveira, K. Haverinen, F. Ginter, J. Nivre, and
C. D. Manning. 2014. Universal Stanford Dependencies: A Cross-linguistic
Typology. In Proceedings of the 9th International Conference on Language
Resources and Evaluation. Reykjavik: European Language Resources Associ-
ation (ELRA). 4585–92.
Marneffe, M., B. MacCartney, and C. D. Manning. 2006. Generating Typed
Dependency Parses from Phrase Structure Parses. Lrec 6(3): 449–54.
Matthiessen, C. M. I. M. 1985. The Systemic Framework in Text Generation:
Nigel. In J. Benson and W. Greaves, eds., Systemic Perspective on Discourse,
Voume 1. Norwood: Ablex. 96–118.
Matthiessen, C. M. I. M. 2015a. Halliday’s Conception of Language as a
Matthiessen, C. M. I. M. 2015b. Register in the Round: Registerial Cartog-

McDonald, D. 1980. Natural Language Production as a Process of Decision Making
under Constraint: Cambridge, MA: MIT Press.
McDonald, D. and R. Woodward-Kron. 2016. Member Roles and Identities in
Online Support Groups: Perspectives from Corpus and Systemic Func-
tional Linguistics. Discourse and Communication 10(2): 157–75.
Neale, A. C. 2002. More Delicate TRANSITIVITY: Extending the PROCESS TYPE for
English to Include Full Semantic Classifications. PhD Thesis, Cardiff
University.
Nivre, J. 2015. Towards a Universal Grammar for Natural Language Process-
ing. In A. Gelbukh, ed., Computational Linguistics and Intelligent Text Process-
ing. Berlin: Springer. 3–16.
O’Donnell, M. 1990. A Dynamic Model of Exchange. Word 41(3): 293–328.
O’Donnell, M. 1993. Reducing Complexity in a Systemic Parser. In Proceed-
ings of the Third International Workshop on Parsing Technologies. Tilburg:
Association for Computational Linguistics. 203–17.
O’Donnell, M. 1994. Sentence Analysis and Generation: A Systemic Perspective.
PhD Thesis, University of Sydney.
O’Donnell, M. 2005. The UAM Systemic Parser. In Proceedings of the
1st Computational Systemic Functional Grammar Conference. Sydney: Univer-
sity of Sydney. 47–55.
O’Donnell, M. 2008. Demonstration of the UAM Corpus Tool for Text and
Image Annotation. In Proceedings of the ACL ’08: HLT Demo Session. Colum-
bus: Association for Computational Linguistics. 13–16.
O’Donnell, M. and J. A. Bateman. 2005. SFL in Computational Contexts:
A Contemporary History. In R. Hasan, C. M. I. M. Matthiessen, and J. J.
Webster, eds., Continuing Discourse on Language: A Functional Perspective,
Oepen, S., D. Flickinger, H. Uszkoreit, and J. Tsujii. 2000. Introduction to
the Special Issue on Efficient Processing with HPSG. Natural Language
Engineering 6(1): 1–14.
O’Halloran, K. L. 2008. Systemic Functional-Multimodal Discourse Analysis
(SF-MDA): Constructing Ideational Meaning Using Language and Visual
Imagery. Visual Communication 7(4): 443–75.
O’Halloran, K. L. 2014. Multimodal Discourse Analysis. In K. Hyland and
B. Paltridge, eds., The Bloomsbury Companion to Discourse Analysis. London:
O’Halloran, K. L. 2015. Multimodal Digital Humanities. In P. P.
Trifonas, ed., International Handbook of Semiotics. Dordrecht: Springer.
389–416.
O’Halloran, K. L., A. Chua, and A. Podlasov. 2014a. The Role of Images
in Social Media Analytics: A Multimodal Digital Humanities Approach.
In D. Machin, ed., Visual Communication. Berlin: Mouton de Gruyter.
565–88.
O’Halloran, K. L., M. E. Kwan Lin, and S. Tan. 2014b. Multimodal Analytics:

Software and Visualization Techniques for Analyzing and Interpreting
Multimodal Data. In C. Jewitt, ed., The Routledge Handbook of Multimodal
Analysis. 2nd ed. London: Routledge. 386–96.
O’Halloran, K. L., S. Tan, P. Wignell, J. A. Bateman, D. Pham, M. Grossman,
and A. Vande Moere. 2016. Interpreting Text and Image Relations
in Violent Extremist Discourse: A Mixed Methods Approach for Big
Data Analytics. Terrorism and Political Violence. DOI: 10.1080/
09546553.2016.1233871
O’Halloran, K. L., S. Tan, D. Pham, J. A. Bateman, and A. Vande Moere. 2018.
A Digital Mixed Methods Research Design: Integrating Multimodal Analy-
sis with Data Mining and Information Visualization for Big Data Analy-
tics. Journal of Mixed Methods Research 12(1): 11–30.
O’Toole, M. 2011. The Language of Displayed Art. 2nd ed. Abingdon:
Routledge.
Peters, S. P. and R. W. Ritchie. 1973. On the Generative Power of Transform-
ational Grammars. Information Sciences 6: 49–83.
Rautaray, S. S. and A. Agrawal. 2015. Vision Based Hand Gesture Recogni-
tion for Human Computer Interaction: A Survey. Artificial Intelligence
Review 43(1): 1–54.
Richardson, M. and P. Domingos. 2006. Markov Logic Networks. Machine
Learning 62(1–2): 107–36.
Schreibman, S., R. Siemens, and J. Unsworth, eds. 2016. A New Companion to
Digital Humanities. 2nd ed. Chichester: Wiley-Blackwell.
Shi, H., C. Jian, and C. Rachuy. 2011. Evaluation of a Unified Dialogue Model
for Human-Computer Interaction. International Journal of Computational
Linguistics and Applications 2(1): 155–73.
Sleator, D. D. K. and D. Temperley. 1993. Parsing English with a Link
Grammar. In Third International Workshop on Parsing Technologies (IWPT).
Tillburg: Association for Computational Linguistics. 277–92. Available
online at: www.link.cs.cmu.edu/link/ftp-site/link-grammar/LG-IWPT93
.pdf. (Last accessed 12/09/2017.)
Socher, R., J. Bauer, C. D. Manning, and A. Y. Ng. 2013. Parsing with
Compositional Vector Grammars. In Proceedings of the 51st Annual Meeting
of the Association for Computational Linguistics. Stroudsburg: Association for
Computational Linguistics. 455–65.
Steedman, M. J. 1993. Categorial Grammar. Lingua 90: 221–58.
Steels, L. 2005. The Emergence and Evolution of Linguistic Structure:
From Lexical to Grammatical Communication Systems. Connection Science
17(3–4): 213–30.
Svensson, P. 2010. The Landscape of Digital Humanities. Digital Humanities
Quarterly 4(1). Available online at: www.digitalhumanities.org/dhq/vol/4/1/
000080/000080.html. (Last accessed 12/09/2017.)
Teich, E., S. Degaetano-Ortlieb, P. Fankhauser, H. Kermes, and
E. Lapshinova-Koltunski. 2016. The Linguistic Construal of Disciplinarity:
A Data Mining Approach Using Register Features. Journal of the Association

for Information Science and Technology (JASIST) 67(7): 1668–78.
Teich, E., E. Hagen, B. Grote, and J. A. Bateman. 1997. From Communicative
Context to Speech: Integrating Dialogue Processing, Speech Production
and Natural Language Generation. Speech Communication 21(1–2): 73–99.
Traum, D. and S. Larsson. 2003. The Information State Approach to
Dialogue Management. In R. Smith and J. van Kuppevelt, eds., Current
and New Directions in Discourse and Dialogue. Dordrecht: Kluwer Academic
Publishers. 325–53.
Tseng, C. 2013. Cohesion in Film: Tracking Film Elements. Basingstoke: Palgrave
Macmillan.
Weerasinghe, R. 1994. Probabilistic Parsing in Systemic Functional Grammar.
PhD Thesis, School of Computing Mathematics, University of Wales
College of Cardiff.
Winograd, T. 1972. Understanding Natural Language. New York: Academic
Press.
XTAG Research Group. 2001. A Lexicalized Tree Adjoining Grammar
for English. IRCS, University of Pennsylvania. Available online at:
www.cis.upenn.edu/~xtag/tech-report/tech-report.html. (Last accessed
12/09/2017.)
Zinn, J. O. and D. McDonald. 2015. Changing Discourses of Risk and Health
Risk. In J. M. Chamberlain, ed., Medicine, Risk, Discourse and Power. London:
23
Clinical Linguistics
Elissa Asp and Jessica de Villiers
In memory of Jonathan Fine (1949–2015).
23.1 Introduction
In this chapter, we first outline the goals and scope of clinical linguistics
and then review the work that has been undertaken using Systemic Func-
tional Linguistics (SFL) models. Our approach to SFL contributions is organ-
ized by diagnostic groups and, within groups, by metafunctions, linguistic
levels (‘strata’ in SFL terms), and domains as relevant. One of our goals will
be to clarify the function(s), domain(s), strata, and disorders which have
(or have not) been addressed by SFL work in clinical linguistics. This forms
the background for assessing existing and potential contributions of SFL to
clinical linguistics and ‘state of the art’.
23.2 Goals and Scope of Clinical Linguistics
Clinical linguistics first appeared as a ‘named’ field of inquiry only in

1981 with the publication of David Crystal’s Clinical Linguistics. Crystal
described work that linguists had been involved with since the 1960s and
set an agenda for clinical linguistics as an emerging applied field. As he
articulated it, clinical linguistics was ‘the application of linguistic science to
the study of communication disability, as encountered in clinical situ-
ations’ (Crystal 1981:1). Broadly, this involved providing explicit, system-
atic, and comprehensive (from phonetics to pragmatics) descriptions of
instances of language use (speech samples) that could inform the work of
With gratitude, as always, to Arnaud and John for their support.
588 ELISSA ASP AND JESSICA DE VILLIERS
clinicians. Its helpfulness was conceptualized as aiding clarification (pro-

viding a coherent terminology for labelling the linguistic manifestations of
speech and language impairments), identification (what communicative
abilities are dysfunctional or atypical?), diagnosis (what linguistic and com-
municative signs are associated with what diagnostic categories?), assess-
ments (to what extent do communicative abilities differ from ‘normal’
usage?), and the design of interventions. All of these depend on linguistic
expertise in providing what Halliday (2005) has subsequently called ‘thick
accounts’ of linguistic and contextual systems of normal use in populations
which the discourse of individuals or groups with a disorder can be related
to. (See Ryle 1968 and/or Geertz 1973 for related concepts of ‘thick descrip-
tions’ in philosophy of language and ethnography respectively.) Thus, clin-
ical linguistics presupposes the availability or creation of reference corpora
and descriptions of them, and of course ethnographies and grammars
designed, or at least adaptable, to describe everyday language use in all its
functional richness. The functional and descriptive work of British lin-
guists, complemented by the ‘ordinary language’ orientation of many of
its leading philosophers of language in the mid-twentieth century, made
Britain a natural environment for the development of clinical linguistics as
practised and articulated by Crystal and colleagues.
Crystal also distinguished the goals of clinical linguistics from, for
instance, neurolinguistics and psycholinguistics. Clinical linguistics was
to focus on the detailed description of linguistic and pragmatic factors
contributing to communicative dysfunction, rather than to use clinical
cases or disorders as evidence for theorizing about language or neuro-
logical processing. That is, it was a novel, practical undertaking aimed at
supporting the work of speech therapists and other clinicians. Neverthe-
less, Perkins (2011) notes that the journal Clinical Linguistics & Phonetics
(CLP) includes articles addressing theoretical aspects of speech production
and perception within its mandate. Its mandate also includes the linguis-
tics and phonetics of disorders of speech (including sign), language, and
communication and pragmatic aspects of these, as well as clinical dialect-
ology and sociolinguistics, research on communication disorders in multi-
lingual populations and in under-researched populations, and languages
other than English.
The CLP mandate excludes studies of ‘normal speech or language devel-
opment’ without clear clinical relevance. However, the type of descriptive
work required for clinical linguistics is highly dependent on descriptions of
‘normal speech or language development’ such as the Survey of English
Dialects and the Nuffield projects describing children’s language, as well as
the substantial descriptions of English represented in large reference gram-
mars (e.g. Quirk et al. 1985; Huddleston and Pullum 2002), functional
grammars such as Halliday’s (1985, 1994; Halliday and Matthiessen 2004),
and the myriad descriptions of specific registers, genres, and text types
undertaken since those early projects.
Clinical Linguistics 589
Crystal (2002, 2013) and Perkins (2011) review the publication history of
CLP and point to accomplishments but also ‘gaps’ in clinical linguistics
research. For example, while the ‘successes’ of clinical linguistics are
reflected by more publications and practitioners over the years, Crystal
(2002, 2013) observes that most work is focused on English: only
12 per cent of articles published between 1987 and 2001 addressed other
languages. The English focus may be because of the well-developed descrip-
tive resources characterizing for English. However, at ICPLA 2016,
28 per cent of presentations addressed languages other than English. So,
the English focus may be shifting. Crystal and Perkins’ reviews also note a
predominance of work (over 60 per cent of all papers) on phonetics and
phonology to the detriment of virtually all other areas of inquiry. Perkins
(2011) attributes this to the case loads for (UK) speech therapists, which are
similarly skewed towards developmental disorders of speech. Crystal also
notes the rarity of work on grammatical and lexical impairments and
especially lamented the even rarer appearance of semantic, discourse, and
pragmatic studies. Of particular interest for our current purpose is Crystal’s
(2013:239–41) call for more ‘routine analysis of text, spoken or written,
from a pragmatic point of view’, with pragmatics defined as ‘the study of
the choices we make when we use language, of the intentions behind those
choices and of the effects that those choices convey’.
Despite the predominance of aspects of phonetics and phonology, the
other areas addressed by clinical linguistics are very broad, and extend
to disorders where speech and language are not primary presenting signs,
but communication may be more or less subtly affected. Thus, the areas
studied in clinical linguistics include fluency disorders (e.g. stuttering,
dysarthria), developmental disorders such as autism spectrum disorder,
(specific) language impairment, speech sound disorder, apraxia of
speech, Down syndrome and Williams syndrome, and acquired disorders
resulting in aphasic and/or motor symptoms from brain injuries, stroke,
the progressive aphasias, tumors, infections, and other causes. However,
there are also studies of discourse or language (dys-)functions arising
from neurodegenerative disorders such as mild cognitive impairment,
Alzheimer’s disease, and Parkinson’s disease, as well as studies addressing
psychiatric disorders such as schizophrenia, anxiety, depression, Atten-
tion Deficit Hyperactivity Disorder (ADHD), and Post-Traumatic Stress
Disorder (PTSD). In some of these, the function of clinical linguistic ana-
lyses is not typically targeted towards speech-language therapy, but vari-
ously towards aiding diagnosis, monitoring treatments and evaluating
therapeutic interventions, improving understanding of how disorders
affect other functional abilities, how such effects may be compensated
for, and also as providing ‘windows’ on cognitive and emotional states of
speakers. Additionally, there is now a large independent field that exam-
ines clinical ‘doctor–patient’ and other interactions. This work is import-
ant but beyond the scope of this review.
23.3 SFL Contributions to Clinical Linguistics
The calls from Crystal – for clinical linguists to engage in more systematic
descriptions of discourse, to take on semantics, grammar, and lexical ana-
lyses more extensively, and especially to consider the choices speakers
make in discourse as indices for understanding ‘why’ speakers exhibit
specific patterns – suggest that SFL models of language and descriptive
techniques should provide excellent frameworks and a toolkit for clinical
linguistics. SFL’s functional grounding, its focus on text/discourse as the
object of linguistics, and its organization of grammars as system networks
that characterize semiotic potential as choices that may be probabilistically
weighted for demographic, contextual, and textual factors recommend it
for clinical work.
23.3.1 SFL Work on Schizophrenia

We begin with the influential contributions of Rochester and Martin (1979),
whose monograph Crazy Talk: A Study of the Discourse of Schizophrenic Speakers
pioneered an approach to describing discourse production in schizophre-
nia, for two reasons. Crazy Talk and related work (e.g. Martin and Rochester
1975; Rochester et al. 1977) is, to our knowledge, the earliest published
research using an SFL construct, ‘cohesion’, for clinical research and it has
been the most influential and translated work.
According to the current Diagnostic and Statistical Manual of Mental
Disorders (American Psychiatric Association (APA) 2013, henceforth DSM-V),
schizophrenia is a unique psychiatric disorder, characterized by at least
two symptoms from the following set: delusions, hallucinations, disor-
ganized speech, grossly abnormal psychomotor behaviour such as catato-
nia, and negative symptoms such as restricted affect or avolition/
asociality. For diagnosis, at least one of the symptoms must be from
the first three (positive) symptoms. The cause of schizophrenia is
unknown. However, during the time of Rochester and Martin’s work,
the then DSM-II (1968) characterized disorders but did not provide diag-
nostics. It referred to ‘incoherence or loosening of associations’ that were
signs of ‘thought disorder’ in disorganized schizophrenia as distinct from
paranoid schizophrenia. It did not explicitly mention language or speech.
Robins and Guze (1970) made the case that the absence of diagnostic
criteria, and indeed of a clear notion of construct validity in psychiatry,
may have contributed to over-diagnosis of schizophrenia. Aboraya et al.
(2005) argue that Robins and Guze’s paper stimulated research to iden-
tify reliable diagnostics as well as to evaluate their construct validity. The
DSM-III (1980) responded to these matters in that it did present diagnos-
tic criteria. However, not until the publication of the DSM-IV (1994) was
‘disorganized speech’ explicitly identified as a diagnostic criterion for
schizophrenia.
Linguistic ‘cohesion’ refers to resources for marking presuppositional

relations in text. In Halliday and Hasan’s (1976) Cohesion in English, cohesive
links, or ties, are established through various kinds of reference such as
the use of pronouns to refer to previously mentioned or contextually
available participants, of ellipsis (reference achieved by the omission of
elements mentioned in a prior clause), and of substitution of non-nominal
elements (e.g. do or so) to refer to previously mentioned events or states.
Cohesion is also achieved through explicit marking of intersentential
logical relations (e.g. however, consequently) and lexical semantic relations
of identity (repetition of lexemes), synonymy, antonymy, and collocation
(the probability of lexical items, such as cogent to co-occur with others such
as argument or evidence) (see Taboada, this volume).
In Crazy Talk, Rochester and Martin examined cohesive ties in the dis-
course of ten schizophrenic speakers diagnosed with thought disorder, ten
without thought disorder, and ten control speakers. The data consisted of
three-minute excerpts taken from half-hour unstructured interviews, car-
toon descriptions and interpretations, and narrative retells. They found
that both groups with schizophrenia used fewer cohesive ties overall than
the control group, and used exophoric references (reference to extra-textual
context) more frequently than control speakers. In interviews, the thought-
disordered group had more lexical cohesion than the non-thought-
disordered group. The authors suggested the high proportion of lexical
cohesion reflected discourse development by associative lexical patterning
rather than by logically or discoursally salient experiential meanings.
McKenna and Oh (2005) observe that although subsequent work has not
consistently replicated Rochester and Martin’s quantitative results, it has
consistently pointed to quantifiable and, in some cases, statistically reliable
differences in how reference and lexical cohesion are used. They suggest the
error patterns indicate disruption of semantic systems as one of the factors
contributing to atypical cohesion, especially in speakers diagnosed with
thought disorder (McKenna and Oh 2005:111–16).
Additionally, Crazy Talk argued for the importance of examining produc-
tions, and urged that full description of schizophrenic discourse was
needed to understand the considerable, yet periodic communication dis-
turbances described as thought disorder. At the same time, it cautioned
against equating striking cohesion patterns with (in)coherence, and sug-
gested that understanding how cohesive selections affect listeners would
require experimental studies focused on listeners (Rochester and Martin
1979:81). Rochester et al. (1977) explored this and also made the point that
trained non-linguist analysts could reliably assess cohesion. Crazy Talk also
highlighted the effects of genre variation on language production. For
instance, unstructured interviews are cognitively more demanding, requir-
ing more creative productions, than narrative retells, where new informa-
tion does not have to be introduced or developed, and this may affect
cohesive patterns (Rochester and Martin 1979:100). Docherty et al. (1996)
subsequently developed the approach into a measure of communication

disturbance for schizophrenia, and Docherty et al. (2003) found most refer-
ential disturbances were stable over time, suggesting a trait more than state
association for cohesive patterns in schizophrenia. Cohesion analysis has
become a main approach for clinical characterizations of discourse produc-
tion in schizophrenia (Kuperberg 2010).
More recently, interpersonal aspects of schizophrenic speech have been
analyzed using SFL techniques (Fine 2006 for overview; Cohen 2011). Cohen
looked at ‘speech functions’ and ‘appraisal patterns’ in the discourse of four
speakers with schizophrenia and four matched control speakers. Speech
functions are grammaticalized primary speech acts (e.g. statements and
questions) through which speakers give or receive goods or information
and realize interactional moves such as initiating or continuing a conversa-
tion (see Berry, this volume). Appraisal is concerned with lexicogrammati-
cal encoding of evaluation, affect, and emotion in discourse (Martin and
White 2005; also see Hood, this volume). Cohen found that the speakers
with schizophrenia sometimes show inconsistent interpersonal positions
by providing continuing moves that contrast with or reject their own prior
moves. Speakers with schizophrenia also produce texts with low precision
and frequent modalization of positive evaluations that suggest a lack of
commitment to their evaluations.
While SFL work on interpersonal contributions to the discourse charac-
teristics of schizophrenic speakers has been limited, it is an important area
given its potential to give insight about negative symptoms such as flat
affect and poverty of speech as well as about (dis-)engagement of partici-
pants in interactions. At present, emotional and social aspects of schizo-
phrenic speakers’ discourse are the focus of intense research initiatives
within the broader clinical community.
23.3.2 Neurodevelopmental Disorders

In DSM-V, neurodevelopmental disorders include intellectual developmen-
tal disorder, communication disorders, autism spectrum disorder, atten-
tion deficit hyperactivity disorder, specific learning disorder, and motor
disorders. Of these, SFL techniques have been applied to autism spectrum
disorders, three communication disorders (specific language impairment,
speech sound disorder, and stuttering), and attention deficit hyperactivity
disorder. Studies have sought to improve functional descriptions of clinical
features, offer reliable ways to categorize clinical impressions, discriminate
between different diagnostic groups, and demonstrate treatment effects.
23.3.2.1 Autism Spectrum Disorder (ASD)

Language and social communication problems are core features of ASD.
Language acquisition is often delayed, and many speakers with ASD have
problems developing functional language. For speakers who develop
functional communication, grammatical and lexical skills are often strong,

but the social use of language poses difficulties (Adams and Bishop 1989).
A number of studies have applied SFL techniques to the discourse of
speakers with an ASD, and two discussions of the application of SFL to the
discourse of ASD (Bartlett et al. 2005; de Villiers 2006b) highlight patterns
in the textual metafunction as particularly relevant.
Cohesion
Baltaxe and D’Angiola (1992) examined cohesion in one-hour discourse
samples of ten children with autism, eight children with specific language
impairment (SLI) and eight typical developing (TD) children, recorded
during semi-structured conversational play. They analyzed both correct
and erroneous uses of cohesive categories and found that all three groups
used cohesive tie types (reference, ellipsis, and conjunction) in the same
descending order of frequency. However, the group with autism made
significantly more errors than the SLI and TD groups. The SLI and TD
groups did not differ from each other significantly.
Fine et al. (1994) studied cohesion in conversational interviews of
speakers with Asperger Syndrome (AS), speakers with high-functioning
autism (HFA), and a control group. Blind coders rated subtypes of exopho-
ric references to the physical environment and endophoric references,
where antecedents are found within the discourse. The study found
speakers with AS produced more unclear references than other groups,
and that speakers with HFA had fewer ties to previous stretches of mutual
conversation than other groups, using more exophoric reference, a pat-
tern that Fine and colleagues suggest indicates a reduced connection to
the co-constructed verbal dialogue. The approach differentiated HFA
from AS, which at the time were separate diagnostic categories in the
autism spectrum.
Intonation
Fine et al. (1991) also looked at stress and intonation (also see Bowcher and
Debashish, this volume) use in speakers with HFA, AS, and a control group.
They compared how the different groups used intonation, and how they
used stress to indicate Given and New information. Specifically, they
marked intonation boundaries, and coded whether they occurred at appro-
priate or inappropriate locations. They coded and quantified the distribu-
tion of stress to signal Given and New information within tone groups
according to six types: unmarked stress – with stress occurring on the last
content word in a tone unit – and five categories of marked stress, including
(a) stress on a content word that is not final; (b) equal stress on two words
within a tone unit; (c) stress on a function word that is final within a tone
unit; (d) marked stress on a function word that is non-final in a tone unit;
and (e) marked stress on a function word that is non-final in a tone unit that
has two stresses (Fine et al. 1991).
The study found no significant differences between groups in the total

number of intonation boundaries and no differences in the frequency of
unmarked stress per 100 tone units. However, the HFA group used
Type A significantly less frequently and used Type C more frequently than
other groups. The AS group used Type B significantly less often than other
groups and also used Type C stress (non-significantly) more frequently
than the control group. No differences were found between groups in the
use of Types D and E. It was noted that Types A and B typically have
the function of contrast with given information, while stress on a function
word, as in Type C, can lead to misunderstanding or misinterpretation.
The findings of this study indicate a pattern of, on the one hand,
reduced use of contrastive stress (Types A and B) by the ASD groups
compared with the control group, and on the other, a more frequent use
of marked stress with unclear interpretation (Type C) by the group with
HFA. Fine and colleagues suggest their findings could reflect problems
with assessing contexts or in recognizing the communicative value of
stress patterns.
Case Studies
Several case studies have highlighted communicative patterns associated
with ASD. Contributions from Fine (in press), de Villiers and Szatmari
(2004) and de Villiers (2005, 2006a, 2006b, 2011) examine semantic patterns
in different functional domains, including those that are characteristic of
pedantic and echolalic speech. De Villiers (2006a, 2009, 2011) considers
aspects of organization, fluency, and information density that interact in
talk that is considered ‘pedantic’. De Villiers (2011) applied ‘phasal analysis’
(e.g. Gregory 2002) to illustrate coherence and interactional strengths that
were difficult to discern in the conversational and narrative discourse of a
boy with ASD. The study included a tri-functional analysis of linguistic
features and considered dimensions such as context, social distance, and
register, as well as the function of specific speech acts in global interaction.
It identified patterns of topic development and interaction that, while
atypical, appeared functionally appropriate from the perspective of the
speaker with ASD, and consistent with his purposes and interests. This
study is one of a small number of micro-analytic studies of conversational
discourse in ASD that highlight communicative strengths and consider
normative expectations and assessments of successful communication in
clinical contexts (see also Asp and de Villiers 2010).
Fine (in press) also examines speech functions and exchange structure
patterns in two clinical data sets to characterize the transitions between
interlocutors’ speech. In one of these, the interactions of eight children
with autism who had echolalia in their speech were studied. Echolalia was
more or less frequent depending on the information demands of the pre-
ceding utterance. Prior turns in which the demands of the interaction were
less clear were more likely to be followed by echolalic responses.
Rating Scale
In a larger study of forty-six children with ASD, de Villiers et al. (2007) used
SFL description and theory combined with neuropsychological evaluation
to develop a scale for assessing conversation skills in ASD. The study
identified atypical intonation, semantic drift, terseness, pedantic speech,
and perseveration as broad areas of conversational behaviour that were
characteristic of speakers across the ASD spectrum. The scale is specific to
ASD, has good inter-rater reliability, and reflects categories that are rela-
tively independent of IQ or language competence based on standard meas-
ures. The scale may be clinically useful in evaluating conversation skills
over time, and treatment response. It also highlights areas of language
behaviour where treatment programmes may focus attention.
23.3.2.2 (Specific) Language Impairment (SLI)

SLI refers to a developmental language disorder in children whose other
abilities are within normal range and in whom there is no other evident
reason for language impairment. Exclusionary criteria are used to establish
the absence of cause. The most frequently described positive features are
impaired morphosyntax and non-word and sentence repetition. The DSM-V
no longer includes a diagnostic category ‘specific language impairment’
because the diagnosis is not supported by epidemiological studies, the
positive signs are not specific, and the exclusionary criteria may negatively
affect research and children’s access to therapy and support (see Reilly et al.
2014 for the rationale). Nevertheless, the term SLI still has wide currency,
and we use it here because the reported studies worked with SLI as a
diagnosis.
Textual Function
Two controlled studies of children with SLI used SFL techniques and found
limited use of textual resources in the discourse of children with SLI.
Analyzing conversation samples of fifty-seven children with SLI and sixty-
seven TD children, Adams and Bishop (1989) included cohesion analysis as
part of a larger index of conversational skills. The cohesion index was lower
for the group with SLI, though a subgroup of fourteen children with
semantic-pragmatic disorder were not limited in their use of cohesion,
but violated turn-taking rules more than usual.
Thomson’s (2005) study of twenty-five children with SLI and twenty-five
TD children analyzed their use of marked and unmarked Theme in per-
sonal narratives and story retells. In SFL, thematic analysis refers to the
organization of information at the onset of clauses in terms of markedness.
For example, declarative clauses are unmarked when they begin with the
subject element in English but marked if they begin with objects. Similarly,
clause onsets may have interpersonal or textual elements such as unfortu-
nately and moreover whose default position is clause-initial, and so thematic-
ally signal unmarked aspects of interpersonal and textual meaning
(see Berry, this volume). The two groups had similar proportions of
marked theme in their discourse. However, the TD group had a signifi-
cantly higher proportion of multiple themes and used theme to develop
their discourse coherently as a linear progression, where new information
in one clause is represented as old thematic information in subsequent
clauses, particularly in the story retell condition. Most of the multiple
themes were a combination of textual and topical elements, and there
were no cases of multiple themes with three elements in the narrative
retells produced by the children in the SLI group. The results suggest that
the children with SLI were not as able as the TD group to exploit the more
varied and linear model of theme in the story retell. Thomson suggests
that restricted access to and/or restrictions in selection of lexicogramma-
tical resources may have been a factor in the reduced production of
multiple themes by children with SLI in the retell condition. She further
suggests that this might be a target for therapy, where scaffolding and
examples could enhance performance.
Speech Sound Disorder

Speech sound disorder is diagnosed in children who have persistent difficulty
producing speech sounds which affects their intelligibility or communication
(DSM-V). The difficulty should not be attributable to other co-existing condi-
tions. It may be associated with phonological, phonetic, or motor systems and
arises from many causes. Müller et al. (2008) present a case study of a 9.8-year-
old child who has difficulty with /r/ production. Their approach is to charac-
terize the phonetic features of the child’s variable attempts at /r/ productions
in different registers (functional varieties) and to build a mini system of these.
By this means, they are able to show that what appear as unrelated produc-
tion errors are in fact related via a consistent feature (velarization). This
common feature is thus identified as a target for speech therapy.
23.3.2.3 Attention Deficit Hyperactivity Disorder (ADHD)

ADHD is defined as ‘a persistent pattern of inattention and/or hyperactivity-
impulsivity that interferes with functioning or development’ (DSM-V)
before age twelve and may continue through adulthood. ADHD has not
generally been associated with problems in linguistic structure, but some
early studies suggested there may be problems using language in context-
ually appropriate ways and a recent meta-analysis provides evidence for
differences in language use in all areas (Korrell et al. 2017). Two SFL-
inspired studies have investigated the discourse of children with ADHD in
relation to register, and one has investigated the effects of treatment with
stimulant medication.
Grammatical Intricacy
Mathers (2005) investigates whether eleven children with ADHD differ
from eleven control children in grammatical intricacy (GI) across text
samples from different registers. GI assesses syntactic complexity by coding

and quantifying the percentage of clause complexes in a text. The control
group’s GI scores were higher than the ADHD group’s scores in written
texts, and the ADHD group showed no difference in GI between spoken and
written modes of the same text, in contrast to the control group. Mathers
suggests that the similarity in GI found between the spoken and written
texts may indicate problems with adjusting style to contexts, and highlights
the importance of register in evaluating language behaviour.
Language and Discourse

Mathers (2006) quantifies a bundle of features commonly used to charac-
terize performance (e.g. type token ratio, clause length, fluency features,
frequency of clause types) together with discourse features such as the
frequency of openings and closings, and moves rated as tangential, unre-
lated, and so on in narrative, procedural, and recount texts and in both
spoken and written modes. The work was informed by SFL in so far as it was
referenced as providing theoretical licence to look at such an amalgam of
features for children whose diagnosis does not include language impair-
ment. Eleven children with ADHD had more tangential and unrelated
meanings in the spoken tasks, and fewer explicit openings and closings in
written tasks, where they also made more spelling and punctuation errors
than (eleven) TD children. However, there were no differences in basic
linguistic features or fluency, except for abandoned clauses. Like Mathers
(2005), the study highlights the limits of conventional metrics that assess
only one text type. It is also useful as an indicator of areas that might be
remedially addressed in teaching children with ADHD.
Treatment Effects
Tannock et al. (1995) use detailed cohesion and speech function analyses to
describe and measure dose-related effects of methylphenidate (MPH) in
ADHD. MPH is a stimulant medication that Tannock et al. note can be
therapeutic, but can also produce a cognitive effect of over-focusing. Con-
versational interactions between two boys and an adult interlocutor were
analyzed to investigate the effect of three different doses of MPH in a
randomized placebo-controlled double-blind single case study. Frequencies
of the main categories of speech function, namely, initiations, responses,
and continuations, were calculated relative to the total number of speech
functions produced by the child, and the proportion of adult questions that
were followed by a response speech function was calculated. Cohesion was
also analyzed.
Overall, MPH decreased talkativeness as measured by total number of
utterances and number of turns. This was reflected in a reduction of the
proportion of speech functions used to initiate interactions, and an increase
in continuations, which were often used to extend a topic. The children also
responded to more of the adult’s questions in the medicated conditions.
The cohesion analysis showed MPH increased the proportion of references

in ways interpreted in the study as decreasing social reciprocity: the chil-
dren used proportionately more unclear endophoric references and fewer
second-person references. They also used more conjunctions such as because
in all three medicated conditions, which indicated increased language
complexity. MPH also increased lexical repetition in one child’s conversa-
tion, which post-hoc was found to be consistent with continuation of topic.
This latter finding suggests a cognitive focusing effect that the authors say
may help with topic continuity, but in the extreme could produce topic
perseveration. In short, the results showed dose-related effects on discourse
related to focusing and over-focusing that were consistent with the cogni-
tive effects expected from MPH. This study was undertaken to investigate
the utility of detailed linguistic analysis in evaluating treatment responses
and is an early exemplar of such work.
23.3.3 Neurocognitive Disorders

In current DSM-V terminology, all non-psychotic disorders that centrally
involve decline from a previously acquired state in one or more cognitive
domains are defined as neurocognitive disorders. The domains include
complex attention, executive function, learning and memory, language,
perceptual–motor function, and social cognition. The designation ‘neuro-
cognitive’ reflects the integrity of the grouping as involving (usually meas-
urable) changes or differences in neural substrates that are not present at
birth or in early development (see Sachdev et al. 2014 for the rationale).
Thus, for instance, the neurodegenerative disorders are grouped together
with traumatic brain injury, stroke, HIV, substance and/or medication use,
and other specified and unspecified causes of neurocognitive dysfunction.
Other disorders (of mood, anxiety, etc.) may involve cognitive domain(s) but
are treated separately on the grounds that neurocognitive dysfunction is
not their primary presenting symptom. Of note for our present purpose,
aphasias and communication disorders arising from any of these causes are
treated as signs and symptoms of neurocognitive disorders. The clustering
of disorders in this section reflects this common thread. Our focus is on
areas where SFL linguists have made contributions, but note that impair-
ment in any of the domains identified under neurocognitive disorders is
likely to have some effect on discourse and communicative abilities.
23.3.3.1 Alzheimer’s Disease (AD)

AD is the most common neurodegenerative disease associated with aging.
Prominent signs are decline in episodic memory (so it may be hard to
answer a question such as What did you eat for dinner yesterday?) and in
executive abilities, limiting capacity for the independent activities of daily
living. Semantic memory is also impaired, and early signs include limited
semantic fluency (measured by timed word generation tasks for specific
categories) and difficulties with picture naming and semantic categoriza-

tion tasks (see Taler and Phillips 2008 for review). Discourse abilities are
also progressively compromised by memory, semantic, and executive def-
icits, with early word-finding difficulties and associated circumlocutions,
‘empty speech’, and hesitations appearing as common signs (see Kempler
1995 for review). In contrast, relatively automatized aspects of syntactic,
phonological, and interactional abilities may be preserved during mild and
moderate phases (Kempler 1995), though complex syntactic constructions
are compromised (Kemper et al. 2001). There is a large literature on lan-
guage and discourse in AD. SFL contributions focus on aspects of discourse
for supporting differential diagnoses, monitoring change, and exploring the
communicative abilities of speakers with AD.
Cohesion
Unsurprisingly, early SFL-inspired work on discourse in AD investigated
cohesion with differential diagnosis as a primary goal. Ripich et al. (2000),
following up on earlier work (Ripich et al. 1983; Ripich and Terrell 1988),
conducted a study of cohesion in the conversation of sixty people with
mild/moderate AD contrasted with forty-seven demographically matched
control participants, and followed a subset of twenty-three participants
with AD for eighteen months. They were interested in differential diagnos-
tic potential compared with healthy elderly, and in the use of cohesion as a
potential monitoring aid. Participants with AD produced more referent
errors than control participants but otherwise used cohesive devices simi-
larly. In the longitudinal follow-up, use of conjunctions and ellipsis became
significantly less frequent, and use of all cohesive devices declined as
participants’ conversational contributions decreased in length and fre-
quency. They conclude that although referencing errors did distinguish
mild/moderate AD participants’ conversation from that of control partici-
pants, the monitoring potential of cohesion analysis in conversational
samples alone is limited by decreasing speech output associated with dis-
ease progression.
The finding that people with AD use cohesive devices similarly to control
participants, but with more referential errors, has been replicated, for
instance, in Dijkstra et al. (2004). However, Chapman et al. (1995) find no
difference in referential errors between people with mild AD and old (>80)
healthy elderly people, though both groups differ from younger (48–78)
controls. Both studies examine cohesion as one of several factors contrib-
uting to discourse coherence, which is consistently found to be impaired
relative to control participants. They evaluate discourse coherence ratings
of local and global continuities and disruptions of topic, though they
operationalize topic analysis differently.
Lock and Armstrong (1997) investigated the potential for differential
diagnosis of people with AD and with anomic aphasia using cohesion
analysis, since word-finding difficulties are common to both. The anomic
aphasic speakers produced twice as many ‘disrupted cohesion ties’ (e.g. use
of pronouns without referents) as people with AD, but they had fewer
misrepresentations of content. Lock and Armstrong suggest that, in the
presence of many other linguistic similarities between the discourses of the
two groups, cohesion analysis could be diagnostically useful.
Transitivity
The breakdown in experiential semantics in AD has been explored in only
one SFL-based case study (although see Asp and de Villiers 2010 for ‘SFL-
informed’ analyses). Mortensen (1992) describes the use of transitivity
options by a woman, L.M., with moderate AD in terms of process types,
participants, and circumstances and the amount of information in phrases.
The approach replicates findings from conventional studies: L.M.’s use of
processes is varied and accompanied by appropriate participant roles, but
her phrase structures are simple and show limited lexical variation which,
together with repetition, incomplete clauses, and pronouns without ante-
cedents limits the informativeness of her discourse to ‘simple core’
information.
Interpersonal Interactions
In a series of case studies, Müller and colleagues have described interper-
sonal and conversational skills of speakers with AD resident in long-term
care homes and have shown that despite cognitive and linguistic limitations,
residents mostly use interpersonal resources effectively in conversations.
For example, analyses of speech functions and moves show a ninety-seven-
year-old man with AD leading a conversation with a younger man, complete
with advice and opinions as befit his role and face needs (Müller and Wilson
2008). Similarly, simple counts of turns, types and numbers of moves and
speech functions, and address terms and naming patterns show extended
conversations (up to forty-five minutes in length) amongst residents in
which they successfully enact social roles, flirt, show affection, and so on
(Müller and Mok 2012; Mok and Müller 2013). While the finding of pre-
served interactional abilities in people with AD is not novel, Müller and
colleagues’ point is that encouraging (and even ‘staging’) interactions
amongst residents in long-term care may improve their quality of life by
reducing their social isolation.
Modalization Battery
Though not presented as systemic functional per se, Asp and de Villiers
(2010) offer an interpersonally grounded ‘modalization battery’ for analyz-
ing and coding epistemic stance in AD discourse. The battery is intended as
AD-specific in foregrounding the resources available for speakers to circum-
navigate the epistemic challenges created by their episodic and semantic
memory impairments. It differs from SFL treatments of appraisal (Martin
and White 2005) in so far as the latter see modalization as almost
exclusively oriented to achieving interpersonal goals, whereas for speakers

with AD, intact interpersonal resources may be recruited as a means of
eliciting support from interlocutors to overcome ideational deficits created
by memory and executive impairments (see also Asp et al. 2006).
23.3.3.2 Acquired Brain Injury (ABI)

Acquired brain injuries (ABI) are comparatively well represented in clinical
linguistic research, in part because of the primary role that speech patholo-
gists have in both the evaluation and care of patients following a stroke or
traumatic brain injury (TBI). Speech pathologists routinely perform assess-
ment and immediate treatment following a brain injury, and are also
involved in longer-term rehabilitation.
Much of the research applying SFL to ABIs has been aimed at isolating
areas of speech and interaction that can be targeted for therapy. There have
also been longitudinal studies designed to evaluate progress or recovery
(e.g. Coelho et al. 1991a), and Togher and colleagues (2004, 2009, 2013) use
SFL analyses to evaluate a treatment for TBI. Studies have focused primarily
on conversation or spoken narratives of people with mild or moderate
aphasia or TBI, and have been explicitly concerned with characterizing
people’s language in more natural tasks and contexts than standardized
testing accommodates.
Traumatic Brain Injury (TBI)

TBI presents assessment challenges in that aphasia batteries often under-
represent functional communication problems (Coelho et al. 1991b; Jorgensen
and Togher 2009). In TBI, resources for experiential meaning are relatively
preserved (Armstrong 2005a), but difficulties have been reported in cohesion
and interpersonal interactions.
Two papers focusing on TBI are notable for explicitly laying out method-
ologies for analyzing discourse from an SFL perspective. Coelho et al.
(1991b) outlines techniques for analyzing intersentential cohesion using
Halliday and Hasan’s (1976) categories, and ‘cohesive adequacy’ using a
procedure outlined in Liles (1985) that divides cohesive ties into three types:
complete (information referred to with the cohesive tie is found easily and
there is no ambiguity); incomplete (information referred to with a tie is not
in the text); and error (a tie refers ambiguously or does not match infor-
mation in the text). Cohesive adequacy is then measured by calculating the
percentage of complete ties relative to all ties.
Emphasizing the potential for use by speech pathologists, Togher (2001)
illustrates the use of SFL at different levels to examine interpersonal mean-
ings in the discourse of people with TBI. She analyzes politeness markers
including mood and modality (Halliday 1994); exchange structure (Berry
1981), including move analysis; and generic structure potential (GSP)
(Ventola 1979; Hasan 1984, 1985), which treats spoken genres as a series
of goal-oriented steps.
Cohesion
Cohesion in TBI has been studied by investigating the number or proportion
of cohesive ties, cohesive adequacy, and cohesive harmony, often in com-
bination with measures of discourse productivity and story grammar
(Coelho et al. 1995). Yet there is not a clear-cut consensus about cohesion
and coherence problems in TBI (Davis and Coelho 2004). Some studies
have found people with closed head injuries (CHI) produce lower percent-
ages of cohesive ties in discourse than control groups (Mentis and Prutting
1987; Coelho et al. 1991a; Hartley and Jensen 1991; Jorgensen and Togher
2009), but others have not found such differences (in narrative tasks)
(Liles et al. 1989; Coelho 2002).
Cohesive adequacy has also been reported to be lower in texts produced
by individuals with CHIs than in those of control participants, with CHI
participants using incomplete ties that control participants did not use
(e.g. Mentis and Prutting 1987), or using inappropriate exophoric references
significantly more frequently than control participants in tasks requiring
frequent use of exophoric references (McDonald 1993). In a longitudinal
case study, Coelho et al. (1991a) looked at story grammar and cohesion in a
picture-based story generation task in two young adults with a CHI. The
approach identified cohesion problems in both participants, whose scores
were below those of control participants, and demonstrated different recov-
ery patterns, with one participant improving over time. However, in a large
study, Coelho (2002) compared fifty-five adults with CHI with forty-seven
non-brain damaged (NBD) adults on story generation and story retelling and
found the groups were not significantly different in cohesive adequacy.
Additionally, Coelho et al. (1995) and Coelho (2002) studied executive func-
tion (EF) in relation to cohesive adequacy in the CHI group just described
and in thirty-two adults with TBI. Both studies used the Wisconsin Card
Sorting Task to measure EF and found no significant correlations between
the EF measure and cohesive adequacy.
McDonald (1993) also looked at ‘cohesive harmony’ using Armstrong’s
(1987) cohesive chain index, a measure of cohesion based on lexical rela-
tions of co-referentiality, co-classification, or co-extension (Halliday and
Hasan 1976; Hasan 1985), but found head injured participants were similar
to controls in this analysis. Interestingly, this study also used rating scales
based on Grice’s (1978) maxims and found that raters perceived the head
injured speakers as sounding disorganized and confused, but these judge-
ments did not correlate with cohesive harmony scores.
The lack of a clear consensus about cohesion findings in TBI may be
attributed to the small sample sizes (some with fewer than four clinical
participants), and even in larger studies, the heterogeneity of TBI, together
with methodological differences, makes studies hard to compare (Coelho
2002). However, one fairly consistent finding has been that the use of ties
varies according to task type (Mentis and Prutting 1987; Liles et al. 1989;
Hartley and Jensen 1991; Coelho et al. 1991b; Davis and Coelho 2004;
Jorgensen and Togher 2009). In particular, researchers have noted that

story generation may be more challenging than story retelling (e.g. Coelho
2002). Coelho (2002) also points out that socioeconomic status is important
for cohesion performance since unskilled workers from both control and
CHI groups perform significantly worse on cohesion than skilled workers or
professionals. Such findings have important implications for sampling
procedures in TBI research.
Interpersonal Analyses
In a substantial body of research, Togher and colleagues have used GSP and ES
analyses to study the interactions of people with TBI in a variety of contexts
and situations. Studies analyze the telephone interactions of individuals with
TBI talking with different interlocutors (Togher et al. 1997a; Togher and Hand
1998) and in different role relationships (Togher et al. 1997b; Togher and
Hand 1998; Guo and Togher 2008), and one compared monologic and jointly
produced discourse in speakers with TBI (Jorgenson and Togher 2009). All
these studies noted some difficulties with aspects of interaction and global
structure in TBI and emphasized the role of communicative partners in
shaping opportunities for communication. For instance, in telephone inter-
actions, service providers adjusted their discourse behaviour when conversing
with people with TBI, by asking for information they already had, by fre-
quently checking information (Togher et al. 1997b), or by giving and asking
for less information than they did with control participants (Guo and Togher
2008:84), all of which limited successful communication.
Subsequently, Togher and colleagues used these results in training and
intervention studies focused on improving the communication of commu-
nicative partners of individuals with TBI, rather than those of TBI partici-
pants, during service encounters (Togher et al. 2004) and everyday
conversations (Togher et al. 2009, 2013). In each of these controlled studies,
the interactional and conversational performance of TBI participants
improved. In the intervention study (Togher et al. 2013), improvements
were maintained six months after training.
Experiential Analysis
Rigaudeau-McKenna (2005) describes language dysfunction in the produc-
tion of clauses based on language sampling of adolescents with TBI and
uninjured children. The study categorized clause simplex and complex
failures in terms of where hesitations, as well as repaired or incomplete
clauses, occur, and what kinds of repairs (if any) are attempted in order to
establish a template for the analysis of ideational dysfunction. This
approach treats dysfluencies as signs of ideational difficulty, rather than
as evidence of executive (e.g. planning) dysfunction.
Aphasia
Aphasia is a generic term for linguistic signs and symptoms of neurological
dysfunction arising commonly from stroke but also from many other
causes that may differentially affect discourse production and/or compre-

hension. One or more of semantics (including but not limited to lexis),
morphology, syntax, phonology, or phonetics may be affected. Difficulties
may affect speech/discourse production, comprehension, or both. Reading
and/or writing may also be affected. Much of the clinical work in aphasia
focuses on ‘fractionating’ the linguistic correlates of aphasic syndromes,
and on developing treatments for specific deficits. SFL work on aphasia has,
in contrast, primarily focused on discourse.
Three reviews of SFL work in aphasia research (Armstrong 2000, 2005a;
Armstrong and Ferguson 2010), as well as illustrative papers on
various methods with attention to the different strata and metafunctions
(Armstrong 1991; Armstrong and Mortensen 2006; Armstrong et al. 2011),
provide a kind of model for SFL research in aphasia. These papers fore-
ground the utility of SFL in addressing the needs for connecting macrolin-
guistics and discourse-level analyses to microlinguistic lexicogrammatical
analyses (Armstrong 2000), assessing language from social and functional
perspectives to facilitate social participation, and assessing abilities rather
than errors (Armstrong 2005a). Armstrong and Ferguson (2010) also pro-
pose a taxonomy of social meanings based on Halliday’s (1994) framework
and provide detailed illustrations of analyses in all metafunctions of apha-
sic discourse. Collectively, the aphasia studies provide a more nuanced
picture of functional language use than is captured with standardized
testing, particularly in the area of interpersonal meaning, highlight new
areas of focus for treatment, and foreground conversation and extended
interactions over monologues to address concerns for quality of life and the
need for assessing function in linguistically complex situations.
Ideational Function
Ideationally, aphasia is characterized by lexicogrammatical disruption such
as word-finding difficulties, reduced syntactic complexity, or other
(morpho-)syntactic problems and restrictions in lexical diversity, although
the extent of lexical and grammatical dysfunction varies with aphasia type
and severity (Armstrong 2000, 2005a). Two SFL studies have explored how
diminished ideational resources in aphasia, reflected in vague or generic
lexis and frequent fillers, may limit participation in complex genres such as
arguments (Armstrong et al. 2013) and written correspondence (Mortensen
2005). In the latter study, aphasic writers were found to use fewer optional
generic elements such as ‘orientation’ than either TBI or control partici-
pants. Mortensen speculates that this reflects lexicogrammatical ideational
limitations rather than limited genre knowledge.
Armstrong (2001) also investigates variation in the use of process types
in elicited recounts of four speakers with fluent aphasia and four control
speakers. Verbs were categorized by superordinate process type as mater-
ial, relational, mental, verbal, or behavioural (Halliday 1994). Two of the
people with aphasia (PWA) had fewer mental and relational processes
than control speakers, resulting in a recount with little or no personal

evaluation or opinion.
Cohesion
Investigations of cohesion in aphasia have been fairly consistent in
reporting poor cohesion (Piehler and Holland 1984; Glosser and Deser
1990; Coelho et al. 1994; Lock and Armstrong 1997; Armstrong 2000;
Olness and Ulatowska 2011; Andreetta et al. 2012). Longitudinal investi-
gations of cohesion and cohesive chains in aphasia discourse have noted
recovery over time (Coelho et al. 1994; Piehler and Holland 1984;
Armstrong 1997).
Additionally, two studies in the 1980s explored cohesive harmony. Bot-
tenberg et al. (1985) compared oral stories of ten PWA to those of control
participants. The study found more variability in the texts of the PWA, as
well as significantly fewer chain interactions. Armstrong (1987) used cohe-
sive harmony to analyze texts produced by three fluent PWA. Lexical chains
were identified and extracted from each text, and the number of lexical
items in the chains was compared to the number of lexical items overall to
create a score or ‘index’. In addition, six listeners rated the texts for coher-
ence. Each of the participants had low scores on the cohesive harmony
index, with almost all the texts falling below 50 per cent, the postulated
threshold for textual coherence. This was significantly positively correlated
with coherence ratings (see also Andreetta et al. 2012). However, Linnik
et al. (2016) observe that five out of six studies that address cohesion and
coherence in aphasia do not find such correlations.
Interpersonal Function
Interpersonal aspects of communication are often fairly intact in aphasia.
Ferguson (1992) studied speech functions in conversational data of five
PWA who were selected to represent a fairly homogenous group and found
they retained access to all options for expressing speech functions and
could modulate their choices by selecting questions instead of commands
in the interest of interpersonal politeness. Mortensen (2005) also found
PWA retained interpersonal resources in her study of written letters.
However, against this positive picture, Armstrong et al. (2011) and
Armstrong and Ulatowska (2007) note that some resources for evaluative
meaning are preserved, yet simplified, in aphasia. Armstrong and
Ulatowska (2007) look at stories about an emotional topic produced by
three PWA and find they use fewer evaluative and emotion words and show
less varied expressions for emotion than control participants. Armstrong
(2005b) reports that some PWA have difficulty with mental and relational
verbs. Five mildly severe to moderately severe PWA use fewer evaluative
relational verbs and fewer mental processes than control participants in an
elicited discourse task. These process types are involved in explicit evalu-
ation and expressions of opinion, and Armstrong suggests the findings,
though preliminary, may indicate difficulties with expressing attitudes and

opinions linguistically, impacting the ability to express identity.
PWA may struggle to participate in more complex communicative
genres and depend on conversational partners for scaffolding of evaluation
(Armstrong et al. 2012, 2013). In Armstrong et al. (2012), an appraisal
analysis (Martin and White 2005) of conversational data found that
resources for evaluation in the PWA were relatively retained. PWA could
contribute opinions during a group conversation and produced a range of
evaluative subtypes. However, the amount of explicit evaluation used
depended on severity of aphasia, and the PWA relied on their conversa-
tional partners to support their evaluations. In a case study of four conver-
sations involving a person with mild aphasia and her non-brain-damaged
partner, Armstrong et al. (2013) found the aphasic speaker used less explicit
evaluation than her partner and her evaluations were more often co-
constructed. She also appeared to struggle when the flow of her discourse
was interrupted with requests for more specificity or other repairs.
23.3.3.3 Childhood-Onset Fluency Disorder (Stuttering)

Stuttering is a relatively common neurodevelopmental disorder that may
persist into adulthood. It is characterized by ‘disturbances in the normal
fluency and time patterning of speech’ and the disturbance is associated
with ‘anxiety about speaking’ (DSM-V). It is often linked to social anxiety
(Iverach and Rapee 2014). We discuss it here, rather than under neuro-
developmental disorders, because SFL-based work on stuttering focuses on
adults, and four of five papers evaluate treatment outcomes, so it nicely
bookends this review of applications of SFL in clinical linguistics, which
began with diagnosis. Two papers from Spencer and colleagues develop
the approach. Spencer et al. (2005) presents two case studies in a pre-/
post-treatment design for adults whose stuttering began in childhood.
Ten-minute telephone conversations recorded before and after treatment
are transcribed and analyzed for modality, theme, and transitivity selec-
tions. Both participants’ stuttering improved post treatment. Post-
treatment, both participants modalized more, and one speaker showed
increased use of continuative themes (words such as well used at the begin-
ning of a sentence). Transitivity was unchanged. The authors suggest that
increase in modality may reflect an increase in speakers’ willingness to
engage in interactions because they are less likely to stutter. Spencer et al.
(2009) contrasts ten adults with stuttering (AWS) with ten adults without
stuttering (ANS) in a monologic task and quantitatively assesses volubility,
grammatical complexity assessed by GI and marked and multiple themes,
and modality (as a marker of social engagement). AWS were less voluble
and used fewer complex constructions and less modality than ANS. Theme
use was similar between groups. This study shows differences in the way
AWS use language and suggests targets (volubility, grammatical complex-
ity, modality) for further evaluation of treatment efficacy.
Lee et al. (2015, 2016a, 2016b) adopt a design that built on Spencer and
colleagues’ work, assessing similar features but adding appraisal and a
longitudinal follow-up. The baseline study (Lee et al. 2015) included
twenty AWS/twenty ANS, and assessed volubility, grammatical complex-
ity (including GI but not Theme), modality, and appraisal. It replicated
Spencer et al.’s (2009) results, adding that AWS also use fewer appraisal
resources for expressing opinions. Post-treatment (Lee et al. 2016a), par-
ticipants stuttered less, and their volubility, grammatical complexity, and
modality use increased. There was also a trend towards increased expres-
sion of affect. At twelve months, Lee et al. (2016b) assessed ten AWS and
six ANS (those available for follow-up). Seven of the ten participants
maintained the therapy effect for reduced stuttering at twelve months,
though the group analysis did not show this. Participants also maintained
increases in other measures, most robustly in modality and appraisal, on
which their scores were not different from control participants. The
interpretation of the results, as in Spencer’s work, is that reductions in
stuttering frequency allow participants to engage more actively in conver-
sations and that these beneficial effects are maintained for most partici-
pants over twelve months.
23.4 Discussion
Clearly, SFL work in clinical linguistics addresses some of the gaps

described by Crystal and Perkins in the introduction in that it has been
largely focused on discourse in clinical groups, it primarily addresses abil-
ities other than phonology and phonetics, it works on descriptions of the
discourse abilities of speakers, and at its best, it seeks explanations
grounded in selection patterns of language and discourse features speakers
use. Linguists doing SFL-based clinical research have also been actively
investigating how speakers’ abilities are modulated and challenged by
different discourse genres and registers and different clinical tasks, recog-
nizing that ‘normal’ language profiles based on standard clinical assess-
ments do not necessarily translate into ‘normal’ discourse abilities.
Viewed from a metafunctional perspective, the areas most intensively
covered (in descending order) are cohesion and, to a lesser extent, theme
at the clause rank in the textual function, and generic structure potential,
move, speech functions, modality, and other appraisal systems in the
interpersonal function. Coherence is commonly discussed in conjunction
with textual and interpersonal patterns. Work on transitivity and even
lexis in experiential function is rare outside studies that include coding
for primary process types as a component of discourse profiles. Work on
syntactic complexity is limited to a few studies that examine grammatical
intricacy and Rigaudeau-McKenna’s (2005) proposals for using fluency
features to interrogate the integrity of clause complexes. Similarly, work
explicitly addressing prosody is uncommon, and so far as we could dis-

cover, work that addresses other aspects of phonology or phonetics is
almost non-existent in SFL-based clinical linguistics.
Viewed from a translational perspective, the most influential work deriv-
ing from SFL approaches is that on cohesion and coherence which
developed out of Rochester and Martin’s (1979) analysis of schizophrenic
discourse. This work not only had a role in changing the diagnostic criteria
for schizophrenia, but because of the attention it drew, cohesion analysis is
now one of the best-known approaches for researchers looking to under-
stand or characterize the discourse of any previously undescribed clinical
group, and it is recommended for clinical assessment of individuals’ dis-
course abilities. Additionally, the variable results on cohesion for some
groups has helped to highlight the need to analyze different genres and a
range of task types. This is important both for research and for clinical
assessments, planning therapy, and identifying areas where training for
communication partners is relevant.
The development of rating scales, as well as the identification of indi-
vidual signs of treatment response and development of batteries to inform
neurocognitive assessments and monitor treatments and changes over
time are also evidence of the value of relatively rich discourse analyses
that inform their creation. We note here that ‘discourse’ is often treated
as a data source from which discourse and linguistic patterns are
extracted by ‘discourse analysis’ with goals of comparing quantified
results between groups, over time, or in response to interventions.
The utility of such work is well recognized. However, approaches which
aim to elucidate meaning-making and messages in individual texts (or
series) in order to come to understand ‘the experience of X’ are also
clinically important. In addition to Müller’s work on conversations of
people with AD and our own attempts to model comprehensive analyses
in AD and ASD (e.g. Asp and de Villiers 2010), the works of Butt et al.
(2010) on dissociation in psychiatric interviews, Körner et al. (2011) on
depression in gay men, and Tebble (2012) on appraisal and depression in
acute care all illustrate how such analyses can be used to help therapists
and clients comprehend and manage their experiences. In a similar vein,
the work of Fine, best illustrated in his Language in Psychiatry: A Handbook of
Clinical Practice (2006), treats SFL-based linguistic analysis of texts and
discourse as an evidence base for teaching clinicians how to listen to and
recognize the linguistic and discourse signs of psychiatric (dys-)function.
Many of the groups addressed do not present with language or communi-
cation disorders, so such work might fall outside the scope of clinical
linguistics narrowly defined. Indeed, this was our rationale for not
reviewing such work in detail here. However, under a broader conceptual-
ization, such work offers a template for investigation of psychiatric dis-
orders where, apart from drugs, interventions are focused on talking
and listening in the clinic, and knowing whether someone is ‘getting
worse or better’ is a clinical judgement. SFL-grounded clinical linguistics

surely has a role to play here.
Assessed from a disorders perspective, apart from the early work on
schizophrenia, SFL-based research has been most concentrated on aphasia
and traumatic brain injuries. This likely reflects the case loads of research-
ers leading this work, who are predominantly speech therapists. Other
areas investigated with SFL include ASDs, ADHD, (S)LI, (adult) stuttering,
AD, and depression. While this suggests fairly broad coverage, for many
areas publication numbers are small and produced by individuals or small
groups. There are also gaps. For example, among neurocognitive disorders,
there appears to be no SFL-based work as yet on the primary progressive
aphasias, mild cognitive impairment, or other neurodegenerative disorders.
Moreover, except for Müller et al. (2008), there is no SFL-based research
on phonological or motor developmental disorders, nor any work that
addresses comprehension in any disorder.
23.5 Conclusion and Future Directions
The contributions and strengths of SFL-informed clinical linguistics

sketched above suggest that important, clinically useful work has been done,
and that there are active research programmes in some of the key areas.
There is also considerable potential for SFL to contribute more to clinical
linguistics. However, for SFL clinical linguistics work to be translatable,
some desiderata should be foregrounded with a view to future directions.
One of these has to do with power. Most of the research studies described
report statistical analysis based on group means for one or more discourse
features of interest. However, in many studies, the actual numbers of
participants are small. Moreover, variability within the studied groups in
terms of etiology, severity of impairments, demographics, and so on may
influence performance on tasks. This combination can mean that research
reports are, at best, treated as pilot work for larger-scale studies. The
reasons for small studies are obvious and not unique to SFL-based work.
Detailed discourse analyses with manual coding for features take time,
money, trained coders, and willing participants. Consequently, larger stud-
ies may be difficult or may take longer than the usual three-year grant
cycles. One alternative is to automate coding, though this often involves
sacrificing theoretically informed measures for more coarse-grained ones.
Another is to develop larger collaborative teams. A third is to embrace the
variability and present case series instead of, or in parallel with, (small)
group comparisons. If supported by careful clinical and demographic char-
acterizations, these can contribute to understanding variability and pattern
and better inform the design of ‘next steps’.
This points to a second desideratum – reporting clinical characteri-
zations. In some published papers, participant characterizations are
underspecified. Moreover, there are few attempts to associate specific

patterns of impairment or ability with, for instance, the results of neuro-
psychological or imaging evaluations. Some lack of specificity is under-
standable where etiology or lesion extent or location is not known, or
where neuropsychological evaluations are not available. However, the
absence of such information makes it difficult to understand or even
reasonably hypothesize about how linguistic and discourse behaviours
might relate to neurocognitive substrates or domains such as memory
and executive functions and their disorders, and so limits translation to
broader clinical and research communities. Addressing such gaps, for
example, through collaboration in larger multidisciplinary teams who
collectively can bring the necessary expertise to complex problems, can
lead to work with more potential for translation to research and clinical
contexts.
Hence, the third desideratum is for SFL clinical linguists to actively
engage with the neurosciences. Linguistic analyses of language and dis-
course data are complementary to computational, imaging, neuropsycho-
logical, and neurobiological approaches in so far as linguistic analyses can
inform decisions about what language and discourse signs to look for and
how to interpret patterns in these. Computational studies attempting to
identify reliable linguistic and discourse signs of prodromal AD, and neuro-
imaging studies investigating the neural systems that support cohesion and
coherence in text processing (e.g. Ferstle and von Cramon 2001; Kuperberg
et al. 2006; Kurczek and Duff 2012) or how the neural systems for motor
processing in Parkinson’s disease link to action verb semantics and use (see
Garcia and Ibáñez 2016 for review) are examples of where SFL has or could
make useful contributions. More broadly in the neurosciences, localization-
ist models drawn from nineteenth-century aphasia research and twentieth-
century concepts of modularity are being superseded by distributed large-
scale functional networks that facilitate top-down and bottom-up process-
ing (e.g. Mesulam 2008), and probabilistic modelling of top-down effects
and of bottom-up input is coming into its own as explanatory of neurocog-
nitive function and dysfunction (e.g. Brown and Kuperberg 2015). SFL’s
functionally oriented, selection-based perspectives on language use, its
theoretical commitment to probabilistic weighting for selection, as well as
the depth of expertise in textual analysis and contextually related interpret-
ation, make it particularly relevant for adaptation to current research on
language disorders in the neurosciences.
In conclusion, SFL has made significant contributions to clinical linguis-
tics. It could offer even more. This is obvious where carefully described
instances and patterns of use may be instrumental not only in assessing
individuals and designing treatments and supports for them (what we
might call ‘Crystal’s vision’ for clinical linguistics addressed primarily to
speech therapists), but also in areas where such descriptions and analyses
may inform other research paradigms and clinical practices. If the goal of
clinical linguistics is to work with others to help people with disorders

that affect language and communication directly or where language and
discourse patterns may be interpreted as indices of state and change, then
we should embrace technologies and expertises that can further the real-
ization of such goals.
References
Aboraya, A., C. France, J. Young, K. Curci, and J. LePage. 2005. The Validity
of Psychiatric Diagnosis Revisited: The Clinician’s Guide to Improve the
Validity of Psychiatric Diagnosis. Psychiatry (Edgmont) 2(9): 48–55.
Adams, C. and D. V. Bishop. 1989. Conversational Characteristics of
Children with Semantic-Pragmatic Disorder. I: Exchange Structure, Turn-
taking, Repairs and Cohesion. British Journal of Disorders of Communication
24(3): 211–39.
American Psychiatric Association. 1968. Diagnostic and Statistical Manual of
Mental Disorders. 2nd ed. Washington, DC: American Psychiatric
Association.
Mental Disorders. 3rd ed. Washington, DC: American Psychiatric
Association.
Mental Disorders. 4th ed. Washington, DC: American Psychiatric
Association.
Mental Disorders. 5th ed. Washington, DC: American Psychiatric
Association.
Andreetta, S., A. Cantagallo, and A. Marini. 2012. Narrative Discourse in
Anomic Aphasia. Neuropsychologia 50: 1787–93.
Armstrong, E. 1987. Cohesive Harmony and Its Significance in Listener
Perception of Coherence. In R. H. Brookshire, ed., Clinical Aphasiology.
Minneapolis: BRK Publishers. 210–15.
Armstrong, E. 1991. The Potential of Cohesion Analysis in the Analysis and
Treatment of Aphasic Discourse. Clinical Linguistics & Phonetics 5(1): 39–51.
Armstrong, E. 1997. A Grammatical Analysis of Aphasic Discourse: Changes in
Meaning-making over Time. PhD Thesis, Macquarie University.
Armstrong, E. 2000. Aphasic Discourse Analysis: The Story so Far. Aphasiol-
ogy 14(9): 875–92.
Armstrong, E. 2001. Connecting Lexical Patterns of Verb Usage with
Discourse Meanings in Aphasia. Aphasiology 15(10–11): 1029–45.
Armstrong, E. 2005a. Language Disorder: A Functional Linguistic Perspec-
tive. Clinical Linguistics & Phonetics 19(3): 137–53.
Armstrong, E. 2005b. Expressing Opinions and Feelings in Aphasia: Linguis-
tic Options. Aphasiology 19(3–5): 285–95.
Armstrong, E. and A. Ferguson. 2010. Language, Meaning, Context, and

Functional Communication. Aphasiology 24(4): 480–96.
Armstrong, E. and L. Mortensen. 2006. Everyday Talk: Its Role in Assess-
ment and Treatment for Individuals with Aphasia. Neurogenic Communi-
cation Disorders (Special Issue: Brain Impairment) 7(3): 175–89.
Armstrong, E. and H. K. Ulatowska. 2007. Making Stories: Evaluative
Language and the Aphasia Experience. Aphasiology 21(6–8): 763–74.
Armstrong, E., N. Ciccone, E. Godecke, and B. Kok. 2011. Monologues and
Dialogues in Aphasia: Some Initial Comparisons. Aphasiology 25(11):
1347–71.
Armstrong, E., L. Mortensen, N. Ciccone, and E. Godecke. 2012. Expressing
Opinions and Feelings in a Conversational Setting. Seminars in Speech and
Language 33(1): 16–26.
Armstrong, E., S. Fox, and R. Wilkinson. 2013. Mild Aphasia: Is This the
Place for an Argument? American Journal of Speech-Language Pathology 22(2):
268–78.
Asp, E. and J. de Villiers. 2010. When Language Breaks Down: Analysing
Discourse in Clinical Contexts. London: Cambridge University Press.
Asp, E., X. Song, and K. Rockwood. 2006a. Self-referential Tags in the
Discourse of People with Alzheimer’s Disease. Brain and Language 97(1):
41–52.
Baltaxe, C. A. M. and N. D’Angiola. 1992. Cohesion in the Discourse Inter-
action of Autistic, Specifically Language-impaired and Normal Children.
Journal of Autism and Developmental Disorders 22: 1–21.
Bartlett, S. C., E. Armstrong, and J. Roberts. 2005. Linguistic Resources of
Individuals with Asperger Syndrome. Clinical Linguistics & Phonetics 19(3):
203–13.
Berry, M. 1981. Systemic Linguistics and Discourse Analysis: A Multi-
Bottenberg, D., M. Lemme, and N. Hedberg. 1985. Analysis of Oral Narra-
tives of Normal and Aphasic Adults. In R. H. Brookshire, ed., Clinical
Aphasiology: Conference Proceedings. Minneapolis: BRK Publishers. 241–7.
Brown, M. and G. R. Kuperberg. 2015. A Hierarchical Generative Frame-
work of Language Processing: Linking Language Perception, Interpret-
ation, and Production Abnormalities in Schizophrenia. Frontiers in
Human Neuroscience 9: 643.
2010. Dissociation, Relatedness, and ‘Cohesive Harmony’. Linguistics and
the Human Sciences 3(3): 263–93.
Chapman, S. B., H. K. Ulatowska, K. King, J. K. Johnson, and D. D. McIntire.
1995. Discourse in Early Alzheimer’s Disease versus Normal Advanced
Aging. American Journal of Speech-Language Pathology 4(4): 124–9.
Coelho, C. A. 2002. Story Narratives of Adults with Closed Head Injury and
Non-Brain-injured adults: Influence of Socioeconomic Status, Elicitation
Task, and Executive Functioning. Journal of Speech Language and Hearing

Research 45(6): 1232–48.
Coelho, C. A., B. Z. Liles, and R. J. Duffy. 1991a. Discourse Analyses with
Closed Head Injured Adults: Evidence for Differing Patterns of Deficits.
Archives of Physical Medicine and Rehabilitation 72: 465–8.
Coelho, C. A., B. Z. Liles, and R. J. Duffy. 1991b. The Use of Discourse
Analyses for the Evaluation of Higher Level Traumatically Brain-injured
Adults. Brain Injury 5(4): 381–92.
Coelho, C. A., B. Z. Liles, R. J. Duffy, J. V. Clarkson, and D. Elia. 1994.
Longitudinal Assessment of Narrative Discourse in a Mildly Aphasic
Adult. Clinical Aphasiology 22: 145–55.
Coelho, C. A., B. Z. Liles, and R. J. Duffy. 1995. Impairments of Discourse
Abilities and Executive Functions in Traumatically Brain-injured Adults.
Brain Injury 9(5): 471–7.
Cohen, I. 2011. The Expression of Schizophrenia through Interpersonal Systems at
the Level of Discourse Semantics. PhD Thesis, Bar-Ilan University.
Crystal, D. 1981. Clinical Linguistics. Vienna: Springer.
Crystal, D. 2002. Clinical Linguistics and Phonetics’ First 15 Years: An
Introductory Comment. Clinical Linguistics & Phonetics 16(7): 487–9.
Crystal, D. 2013. Clinical Linguistics: Conversational Reflections. Clinical
Linguistics & Phonetics 27(4): 236–43.
Davis, G. A. and C. A. Coelho. 2004. Referential Cohesion and Logical
Coherence of Narration after Closed Head Injury. Brain and Language 89:
508–23.
de Villiers, J. 2005. Experientially-based Narration in Autism: A Case Study.
In A. Makkai, W. J. Sullivan, and A. R. Lommel, eds., LACUS Forum XXXI.
Houston: LACUS. 215–25.
de Villiers, J. 2006a. Syntactic and Semantic Patterns of Pedantic Speech in
Asperger’s Syndrome. In S. Hwang, W. J. Sullivan, and A. Lommel, eds.,
LACUS Forum XXXII. Houston: LACUS. 81–9.
de Villiers, J. 2006b. Discourse Analysis in Autism Spectrum Disorder.
Linguistics and the Human Sciences 1(2): 245–60.
de Villiers, J. 2009. Use of Um and Uh in spontaneous speaking in Autism
Spectrum Disorder. In P. Sutcliffe, W. J. Sullivan, and A. R. Lommel, eds.,
LACUS Forum XXXVI. Houston: LACUS, 101–10.
de Villiers, J. 2011. I Saw the Yellowish Going South: Narrative Discourse in
Autism Spectrum Disorder. Belgian Journal of Linguistics, Cognitive and
Empirical Pragmatics: Issues and Perspectives 25: 3–29.
de Villiers, J. and P. Szatmari. 2004. Message Organization in Autism
Spectrum Disorder. In G. D. Fulton, W. J. Sullivan, and A. R. Lommel,
eds., LACUS Forum XXX. Houston: LACUS. 207–14.
de Villiers, J., J. Fine, G. Ginsberg, L. Vaccarella, and P. Szatmari. 2007. Brief
Report: A Scale for Rating Conversational Impairment in Autism Spec-
trum Disorder. Journal of Autism and Developmental Disorders 37(7):
1375–80.
Dijkstra, K., M. S. Bourgeois, R. S. Allen, and L. D. Burgioc. 2004. Conversa-

tional Coherence: Discourse Analysis of Older Adults with and without
Dementia. Journal of Neurolinguistics 17: 263–83.
Docherty, N. M., M. DeRosa, and N. C. Andreasen. 1996. Communication
Disturbances in Schizophrenia and Mania. Archives of General Psychiatry
53(4): 358–64.
Docherty, N. M., A. S. Cohen, T. M. Nienow, T. J. Dinezo, and R. E. Dangel-
maier. 2003. Stability of Formal Thought Disorder and Referential Com-
munication Disturbances in Schizophrenia. Journal of Abnormal Psychology
112(3): 469–75.
Ferguson, A. 1992. Interpersonal Aspects of Aphasic Conversation. Journal of
Neurolinguistics 7(4): 277–94.
Ferstl, E. C. and D. Y. von Cramon. 2001. The Role of Coherence and
Cohesion in Text Comprehension: An Event-related fMRI Study. Cognitive
Brain Research 11(3): 325–430.
Fine, J. 2006. Language in Psychiatry: A Handbook of Clinical Practice. Sheffield:
Equinox.
Fine, J. in press. Modeling Transitions in the Language of Psychiatric Dis-
orders. In J. de Villiers, W. Sullivan, and D. Mailman, eds., LACUS Forum
XLI. Houston: LACUS.
Fine, J., G. Bartolucci, G. Ginsberg, and P. Szatmari. 1991. The Use of
Intonation to Communicate in Pervasive Developmental Disorders.
Journal of Child Psychiatry and Psychology 32(5): 771–82.
Fine, J., G. Bartolucci, P. Szatmari, and G. Ginsberg. 1994. Cohesive Dis-
course in Pervasive Developmental Disorders. Journal of Autism and Devel-
opmental Disorders 24(3): 315–29.
Garcia, A. M. and A. Ibáñez. 2016. Processes and Verbs of Doing, in
the Brain: Theoretical Implications for Systemic Functional Linguistics.
Functions of Language 23(3): 305–35.
Geertz, C. 1973. The Interpretation of Cultures: Selected Essays. New York: Basic
Books.
Glosser, G. and T. Deser. 1990. Patterns of Discourse Production among
Neurological Patients with Fluent Language Disorders. Brain and Language
40: 67–88.
Gregory, M. 2002. Phasal Analysis within Communication Linguistics:
Two Contrastive Discourses. In P. Fries, M. Cummings, D. Lockwood,
and W. Spruiell, eds., Relations and Functions within and around Language.
Grice, H. P. 1978. Further Notes on Logic and Conversation. In P. Cole, ed.,
Syntax and Semantics, Volume 9: Pragmatics. New York: Academic Press.
183–97.
Guo, Y. E. and L. Togher. 2008. The Impact of Dysarthria on Everyday
Communication after Traumatic Brain Injury: A Pilot Study. Brain Injury
22(1): 83–97.

Arnold.
Edward Arnold.
Halliday, M. A. K. 2005. A Note on Systemic Functional Linguistics and the
Study of Language Disorders. Clinical Linguistics & Phonetics 19(3): 133–5.
Halliday, M. A. K., and C. M. I. M. Matthiessen. 2004. Introduction to Func-
tional Grammar. 3rd ed. London: Edward Arnold.
Hartley, L. L. and P. Jensen. 1991. Narrative and Procedural Discourse after
Closed Head Injury. Brain Injury 5(3): 267–85.
ation. 181–219.
Hasan, R. 1985. The Texture of a Text. In M. A. K. Halliday, and R. Hasan,
eds., Language, Context and Text: Aspects of Language in a Social-semiotic
Perspective. Victoria: Deakin University Press. 70–96.
Huddleston, R. and G. K. Pullum. 2002. The Cambridge Grammar of the English
Language. Cambridge: Cambridge University Press.
Iverach, L. and R. Rapee. 2014. Social Anxiety Disorder and Stuttering:
Current Status and Future Directions. Journal of Fluency Disorders 40:
69–82.
Jorgensen, M. and L. Togher. 2009. Narrative after Traumatic Brain Injury:
A Comparison of Monologic and Jointly-produced Discourse. Brain Injury
23(9): 727–40.
Kemper, S., M. Thompson, and J. Marquis. 2001. Longitudinal Change in
Language Production: Effects of Aging and Dementia on Grammatical
Complexity and Semantic Content. Psychology and Aging 16(4): 600–14.
Kempler, D. 1995. Language Changes in Dementia of the Alzheimer Type. In
R. Lubinski, ed., Dementia and Communication: Research and Clinical Implica-
tions. San Diego: Singular Publishing Group. 98–114.
Körner, H., C. Newman, M. Limin, M. R. Kidd, D. Saltman, and S. Kippax.
2011. ‘The Black Dog Just Came and Sat on my Face and Built a
Kennel’: Gay Men Making Sense of ‘Depression’. Health (London) 15(4):
417–36.
Korrel, H., K. L. Mueller, T. Silk, V. Anderson, and E. Sciberras. 2017.
Research Review: Language Problems in Children with Attention-Deficit
Hyperactivity Disorder: A Systematic Meta-analytic Review. Journal of
Child Psychology and Psychiatry 58(6): 640–54.
Kuperberg, G. R. 2010. Language in Schizophrenia Part 1: An Introduction.
Language and Linguistics Compass 4(8): 576–89.
Kuperberg, G. R., B. M. Lakshmanan, D. N. Caplan, and P. J. Holcomb. 2006.
Making Sense of Discourse: An fMRI Study of Causal Inferencing across
Sentences. Neuroimage 33(1): 343–61.
Kurczek, J. and M. C. Duff. 2012. Intact Discourse Cohesion and Coherence

Following Bilateral Ventromedial Prefrontal Cortex. Brain and Language
123(3): 222–7.
Lee, A., O. van Dulm, M. Robb, and T. Ormond. 2015. Communication
Restriction in Adults Who Stutter. Clinical Linguistics & Phonetics 29(7):
536–56.
Lee, A., M. Robb, O. van Dulm, and T. Ormond. 2016a. Communication
Restriction in Adults Who Stutter: Part II. Clinical Linguistics & Phonetics
30(7): 546–67.
Lee, A. S., M. Robb, O. van Dulm, and T. Ormond, 2016b. Communication
Restriction in Adults Who Stutter: Part III. Clinical Linguistics & Phonetics
30(11): 911–24.
Liles, B. Z. 1985. Narrative Ability in Normal and Language-disordered
Children. Journal of Speech and Hearing Research 23: 123–33.
Liles, B. Z., C. A. Coelho, R. J. Duffy, and M. R. Zalagens. 1989. Effects
of Elicitation Procedures on the Narratives of Normal and Closed Head-
injured Adults. Journal of Speech and Hearing Disorders 54: 356–66.
Linnik, A., R. Bastiaanse, and B. Hӧhle. 2016. Discourse Production in
Aphasia: A Current Review of Theoretical and Methodological Challenges.
Aphasiology 30(7): 765–800.
Lock, S. and L. Armstrong. 1997. Cohesion Analysis of the Expository
Discourse of Normal, Fluent Aphasic and Demented Adults: A Role in
Differential Diagnosis? Clinical Linguistics & Phonetics 11(4): 299–317.
Martin, J. R. and S. R. Rochester. 1975. Cohesion and Reference in
Schizophrenic Speech. In A. Makkai and V. B. Makkai, eds., The First
LACUS Forum. Columbia: Hornbeam Press. 302–11.
Martin, J. R. and P. R. White. 2005. The Language of Evaluation: Appraisal in
English. Basingstoke: Palgrave Macmillan.
Mathers, M. E. 2005. Some Evidence for Distinctive Language Use by Chil-
dren with Attention Deficit Hyperactivity Disorder. Clinical Linguistics &
Phonetics 19(3): 215–25.
Mathers, M. E. 2006. Aspects of Language in Children with ADHD: Applying
Functional Analyses to Explore Language Use. Journal of Attention Disorders
9: 523–33.
McDonald, S. 1993. Pragmatic Language Skills after Closed Head Injury:
Ability to Meet the Informational Needs of the Listener. Brain and Lan-
guage 44(1): 28–46.
McKenna, P. and T. Oh. 2005. Schizophrenic Speech. Cambridge: Cambridge
University Press.
Mentis, M. and C. A. Prutting. 1987. Cohesion in the Discourse of
Normal and Head Injured Adults. Journal of Speech and Hearing Research
30: 88–98.
Mesulam, M. 2008. Representation, Inference, and Transcendent Encoding
in Neurocognitive Networks of the Human Brain. Annals of Neurology
64(4): 367–78.
Mok, Z. and N. Müller. 2013. Staging Casual Conversations for People with
Dementia. Dementia 13(6): 834–53.
Mortensen, L. 1992. A Transitivity Analysis of Discourse in Dementia of the
Alzheimer’s Type. Journal of Neurolinguistics 7(4): 309–21.
Mortensen, L. 2005. Written Discourse and Acquired Brain Impairment:
Evaluation of Structural and Semantic Features of Personal Letters from a
Systemic Functional Linguistic Perspective. Clinical Linguistics & Phonetics
19(3): 227–47.
Müller, N., M. J. Ball, and B. Rutter. 2008. An Idiosyncratic Case of /r/
Disorder: Application of Principles from Systemic Phonology and Sys-
temic Functional Linguistics. Asia Pacific Journal of Speech, Language and
Hearing 11(4): 269–81.
Müller, N. and Z. Mok. 2012. Applying Systemic Functional Linguistics
to Conversations with Dementia: The Linguistic Construction of Rela-
tionships between Participants. Seminars in Speech and Language 33(1):
5–15.
Müller, N. and B. T. Wilson. 2008. Collaborative Role Construction in a
Conversation with Dementia: An Application of Systemic Functional
Linguistics. Clinical Linguistics and Phonetics 22: 767–74.
Olness, G. S. and H. K. Ulatowska. 2011. Personal Narratives in Aphasia:
Coherence in the Context of Use. Aphasiology 25: 1393–413.
Perkins, M. 2011. Clinical Linguistics: Its Past, Present and Future. Clinical
Linguistics & Phonetics 25(11–12): 922–7.
Piehler, M. F. and A. L. Holland. 1984. Cohesion in Aphasic Language. In
R. H. Brookshire, ed., Clinical Aphasiology: Conference Proceedings.
Volume 14. Seabrook Island: BRK Publishers. 208–14.
Quirk, R., S. Greenbaum, G. Leech, and J. Svartvik. 1985. A Comprehensive
Grammar of the English Language. London: Longman.
Reilly, S., B. Tomblin, J. Law, C. McKean, F. K. Mensah, A. Morgan, S. Gold-
feld, J. M. Nicolson, and M. Wake. 2014. Specific Language Impairment:
A Convenient Label for Whom? International Journal of Language & Commu-
nication Disorders 49(4): 416–51.
Rigaudeau-McKenna, B. 2005. Towards an Analysis of Dysfunctional Gram-
mar. Clinical Linguistics & Phonetics 19(3): 155–74.
Ripich, D. and B. Terrell. 1988. Patterns of Discourse Cohesion and Coher-
ence in Alzheimer’s Disease. Journal of Speech and Hearing Disorders 53:
8–15.
Ripich, D. N., B. Y. Terrell, and F. Spinelli. 1983. Discourse Cohesion in
Senile Dementia of the Alzheimer Type. In R. H. Brookshire, ed., Clinical
Aphasiology Conference Proceedings. Volume 13. Minneapolis: BRK Publish-
ers. 316–21.
Ripich, D. N., B. D. Carpenter, and E. W. Ziol. 2000. Conversational Cohe-
sion Patterns in Men and Women with Alzheimer’s Disease: A Longitu-
dinal Study. International Journal of Language & Communication Disorders
35(1): 49–64.
Robins, E. and S. B. Guze. 1970. Establishment of Diagnostic Validity in

Psychiatric Illness: Its Application to Schizophrenia. American Journal of
Psychiatry 126: 983–7.
Rochester, S., J. R. Martin, and S. Thurston. 1977. Thought-process
Disorder in Schizophrenia: The Listener’s Task. Brain and Language 4:
95–114.
Rochester, S. and J. R. Martin. 1979. Crazy Talk: A Study of the Discourse of
Schizophrenic Speakers. New York: Plenum.
Ryle, G. 1968. The Thinking of Thoughts: What is “Le Penseur” Doing? In
Collected Papers, Vol. 2: Collected Essays 1929–1968. New York: Barnes and
Noble.
Sachdev, P. S., D. Blacker, D. G. Blazer, M. Ganguli, D. V. Jeste, J. S. Paulsen,
and R. C. Petersen. 2014. Classifying Neurocognitive Disorders: The DSM-
5 Approach. Nature Reviews Neurology 10(11): 634–42.
Spencer, E., A. Packman, M. Onslow, and A. Ferguson. 2005. A Preliminary
Investigation of the Impact of Stuttering on Language Use. Clinical Linguis-
tics & Phonetics 19(3): 191–201.
Spencer, E., A. Packman, M. Onslow, and A. Ferguson. 2009. The Effect
of Stuttering on Communication: A Preliminary Investigation. Clinical
Linguistics & Phonetics 23(7): 473–88.
Taler, V. and N. A. Phillips. 2008. Language Performance in Alzheimer’s
Disease and Mild Cognitive Impairment: A Comparative Review. Journal of
Clinical and Experimental Neuropsychology 30(5): 501–56.
Tannock, R., J. Fine, T. Heintz, and R. J. Schachar. 1995. A Linguistic
Approach Detects Stimulant Effects in Two Children with Attention-
Deficit Hyperactivity Disorder. Journal of Child and Adolescent Psychophar-
macology 5(3): 177–89.
Tebble, H. 2012. Subjectivity in the Discourse of Depressed Acute Care
Hospital Patients. In N. Baumgarten, I. Du Bois, and J. House, eds.,
Subjectivity in Language and in Discourse. Studies in Pragmatics, Volume
10. Bingley: Emerald. 115–35.
Thomson, J. 2005. Theme Analysis of Narratives Produced by Children with
and without Specific Language Impairment. Clinical Linguistics & Phonetics
19(3): 175–90.
Togher, L. 2001. Discourse Sampling in the 21st Century. Journal of Communi-
cation Disorders 34: 131–50.
Togher, L. and L. Hand. 1998. Use of Politeness Markers with Different
Communication Partners: An Investigation of Five Subjects with Trau-
matic Brain Injury. Aphasiology 12(7/8): 491–504.
Togher, L., L. Hand, and C. Code. 1997a. Analyzing Discourse in the Trau-
matic Brain Injury Population: Telephone Interactions with Different
Communication Partners. Brain Injury 11(3): 169–89.
Togher, L., L. Hand, and C. Code. 1997b. Measuring Service Encounters
with the Traumatic Brain Injury Population. Aphasiology 11(4–5):
491–504.
Togher, L., S. McDonald, C. Code, and C. Grant. 2004. Training Communi-

cation Partners of People with Traumatic Brain Injury: A Randomised
Controlled Trial. Aphasiology 18(4): 313–35.
Togher, L., S. McDonald, R. Tate, E. Power, and R. Rietdijk. 2009. Training
Communication Partners of People with Traumatic Brain Injury:
Reporting the Protocol for a Clinical Trial. Brain Impairment 10(2):
188–204.
Togher, L., S. McDonald, R. Tate, E. Power, and R. Rietdijk. 2013. Training
Communication Partners of People with Severe Traumatic Brain Injury
Improves Everyday Conversations: A Multicenter Single Blind Clinical
Trial. Journal of Rehabilitation Medicine 45: 637–45.
Ventola, E. 1979. The Structure of Casual Conversation in English. Journal of
24
Language and Science,
Language in Science,
and Linguistics as Science
M. A. K. Halliday† and David G. Butt
24.1 Introduction
This chapter highlights various correspondences between historical and

contextual changes affecting enquiries into nature, and the ways such
enquiries took up meaning potential offered by choices in the structure of
language. Our main focus, and source of illustration, is vernacular English
after 1600, although we do discuss the legacy of classical forms (rhetorical
and morphological structures) that were the basis of Aristotelian and later
Latin authority. Like other investigators in the field (see, in particular, the
quantitative and qualitative study by Banks 2008), we take the period
1600–1700 as the ‘watershed’ of change and development when speech
and written forms had to respond to new semantic pressures. Beside the
shift between Latin and exposition and argument in vernacular languages,
there were registerial pressures. These created a vector character in the
drift of English grammar, a drift that can be seen in an immediate shift in
the contexts of enquiry, and in the evolving and intensifying patterns of
change that have influenced the idiom across sciences and other registers.
There are three issues taken up in this chapter:
(1) the role of language in the development and evolution of science;

(2) the current character of verbal science, with its dramatic ‘swerve’ into
a favoured pattern for recoding experience and for reconstruing
common-sense reasoning; and
(3) the paradoxical place of linguistics and its techniques in the history
and present state of scientific discourse.
In relation to the broad scope of point (1), we introduce brief historical

reflections on the changing contexts within which meanings were made in
periods of pre-science, or protoscience. We refer to the way different social
†
2018
Language and Science 621
relations had a bearing on the classical Greek versions of what we now refer
to as science – e.g. cosmology; medicine; physical theory. We note here the
enormous project of Geoffrey Lloyd (1970, 2002) in investigating the roots
of sciences in classical Greece. We also note the insights of Edward
Grant (1996, 2007) and others in relation to the actual breadth of ‘natural
philosophy’ as European science in the Latin Middle Ages.
But our focus is on the linguistic patterns that are produced by the shift
towards enquiry articulated and argued in the vernacular languages of West-
ern Europe, and more precisely by the shift ‘within’ English in the seventeenth
century. This is when science undergoes a transition in the contexts in which it
is conducted, and when there is a shift in the relation between novelty,
knowledge, and authority. For instance, scientists moved away from the
assumption that science was just the rediscovery of what had been previously
known, but then hidden, by unimpeachable authorities in eras of revelation.
In addressing our second question, we take up the specifics of what can
be regarded as a grammatical ‘drift’ in English, as described in Sapir’s work
(1921:154–5). It constitutes a convergence of features that, together, can be
thought of as an instance of ‘configurative rapport’ – a notion proposed by
Whorf (1956) for an alignment of grammatical categories which appear to
facilitate a specific variety of socio-semantic activity, a register in this case.
The main purpose of Section 24.2 is to illuminate the features of that drift
and their characteristic direction with respect to the purposes of science
from the seventeenth to twenty-first century.
Our third goal is to consider the place of linguistics and its terminologies
and arguments – our language for talking about language – within the drift
discussed in Section 24.3, ‘Language in Science’. As summarized by Wino-
grad (1983), linguistics has hunted down and borrowed its terminology from
diverse cultural activities: for example, from law; chemistry; atomic physics;
mathematics; logic; philosophy; music; and literary theory. Linguistics is not
unique in these analogical borrowings; but it may be unusual in the degree
to which linguistic thought has been controlled by the metaphors drawn
from outside the specific works of the subject (consider the role of logic as an
influence on semantics). In addition, there is in Western science a double
fold in terminology: as Halliday (2004c) points out, we use and mix both
Greek and Latin terminologies, with Greek typically the term that is higher
up in a scale of abstraction (see Section 24.2 below).
At different points in this discussion, we elaborate our views of an
important analogy between ‘verbal science’ and ‘verbal art’. In both verbal
art and verbal science, there is a ‘sideways’ move in the way the practitioner
utilizes the linguistic system: there is a step into an established yet
malleable metaphorical architecture, a tradition that presents opportunity,
constraint, and challenge. The opportunity is in the potential for
innovation – adding to the range of representations and their relevance to
linguistic activities. The constraint is in the formal alignment with previous
practice – how newly introduced metaphors are connected with those
already established and legitimized by a tradition (either formalities
622 M . A . K . H A L L I DAY A N D DAV I D G . B U T T
of knowledge – ‘scientia’ – or forms and conventions of artistry). The

challenge is in whether the verbal arrangements appear to connect with
our immediate world. In this analogy between verbalized science and verbal
art, we draw attention to the patterning of patterns referred to in Hasan’s
(1985) stylistics as an inter-level of ‘symbolic articulation’. In both
Halliday’s view of metaphorization through the grammar in verbal science
and in Hasan’s view of verbal art, the habitual patterns of the conventional
linguistic system are reworked to serve purposes at the limits (and beyond)
of current collective understandings.
24.2 The Development and Evolution of Science:

Historical and Contextual Reflections
24.2.1 Historical Reflections

Science has extended the resources of language by making novel demands
on the meaning potential of the communities in which the practices of
science have developed. At the same time, we can say that particular
resources in the grammars of languages have been peculiarly enabling to
those working to extend scientific explanations of nature. With each new
paradigm of explanation, from ‘pre-scientific’ (mythic) to natural philoso-
phy, classical physics, relativity and quantum theory, and on to the
emphasis of physics and complexity in the evolutionary synthesis of life
sciences, new semantic pressures have had to be accommodated by new
linguistic arrangements. This favoured path to making sense in science may
be interpreted as a form of metaphor, what Halliday has called ‘grammat-
ical metaphor’. In its most schematic approximation, the grammatical
metaphor of sciences has a consistent tendency, namely: to bring together
an ensemble of resources into verbal equations, the terms of which extend
qualifying meanings into greater and greater specifications within nominal
groups. Such compacting has been typically for the purposes of:
(a) distilling the complexity of phenomena in a defining clause; and

(b) summarizing expressions in the textual unfolding of scientific
reasoning, a placeholder in the chain of complex metaphoric ‘things’
that supply the constant thread, the topical consistency.
This consistency is essential, particularly when the ‘thing’ is a kind of

fiction, a placeholder in a total theory which itself is a thread of fictions
(for example, the way ‘infinitesimals’ are a fiction or metaphor in the
application of calculus).
During the seventeenth century, vernacular languages in Europe, under
this pressure from new forms of structured enquiry and debate, showed the
effects of a new semantic pressure – the exposition of a logical (i.e. causal)
sequence within a rhetorical posture of depersonalized objectivity. Halliday
(2004a) has tracked the configuration of features that realize this ‘drift’ in
discourse, this shift in semantic weight, across the grammatical resources

of English. From Chaucer’s explanation of the astrolabe (to his young son)
in Middle English through to the texture of scientific and technical registers
in the twenty-first century, there is an intensifying syndrome of what is
characterized through the term ‘grammatical metaphor’. Grammatical
metaphor is a reallocation of responsibility within and across clauses
rather than a marked adoption of a lexical item. The ensuing clausal forms
of verbal equation in a language began to provide a frame for the four
most important moves of scientific rhetoric – (a) defining; (b) stating class
membership (cladistics, e.g. Hoenigswald and Wiener 1987); (c) measuring
(e.g. Crosby 1996); and (d) exploring analogy (and homology). Other textual
resources were folded into this evolution of scientific discourse, and merit
consideration in their own terms. These include the meaning potential of
the definite article (according to Snell 1953), which in Greek led felicitously
to abstract questions concerning the essential character of matter and
to human observations on the phenomenal and the moral dimensions of
experience: viz. What is ‘the’ good? Furthermore, signs of definiteness also
mediate between the abstract definition of a class and the individual
instance (instantiation). There are also generic patterns that have emerged
in a culture and which may have favoured certain practices. These patterns
became significant in the adoption of a scientific stance with respect
to what we now think of as separate studies, for example: studies of
perception, or of time, or of cosmology, and so on (see Snell 1953; Kline
1953:310–11).
Different phases in human learning (common-sense – educational –
technical – theoretical) all make their special demands on reasoning and
persuasion in language. This is further differentiated by the division of
semantic labour into distinct registers, at least as communities create novel
niches of argument, problem solving, and discourse. Nevertheless, we argue
(see Section 24.3) that verbal science has created a domain of metaphoriza-
tion different from, but analogous to, the re-articulation of human experi-
ence in verbal art. This last claim may seem counter-intuitive. But scientists
have created a considerable recoding of common-sense speech into an
idiom of interlocking ‘fictions’ that, in toto, allow their specialist commu-
nities to intervene in, and predict, the transactions between different forms
of matter, different phenomena, and different processes.
Verbal science and verbal art both demand metaphorical constructions of
meaning in all of their instances or utterances. Furthermore, these diver-
gent forms of creativity draw upon – albeit in different ways – a pre-existing
metaphoric architecture with which scientists and artists must seek certain
conventional alignments and, if possible, some novel extension. The fictions
of verbal art are no less efficacious to the living of life than the tools of
science; but they are different in their direction (or vector character).
For example, they do deal with judgement and proportion, but in relation
to moral and existential complexities. In verbal science, quantities and
proportions are invoked to explore the transactions between forms of

matter and to capture the character and consequences of processes.
24.2.2 Contextual Changes

According to Grant, there were three crucial contextual factors in the
development of science between the Middle Ages and the seventeenth
century in Western Europe: the availability of new translations; the role
of universities; and the existence of a class of theologians trained equally in
natural philosophy (Grant 1996:171–2). While learning had been expressed
and retained in Arabic communities (Bala 2010) as well as through certain
recovered classical texts of Latin and Greek in Europe, knowledge under-
went a number of further, revolutionary, changes in the seventeenth cen-
tury (see Dolnick 2011). Thinking of these changes in terms of Halliday’s
contextual parameters and metafunctions, we can suggest the following:
(i) experimentation became a more widely accepted approach to enquiry

(field and ideational meaning);
(ii) authority (including religious authority) lost ground to the value of
novel science, along with a fascination for demonstrations (tenor and
interpersonal meaning; mode/medium);
(iii) scientific clubs, societies, and forms of patronage created new forums
for discourse in ‘natural philosophy’, as well as a new style of enter-
tainment (hence, tenor related);
(iv) nascent journals gradually took over from the epistolary communi-
cations of the past, and such journals were indicative of a desire
to declare findings publicly rather than to hide discoveries (mode/
channel and textual metafunction);
(v) new symbolic tools (e.g. the calculus of Leibniz, or Newton’s fluxions)
created the potential for dealing with change in the natural world (e.g.
motion: hence, field and experiential meaning; and mode/medium
related; that is, new forms of semiosis working with wordings);
(vi) certain beliefs and assumptions diminished, especially that all truths
had been revealed to ‘worthies’ in the past (e.g. Aristotle, Ptolemy)
and then hidden in coded messages in their texts, or in the Bible
(tenor and authority related – note that Isaac Newton, who wrote half
a million words on exploring the alchemy of the ‘philosopher’s
stone’, was referred to, in 1946, by a shocked Maynard-Keynes as
the ‘last Babylonian’ rather than the first of modern scientists);
(vii) knowledge moved away from the status of metaphysical ‘mystery’
(originally a word meaning ‘craft’) and became associated with the
technical innovation of the eighteenth century (mode/channel and
tenor related). This shift is exemplified by the explosion of practical
genius from the collaborations of the Lunar Society in Birmingham
(Uglow 2003).
From this drift in the contexts of scientific debate, we can see the conse-
quences in new pressures on the discourses of science. With such new
pressures to explain the world, it is not difficult to understand that scien-
tists have often expressed dissatisfaction with the forms of language they
have inherited and, in some cases, dissatisfaction with the idioms that
they have (unconsciously) created for themselves. Debate has often con-
cerned precision in language and drawn in major figures like Leibniz and
Bentham. Bentham extolled the solidity of nouns and decried the slipperi-
ness of verbs (Kline 1953:312). Bertrand Russell in the twentieth century, as
well as the logical positivist movement and the Vienna Circle, still worked
for a similar purging of misleading grammatical constructions, namely,
those which looked like propositions but which implied the existence of
the non-existent: e.g. The king of France is bald. There has been, since the
Greeks, a preoccupation with ‘what is’, with ontology and the allocation of
‘be-ing’ (see Kahn 2009).
In both useful and eccentric ways, scientists have had much to say about
the status of language in science: for instance, Einstein, Heisenberg (1958),
and Bohm (1980) in physics; Waddington (1977) and Medawar (1984) in
biology; Whitehead in philosophy/logic; both Whorf and Chomsky (in dif-
ferent ways) in relation to grammar; and Firth in phonology and in linguis-
tics more generally (see Firth 1957, 1968; and also Butt, this volume).
The views are polarized: poetry was dismissed as ‘ingenious fiction’
(by Newton’s teacher Barrow); but in the twentieth century, the poet
Shelley’s sense of imagination (poiesis) was carefully aligned with the innov-
ations of science by Medawar – a biologist with little patience for anything
less than a strong Popperian approach to falsification (Medawar 1984:50–2).
Waddington (1977) and Bohm (in biology and physics respectively), and
Whitehead (logic and philosophy) have all expressed Whorfian-like con-
cerns about the narrowness of scientific habits of representation, for
instance, in relation to the inadequate representation of ‘process’. By con-
trast, Heisenberg was disgusted by the rival version of quantum mechanics
proposed by Schrödinger because it moved away from mathematical for-
malism to an unworthy tendency to visualization (Cassidy 1992:215).
By contrast, in Crosby’s study of measurement – in relation to the
development of sciences and European imperial expansions – it is to the
visualization of knowledge that the colonialist dominance of Europe is
attributed – from ballistics to bookkeeping in commerce (Crosby 1996).
Halliday has discussed two technical problems that are relevant to the
role of natural language as an instrument of science, and as an object of
scientific enquiry – hence, to the two main issues of this chapter. Halliday
(2002) notes that, firstly, the meanings of grammatical categories are
‘ineffable’, or unresolvable to any origin in a non-semantic plane of experi-
ence. That is, they cannot be referred to anything but another version of
themselves. And, second, that the linguistic categories which have to do
service in a metalinguistic statement are dominated by traditions which are
themselves poor representations for the flow of natural spoken discourse.

In response to the suggestion by the physicist David Bohm (1980) that, to
capture change, science needed a rheomode, or language of ‘flow’, Halliday
(2003) argues that the choreography of natural conversation better fulfilled
that fluid characteristic than did the compressions of the nominalized
texture in the drift of science publications.
While scientists do emphasize process in material processes (viz. Bohm’s
linguistic experiment with word building), their linguistic and mathemat-
ical idioms appear to conjure with ever more fantastic things, known
only by the subtle perturbations they produce when humans measure the
flow of experience (viz. neutrinos; Higgs boson particles; gravitational
waves; black holes; parallel universes; dark matter, etc.). Black holes, for
example, were inferred in 1939 from mathematics (itself a semiotic and
metaphorical edifice); but the idea has had a long gestation period before
recently receiving the full acceptance of physicists and astronomers (now
fully ‘thingified’, and considered the source of a further latent phenom-
enon: gravitational waves).
Despite the new forms of instrumentation, and the intensity of communi-
cations and publications, it is important to appreciate the role of personal
exchanges, conversations, and even the face-to-face exchanges of meanings in
the development of sciences. Latin functioned as a lingua franca amongst
‘natural philosophers’ of Europe. Greek and Arabic had flourished around
the Mediterranean centres of commerce and learning. Greek continued to be
a dominant medium of learning throughout the Roman Republic and
Empire. Milton, whom the geneticist Steve Jones described as maybe ‘the last
man to have known everything’ (Jones 2000:xxxviii), talked with Galileo face
to face in Italy. Much of modern quantum theory was argued out in walking
around the streets of Copenhagen: for example, the debate over the com-
pleteness of Bohr’s theory as tested by the EPR thought experiment. Darwin
was a confidante of all the leading naturalists of England, from Gould to Lyell
(in geology); and so Darwin’s writing in On the Origin of Species bears the
hallmarks of the twenty years of personal conversations which prefigured
the lucid, persuasive character of that astounding intimation with its reader.
24.3 Language in Science: The ‘Drift’ of English
At present, language is having to accommodate to (let us call them) twenty-

first-century ways of thinking, as knowledge shifts from the largely discip-
linary base of the last century to a more ‘thematic’ approach to problems
(see Shapiro 2011). There is, then, a broad front of scientific themes at the
same time that specialization has become even more rigorous through
new understandings, often provoked by finer scales of investigation, by
dynamic imaging, and by the ‘mining’ of archives. Consider, for example,
the potential for seeing the trillion-frame-per-second camera at MIT.
Themes to reflect upon include:
(i) population thinking, which Darwin mistakenly sought in the work

of Malthus (based as it was on misleading data from America). An
adequate sample is an assumption of all rational activities, with an
emphasis on quantitative results and a representative corpus (now
‘big data’);
(ii) consciousness thinking, which is neurocognitive and needs rela-
tional networks (see Lamb 2013; Garcia 2013; Strogatz 2003) and the
evaluation of networks;
(iii) systems thinking, which is paradigmatic and multilevelled, and
needs complexity theory to construe the vastness of combinatorial
potential;
(iv) evolutionary thinking, which is diachronic and restores our need to
reconcile synchrony with our histories;
(v) information thinking, which is increasingly relevant to the infor-
mational exchanges of supposedly ‘simple’ life forms, encompasses
the probabilistic modelling of populations within us and beyond us,
and consequently demands that we seek a more general theory of
meaning (viz. Shapiro 2011; Josephson 2002; Hoffmeyer 2010; Davies
and Gregerson 2010; and the extended overview by Deacon 2011).
Register variation may be called upon to explain the variation in the

ensembles of semiotic modes that scientists find in these different (though
overlapping) contexts. Shapiro (2011), for example, notes that there are
loosely bounded ‘cassettes’ of information awaiting the demands of experi-
ential pressures before they adopt a specific coding/form (e.g. ‘systems of
domains’; ‘domain swapping’; ‘distributed genome’ across bacteria; and
‘genomic islands’). This horizontal transfer of genetic potential to exploit
a ‘specific ecological opportunity’ sounds very much like ‘biological mean-
ing potential’, perhaps analogous to the registerial potential that social
beings invoke to fulfil relatively distinct conventional social purposes,
always with some ‘in the moment’ fashioning of systemic probabilities
(responding to variation in the ‘instance’). With instantial adaptation in
language, modifications may be ‘chosen’ in any or all of the patterns of the
four metafunctions: changes in experiential, logical, interpersonal, and
textual systems of meaning. Hence, the ways we write and talk about
sciences ‘drift’ or evolve.
Through both language and cells, the contingencies of living are worked
into systems of life. This is not just through fuzziness or hybridity between
categories, but is due to organization by which every instance of behaviour
exhibits creative variation. For example, the influence of some lived experi-
ences has been seen to be expressed through epigenetics, thereby redirect-
ing the overly deterministic tendencies of the earlier ‘READ ONLY’ genetics
of the Crick era (viz. Crick dogma, 1958, and its revisions, recounted in
Shapiro 2011:24–5).
Let us briefly summarize this part of the argument. Language – every

human language – is a stratified system in which the content plane is split
into a semantics, interfacing with the world of human experience (and of
human social relationships), and a grammar, which is a purely abstract
level of organization; the two are coupled through a relation of congruence,
but they can be decoupled and recoupled in other ways (which we are
calling ‘grammatical metaphor’). This gives the system indefinitely large
semogenic power, because new meaning is created at the intersection of the
congruent and the metaphoric categories (‘semantic junction’). Such recou-
plings can be combined according to different criteria of world building via
word building. Where the purpose or ‘end’ of meaning is constrained by the
expectation of a successful prediction, we have the naturalism of scientific
metaphor, no matter how fanciful the tools by which the prediction is
reached (viz. Bishop Berkeley’s satirizing of the ideas of calculus (Berkeley
2002)). When the ‘end’ of a text is unconstrained by naturalism, but is
judged by whether it constitutes a discovery of feeling, a novel experience,
or a recognition of a hitherto unarticulated experiential state, we may
bestow upon it the status of verbal art. Writers, in the human sciences,
for example, may also achieve in their expositions some of the character of
verbal art. And the exposition of Lucretius and the perspectivism (percep-
tual relativism) of Wallace Stevens demonstrate that poets may extend the
naturalism of sciences.
This metaphorical potential seems to be exploited particularly at
moments of major change in the human condition. We find one such
reconstrual of experience in the languages of the Iron Age cultures of
the Eurasian continent (of which classical Greek was one), which evolved
discourses of measurement and calculation, and ordered sets of abstract,
technical terms – the registers of mathematics and science (see Dijkster-
huis 1961: Part 3). This grammar was carried over through classical
and medieval Latin, and also, with a significant detour via Syriac and
Arabic (Bala 2010), into the national languages of modern Europe (Crosby
1996).
A further reconstrual then took place in the ‘modern’ period, with the
evolution of the discourses of experimental science from Galileo and
Newton onwards; and it is this secondary reconstrual – a drift or syn-
drome of tendencies – that we wish to depict in the present chapter. This
new semiotic potential provided the foundation for our discipline-based
organization of technical knowledge. In fact, our sense of modernity and
the ways we think about and transact with experience may be a function
of this syndrome of changes (see Flynn 2007 and elsewhere on modernity
and IQ). Let us consider an example from the language of physical science
that Halliday (2004b) has set out in thirteen detailed steps. In this section
of the chapter, it is most practical to take Halliday’s illustrations in full.
The first illustration falls within the ‘moderate’ range of the background
drift to a ‘favoured’ clause pattern, a paragraph from thirty years ago
and drawn from the public, non-specialist literature of science, Scientific

American, December 1987:
The rate of crack growth depends not only on the chemical environment but
also on the magnitude of the applied stress. The development of a complete
model of the kinetics of fracture requires an understanding of how stress
accelerates the bond-rupture reaction.
In the absence of stress, silica reacts very slowly with water.
(Michalske and Bunker 1987)
Here are instances of some of the features that express the syndrome
referred to above:
(1)the expression rate of growth, a nominal group having as Head/Thing

the word rate which is the name of an attribute of a process, in this
case a variable attribute: thus rate is agnate to how quickly?;
(2) the expression crack growth, a nominal group having as Head/Thing
the word growth which is the name of a process, agnate to (it) grows;
and as Classifier the word crack which is the name of an attribute
resulting from a process, agnate to cracked (e.g. the glass is cracked), as
well as of the process itself, agnate to (the glass) has cracked; crack growth
as a whole agnate to cracks grow;
(3) the nominal group the rate of crack growth, having as Qualifier the
prepositional phrase of crack growth; this phrase is agnate to a qualify-
ing clause (the rate) at which cracks grow;
(4) the function of the rate of crack growth as Theme in the clause; the
clause itself being initial, and hence thematic, in the paragraph;
(5) the finite verbal group depends on expressing the relationship
between two things, ‘a depends on x’: a form of causal relationship
comparable to is determined by;
(6) the expression the magnitude of the applied stress: see points (1) and
(3) above; its function as culminative in the clause (i.e. in the
unmarked position for New information);
(7) the iterated rankshift (nominal group in prepositional phrase in nom-
inal group etc.) in the development [ of [a complete model [for [the kinetics
[of [fracture ] ] ] ] ] ];
(8) the finite verbal group requires expressing the relationship between
two things, development – requires – understanding (see point (5) above);
(9) the parallelism between (rate of ) growth – depends on – (magnitude of )
stress and development – requires – understanding, but contrasting in that
the former expresses an external relationship (third person, ‘in rebus’:
‘if (this) is stressed, (that) will grow’), while the latter expresses an internal
relationship (first-and-second person, ‘in verbis’: ‘if (we) want to model,
(we) must understand’) (see Halliday and Hasan 1976:240–1);
(10) the expression an understanding of how . . ., with the noun functioning
as Head/Thing being the name of a mental process: agnate to (we) must
understand; and with the projected clause how stress accelerates . . . func-
tioning, by rank shift, in the Qualifier;
(11) the clause stress accelerates the bond-rupture reaction, with finite verbal
group accelerates as the relationship between two things which are
themselves processes: one brings about a change in an attribute of the
other, agnate to makes . . . happen more quickly;
(12) the simple structure of each clause (three elements only: nominal
group + verbal group + nominal group / prepositional phrase) and the
simple structure of each sentence (one clause only);
(13) the relation of all these features to what has gone before in the
discourse.
In the most general terms, a grammar construes experience as process, in

the form of a grammatical unit, a clause. Each process, in turn, is construed
as a configuration, in the form of a grammatical structure; the components
of this configuration are (1) the process itself, (2) certain entities that
participate in the process, and (3) various circumstantial elements that
are associated with it; these are construed in the form of grammatical
classes, the verbal, the nominal, and some more or less distinct third type.
Then, one process may be construed as being related to another, by some
form of grammatical conjunction.
The way things are is the way our grammar tells us that they are. In the
normal course of events, we do not problematize this construal; it is our
‘taken for granted reality’, and we do not reflect on why the grammar
theorizes experience the way it does or whether it could have been done
in some other way. If we do reflect, we are likely still to appeal to a sense of
what is natural. We might reason that, as long as to our perceptions things
stay just as they are, we do not ‘experience’ them; experience begins when
the organism becomes aware of some change taking place in its environ-
ment (or in itself ). Hence the grammar construes experience around the
category of ‘process’: a process typically represents some sort of change, of
which staying the same – not changing – becomes just the limiting case.
But sorting out a process of change from the entities that remain in
existence throughout and despite the change (let alone from other phenom-
ena that are seen as circumstantial to it) is already a major enterprise of
semiotic construction.
At the same time as construing instances of human experience, the
grammar also has to construe itself, by creating a flow of discourse. This
is often referred to as ‘information flow’; but this term – as always! –
privileges the ideational meaning, whereas the discursive flow is interper-
sonal as well as ideational. It is as if the grammar was creating a parallel
current of semiosis that interpenetrates with and provides a channel for the
mapping of ideational and interpersonal meanings. The metafunctional
component of the grammar that engenders this flow of discourse is the
textual (see Martin 1992: Chapter 6; Matthiessen 1995).
In Halliday’s comparison of a scientific explanation from three eras – that

of Chaucer on the astrolabe, to his young son; of Newton (1998) on
‘Opticks’; and Clerk-Maxwell (at the outset of modern physical theory) –
certain tendencies were dramatic. Such tendencies could be summed up
under five generalizing headings:
(1) expanding the noun as a taxonomic resource: this was the goal of
language planning in the 1600s, especially in England and France (see
Salmon 1979);
(2) transcategorizing processes and qualities into nouns, relators into
verbs, etc., with resulting semantic junction;
(3) compacting pieces of the argument to function (e.g. as Theme of the
clause) in an ‘information flow’ of logical reasoning;
(4) distilling the outcomes of (2) and (3) to create technical taxonomies of
abstract, virtual entities;
(5) theorizing by reconstruing of experience as in (1) to (4), with a
‘favoured clause type’ in which virtual entities participate in virtual
processes based on logical-semantic relations that bestow a linear and
causal plausibility on the phenomena under investigation (see Halliday
and Matthiessen 2014).
Let us consider a random snapshot of the current intensities of even the

general release and discussion of scientific developments (published in the
magazine of the American Association for the Advancement of Science,
July 1, 2016). The first example comes from a study of early hominin teeth:
We propose that although a flat-toothed grinding platform may be subopti-

mal for fracturing tough foods relative to a bladed (crested) morphology, it is
a better solution than the smaller, less thickly enameled molars of A./P.
Boisei’s predecessor.
(Ungar and Hlusko 2016:29)
This article pursues a broad, extremely significant theme in evolution –

Does function follow form? Are there different forms for the same func-
tion? We can also see the teleonomic issues of species having a solution to a
problem and the grading of what is the optimal form in relation to that
putative ‘problem’ in natural selection. The glass crack growth rate of our
previous example was subject to being directly observed (even if by a
camera). But the ‘optimality’ in this passage is a product of complex
inference and measurements, including the correlation of the phenomena
measured with other abstract inferences, especially concerning the rate of
evolutionary change in hominin ancestry: a path of least resistance ‘defined
by the structure of the underlying genetic covariance’ (Ungar and Hlusko
2016:30).
This is an interlocking of established rhetorical ‘figures’ at the level of
semantics – essentially the distillations of previously achieved progressions
(of argument) – with the lexicogrammatical syndrome discussed above. It is
also usefully illustrated in Cameron and Goldberg’s (2016) article in the

same journal issue on central nervous system repair. Here it opens with
three nominal groups that assume the established arguments of specialists:
(1) A developmental loss of intrinsic reparative capacity and (2) the inhibitory
environment in injury and disease contribute to (3) regenerative failure in
the central nervous system (CNS).
(Cameron and Goldberg 2016:30, emphases added)
What needs to be noted in English, additionally, are the morphological

extensions of classical elements that become a barrier to construal in the
taxonomic architectures that are built in specialist branches of sciences. For
instance, we are directed to researchers who have identified a pathway:
a conserved injury response pathway [in innate immune cells] [[that is

mediated by | mesencephalic astrocyte-derived neurotrophic factor (MANF)
|, a macrophage-dependent, prosurvival signaling molecule | ]].
(Cameron and Goldberg 2016:30, emphases and bracketing added)
The pathway as Head noun is qualified by a nominal group (Head word: cells),
and then qualified separately by a clause with new nominal (Head noun:
factor), which is itself extended by a group which elaborates factor with
another complex group with the Head noun molecule. The management of
the web of life demands an isomorphic web of intricate verbal taxonomy,
each node of which is legitimated by logical relations (which are often
muted or neutralized with forms like involves, mediates, contributes to, emerges,
depends on, comes mainly from, etc.).1
The grammar, as a stratified system of contexts, semantics, and lexico-
grammar, sets up categories and cross-stratal relationships which have the
effect of transforming experience into meaning. In creating a formal dis-
tinction such as that between verb and noun, the grammar is theorizing
about processes: that a distinction can be made, of a very general kind,
between two facets – the process itself, and entities that are involved in it.
But, as remarked above, since the grammar has the power of construing,
by the same token (that is, by virtue of being stratified) it can also decon-
strue, and reconstrue along different lines. Since stratification involves
mapping meanings onto forms, ‘process’ onto verbal, and ‘participant’ onto
nominal, it also allows remapping – say, of ‘process’ onto a nominal form: a
process can be re-packaged as a nominal group so that a bus driver does not
drive but owns the process: viz. the driver’s overrapid downhill driving of the
bus. The experience has now been retransformed – in other words, it
has undergone a process of metaphor. It can now be predicated with a
1
A crucial point in the next decades of science teaching and learning may be that such taxonomies in Chinese, for
example, do not take their morphology from an arcane, classical, synthetic tradition. A fifteen-year-old student learning
science in China deals with combinations of common-sense meanings in trying to understand the complexifications of
biology or physics. We may need to take stock of the language of science teaching along with its current creative
visualizations and online presentations.
judgement, a measurement, or an attribution. A stratified system has

inherent metaphoric power.
As widely noted, English and other scientific codes are exploiting an
established resource, the grammar’s potential for nominalizing: turning
verbs and adjectives into nouns, as in these prototypical examples from
ancient Greek:
(1) verb: active (Actor) noun: ‘one who / that which . . .-s’
ποιέω make ! ποιητής maker
πράσσω do ! πράκτωρ doer
(2) verb: passive (Goal) noun ‘that which is . . .-n’
be made ! ποίημα thing made
be done ! πρᾶγμα thing done, deed
(3) verb: middle (Medium) noun ‘. . .-ing’ (abstract)
make ! ποίησις making
do ! πρᾶξις doing, action
(4) adjective: noun of quality / degree ‘being . . .; . . .?’
μέγας big ! μέγαθος size; greatness
βαθύς deep ! βάθος depth, deepness; altitude
The Greek forms provided the model for scientific terminology in Europe.
In scientific discourses, the semiotic power of referring is being further
exploited so as to create technical taxonomies: constructs of virtual objects
that represent the intensification of experience (typically experience that
has itself been enriched by human engineering, in the form of experiment
and/or measurement). The semiotic power of expanding begins by relating
one process to another by a logical-semantic option. This potential can be
exploited so as to create chains of reasoning: drawing conclusions from
observation (often observation of experimental data) and construing a line
of argument leading on from one step to the next.
Grammatically, these two discursive processes, which lead out of the
daily language into a mode of systematic theory, both depend first and
foremost on the same basic resource: the metaphoric compression of a
clausal into a nominal (thingifying) mode of construal (see Figure 24.1
and Table 24.1). The novel ‘things’ of science populate a world of semiotic
artifice, no less than is true for the metaphoric strategies of verbal art. The
depiction of reality in both cases is through an oblique, derivative semiotic
technique. They start from two variants of experience, however. Verbal art
might find its conventions in clusters of metaphors (consider, for example,
the European Middle Ages and the ‘expected’ values expressed in the
motifs: rose, thorn, blood, and knight errant).
But science, typically as natural philosophy, has had to thingify by creat-
ing a more plausible architecture of smaller components, with or without a
God as ‘prime mover’. Reduction has been the dominant pathway of sci-
ence, whether the existence and relevance of the particular elements pro-
posed could be supported by evidence or not. This semiotic vector has
Figure 24.1 Typology of grammatical metaphors (from Halliday 2004a:41–2)
reached an apogee in atomic physics where the formalism of the math-

ematics has been argued as the limit of what is ‘knowable’ (viz. Heisenberg,
in particular, his rejection of Schrödinger’s wave mechanics ‘picture’ of
quantum phenomena in Cassidy 1992:215). With the entanglement of
electrons – that is, what Einstein called ‘spooky action at a distance’ –
now supported by experiment (Aspect et al. 1982), and with new infor-
mation drawn from quantum ‘discord’ (Brooks 2016a:94), the metaphors
of physics have attained a new distance from common sense, as well as
from some central tenets of the architecture of physics (e.g. Is light still the
speed limit of the cosmos?). In parallel, developments in biology have
replaced simplistic interpretations of the emergence of the higher levels
of organization in goal-directed living processes (Deacon 2011: Chapters 1
and 17).
From Newton’s Opticks, we can illustrate how the full semogenic potential
of metaphoric nominalization in the grammar is opened up. A process,
such as ‘move’, is observed, generalized, and then theorized about, so that it
becomes the virtual entity ‘motion’. As a noun, it now has its own potential
(a) for participating in other processes, as in: ‘The Rays of Light, whether
they be very small Bodies projected, or only Motion or Force propagated, are
moved in right Lines’ (Newton 1998:268); and (b) for being expanded into a
taxonomy, such as linear motion, orbital motion, parabolic motion, peri-
odic motion. Semantically, motion realizes the junction of two features:
(i) that of ‘process’, the category meaning of the congruent form ‘move;’
and (ii) that of ‘entity’ or ‘thing’, which is the category meaning of the class
‘noun’ of motion. This kind of semantic junction is what is meant by saying
that the meaning of the term is ‘condensed’. But technicality involves more
than the condensation of ideational semantic features (Martin 1992).
Table 24.1 The ‘general drift’ of grammatical metaphor (from Halliday

2004a:42)
key: semantic element grammatical class
grammatical function example
1. quality entity adjective noun

Epithet Thing unstable = instability
2. process entity verb noun

(i) Event = Thing transform = transformation
(ii) Auxiliary = Thing:
(tense) will/going to = prospect
(phase) try to = attempt
(modality) can/could = possibility, potential
3. circumstances entity preposition noun

Minor Process = Thing with = accompaniment; to = destination
4. relator entity conjunction noun
Conjunctive = Thing so = cause/proof; if = condition
5. process quality verb adjective

(i) Event = Epithet [poverty] is increasing = increasing [poverty]
(ii) Auxiliary =
(tense) was/used to = previous
(phase) begin to = initial
(modality) must/will [always] = constant
6. circumstance quality adverb/prepositional phrase adjective*

(i) Manner = Epithet [decided] hastily = hasty [decision]
(ii) other = Epithet [argued] for a long time = lengthy [argument]
(iii) other = Classifier [cracked] on the surface surface cracks
*or noun; cf. mammal [cells]/mammalian [cells]
7. relator quality conjunction adjective

Conjunctive = Epithet then = subsequent; so = resulting
Table 24.1 (cont.)

8. circumstance process be / go + preposition verb
Minor Process = Process be about = concern; be instead of = replace
9. relator process conjunction verb

Conjunctive = Event then = follow; so = cause; and = complement
10. relator circumstance conjunction preposition/-al group

Conjunctive = Minor Process when = in times of/in ... times
if = under conditions of/under ... conditions
11. [zero] entity = the phenomenon of ...
12. [zero] process = ... occurs/ensues
13. entity [expansion] noun [various] (in env.1, 2 above)

Head = Modifier the government [decided] = the government’s
[decision], [a/the decision] of/by the
government, [a] government(al) [decision]
the government [couldn’t decide/was

indecisive] = the government’s [indecision],
[the indecision] of the government,
government(al) indecision
The term ‘motion’ is now functioning as a theoretical abstraction, part of a

metataxonomy – a theory which has its own taxonomic structure as a
(semi-)designed semiotic system (for scientific theories as semiotic systems,
see Lemke 1990). Martin refers to this semantic process as distillation
(Halliday and Martin 1993). We can get a slight sense of the gradual
‘distilling’ effect of progressive nominalization from a simple morphosyn-
tactic sequence in English such as the following:
moves – is moving – a moving – movement – motion

planets move – the planet is moving – a moving planet – the planet’s moving – the
movement of planets – planetary motion
culminating perhaps in the Greek kinesis. Note that the most distilled terms
in English at its most theoretical level tend to be those from Greek (e.g.
ornitho- for ‘bird’).
Let us consider then:
(1) What is the payoff? That is, what effect has such reconstrual on the
construction of the discourse?
(2) What different kinds of metaphorical shift take place, and is there any
general principle lying behind them?
(3) What are the systemic consequences? To put this in other terms, in
what way is ‘regrammaticizing’ experience also ‘resemanticizing’ it?
Again, we can turn to a relatively straightforward example:
If electrons weren’t absolutely indistinguishable (1), two hydrogen atoms

would form a much more weakly bound molecule [[than they actually do]]
(2). The absolute indistinguishability of the electrons in the two atoms gives
rise to an ‘extra’ attractive force between them (3).
(Layzer 1990:61, bracketing and numbers added)
Here the grammatical metaphor has a discursive function: it carries for-

ward the momentum of the argument. The combination of metaphoric
features in clause (3) is what we can regard as the ‘favoured clause type’
of English scientific writing. It is a fuzzy type; but we could perhaps
characterize it as follows:
• semantic: sequence of two figures, linked by a logical-semantic relation

• grammatical [congruent]: nexus of two clauses, with Relator/conjunc-
tion in
• secondary clause (optionally also in primary clause) [metaphoric]: one
clause, ‘relational: identifying/intensive, circumstantial or possessive’,
of three elements:
• Identified + Process + Identifier nominal group verbal group nominal
group.
In saying that these are of the ‘favoured clause type’, this is not asserting
that they are the most frequent (there would be no sensible way of estimat-
ing this; at the least, they are certainly very common). But they are the most
critical in the semantic load that they carry in developing scientific argu-
ment. What is interesting about them is that their clause structure is
extremely simple: typically one nominal group plus one verbal group plus
a second nominal group or else a prepositional phrase. But packed into this
structure there may be a very high density of lexical matter; again, compare
the tendency of spoken discourse with the ten lexical words to one verbal
group within the third clause (referred to by the number [3] as cited above).
So, a scientific theory is a specialized, semi-designed subsystem of a
natural language; constructing such a theory is an exercise in lexicogram-
mar. Science and technology are (like other human endeavours) at one and
the same time both material and semiotic practices; knowledge advances
through the combination of new techniques with new meanings. Thus
‘reconstruing experience’ is not merely rewording (regrammaticizing); it
is also resemanticizing. The languages of science are not saying the same
things in different ways (although they may be appropriated for this pur-
pose by others wishing to exploit their prestige and power). What is
brought into being in this reconstrual is a new construction of knowledge;

and hence, a new ideology.
What may have been enabling in one era, however, can become con-
straining in another. The noun-based discourse, which made possible the
moves into measurement and rational argumentation, comes at the cost of
reducing all flux and variation into a stable state and all phenomena into
species of virtual, idealized ‘things’.2 Scientific English and Standard Aver-
age European (the SAE of Whorf’s reflections on the configuration of
categories across explanations in dominant European languages) have
become a static framework for a labile world, a universal inventory instead
of a dynamic, protean cosmos. Halliday (2002) emphasizes this restriction
in the representations of language, much as did Firth in relation to the
variable prosodic (legato) effects that rendered speech more realistically
than did the ‘the rack’ of artificially isolated phonemes. Halliday (2003)
emphasizes the more process-oriented architecture of common speech and
dialogue as a foil to the static Dorian mode that has emerged from science
discourses.
24.4 Linguistics in the Sciences
With our third question put forward at the outset of the article, we empha-
size how scientific thinking must include thinking about language. Much
as an observation requires understanding of light and the effect of lenses,
language is not only an object of enquiry for natural sciences, it is the
medium and channel by which the knowing is secured for the thinking of
scientists. While mathematics has come to be a lingua franca or, perhaps
more narrowly, an esperanto for science (with its own registerial or generic
variants), it is still a requirement of verbal science that quantification
be demonstrably relevant to verbal explanation: i.e. that numbers be
explained. While linguistic sense may need to introduce the role of numer-
ical sense as protection against certain forms of nonsense, all claims come
back to the court of collective consciousness – the language of a
community.
Where there appear to be exceptions – for instance, in Heisenberg’s
formalism in quantum theory, and in quantum entanglement – human
thinking remains restless for a consistency with our common speech and
its common-sense reasonings (see Brooks 2016b; Ananthaswamy 2013). In
quantum weirdness, for example, we might say that the semogenic
resources have been extended to create a new complex meaning; and this
meaning can be thought of, as mentioned above, as an extension of the
texture of language. A comparison with the higher-order process that we
2
See Lucretius’ (1992) 7,000 line poem, De Rerum Natura (On the Nature of Things); and also Mann (2000) on
Aristotle, The Discovery of Things.
call ‘verbal art’ arises at this point in our discussion. Through the tradition
of specialist discussions and mutual exchanges – sometimes face to face –
both scientists and poets create semiotic novelty by reworking the meaning
potential of the collective. In some cases, new symbolic resources are
introduced – for instance, the uncommon sense of calculus with Leibniz’s
infinitesimals and Newton’s fluxions; more recently with the counter-
intuitive behaviours of the very big and the very small in physics. The
physicist Bohm asks us to become conscious of the symbolic process of
making statements in science much in the way that the Russian Formalist
critics of the early twentieth century spoke of defamiliarizing our habitual
modes of describing experience (Bohm 1980:32).
Certain developments may be characterized as forms of semantic canal-
ization (analogous to the biologist’s chreods in an epigenetic landscape
(Waddington 1977:106–7)). These might be efficacious ways of saying novel
ideas as the result of the latent potential of a habitual speech pattern, or a
generic structure. An example may be the power of the definite article in the
grammar of classical Greek, mentioned above. Along with the semantic
‘figure’ of dialogic enquiry (viz. Socrates/Plato), the definite article facilitates
broad, abstract questioning: the essentialist, generalizing, and defining
orientation of sciences (as well as the potentially misleading reification, we
should add). As pointed out by Snell (1953), a Greek philosopher could easily
take discourse from the concrete behaviours of those around to the question
of what defined ‘the good’, and hence on to essentialist discussions of good,
beauty, freedom, power. Kappagoda (2004) gives another important
instance of what might be called a semiotic canalization – a semanticization
of a mode of understanding previously outside the habits of the community.
He shows how Thucydides applies a forensic perspective to the sequence of
events in the plague in Athens. He argues that such a potential grew out
from the metaphoric form As to the leaves of Autumn are human beings.
While there has been a shared origin in the measurement of language in
verse and the measures of music, with Pythagoras first showing the relation
between measurement and human perception (the harmonic scale based on
systematic divisions of string lengths), we will focus on the emergence of
evolutionary ideas in Europe to review the paradoxical status of language
studies, their hidden role at the forefront of what are tendentiously referred
to as the ‘hard’ sciences. As suggested above, the innovations of such sciences
were prepared for by progressive cultural increments in other ideational
domains, domains today not necessarily thought of as objective sciences.
The genealogical ideas of Indo-European and other language families
were adumbrated by William Jones, and authorities preceding Jones, more
than a century before the publication of Darwin’s ‘descent with modifica-
tion’ – the genealogical principle in nature. We would emphasize the
important parallel that Darwin urged between a hypothetical knowledge
of the ancestry of all languages, with all the intervening dialectal variants,
and the ‘economy of nature’, with its slow variants in which ‘natura non
facit saltum’ (‘nature does not make leaps’), and the ‘community of des-
cent’ in relation to species and morphology (Darwin 1998:562–4). Without
any knowledge of genes, the language analogy became Darwin’s most
concrete instance of a mechanism of the genealogical principle. This is
particularly because, in the opening chapters of On the Origin of Species,
Darwin takes such care in distinguishing between selection under nature
and selection by humans, for instance, by animal breeders. Dawkins too, in
his explanation for children of crucial rational ideas – The Magic of Reality –
emphasizes the analogy between natural species and languages (Dawkins
and McKean 2011). But both Darwin and Dawkins pull up short of claiming
a homology, not just an analogy.
Research by Dunn et al. (2011) used mathematical methods for what the
Prague School referred to as ‘characterology’ – a comparison across typo-
logical features between languages. The authors offered the same Darwin-
ian conclusion: the ancestry of a language is the best predictor of the
affinities and contrasts between languages. It is therefore curious to us,
given the role of proto-social sciences in the crowning works of physics and
biology, why so much about language has been excluded from the scientific
conversation. Bohr himself urged the analogy between an anthropologist
disturbing the community under observation and the physicist intervening
in a subatomic system (Bohr 1961); and the literary potential of point of
view, relativity, and complementarity have been utilized by writers and
critics (see, for example, the broad cultural study by Albright 1997). There
was nascent ‘population thinking’ in Defoe’s A Journal of the Plague Year (see
Defoe 2001) as there is a non-fiction journal in Mailer’s Armies of the Night
(1968), and much about point of view and the philosophies of physics and of
rhetoric are illustrated in Sterne’s The Life and Opinions of Tristram Shandy
(1759–1767) (hence the fascination the work produced among the Russian
Formalists, who sought to make a verbal science of the study of verbal art).
The forensic analysis of social complexity in the opening chapter of Mid-
dlemarch (Eliot 1871), and the forensic criticism of Napoleon’s military
strategies in the epilogue of Tolstoy’s War and Peace (1869) tend to support
the idea argued by Snell (1953): namely, that the very forms of literary
genres supplied modes of transitional thinking in the development of
classical Greek science. Since the mobilization of language to represent
human purposes necessarily draws one into “choosing” one thing . . . rather
than the other’ (Halliday 2005), verbal art is a laboratory for thought
experiments on relativity and physics/physis (Butt 2007). Consider how
prose accounts of experience and reported speech commit us to choices of
point of view and deixis.
One problem for the development of any science of language was and is
the ineffability of grammatical categories cited above (Halliday 2002).
While the earliest Greek and Latin grammarians purported to be offering
a theory of natural kinds in the classifications of verbs and nouns, and
other parts of discourse (viz. the very word ‘genres’, or types), it becomes
clear that a peculiar circularity is involved in ‘turning language back on

itself’ (Firth 1957). This process becomes alarming as one compares lan-
guages across language families and cultures of greater semiotic distance
(Hasan 2011: Chapter 3). Saussure’s (1959) notion of ‘valeur’ – the complex
of relations that a linguistic term enters into in a hypothetical total, syn-
chronic system – applies twice over in what has come to be called a
‘metalanguage’ (a language for describing language). Not only are language
categories ‘adrift’ and relativistic, languages themselves both encourage
and challenge deeply held universalist assumptions. A special instance
was the ergative case marking of languages in the Caucasus, also in
Basque, and later found to be central to the description of Australian lan-
guages (Yallop 1982:90–7; and in the memoir by Dixon 2011:100–3). The
intervention of Stalin in the debates surrounding ergativity, along with the
criticism levelled at Russian Formalists for their aims of treating text from
within a ‘systemo-functional’ science (Steiner 1984:94), demonstrate how
volatile the topics of language sciences have been for monovocal ideologies.
There is an intriguing comparison here between this diachronic/
synchronic opposition in linguistics and the problem of intervening stages
in Darwin’s account of species. Evolution is the diachronic theory par
excellence: it replaces an unrealistic, unnatural account of the emergence
of the current state of the ‘economy of nature’. In an apparent contrast,
Saussure (1959) rejected the unrealistic, irrelevant role of diachrony to the
speaker’s ‘langue’, even though linguists did have a practical sense of
dialects as natural intervening stages in the evolution from Proto-Indo
European (PIE) to the vernacular forms (i.e. into the languages of continen-
tal Europe). Saussure himself had become famous for his prediction of the
need of a ‘sonant co-efficient’ in the system of PIE – a prediction confirmed
fifty-three years later in the study of Hittite (Harris 1988:40). Ironically,
evolutionary origins of language became a prohibited topic in linguistics,
even though atomistic mechanisms of change across Indo-European lan-
guages were understood with the reliability of linear statements in chemis-
try. Furthermore, this was precisely the time when the genetic basis of
change in living systems was being uncovered in the re-emergence of the
controlled experiments of Mendel (in 1906, the very year of Saussure’s
(1959) first lectures on the Course and one year after Einstein’s ‘annus
mirabilis’).
The point here is that linguistics might have been seen as an exo-somatic
exemplum of bio-semiotic organization, in particular, in relation to the
genealogical (but, ultimately, not merely genetic) interpretation of natural
forms. Linguistics had its own theories of morphology, and its accounts of
how these forms changed, in particular, how they had transformed under
the specific historical conditions of Europe (viz. the Roman Empire: Ostler
2005, 2007). While Tynjanov and Jakobson (1978) foresaw the necessity of a
systems and functions approach to text variation (Steiner 1984), and Jakob-
son (1973) made much of the potential analogy, the critique of semantics
from logicians and mathematicians in the Vienna Circle and at Cambridge

(e.g. Tarski, Russell, Ayer, and the early Wittgenstein) meant that the
connections were lost. This is to say, science missed the opportunity to
see the meaning systems we create on a continuum with the bio-semiotic
systems that create us – as pointed out to me by J. Cartmill (personal
communication), in the writings of von Uexküll (see Favereau 2010).
There was a period, then, when linguistics could have been productively
viewed as a harbinger of change in science. This was part of the zeitgeist
around the brothers von Humboldt (see von Humboldt 1988); it is in the
movement of Boas and Malinowski from their backgrounds in physics
and mathematics over to the focus on humans in nature; and it is in
Jakobson’s observation that processes of syntax (like neutralization) were
characteristic of ‘the syntax of DNA’ (Jakobson 1973:51). How then, we need
to ask, did this propitious state of nascent interdisciplinary connection
become a source of such a disciplinary ‘stand-off’, with primatologist Lea-
key (1994) noting in wonder that ‘linguists have become the last defenders
of the Rubicon between humans and the rest of nature’? Similarly shocking
is the statement by Shapiro, who after fifty years or more in microbiological
research, notes that there had been a place for semiotics and linguistics in
the information-systems perspective of biology, but that now, for linguists
to be readmitted, they would need to ‘temper’ their thinking based on ‘20th
Century genetic determinism’ (Shapiro 2011:146). How then did the study
of language, and of linguistics, become so insulated from the broader
conversation of the human sciences? How has it taken on the posture of
genetic ‘fundamentalism’ in crucial conversations between scientists?
The crucial issue for us appears to be that Chomsky took his framework
from logic and philosophy, thus cutting linguistics off from its mainstream
scientific inheritance: observational evidence. Syntax and grammaticality
were, in the Chomskyan approach, described in starred and unstarred
strings, like WFF (well-formed formulae) in logic; criteria of formality
prescribed that transformations were meaning-preserving, and that this
meaning was based only on truth conditions; syntactic deep structure
preceded meaning or semantic categories; instances of recorded language
were not relevant, even seen as misleading, since every native speaker was
an intuitive master of the boundary between grammatical and ungrammat-
ical *starred strings; data gathering was merely ‘butterfly collecting’; and
universal grammar came to look more like English, with Pro-drop being
applied to any language which did not exhibit the English pattern of
insistence on grammatical subject (i.e. the majority of the world’s lan-
guages). In this imbroglio, the ‘poverty of the stimulus’ was taken to be
self-evident; empiricism was a benighted doctrine; mind was a Cartesian
mystery about which no one in neuroscience had the slightest ‘interesting’
idea (i.e. about ‘what happens when 100b. neurons are packed into some-
thing the size of a basketball’). Yet, on the other hand, single neurons could
be associated with a single universal idea like that of triangleness. Other
notions included the idea that Darwinism was suspiciously circular as an

explanation; language may have emerged relatively recently in human
history; artificial intelligence was irrelevant to understanding language or
brains; functional explanations, for example, of the heart, serve little pur-
pose (e.g. Chomsky 1975:59); and studies of animal capacities for language
are irrelevant. All these were forcefully transmitted to audiences and read-
erships in the idiom of: we now know . . .; there is massive evidence for x; the
notion y is uninteresting; and no one has the faintest idea about . . . These positions
are reiterated in interviews (recently in Chomsky 2012) and, for a very
recent example, in the tenor of the attacks upon the twenty-five years of
work on Piraha, in the Amazon region, by Daniel Everett. Everett’s situation
is explored through interviews in the television documentary The Grammar
of Happiness, and variously in public debates and in Everett’s books, the
most general of which is Everett (2013).
Unfortunately, the rhetorical investment in Chomsky’s Universal Gram-
mar has run so much further than the support afforded by its base of
inferential evidence. Yet a consequence has been that to mention a science
of linguistics has been, for decades now, taken to be a reference to Choms-
ky’s claims, in particular, for psychologists, who found psycholinguistics a
congenial professional designation, and for non-linguists, who knew
Chomsky through the collateral fame of his stands against the foreign
policies of the US government. The problems with this grafting of grammar
onto a formal, logic-based architecture are set out in the close examination
by, for example, Ellis (1993); and the consequences for semantics have been
taken up by, for example, Seuren (1998). In a recent publication: Why only
us? (Berwick and Chomsky 2016), there is an attempt to reconcile the tenets
of the Chomskyan programme with aspects of current biological discus-
sion. In our view, the influence from a formalist, semantics-free, autono-
mous syntax has had a narrowing effect on the recent history of knowledge.
The opportunity of the current era is that we have now a new conver-
gence of ideas about information, ideas that are ‘visible’ and able to be
elaborated. This is not just a reference to the revolution in the variety of
techniques for brain and body imaging. Internet users can now casually call
up the once unthinkable sample sizes from corpus sites and archives.
Beside the trillion-frame-per-second camera for tracking photons and the
nanoscale activities of molecular processes in living systems, the public can
see the WMAP of the cosmos only 300,000 years after the Big Inflation
(based on the dispersion of heat). Shapiro (2011:132) refers to the virosphere
as the ‘Research & Development’ centre of nature with coalitions and
innovations or ‘rewrites’ of gene sequences occurring in multiple gener-
ations over very short human time (viz. eight hours). He further urges that
‘life requires cognition at all levels’ (Shapiro 2011:7). Fundamental to his
claim is the notion of choice or ‘decision points’, and a fresh approach
across biological and social sciences (Shapiro 2011:146). There are now signs
of a new evolutionary synthesis in which a synchronic view of human
systems is reconciled with evidence from deep evolutionary time, for

instance, in the work of Panksepp and Biven (2012) and Porges (2011), and
integrated into the highly semantic, psychotherapeutic theory of the ‘self’
in the writings of Meares (2012) and Korner (2015).
At the scale of linguistic behaviour, we can see Halliday’s fundamental
metaphor of choice and Sapir’s emphasis on unconscious patterning played
out in electronic channels. The semiotic version of ‘descent with modifica-
tion’ is now even more palpable as the concepts of dialect, register, idiolect,
and accent are accessible as innumerable samples in a burgeoning archive
of ‘snapshots’ – synchronic moments that can be lined up as a diachronic
series which reveal an arc of change. This is not just a graph in the Google
Ngram Viewer, mapping the uptake of a word in written texts, but the fact
that dialects can be tracked at different SES and geographic parameters (see,
for example, Hasan 2009; and inventions like thinkative are all archived for
the convenience of the twenty-first-century Henry Sweet).
Along the contextual parameters of field, tenor, and mode, language
samples are being stored in forms that mean regularity and change in
language need not be theoretical placeholders or acts of faith. The plausible
modelling of process in every discipline, so long a problem of scale, evi-
dence, and mechanism (i.e. as to what motivates change) is now tractable
and practical in linguistics. Even the metaphors of ‘connection’ (viz. ‘net-
works’ in Matthiessen 1995, 2015; Lamb 2013; Strogatz 2003; and ‘connec-
tome’ in Seung 2012) and of ‘information’ (e.g. Davies and Gregersen 2010)
suggest a new potential in the way the systems that humans create may
inform on the transdisciplinary themes of complexity. In this era of cellular
functions for self-modification and of cognitive networks, the ‘emphasis is
systemic rather than atomistic and information based rather than stochas-
tic’ (Shapiro 2011:146).
With sciences moving towards more integrated patterns of knowledge, in
particular, around the motifs of information, systems, and functions, the
study of language could be an integrating force, with linguistics underwrit-
ing a general science of meaning – the original aim of both Saussure and
C. S. Peirce. For progress to be made in this direction, there are two critical
distinctions to be observed. One is that between causal systems and realiza-
tional systems: between the relation of cause and effect and the relation of
token and value. The other is that between realization and instantiation.
Instantiation is the perspectival distance between the observer and the
object under observation; realization is a function of the multi-stratal
semiotic architecture of language. Neither term can be ‘actualized’ without
the other; but the distinction enables us to escape from the idealization that
is built into a philosophical modelling of language and to engage with the
significant object of our investigation, which is language in contexts of use.
This can take linguistics back to the mainstream of scientific thinking, to
share in the enterprise of a general theory of meaning and in exploring the
biological emergence of value.
A Note on the Co-Authorship of This Chapter
‘Was this the last article that Michael penned? Would this be the last that
he published?’ When the editors asked these two reasonable questions
following Michael Halliday’s death on 15 April 2018, I was concerned that
Michael’s career of assiduous attention to his own textual crafting of
articles might be diminished by my role as a co-author. Certainly, Michael
launched the chapter with typed pages of its plan and coverage of subject
matter; and, yes, for a number of years, Michael and I had discussed
linguistics as a science; the language of science; evolutionary theory; the
semiotic character of debate in physics; the relation between recent neuro-
science and the claims of cognitive science; and the connection between
Ruqaiya Hasan’s work in verbal art and his own work in what he called
‘verbal science’. Ultimately, however, as Michael neared the end of his life,
he was not sufficiently robust to complete the work as he would normally
do – as the keen-eyed writer and editor we knew him to be. The process
made me acutely aware of my own shortcomings as a co-author.
In fact, after many attempts to reduce and represent Michael’s articles on
‘grammatical metaphor’, I had a fresh sense of the profound nature of his
mode of thought and expression: each article added new angles and set out
abstract meanings with the concreteness of the great teacher that he was.
In the end, I opted to include passages from his work rather than mangle
his expository strategies.
By March 2016, Michael had completed the preface to the translation of
some of his papers into Spanish (published 2017). He also penned a series of
gnomic headings concerning the integration of his theory of metaphor with
Ruqaiya’s work, in particular, in relation to the development of higher-
order thinking in a culture. I will pass on to his readers what I can develop
of some challenging words and instructions that he left through his con-
versations, when I can do them justice. For now, as I reflect on Michael’s
lexicogrammatical choices, I see the fulfilment of a goal attributed by Luria
to Marx: the importance of ‘ascending to the concrete’. To end on an
instance, recall:
There can be no semiotic act that leaves the world exactly as it was before.
(Halliday 2002(1994): 2.254)
David G. Butt
References
Albright, D. 1997. Quantum Poetics: Yeats, Pound, Eliot, and the Science of
Modernism. Cambridge: Cambridge University Press.
Ananthaswamy, A. 2013. Quantum Shadows. New Scientist 217(2898):
36–9.
Aspect, A., P. Grangier, and G. Roger. 1982. Experimental Realization of

Einstein-Podolsky-Rosen-Bohm Gedanken Experiment: A New Violation
of Bell’s Inequalities. Physical Review Letters 49(2): 91–4.
Banks, D. 2008 The Development of Scientific Writing: Linguistic Features and
Historical Context. Sheffield: Equinox.
Bala, A. 2010. The Dialogue of Civilisations in the Birth of Modern Science. London:
Palgrave Macmillan.
Berkeley, G. 2002. The Analyst: A Discourse Addressed to an Infidel Mathemat-
ician. London: Tonson in the Strand.
Berwick, R. C. and N. Chomsky. 2016. Why Only Us: Language and Evolution.
Cambridge, MA: MIT Press.
Bohm, D. 1980. Wholeness and the Implicate Order. London: Routledge.
Bohr, N. 1961. Natural Philosophy and Human Culture. Address given in 1938,
reprinted in Atomic Physics and Human Knowledge. New York: Science
Editions.
Brooks, M. 2016a. How Weird Do You Want It? New Scientist: The Collection
3(3): 92–5.
Brooks, M. 2016b. Matter of Interpretation. New Scientist: The Collection 3(3):
16–19.
Butt, D. G. 2007. Thought Experiments in Verbal Art: Examples from
Modernism. In D. R. Miller and M. Turci, eds., Language and Verbal Art
Revisited: Linguistic Approaches to the Study of Literature. Sheffield: Equinox.
68–96.
Cameron, E. G. and J. L. Goldberg. 2016. Promoting CNS Repair: What
Influences Glial and Neuronal Response to Neurodegeneration? AAAS
353(6294): 30–1.
Cassidy, D. C. 1992. Uncertainty: The Life and Science of Werner Heisenberg.
New York: W. H. Freeman.
Chomsky, N. 1975. Reflections on Language. New York: Pantheon Books.
Chomsky, N. 2012. The Science of Language: Interviews with James McGilvray.
Crosby, A. W. 1996. The Measure of Reality: Quantification and Western Society
1250–1600. Cambridge: Cambridge University Press.
Darwin, C. 1998. The Origin of Species by Means of Natural Selection or the
Preservation of Favoured Races in the Struggle for Life. New York: The Modern
Library.
Davies, P. and N. H. Gregersen, eds. 2010. Information and the Nature of Reality:
From Physics to Metaphysics. Cambridge: Cambridge University Press.
Dawkins, R. and D. McKean, D. 2011. The Magic of Reality: How We Know
What’s Really True. London: Bantam Books.
Deacon, T. 2011. Incomplete Nature: How Mind Emerged from Matter. London:
W.W. Norton.
Defoe, D. 2001. A Journal of the Plague Year. London: Dover Publications.
Dijksterhuis, E. J. 1961. The Mechanization of the World Picture: Pythagoras to
Newton. Princeton: Princeton University Press.
Dixon, R. M. W. 2011. I Am A Linguist. Leiden: Brill.

Dolnick, E. 2011. The Clockwork Universe: Isaac Newton, the Royal Society, and the
Birth of the Modern World. New York: Harper Collins.
Dunn, M., S. J. Greenhill, S. C. Levinson, and R. D. Gray. 2011. Evolved
Structure of Language Shows Lineage-specific Trends in Word-order
Universals. Nature 473(7345): 79–82.
Eliot, G. 1871. Middlemarch: A Study of Provincial Life. Edinburgh and London:
William Blackwood and Sons.
Ellis, J. 1993. Language, Thought, and Logic. Evanston: Northwestern Univer-
sity Press.
Everett, D. 2013. Language: The Cultural Tool. London: Profile Books.
Favareau, D., ed. 2010. Essential Readings in Biosemiotics: Anthology and Com-
mentary. Dordrecht: Springer.
Press.
Firth, J. R. 1968. Linguistic Analysis as a Study of Meaning. In F. R.
Palmer, ed. Selected Papers of J. R. Firth 1952–59. London: Longman.
12–26.
Flynn, J. R. 2007. What is Intelligence? Cambridge: Cambridge University
Press.
Garcia, A. M. 2013. Relational Network Theory as a Bridge between Linguis-
tics and Neuroscience: An Interview with Professor Sydney Lamb. Linguis-
tics and the Human Sciences 8(1): 3–27.
Grant, E. 1996. The Foundations of Modern Science in the Middle Ages: Their Religious,
Institutional, and Intellectual Contexts. Cambridge: Cambridge University Press.
Grant, E. 2007. A History of Natural Philosophy: From the Ancient World to the
Nineteenth Century. Cambridge: Cambridge University Press.
Halliday, M. A. K. 2002. On the Ineffability of Grammatical Categories. In
J. J. Webster, ed., The Collected Works of M. A. K. Halliday, Volume 1: On
Grammar. London: Continuum. 291–322.
Halliday, M. A. K. 2003. Language and the Order of Nature. In J. J. Webster,
ed., The Collected Works of M. A. K. Halliday, Volume 3: On Language and
Halliday, M. A. K. 2004a. Language and Knowledge: The ‘Unpacking’ of
Text. In J. J. Webster, ed., The Collected Works of M. A. K. Halliday, Volume
5: The Language of Science. London: Continuum. 24–48.
Halliday, M. A. K. 2004b. On the Language of Physical Science. In J. J.
Webster, ed., The Collected Works of M. A. K. Halliday, Volume 5: The Language
of Science. London: Continuum. 140–58.
Halliday, M. A. K. 2004c. Things and Relations: Regrammaticizing Experience
as Technical Knowledge. In J. J. Webster, ed., The Collected Works of M. A.
K. Halliday, Volume 5: The Language of Science. London: Continuum. 49–101.
Halliday, M. A. K. 2005. Towards Probabilistic Interpretations. In J. J.
Webster, ed., Computational and Quantitative Studies. London: Continuum.
42–62.

sive Power. Pittsburgh: University of Pittsburgh Press.
Functional Grammar. 4th ed. Oxford: Oxford University Press.
Harris, R. 1988. Language, Saussure and Wittgenstein. London: Routledge.
Hasan, R. 1985. Linguistics, Language and Verbal Art. Geelong: Deakin Univer-
sity Press.
Hasan, R. 2011. The Collected Works of Ruqaiya Hasan, Volume 3: Language and
Education: Learning and Teaching in Society. Sheffield: Equinox.
Heisenberg, W. 1958. Physics and Philosophy: The Revolution in Modern Science.
New York: Harper.
Hoenigswald, H. M. and L. F. Wiener, eds. 1987. Biological Metaphor and
Cladistic Classification: An Interdisciplinary Perspective. Philadelphia: Univer-
sity of Pennsylvania Press.
Hoffmeyer, J. 2010. Semiotic Freedom: An Emerging Force. In P. Davies
and N. H. Gregersen, eds., Information and the Nature of Reality:
From Physics to Metaphysics. Cambridge: Cambridge University Press.
236–60.
Jakobson, R. 1973. Main Trends in the Science of Language. London: George
Allen and Unwin.
Jones, S. 2000. Almost like a Whale: The Origin of Species Updated. London:
Anchor.
Josephson, B. 2002. Beyond Quantum Theory: A Realist Psycho-biological
Interpretation of Reality Revisited. Biosystems 64: 43–5.
Kahn, C. 2009. Essays on Being. Oxford: Oxford University Press.
Kappagoda, A. 2004. Semiosis as the Sixth Sense: Theorising the Unperceived in
Ancient Greek. PhD Thesis, Macquarie University.
Kline, M. 1953. Mathematics in Western Culture. London: George Allen and
Unwin.
Korner, A. 2015. Analogical Fit: Dynamic Relatedness in the Psychotherapeutic
Setting (with Reference to Language, Autonomic Response, and Change in Self
State). PhD Thesis, Macquarie University.
Lamb, S. 2013. Systemic Networks, Relational Networks and Choice. In
L. Fontaine, T. Bartlett, and G. O’Grady, eds. Systemic Functional Linguistics:
Exploring Choice. Cambridge: Cambridge University Press. 137–60.
Layzer, D. Cosmogenesis: The Growth of Order in the Universe. Oxford: Oxford
University Press.
Leakey, R. 1994. The Origins of Humankind. London: Phoenix.
Lemke, J. L. 1990. Talking Science: Language, Learning, and Values. Norwood:
Ablex.
Lloyd, G. E. R. 1970. Early Greek Science: Thales to Aristotle. New York: W.W.
Norton.
Lloyd, G. E. R. 2002. The Ambitions of Curiosity: Understanding the World in

Ancient Greece and China. Cambridge: Cambridge University Press.
Lucretius. 1992. De Rerum Natura. Translated by W. H. D. Rouse, revised by
M. F. Smith. Cambridge, MA: Harvard University Press.
Mailer, N. 1968. Armies of the Night: History as a Novel/The Novel as History.
New York: New American Library.
Mann, W. 2000. The Discovery of Things: Aristotle’s Categories and Their Context.
Princeton: Princeton University Press.
Benjamins.
Matthiessen, C. M. I. M. 1995. Lexicogrammatical Cartography. Tokyo: Inter-
national Language Sciences Publishers.
Matthiessen, C. M. I. M. 2015. Halliday on Language. In J. J. Webster, ed.,
137–202.
Meares, R. 2012. A Dissociation Model of Borderline Personality Disorder.
New York: W.W. Norton.
Medawar, P. 1984. The Limits of Science. Oxford: Oxford University Press.
Michalske, T. A. and B. C. Bunker. 1987. The Fracturing of Glass. Scientific
American (December): 122–9.
Newton, I. 1998. Opticks: A Treatise of the Reflexions, Refractions, Inflexions and
Colours of Light: Also Two Treatises of the Species and Magnitude of Curvilinear
Figures. Commentary by N. Humez. Palo Alto: Octavo.
Ostler, N. 2005. Empires of the Word: A Language History of the World. London:
HarperCollins.
Ostler, N. 2007. Ad Infinitum: A Biography of Latin. London: HarperCollins.
Panksepp, A. and L. Biven. 2012. The Archaeology of Mind: Neuroevolutionary
Origins of Human Emotions. New York: W.W. Norton.
Porges, S. W. 2011. The Polyvagal Theory: Neurophysiological Foundations of
Emotions, Attachment, Communication, and Self-Regulation. New York: W.W.
Norton.
Salmon, V. 1979. The Study of Language in Seventeenth Century England. Amster-
Sapir, E. 1921. Language. New York: Harcourt, Brace and Company.
Saussure, F. 1959. Course in General Linguistics. Translated by W. Baskin.
London: Peter Owen Ltd.
Seung, S. 2012. Connectome: How the Brain’s Wiring Makes Us Who We Are.
New York: Houghton Mifflin Harcourt Publishing Company.
Seuren, P. 1998. Western Linguistics: An Historical Introduction. Oxford:
Blackwell Publishers.
Shapiro, J. A. 2011. Evolution: A View from the 21st Century. Upper Saddle
River: FT Press Science.
Snell, B. 1953. The Discovery of the Mind: The Greek Origins of European Thought.
Translated by T. G. Rosenmeyer. Cambridge, MA: Harvard University
Press.
Steiner, P. 1984. Russian Formalism: A Metapoetics. Ithaca: Cornell University

Press.
Sterne, L. 1759–1767. The Life and Opinions of Tristam Shandy, Gentleman. 9 vols.
York.
Strogatz, S 2003. Sync: The Emerging Science of Spontaneous Order. New York:
Hyperion.
Tolstoy, L. 1869. War and Peace. London: Heron Books.
Tynjanov, J. and R. Jakobson. 1978. Problems in the Study of Literature and
Language. In L. Matejka and K. Pomorska, eds., Readings in Russian Poetics:
Formalist and Structuralist Views. Cambridge, MA: MIT Press.
Uglow, J. 2003. The Lunar Men: The Friends Who Made the Future 1730–1810.
London: Faber & Faber.
Ungar, P. S. and L. J. Hlusko. 2016. The Evolutionary Path of Least
Resistance: Evolution Favored Teeth with Thicker Enamel over Sharply
Crested Teeth in Hominins Confronted with Tough Diets. AAAS
353(6294): 29–30.
Yallop, C. 1982. Australian Aboriginal Languages. London: André Deutsch.
von Humboldt, Wilhelm. 1988. The Diversity of Human Language Structure and
Its Influence on the Mental Development of Mankind. Cambridge: Cambridge
University Press.
Waddington, C. H. 1977. Tools for Thought. Frogmore: St Albans, Paladin.
Winograd, T. 1983. Language as a Cognitive Process, Volume 1: Syntax. Reading:
Addison-Wesley.
Lee Whorf. Edited by J. B. Carroll. Cambridge, MA: MIT Press.
25
Language and Medicine
Alison Rotha Moore
25.1 Introduction
From early in its development Systemic Functional Linguistics (SFL) has

drawn on and contributed to the study of medical discourse (e.g. Halliday
et al. 2007) and it was in the context of language and medicine in the
1970s that J. R. Martin first developed his work on discourse semantics
(Martin 2014). Since then there has been a steady stream of theses and
articles using SFL concepts and techniques to study the language of medi-
cine. Yet, to the best of my knowledge, this work has never been brought
together into a monograph, special issue, or edited collection, and
research in this area has not yet gained the profile of a recognized special-
ism within SFL in the way that fields of application such as education or
child language development have done. An important consideration here
is that research on medical discourse has lacked the coordination of other
major applications of SFL, which has affected the amount of work done,
the degree to which initial projects are followed up, and the impact of
such work and its visibility.
This chapter offers a profile of work on language and medicine
informed by SFL theory. It proceeds by outlining the health problems
and settings on which SFL-based studies have focused, reviewing the
theoretical and descriptive tools used, and considering how the role of
language in healthcare (and in health more broadly) has been conceptual-
ized. A research example is given, and the chapter critically reflects on
what has been achieved so far and the potential for this field to develop
into a more strategically coordinated application of Systemic Functional
Linguistics that makes a substantial contribution to improving health and
healthcare.
652 A L I SO N R OT H A M O O R E
25.2 Sites of Engagement and Points of Intervention
There are well over 100 publications and theses on language and medicine
that use SFL principles, covering many healthcare contexts and analytical
foci. A good way of grouping such publications is around their stated or
implied ‘intervention points’ – the physical or conceptual places within a
biological or institutional system where pressure can be applied to disrupt
existing function and promote change (Reinsborough and Canning 2010).
The present review starts outside the health system, with the everyday
construal of pain, then moves to the ‘core’ of the healthcare system, namely,
spoken interaction between clinicians and patients. We then move to con-
texts that support and/or shape this core, such as interpreting, and inter-
actions within clinical teams. Finally, we consider broader institutional and
cultural contexts in which healthcare is situated, suggesting there is
untapped potential for SFL here.
This notion of intervention is important for evaluating the impact of SFL
medical linguistics and considering where future efforts are best directed
because healthcare is a relatively weak determinant of health (see Figure 25.1),
perhaps as low as 15 per cent (McGinnis et al. 2002). Note that some projects
discussed involve multiple sites of engagement/intervention.
25.2.1 Everyday Construal of Experience

In foundational work on SFL, Halliday foreshadows medicine as an import-
ant domain for institutional linguistics and register (see Halliday et al.
Figure 25.1: The main determinants of health (after Dahlgren and Whitehead 2007)
Language and Medicine 653
2007), but it is not until 1998 that he contributes an extended discussion of

medical language. Using evidence from the first COBUILD corpus, a short
text, and arrays of typical spoken expressions, Halliday shows how pain is
multiply categorized – as a quality (sore tummy), a thing (tummyache), and
various kinds of process (my tummy aches, is giving me trouble) (Halliday
2005:306).
Halliday’s argument proceeds by comparing possible ways of represent-
ing pain in English with typical representations according to the corpus
(albeit of written texts), considering why I have a headache is more frequent
than My head aches/is aching/hurts. Interestingly, similar patterns are reported
across disparate languages including French, Russian, and Chinese: present-
ing the self (I/me) as clause Theme construes the whole person as the
‘setting out point’ for the experience of headache. Several follow-up cross-
linguistic studies on pain include Greek (Lascaratou 2007), Japanese (Hori
2006), German (Overlach 2008), and Italian (Bacchini 2012). More generally,
this kind of analysis, where varying ‘ways of saying’ are interrogated
lexicogrammatically, has been taken up by numerous scholars studying
clinical consultations.
On the question of an interventionist medical linguistics, Halliday
(2005:307) seems ambivalent: ‘Whether by analysing the grammar we
could in any way contribute to the practical alleviation and management
of pain I do not know.’ However, he also stresses that the ‘boundary
between the semiotic and the material worlds is by no means totally
impermeable’ (Halliday 2005:307), raising the idea that interlocutors might
helpfully reconstrue pain, as a form of ‘logotherapy’ (Halliday 2005:311).
This idea is central to the burgeoning narrative therapy/narrative medicine
movement (Charon 2007), including the way that patients’ construals are
taken up in clinical reasoning.
25.2.2 The Clinician–Patient Interface

Interaction between patients and clinicians is the area of language and
medicine most studied within SFL, with at least sixty publications since
the 1980s.
In this context – simplifying greatly – patients and clinicians must
communicate effectively in order to appropriately plan and implement
medical treatment and preventive measures. Most patient dissatisfaction
with clinicians concerns their communication and interpersonal skills, not
their medical knowledge and abilities (Slade et al. 2008). Poor communi-
cation between patients and clinicians leads to medical error and, far too
often, considerable patient distress (Vincent and Coulter 2002). Effective
communication and patient involvement in decision-making can improve
treatment decisions, treatment adherence, and patient health outcomes
(e.g. Kaplan et al. 1989), although it can also lead patients to ‘rational
non-compliance’ (Donovan and Blake 1992) or refusing recommended
treatment (Moore et al. 2001; Moore 2004). Some health outcomes may
stem directly from the therapeutic value of the communication itself in
physical conditions (Street et al. 2009) as well as in the context of
psychotherapy.
Probably the first use of SFL for extended analyses of clinician–patient
interaction was Mishler’s Discourse of Medicine (1984). Critiquing the then
mainstream quantitative methods of studying clinical interaction, Mishler
argues that they ignore the problems of transforming speech to written
transcripts as ‘data’, which tends to strip away meaning in a quest for
objectivity. Mishler, a social psychologist, uses all three metafunctions of
SFL but focuses in particular on cohesion, adapting Halliday and Hasan’s
(1976) approach to suit dialogue. He identifies clusters of structural, seman-
tic, and grammatical patterns that tend to be treated as routine and
unproblematic in medical interaction, which he identifies as ‘the voice of
medicine’. These contrast with and are typically used to interrupt ways of
speaking known as ‘the voice of the lifeworld’, which is dominated by
features such as temporal rather than causal organization. Mishler’s work,
which largely aimed to get clinicians to encourage not silence ‘the voice of
the lifeworld’, has had a strong influence on how clinical interaction is
studied and taught, in no small part due to Mishler’s position and influence
at Harvard Medical School.
Close on Mishler’s heels was Cassell’s Talking with Patients (1985). Cassell
stresses that clinicians need a solid grounding in the systems of language
underlying clinical communication before ‘communication skills’ are
taught, just as one would ‘never dream of teaching physical diagnosis to
students lacking a background in anatomy and pathology’ (Cassell 1985:5).
Having recorded hundreds of hours of consultations, Cassell exhorts clin-
icians to study their own dialogue with patients and learn to spot subtle
features, such as how ‘people shift to impersonal pronouns when they
describe their illnesses or unpleasant events’ (Cassell 1985:8) and how
patients attach meaning to symptoms and illnesses.
To my knowledge, the first description of generic structure for medical
consultations was given by Halliday (2005), comprising ‘opening’, ‘investi-
gation’, ‘examination’, ‘diagnosis’, and ‘suggested treatment’, which
includes ‘negotiation’ and ‘reassurance’ (a structure for his single example
text, not a Generic Structure Potential). Interestingly, the ‘treatment phase’
is seen as a typical manifestation of the complex power relationship
between professionals and clients, with its grammatical shifts in mood
and modality. Such an account can be linked with Halliday’s discussions
elsewhere of the relation between domains of activity and registerial
boundaries, or points on the cline of instantiation, since it emphasizes
similarities in the registerial settings of medical and other professional
discourse (see Moore 2017).
Following in this vein, much of the SFL literature on medicine and
language describes generic structure and/or explores patients’ construals
of their experience and values, and how these are taken up or not by
clinicians in consultation settings. This body of work includes descriptive
studies of specific sites (e.g. general practice, emergency medicine) and
conditions (e.g. HIV disease) and has generated interventions such as prac-
tical handbooks and other professional development material for clinicians.
25.2.2.1 General Practice

Most SFL studies of general practice focus on the tenor of consultations. For
instance, using four consultations, Thompson (1999:101) examines how GPs
‘act the part’ linguistically, balancing ‘superiority and humanity’ such as
through the prominent use of ‘declarative questions’ wherein the doctor
both gives information and seeks confirmation.
In Thompson’s study doctors used marked ellipsis more frequently than
patients. Thompson interprets this as a textual resource realizing not just
mode but also tenor, since it construes informality and familiarity by
evoking co-operation, but also hierarchy by virtue of speaker difference.
A rare example of multimodal analysis of clinical interaction is offered
by Thwaite (2015), drawing on a registerially varied video corpus designed
for TESOL contexts. Profiling one GP consultation, Thwaite shows, for
example, that the doctor speaks for 49 per cent of video time, whereas
the patient speaks for 21 per cent (the remainder is silence). While speak-
ing, the doctor looked directly at the patient (direct gaze) for 37 per cent of
the video time, whereas patient direct gaze lasted 16 per cent of the video
time. Patterns are compared with other registers studied (e.g. lawyer–
client interview).
An intonation analysis is also presented showing that the doctor uses all
five primary tones, whereas the patient uses no Tone 5. Since Tone 5 conveys
meanings such as ‘You may not realize this but it turns out to be the case’,
these differences arguably reflect participant roles in the context.1
Thwaite’s preliminary results indicate the potential of multimodal video
analysis for clinically relevant SFL research.
Three recent SFL-oriented studies try to address how empathy is realized
linguistically, and how empathic communication can be taught or sup-
ported. Pounds (2011) offers a model of the language resources available
for empathic expression, drawing on appraisal systems in English (Martin
and White 2005). His aim is to provide doctors with a flexible resource for
controlling their construal of empathy, rather than just a few key phrases
they can insert into consultations. His thesis does not however apply the
model to a corpus of texts.
Appraisal is also used in Watson’s empirical analysis of empathy (2012),
but this study also examines phonological features (particularly intonation)
that construe affiliation and bonding (Martin 2004). A key finding is that
1
For additional analysis of this data, see Halliday and Greaves (2008:80–94).
GPs and patients bond over the patient’s values rather than the doctor’s,
although other studies suggest that this might depend on the presenting
condition and length of clinical relationship (Moore 2004).
Patient-initiated humour is a further resource for building empathy
(Eggins 2014). Whereas clinical discourse can stigmatize patients or treat
them as ‘non-persons’ (Goffman 1963), patient-initiated humour encour-
ages clinicians to depart from the ‘professional’ script and use more inclu-
sive, egalitarian modes of everyday interaction. Eggins’ subjects were
hospital inpatients, but her findings probably extend to primary care set-
tings. Patient-initiated humour seems to differ in function from humour
between clinicians, which can promote solidarity or enforce existing hier-
archies (see Eggins and Slade 2015).
25.2.2.2 HIV Medicine

SFL-based studies of HIV discourse have contributed to a social research
response to HIV/AIDS in Australia. While still concerned with tenor, these
studies have also examined agentivity and technicality, and illuminated
relations between contextual patterning and linguistic patterning.
Moore et al. (2001) draw on transitivity, cohesion, and implicature pat-
terns to show how the technical term viral load is multiply coded – as a
biological property of the HIV-positive body, and as an indicator of treat-
ment effectiveness, patient compliance, and overall wellness. In practice, it
is the discursive alignment of patient and doctor regarding what such
codings index (biological, clinical, lifeworld) that determines how techni-
cality moves treatment forward (or not). Consultations between HIV
doctors and patients can look superficially like conversations between
clinicians, but patient expertise in clinical reasoning can be overestimated.
One recommendation is that the ability to recognize and flag discursive
shifts be considered a central component of doctors’ professional expertise.
In related research, Moore (2004, 2005, forthcoming) gives a multi-stratal
account of joint decision-making in HIV medicine, and critiques tools for
the semantic-level modelling of agency such as Hasan’s (1985) cline of
dynamism, van Leeuwen’s (1995) socio-semantic networks, and the use of
transitivity/ergativity as exhausting the textual analysis of agentivity.
Results show how important it is to model semiotic agency (action affecting
others through processes of sensing and saying), and that joint decision-
making is more likely to occur where doctors and patients construe each
other as semiotic agents, building on reciprocal expertise. Additional foci
include the therapeutic construal of identity, the linguistics of ‘compliance’
and its lack of fit with patient-centred medicine, and the analysis of com-
plex contexts (Candlin et al. 1998; McInnes et al. 2001).
25.2.2.3 Oncology
Several studies have used SFL to explore clinical interaction in oncology
settings, including breast cancer (Lobb et al. 2006; Moore and Butt 2004;
see example in Section 25.3), colorectal cancer (Jordens 2002), ovarian

cancer (Jordens et al. 2010), melanoma (Williams 2014), and oncological
palliative care (Karimi et al. 2018). Although interpersonal features are
included in these studies, they also often focus on ideational aspects such
as technicality (e.g. genes, mutation, equipoise) and construing agency.
Some of this research examines how oncologists present clinical trials as
treatment options (Brown et al. 2004) with results used in professional
development (e.g. Brown et al. 2007), influencing discursive approaches
beyond SFL (Brown 2014).
An important contribution (based on interviews not consultations) is
Jordens’ analysis of cancer illness narratives (Jordens et al. 2001; Jordens
2002), which interrogates complexity in healthcare discourse and its social
significance, using Martin and Plum’s (1997) narrative types. Against the
dominant view (Frank 1995), Jordens argues that patients with the greatest
life disruption have the most complex and, in some ways, the most tightly
organized – rather than chaotic – narratives (see Henderson-Brooks (2006a,
2006b) and Butt et al. (2010) on complexity in psychotherapeutic discourse).
Jordens et al.’s (2010) research has informed Cancer Australia’s policy on
post-treatment surveillance in ovarian cancer. They use Foucault’s notion of
the medical gaze and methods from Moore et al. (2001) to critique CA125
testing (a serum marker used to check for recurrence). Like viral load in
HIV, the various meanings of CA125 play out in ways that can undermine
shared decision-making, and increase women’s anxiety, without clear evi-
dence that testing improves survival.
Turning to melanoma, social stratification is the focus of current multi-
disciplinary research (Williams 2014). Departing from ‘health literacy’
explanations, Williams uses semantic variation (Hasan 2009a) to explore
why patient socioeconomic status (SES) influences treatment success for
melanoma, where incidence is greater in high-SES groups, but mortality is
greater in low-SES groups.
25.2.2.4 Emergency Medicine

Reporting on two collaborative projects on emergency medicine, Matthies-
sen (2013) gauges the potential for healthcare to become a major site of
application for Halliday’s ‘appliable linguistics’. Drawing on Hydén (1997)
on the clinical gaze, and Halliday’s (2002) orders of system complexity,
Matthiessen argues that it is no longer enough for patients to be seen as
persons: they must also be seen as ‘meaners’, located in networks of mean-
ers and negotiated meanings. Such a framing helps bring out the interactive
complexity of emergency medicine, especially in multilingual settings, and
seems crucial to evaluating how ‘patient-centred’ care works in practice
(see Karimi et al. 2018).
Whereas Matthiessen (2013) emphasizes work conducted in Hong Kong,
related research investigates emergency communication in Australia. Slade
et al. (2015b) is a collection of papers demonstrating that effective
communication is the best way of controlling ‘potential risk points’. One

key finding from these collaborations is that choices in thematic signpost-
ing (an aspect of the textual metafunction) can increase patient involve-
ment, thus increasing opportunities for assuring clinician–patient
alignment; by contrast, poor thematic signposting can increase the risk of
both communicative errors and medical errors (Herke et al. 2008).
25.2.2.5 End-of-Life Care

SFL work on palliative care has informed Australian national guidelines on
communicating about end-of-life issues (Clayton et al. 2007). Whereas
health communication literature typically advises using ‘open questions’,
in Moore (2015) doctors appeared to best facilitate discussion – without
forcing unwanted discussion – by using iterative sequences featuring cer-
tain choices from Hasan’s (2009a) Demand Information semantic network,
namely ask, verify, apprize, and probe questions, with the additional features
prefacing and assumptive strategically used at certain points. Another key
resource for eliciting discussion was graduated evaluation, particularly in
nominal groups, e.g. issues > concerns > worries > fears (Tuckwell and Moore
in preparation).
Driscoll (2012) uses transitivity analysis and Hasan’s (1985) cline of dyna-
mism to explore the ‘voice of medicine’ and the ‘voice of everyday life’
(after Mishler 1984) in patient interviews and advice websites about ter-
minal illness. Advice texts constructed patients as wanting information on
their illness, care and support, and certain living activities, whereas in the
interview data, what patients said they wanted included people, certain
qualities in their care (e.g. kindness), and – importantly – to avoid treat-
ment or certain treatments. Patients did not refer to wanting to discuss
their illness.
Karimi et al. (2018) use Hasan’s contextual system networks (e.g. Hasan
2009b) to show how the medical oncologist’s role in advanced cancer care is
multifaceted and complex, calibrating shifting roles against specific textual
properties. For example, as consultations move closer towards the end of a
patient’s life, the turn length and ‘semantic work’ of the patient appears to
increase and that of the oncologist decreases: these changes are explained
as contextual reconfigurations, including changes from specialized to quo-
tidian field, as the agentive role moves to therapist–client.
25.2.2.6 Nursing
Several studies discussed here use SFL tools to analyze nurse–patient inter-
action (e.g. Chandler et al. 2015; Eggins et al. 2016; Kealley et al. 2004;
Kealley 2007; Slade et al. 2015a; Wyer et al. 2017). Additionally, Candlin
(2000, 2002) uses Hasan’s cline of dynamism and coins the term ‘compre-
hensive coherence’ to describe how superficially ‘casual’ conversations
with patients constitute professional nursing expertise. Lassen and
Strunck (2011) show how nurses invoke a ‘positive’ discourse (Martin and
Rose 2003), disrupting national stereotypes that exclude ethnic minority

patients and frame them as an expensive burden on the Danish
health system.
25.2.2.7 Sexual and Reproductive Health

Video observation and interviews conducted in Family Planning (FP) clinics
(Slade et al. 2009; de Silva Joyce et al. 2015) suggest that this context
features a particularly high level of effective communication, including
strong congruence between messages given and received by both doctors
and clients,2 and high levels of satisfaction, although FP consultations are
admittedly much longer than average GP consultations.
One particularly interesting finding is that women’s reasons for
attending Family Planning involve both reduced and increased social dis-
tance – talking to female specialists about ‘female issues’ configures per-
haps surprisingly with the ‘anonymity’ clients feel they cannot get with
family GPs. In addition, family planning exemplifies delicate register vari-
ation: here socio-semantic processes of ‘sharing’ are frequent (Slade et al.
2009), possibly unlike medicine more generally (Matthiessen 2013).
25.2.2.8 Mental Health

There is a long tradition of SFL research into mental health discourse. Often
scholars are concerned not only with examining language as ‘a symptom
and a resource’ (Matthiessen 2013) for treating a specific illness/disorder,
but also with the extent to which mental health problems are etiologically
related to specific ways of interacting and meaning-making. Much of the
SFL work conducted on mental health has been integrated into clinical
practice and theory. The language used by schizophrenia patients is one
of the earliest SFL studies of any medical context (Rochester et al. 1977;
Rochester and Martin 1979; also see Asp and De Villiers, this volume).
Building on this work is an ongoing collaboration between SFL scholars
and authors and practitioners of the Conversational Model of psychother-
apy. Particular focus has been on Borderline Personality Disorder (BPD),
where patients ‘struggle in establishing a border between themselves and
significant others, which is itself fundamental to a deeper construal of their
own existence’ (Henderson-Brooks 2006a:1). Outputs of this collaboration
include several honours and PhD theses by linguists (e.g. Henderson-Brooks
2006a, 2006b; Khoo 2013) and by psychiatrists/psychotherapists (Korner
2015), plus research articles (e.g. Butt et al. 2010) and, importantly, contri-
butions to a clinical practice manual on Borderline Personality Disorder
(BPD), as well as to curricula for postgraduate degrees in psychotherapy
(e.g. Butt et al. 2012).
2
Women attending FP clinics are called ‘clients’, which is arguably consistent with values of patient/client autonomy and
feminism that inform sexual and reproductive healthcare.
Cohesion and cohesive harmony (Hasan 1984) have been key to this work,
with Butt et al. (2010) exploring the semantic fragmentation and fusion
that characterizes dissociative episodes among patients with BPD,
and tracking their possible resolution in therapeutic interaction. Butt’s
linguistic concepts of motivated selection, semantic drift, and instantial
weight have been deployed, and analogies between psychotherapy, verbal
art, and science are drawn (Butt et al. 2013). The linguistic notion of
cohesion has become a central metaphor in the Conversational Model’s
theory of BPD and how it arises (Meares et al. 2013).
Khoo (2013) takes a closer look at cohesion in her study of psychothera-
peutic discourse, including the relative merits of quantitative and qualita-
tive analyses, and the iconicity of cohesive harmony. She gives examples of
texts with poor numerical measures of cohesion that are judged more
therapeutically valuable than others with high scores (Khoo 2016).
Using a multi-stratal approach and fine-grained analyses of agency and
appraisal, Henderson-Brooks (2006a, 2006b) examines claims about three
conversation types observed in consultations with patients with BPD.
These linguistically distinguishable text types represent shifts between an
alienated or truncated self, construed through negative capacity, little
agency, and ineffectual verbal action (Chronicles and Scripts), and an
expanded self, construed through features such as real and hypothetical
action on others, positive mental action and verbal action (Narratives).
Other clinical concepts such as the contrastive ‘linear/non-linear speech’
are associated with logico-semantic complexity.
The appraisal system has also been used to explore the extent and nature
of depression in hospitalized patients via their discourse semantics. Using
interview data, Tebble (2012) concluded that familiarizing clinical
staff with key appraisal systems could help identify undiagnosed depres-
sion among inpatients, with the aim of improving their treatment experi-
ence, prognosis, and quality of life. Related research by Caldwell et al.
(2006) reports on appraisals of well-being among a non-depressed compari-
son group.
Korner (2015) draws on SFL and anthropomorphic measurement (heart
rate, skin conductivity, etc.) to examine ‘self’ and ‘person’ as the embodied
flux of feeling in a symbolic, acculturated personal context, or what he calls
a system of self and other in psychotherapeutic discourse.
‘Formulation’ (synopses of a patient’s presenting condition) is another
aspect of psychiatry studied using SFL tools. Formulations produced within
intrapsychic models have been found to be more highly nominalized
than those produced within intersubjective models, reducing the sense of
patient agency, and representing patients as being influenced by ‘unseen’
forces (Korner et al. 2010). Walsh et al. (2016a) address the need for teach-
ing the genre of formulation to mental health professionals. They examine
lexical relations, nominalization, and conjunctions, showing how clin-
icians’ talk shapes their developing understanding into a logical
formulation. In a related paper, Walsh et al. (2016b) consider how mental

health patients are represented differently in handovers, particularly in
terms of transitivity roles, being construed as objects and beneficiaries
rather than agents or actors (see Eggins et al. 2016). The authors conclude
that such representations need to change, just as handovers should include
patients as interlocutors (Walsh et al. 2016c).
There is a substantial body of SFL-informed research on couples counsel-
ling (e.g. Muntigl 2004, 2006; Muntigl et al. 2013). In one of the few SFL-
based health discourse monographs, Muntigl (2004) reports client change
over six sessions of narrative therapy. In Muntigl’s analysis, clients initially
produce recounts (after Martin and Plum 1997) as a means of problem
identification, then get scaffolded by therapists into a more expository
mode, which foregrounds causal relations and mental projection. Like
Moore’s (2004) HIV study, the construal of semiotic agency is particularly
salient here (X has got you thinking/put that in your head, etc.). Finally, in the
‘developed semiotic repertoire’, clients return to narrative mode but now
include complication and resolution and dispersed evaluation. Muntigl
(2006) explores the concept of ‘macrogenre’ using counselling data, and
Muntigl et al. (2013) identifies resources through which therapists and
clients achieve affiliation.
Although research on spoken language is dominant here, the written
linguistic correlates of mental states have also been examined. Nagar and
Fine (2013) report that subjects with current depression used more elabor-
ation, more extension, and less enhancement than previously or never
depressed subjects in a free writing task. Severity of current and lifetime
depression was associated with the extent of this preference, which Nagar
and Fine interpret in terms of impaired concentration. An alternative
explanation might see these patterns as semantically motivated – constru-
ing a factive, unchanging world, on the one hand, or a world of cause and
effect and different subjective perspectives, on the other (see Henderson-
Brooks 2006a).
25.2.3 Mediating the Clinician–Patient Interface

Work outlined in this section aims to transform processes or objects that
mediate how doctors and patients interact – such as medicines information
leaflets, question prompt sheets, and using interpreters.
25.2.3.1 Medical Interpreting

Studies of healthcare interpreting constitute one of the largest bodies of SFL
work on language and medicine, much of it by Tebble and her students
(including Tebble 1993, 1996a, 1996b, 1998, 1999, 2003, 2008, 2012, 2014;
Hirsh 2001; Caldwell et al. 2006; Willis 2001). This work has informed
curricula for interpreters (Tebble 1996b) and for training physicians who
work with them (Tebble 1998, 2003).
Drawing on Hasan (1996) and Sinclair and Coulthard (1975), Tebble

provides a ranked scale of discourse structures for spoken discourse. At
the ‘top’ level, generic structure potential (GSP) of interpreted professional
consultations3 is given as follows (Tebble 2008:152):
GreetingsÎntroductions^(Contract)^Stating/Eliciting ProblemÂscertaining
Facts^(Diagnosing Facts)^Stating Resolution/Exposition^(Decision by Client)
^Clarifying Residual Matters^Conclusion^Farewell
Such a model allows interpreters to map their ‘location’ and pace their
energy around the critical parts of the consultation.
Interpersonal meanings are a focus in interpreting, and Tebble (1999)
theorizes the teaching of interpreters to ‘read’ the tenor of physicians’
consulting styles, using ‘Exposition’ moves in two specialisms. Appraisal
has been deployed (Willis 2001; Hirsh 2001) including studies of depressed
patients (e.g. Tebble 2012). A German study (Bührig 2004) highlights the
textual function, showing how a doctor and an untrained interpreter used
different ‘linguistic action patterns’ for obtaining informed consent (see
also Torsello 1997).
25.2.3.2 Written Information for Patients and Carers

Written information – often still presented to patients on paper – plays an
important role in mediating face-to-face clinical communication. Clerehan
(2014) offers an excellent overview which points out that, for all its dynamic
complexity, the patient’s story in consultations remains their own, but
written material testifies to ‘the commonness of the disease experience,
implying appropriation and “generification” of their story by the doctor’
(Clerehan 2014:212). Patient engagement with such material remains a
complex and under-researched phenomenon. Very little research has
involved linguistic methods or considered culturally and linguistically
diverse groups, with developers and researchers relying largely on readabil-
ity scores and ‘industry standard’ checklists that often correlate poorly with
patient-reported effectiveness (Clerehan 2014).
Alternative approaches include Clerehan and Buchbinder’s (2006) analy-
sis of eighteen patient information leaflets. Their Evaluative Linguistic
Framework (ELF) (see also Clerehan et al. 2005; Hirsh et al. 2009) has been
used to compare medicine labelling in two countries (Connor et al. 2008),
and to study decision aids, patient package inserts, and consent forms in
Australia, Denmark, and Norway (e.g. Askehave and Zethsen 2003).
Similarly, Moore and colleagues (Moore and Wegener 2010; Aslani et al.
2010) find that Consumer Medicines Information leaflets (CMIs) have
unusual and uncomfortable combinations of field, tenor, and mode, argu-
ably inconsistent with shared decision-making. Recommendations include
equipping writers with concepts around context and its textual realization,
3
Tebble extends her model to ‘dialogue interpreting’ including legal and bureacratic contexts.
so they can control tenor as field varies within texts, using an authoritarian
tone when necessary (Do not take this medicine if you are pregnant), but not
across all sections. Compared with CMIs from other English-speaking
nations, Australian CMIs showed disrupted cohesive harmony (Moore and
Wegener 2010; Moore 2010). This is partly because writers assemble CMIs
from pre-written paragraphs, but the implicit goals of such texts constitute
another factor; these goals include protecting drug companies (who develop
such documents) from legal harm. While their findings informed docu-
ment redesign (Aslani et al. 2010), one aspect considered too controversial
to include in the national recommendations was the observation that
patients/readers interpreted statements of drug purpose (e.g. Lipitor is used
in people with high cholesterol who have high blood pressure and coronary heart
disease or are at risk of a stroke . . .) as statements of likely benefit, thus
overestimating the chance of personal health benefit (Moore 2010).
Written information can also be about healthcare processes. Kealley et al.
(2004) examine a pamphlet aimed at empowering patients and relatives in a
critical care unit to be active in the healthcare process. Contrary to its
purpose, the pamphlet depicts staff as retaining great authority in a way
that restricted relatives’ actions and interactions, thus reinforcing passive
and compliant behaviour among relatives and patients. This study is one of
the few that does not assume the neutrality of ‘information’ for patients
and relatives and uses linguistic concepts to explore its value.
Eckkrammer (2004) examines medical self-counselling texts and hyper-
texts, showing that layers of intersemiosis were already present in late
fifteenth-century texts, and discussing the specific affordances of hypertext
for this register. One finding of interest is that most diagrams in self-
counselling texts had an illustrative function only.
25.2.3.3 Decision Aids

An increasingly used mediation of clinician–patient dialogue is the patient
decision aid – a multimodal discourse technology for supporting shared
decision-making around treatment and testing. See Section 25.3 for an
extended example from genetic counselling for breast cancer (Lobb et al.
2006).
In the context of colorectal cancer screening, Smith et al. (2008) draw on
Clerehan’s work to tailor a decision aid for low-literacy patients. Although
both high- and low-literacy groups preferred the revised design, the low-
literacy participants felt the information was not directive enough and
appeared unfamiliar with metadiscourse around informed choice.
Decision aids now include online interfaces that map individual patient
characteristics onto large data sets to customize prognosis and treatment
recommendations, such as deciding about chemotherapy for cancer (e.g.
Predict n.d.). Many are designed for clinicians but are used in consultations
with patients or by patients alone. One of their effects is to widely expand
the degree and types of distributed agency in the consulting room. Little
linguistic research has been done on this, but see Bloor (2016) on potentially
misleading construals of terminal illness in online prognostic information.
25.2.3.4 Social Media

Using a ‘big data’ approach, McDonald (2016) and McDonald and
Woodward-Kron (2016) have explored interaction in an online mental
health support group (approximately 6,000 members generating eight mil-
lion words of text). They show how users comply with the discursive norms
of the target group over time, how this compliance is realized through
mood and transitivity choices, and how such transformation is likely to
be therapeutic itself and support better clinical interaction and outcomes.
These observations resonate with Moore (2004) on viral load and identity in
HIV, and Fleischman (1999) on identity in obscure conditions. McDonald’s
additional findings include a shift from seeking information to providing
social support as users gain experience in the group, during which ‘socio-
semiotic processes’ move from Matthiessen’s (2013) ‘sharing and reporting’
to ‘expounding and recommending’ (see Bowcher, this volume).
25.2.4 Healthcare as System and Institution

The small but growing amount of research linking clinical communication
to ‘hard’ health outcomes confirms the importance of detailed descriptions
of consultations and non-clinical interactions around health topics, using
functional models of language such as SFL. However, it is important that
such research does more than simply ‘tweak at the margins’ of practices
and systems.
Iedema (2006, 2007) points out that medicine has increasingly become
accountable to other professions such as nursing, administration, and IT
specialists, as well as to healthcare ‘consumers’ and their caregivers. This
means that medical discourse should not be studied in vacuo. Iedema criti-
cizes the separation of studies of doctor–patient interaction, on the one
hand, and of medical documentation (and policies), on the other, arguing
that the ‘medical dependencies’ that shape clinical interaction need to be
treated as part of the discourse analysis ‘proper’. In other words, profes-
sional expertise and judgement is not left to an individual’s understanding
of best practice – as a nurse, doctor, or other clinician – but is governed by
institutional agency (Candlin and Candlin 2002; also see Sarangi and
Roberts 1999).
25.2.4.1 Adverse Outcomes

Increasingly, health discourse research has been concerned with reducing
medical error and promoting patient and staff safety, often with substantial
impact on practice.
One intervention with the potential to substantially reduce adverse out-
comes is to improve clinical handovers (ACSQHC 2010) – where patients are
moved symbolically and/or physically between nodes of responsibility in

the health system. Health systems worldwide have made considerable
efforts here, largely through standardized protocols such as the iSBAR tool,
but scant improvement has occurred (Slade and Eggins 2015). One explan-
ation is that protocols such as iSBAR fail to treat communication as inher-
ently interactive (Eggins and Slade 2012).
While some authors suggest standardization itself may be the problem
(Patterson 2007; also see Butt et al. 2016), Slade and colleagues argue that
standardizing can be helpful (McGregor et al. 2011), and it is the attempt to
describe a staged, dialogic genre without the appropriate theoretical under-
standing of language that is the problem. This, sadly, is a recurring theme
across different areas of medical discourse: despite more than fifty years of
ethnomethodology and sociolinguistics of health, tools for communication
in healthcare rarely draw on appropriate resources to model genre and are
often still caught in a ‘representational bias’ (see Moore 2004). Nevertheless,
in the handover context, communication training via functional linguistic
models has been able to produce behavioural change (Slade and Eggins
2015:198).
A recent volume (Eggins et al. 2016) based on 829 audio and video
recordings in Australian hospitals includes examination of bedside nurs-
ing handovers (Eggins and Slade 2016), emergency department shift hand-
overs (McGregor and Lee 2016), inter-hospital transfers (Geddes et al.
2016), and mental health handovers (Walsh et al. 2016b, 2016c) among
others, along with instructional resources for health professionals.
A further topic explored by this team is the role of humour in handovers
and in healthcare more generally (Eggins and Slade 2012; Eggins 2014).
Handovers in the multilingual context of nursing in Pakistan, complicated
by various factors including low levels of literacy in English (which is the
language of hospital records), are the subject of new SFL-related research
(Mahboob 2017).
When errors or near misses do occur, the way that healthcare systems
respond is crucial: work by Iedema and colleagues has had substantial
impact here. Their research for the Australian Commission on Safety and
Quality in Health Care (Iedema et al. 2008) was instrumental in achieving
ministerial endorsement of a national system for critical incident reporting,
helping to change the culture around medical error in Australia from one
where clinicians were advised not to talk to patients or families when
things went wrong, to one of much greater transparency and reflexivity,
although healthcare staff still need to ‘learn to be sorry on an organiza-
tional basis’ (Iedema et al. 2009:266). Related work documenting patients’
experience of adverse events has helped convince health governance
bodies that patients want and need explanations from clinicians when
things go wrong (Iedema et al. 2011), and also that patient experience can
itself provide crucial missing information about adverse events (Walton
et al. 2017).
Iedema’s research group has also transformed handovers from ambu-

lance officers to emergency intake nurses (Iedema et al. 2012), and has
recommended new genres of family conferencing within critical and end-
of-life care (Sorensen and Iedema 2006). Current work addresses infection
control (Wyer et al. 2017). The group’s innovations in video ethnography
are well recognized among healthcare researchers. For example, Wyer et al.
(2017) have nurses view video footage of their interactions with patients,
along with footage of patients analyzing the videos of their infection-risks,
giving affective views of infection control systems. This is a good example of
research that does not treat clinical interaction as ‘quarantined’ from insti-
tutional policy.
25.2.4.2 Clinical Teams

A small number of studies using SFL have examined clinical interactions
that do not involve patients as interlocutors or readers. Santiago et al.
(2011) studied Medical Emergency Teams (MET) in Australian hospitals –
itinerant teams of clinicians who provide emergency care and high-risk
patient identification outside the walls of ICUs. Drawing on Hasan’s (1996)
‘generic structure potential’, this research shows that there is substantial
variation in the nature of MET interactions and activities in different
hospitals.
Routine interaction between members of surgical teams has been the
focus of a major Australian study. Through this project, linguistics and
semiotics have contributed a way of understanding surgery – and health-
care more generally – as a highly complex ‘realizational system’ (Butt
2008; Butt et al. 2016). Analysis of language, gaze, and body alignment
patterns in surgical interaction has supported arguments for a registe-
rially sensitive approach to proxemics using SFL principles (Moore 2006)
and for a ‘language’ of surgery (Cartmill and Butt 2016). Body alignment
between senior and trainee surgeons has been shown to contribute
crucially to the construal and negotiation of agency in the surgical
process, and to the phasing and layering of professional and pedagogical
activities (Moore et al. 2010; Moore 2016). The study also explores
the operationalization of Halliday’s distinct notion of register, critically
engaging with Hasan’s system of Message Semantics (see Moore’s case
study in Lukin et al. 2011; also see Moore 2016). Analyses show how
senior surgeons use subtle variations in command type to control the
phasing of surgery and to accomplish critical moments such as ‘swap-
ping sides’ with their trainees. Recommendations include the need to
make such interpersonal and registerial competence a component of
professional expertise. Under the leadership of the surgeon researcher
on the project, this approach has become a cornerstone of the surgical
training programme established as part of a new medical school and
hospital at Macquarie University.
25.2.4.3 Medical Informatics and Computational Linguistics

A number of studies aim to characterize medical registers, including syn-
chronic and diachronic variation in medical language. Although not
designed to inform activities to improve health or health care, they never-
theless help understand the medical/health context and how underlying
societal changes may be linked to changes in priorities and practices in
health. These include Martinez-Insua and Perez-Guerra’s (2015) study of
Theme patterns in early Modern English medical texts, using a two-million-
word corpus from 1500–1700, showing variation in theme type as a func-
tion of tenor, and Zinn and McDonald’s (2016) corpus-based transitivity and
mood analysis of 1.9 million articles from the New York Times between
1964 and 2014, which found a growing incidence of meanings around risk
in health journalism accompanied by increasing reference to scientific
expertise and increased individualism.
Van Moll and O’Donnell (2004) have demonstrated computer recognition
of generic structure in medical discourse, using medical discharge notices
(MDNs) as their primary example genre, which they consider a subtype of
business letter. Their work provides an interesting example of the interde-
pendence of different genres within a single professional domain.
25.2.4.4 Material Settings in Healthcare

Innovative work on space and medicine, based on SFL principles, has been
conducted by Stenglin and Foureur (2013). Stenglin’s scale of the degree of
boundedness that architectural spaces create was used to understand the
types of birth spaces that help women feel safe and secure. Their study
offers ways of ‘designing out fear’ to increase the likelihood of normal
birth in the context of worryingly high rates of Caesarian delivery in
Western settings.
25.2.4.5 Medical Education, Training, and Research

Scholars using SFL tools have made substantial contributions to medical
education, including the challenges arising from culturally and linguistic-
ally diverse patient and health worker groups.
For example, focusing on register, Woodward-Kron et al. (2014) have
developed curricula and award-winning multimedia resources around com-
munication skills for International Medical Graduates (IMGs). Woodward-
Kron (2016) shows how IMGs partially deploy the discursive patterns associ-
ated with patient-centred communication that are expected of Australian
medical trainees; Pryor and Woodward-Kron (2014) examine IMGs’ tele-
phone consultations with senior doctors; and Woodward-Kron and Elder
(2016) discuss language testing for internationally trained clinicians.
Elsewhere, this group reveals that, among culturally and linguistically
diverse (CALD) groups, consenting to a clinical trial means ‘family consent’,
so ethics committees must allow novel discourse processes if CALD research
participation is to improve (Woodward-Kron et al. 2016). Woodward-Kron’s
group has published widely on healthcare communication, including a

systematic review of communication skills training outcomes, which they
argue are limited by a close focus on behavioural outcomes at the expense
of understanding language and communication (Denniston et al. 2017).
Public health postgraduate education has been studied by Lander and
colleagues. For example, Lander (2014) evaluates asynchronous online dis-
cussions and concludes that ambiguity in tenor roles contributes to student
dissatisfaction with such a course component – a finding that is generaliz-
able to tertiary online discussions as a whole. Lander et al. (2010) discuss the
benefits of online delivery for teaching clinical safety.
Analyses of student writing include studies of medical students’ know-
ledge and reasoning in examination texts (Fraser and Gannon 2003) and of
the degree of reflection in students’ essays about dissecting their first
cadaver (Chan and Shum 2011). Analysis of spoken classroom discourse
includes Chang’s (2017) study of textual and multimodal aspects of English
Medium Instruction in Chinese Medical Sciences classes.
Several SFL-informed studies have refined our understanding of the med-
ical research article as a genre, identifying ten to fifteen rhetorical moves
and some of the lexicogrammatical features that constitute them, and
delineating obligatory from mandatory moves (Fryer 2012; Nwogu 1990;
Nwogu and Bloor 1991). Nwogu in particular focuses on thematic progres-
sion and cohesion. Because English is used extensively for publishing med-
ical research, these studies orient themselves to explicit instruction in
research writing for medical graduates whose first language is not English,
while other work analyzes medical research articles for finer insights into
register variation itself (e.g. Biber and Finegan 1994).
A sobering finding in the medical education literature is that, as medical
students pass through their degrees, they generally lose faith in social
approaches to medicine that include sensitivity to patients’ life contexts –
and male students show greater dismissal of social approaches than female
students (Woloschuk et al. 2004). SFL-based research could help explain this
problem, including how pedagogic interactions erode students’ initial
values.
25.2.4.6 Shaping Illness, Treatment, and Prevention:

Culture, Environment, Social Marketing
SFL research is now appearing on cultures and practices outside healthcare
delivery that have a strong influence on health, including preventive
medicine.
In the UK, Brookes and Harvey (2015) critique the fear-inducing, com-
mercially funded public health campaigns that raise public awareness of
adult-onset diabetes but fail to address factors such as the price and avail-
ability of unhealthy food, and work cultures that make exercising difficult.
Elsewhere, these authors critically compare multimodal discourses promot-
ing breastfeeding and baby formula (Brookes et al. 2016), an area also
addressed by Sheehan and Bowcher (2017). Prevention is the focus of

Körner et al.’s (2004) transitivity analysis of how HIV+ interview respond-
ents constructed their sexual partners rather than themselves as agentive
with respect to the ‘unsafe sex’ which caused their infections.
Harvey and colleagues also critique commercialization and commodifica-
tion in the National Health Service, medicalization in UK/Western society, and
the pharmaceutical industry’s role in these processes (e.g. Harvey 2013),
resonating with Australian work on direct-to-consumer advertising of pre-
scription pharmaceuticals (Mackenzie et al. 2007) and consumer medicines
information (Aslani et al. 2010; Moore 2010). Teenagers’ use of the Internet
for information on sexual health/identity (Harvey et al. 2007), depression
(Harvey 2012), and anorexia (Mullany et al. 2015) has also been studied.
Other SFL research which interrogates the cultural emergence and trans-
formation of medicine includes Kappagoda’s (2004) study of the co-
evolution of science and medicine with grammatical and semantic
resources in ancient Greek, and of systems of meaning in contemporary
epidemiology (Moore and Grossman 2003) and evidence-based medicine
(Moore 2007). Additionally, Körner and Treloar (2006) examine representa-
tions of people with HIV and Hepatitis C in medical journal editorials.
This section suggests there is scope for SFL to target issues that account
for a larger component of the ‘burden of illness’ than clinical communi-
cation. One underexplored area, following Sontag (1978) and Fleischmann
(1999), is the metaphors that drive and divide health discourses and the
policies that flow from these, e.g. ‘war on drugs’ discourses that obscure
how some industry research arguably aims to increase addiction in the
community (Neil 2017). Other opportunities include the work of lay carers
and the relation between health policies and clinical interaction.
Structural issues too important to ignore include the persistent health
gaps between privileged and less privileged groups, and the mechanisms for
these on which SFL has had very little to say. We might also critically
examine discourses through which we explain such patterns as ‘equity
gaps’, rather than perhaps ‘exploitation inherent in our social structures’
(Hage 2017), as well as the visual semantics through which health deter-
minants are modelled, as in Figure 25.1.
25.3 Exemplifying SFL Analysis of Language and Medicine
Before concluding this chapter I briefly illustrate one SFL approach to

language and medicine from a collaboration with clinicians on genetic
counselling for women with a family history of breast cancer (Butt 2006;
Lobb et al. 2006; Moore and Butt 2004). The study combines contextual,
semantic, and multimodal analysis (text and image relations) with recep-
tion studies, and connects analysis to contestable intellectual and political
agendas in medicine.
This collaboration began with a conundrum: research had shown that

women from families with a high risk of breast cancer persistently over-
estimate risk: linguistic consultants were invited to analyze twenty tran-
scribed consultations to help explain why risk was so misunderstood, even
after specialist counselling (Lobb et al. 2006).
The study produced a map of the typical discourse strategies used in
breast cancer counselling, including semantic networks that set out how
critical choices in generic phases were realized (after Hasan 1996).
A three-phase broad generic structure potential was derived from the
twenty consultations:
Who are you? ^ What is a gene? ^ [What forecast? (^Go back and talk)]
In Phase 1 the counsellor establishes the patient’s values, information

thresholds, anxiety, and general risk category from family history and
previous tests. In Phase 2 the counsellor explains the scientific basis of risk
by setting out dependencies between the main concepts, including gene,
chromosome, DNA, replication, mutation, purpose, and penetrance, along with
statistical interpretations. Phase 3 is a consideration of how these general-
ities apply to ‘you’ in terms of risk, knowledge, and possible action. Phase 4,
an optional phase which depends on the answers in Phase 3, sends the
presenting woman back to their mother, sister, daughter, etc., to learn
more about the family and consider having a test.
This illustration focuses on Phase 2, coined the ‘Genes Talkfest’ by clin-
icians. A range of productive interpersonal, textual, and experiential strat-
egies were used in this challenging discursive context. Yet, although
counsellors typically gave accurate statistics when explicitly estimating risk,
they also produced ‘latent patterning’ (Butt 1988) that, arguably, over-
emphasized genetic causality and individual cancer risk (Moore and Butt
2004; Lobb et al. 2006). The healthcare literature is clear that risk perception
is an emotional issue as well as a rational one, but very little literature
acknowledges that implicit meanings also operate in the rational (experien-
tial) domain, shaping participants’ understanding of risk in an inappropriate
way.4 Consider the following excerpt from Consultation 82, Turn 161.
Counsellor: When we talk about an inherited tendency to breast cancer,

what we’re talking about really are genes that are passed through the
families. And in some cases seem to cause breast cancer (Patient: mm). So a
gene, well the way I describe a gene is a bit like an instruction to the body
and we’ve got thousands of these instructions that determine hair colour
and eye colour and they influence our height and our weight. And we’ve also
got a set of genes or instructions that are involved in preventing cancer from
happening. (Patient: mm) So what they normally do is act a bit like a brake
in the body. And we know that some of these genes are involved in prevent-
ing breast cancer from occurring. They don’t completely prevent it
4
‘Framing effects’ for risk statistics (Tversky and Kahneman 1981) are, however, well known within health
communication research.
completely but um they certainly make a difference to the chance that

somebody gets breast cancer. [turn continues]
Five key things characterize this excerpt, each indexing a dimension of

meaningful choice at the level of context of situation, arguably at Butt’s
(2004) level of ‘move’; see Figure 25.2.
• The rationale for this move is posed by the counsellor rather than the
patient5 (which we could perhaps gloss as you need to understand ‘inherited
tendency’ in genetic terms);
• The move comes as a long monologic bloc (punctuated by patient
backchannelling);
• The conceptual sequence of the move focuses on biological processes
rather than, say, social or statistical groupings;
• The move draws heavily on naturalizing metaphors (instructions and
brakes, whose functions are commonly known) but does not de-automatize
or limit these metaphors; importantly brake failure in a car invariably
generates some observable problem, preferably a gentle roll but often a
fatal crash, whereas carrying BrCa1 or BrCa2 mutations only leads to
breast cancer in approximately 40 to 85 per cent of women, and to early
death only in a proportion of those, so the analogy is problematic;
• Semantic tendencies in the spoken move are reinforced by the images
used (see below).
At the level of abstraction below ‘Move’ (see Figure 25.3 – roughly the
stratum/rank of Butt’s (2000) ‘Argument’, Halliday and Matthiessen’s
(1999) ‘Sequence’, and Cloran’s (1994) ‘Rhetorical Unit’), the most crucial
features are as follows.
• The ‘rational strategy’ of the explanation begins with analogy, indicating

what a gene is ‘like’; in greater delicacy this can be specified as an example
of Wittgenstein’s ‘family resemblance’ – that is, similarities that are not
the result of a common factor (brakes and genes and instructions);
• One alternative rational strategy, which appears in the latter arguments
about hair colour and genes, is to use an explicit ‘x causes y’ strategy;
• Other possible strategies are not taken up, such as instantiation or
taxonomy: utterances such as a gene is a bunch of chemicals the body uses
in making proteins and other molecules are not found.
Taken together, these choices of context- and semantic-level patterning in

the Genes Talkfest stress the agentivity of genes in causing and preventing
cancer and draw attention away from environmental and non-genetic her-
editary aspects of cancer and its prevention – meanings that could ‘hose
down’ the scare factor of genes. Such implicit semantic patterns appear to
strongly affect risk perception despite explicit caveats.
5
The woman in this consult is not strictly speaking a patient, but this term is used for clarity rather than ‘woman’ or
‘woman from a high breast cancer risk family’, etc.
Network I: Key options in the Genes Talkfest Move (Phase 2 in Breast Cancer Genetic Counselling GSP)
David Butt and Alison Moore, CLSL, Macquarie University 2001
Rationale for what is coming:
Why you need to know/hear this Posed by Patient
Framing Posed by Doctor/Counsellor

Response to question
Precipitated by explicit reference to risk statistic
Precipitated by causal chain in 'process'

Presented as a background interlude
Unprompted, rhetorical question
Organised around Q & A
Monologic Checks by P
Seeks evidence of understanding thru feedback
Explanation Organised as single bloc Punctuated Checks by D/C
(but not by Q &A) Formality only
Checks by both
Fortuitously Degrees of technicality demanded
Organised in stages/steps (from anywhere to anywhere) e.g. I'll start with something you're probably familiar with
By topic By a taxonomy or dependency of concepts

Predetermined e.g. You need to understand this before I can explain that
By likelihood (probability; unusualness)
Social elements (family, community, "you" "us")
Conceptual Taxonomic
Sequence Biological elements (cells, genes, etc.)
Biological process
Naturalise analogy only (e.g., a gene is a little bit like a brake; lightning doesn't strike twice)
Pincer De-automatize - make strange (e.g., actually lightning can sometimes strike twice)
Both naturalise and de-automatize
Modality Speech only (single mode) Pedagogically focused (including modifiable)

Adjuncts to speech only
Multimodal presentation (e.g., website follow-up) Ancillary and not focused on the specifics of patients
Other modes in the nucleus of explanation
Figure 25.2 Network of key contextual options in Phase 2 of genetic counselling (after Moore and Butt 2004)
Figure 25.3 Network fragment: key semantic options in ‘rational strategy’ within Phase 2 of genetic counselling (after Moore and Butt 2004)
Figure 25.4a Original image used in decision aid for genetic counselling (after Lobb
et al. 2002)
Figure 25.4b Revised image used in decision aid for genetic counselling (after Lobb
et al. 2002)
A final point concerns text–image intersemiosis. Images used in decision

aids for genetic counselling also adopted a rational strategy that arguably
overemphasizes the link between genes and cancer in a highly implicit way.
Figure 25.4a shows a widely used diagram in which the x-axis conflated
time and cancer risk. A highly plausible interpretation of this image is that
if a woman has an inherited faulty gene, she will inevitably get breast
cancer, since breast cancer is, graphically, the only outcome. While it
may technically be true that anyone who lives long enough will get some
form of cancer, this diagram is misleading and was amended to include
alternative endpoints of ‘no cancer’ (see Figure 25.4b).
25.4 Concluding Comments
As this chapter indicates, SFL has generated a significant body of research

on language and medicine. Areas where impact has been reported include
mental health services, cancer care, HIV, emergency, surgery, handovers
in hospital departments, critical incident reporting, written medicines
information, and health curricula, confirming that most SFL research on
health focuses on spoken communication. This chapter also confirms
that interpersonal meanings have been given most attention. Four iden-
tifiable impacts on practice stand out, namely, the introduction of critical
incident reporting in all Australian states based partly on Iedema et al.
(2008); the withdrawal of CA125 testing in ovarian cancer surveillance in
New South Wales, Australia, following Jordens et al. (2010) and other
research; the National Accreditation Authority for Translators and Inter-
preters in Australia (NAATI) curriculum based on Tebble’s cumulative
work (especially 1996b); and Woodward-Kron’s influence on curricula for
International Medical Graduates in Australia (e.g. Woodward-Kron et al.
2011, 2014).
Following on from Halliday et al.’s (2007) imagining of SFL-based
logotherapy, applying a functional linguistic lens to healthcare is
helping to show how and why some forms of dialogue may be thera-
peutic, and also helping to identify those particular patterns of speak-
ing that produce developmental and restorative effects (e.g. Butt et al.
2012). As a test-bed for SFL, research on language and medicine has
sharpened our understanding of linguistic complexity, including struc-
tural complexity within and between genres, semantic counterpoint
and its ensemble effects, and complex contexts, fleshing out some
pictures and contradicting some received views. Hospitals and other
medical settings have proven excellent examples of complex realiza-
tional systems.
Given that SFL is a theory that grounds meaning in social structure and
culture, there has been surprisingly little attention paid to so-called ‘struc-
tural’ barriers to health and health equality, though this appears to be
changing. However, the health impact of structural and preventive health
measures themselves can be hard to gauge, let alone the effect of discourse
research on such measures, so it will remain a challenge for health linguists
to defend such work in a ‘research impact’ era.
Importantly, SFL’s capacity for explaining and addressing community
problems in terms of class, consciousness, and code – where relevant –
has so far been underutilized but offers potential. Increased attention here
would link healthcare findings to SFL work on other registers and policy
issues. SFL may provide more productive explanations of differences in
morbidity and mortality between SES groups, differences that have too
readily been couched in terms of health literacy, but may really be about
coding orientation and shared cultural capital.
Compared with other fields of application in SFL, healthcare research has

been relatively loosely organized. The sense of a ‘shared problem’ is not as
strong as in, say, educational semiotics or the study of children’s language
development. This may be related to differences in tenor relations between
medicine and linguistics: for medical practitioners, linguistics is an
unlikely technical authority/resource, whereas it can more easily serve this
role in other contexts such as education. The inclusion of a chapter on
medicine and linguistics in the present handbook is a good start towards
building stronger networks around SFL and health discourse.
Although this chapter started out as a discussion of SFL work on
language and medicine, it should now be clear why I prefer a broader
focus on health and promote the notion of SFL health linguistics as a
coherent field of application. It is hoped that this apparent growth area
will strengthen its capacity to provide timely and actionable results to
those involved in health policy and healthcare delivery, and will con-
tinue to extend its reach to domains outside healthcare that impact on
health and well-being. An important strategy for achieving these goals is
for health linguistics to deepen its critical engagement with the theoret-
ical concepts and descriptive systems of SFL, as well as with other
linguistic and social-theoretical approaches generating fruitful work in
this field.
References
ACSQHC. 2010. OSSIE Guide to Clinical Handover Improvement. Sydney: Austra-

lian Commission for Safety and Quality in Health Care. Available online
at: www.safetyandquality.gov.au/our-work/clinical-communications/clin
ical-handover/ossie-guide. (Last accessed 26/09/2017.)
Askehave, I. and K. K. Zethsen. 2003. Communication Barriers in Public
Discourse: The Patient Package Insert. Document Design 4: 22–41.
Aslani, P., K. Hamrosi, E. Feletto, P. Knapp, B. Parkinson, J. Hughes, L.
Nissen, A. R. Moore, and D. K. Raynor. 2010. Investigating Consumer Medi-
cine Information (CMI) Report: Report to the Pharmacy Guild of Australia and
Department of Health and Ageing. Available online at: http://6cpa.com.au/
resources/fourth-agreement/investigating-consumer-medicine-information-
i-cmi-project. (Last accessed 26/09/2017.)
Bacchini, S. 2012. Telling Pain: A Study of the Linguistic Encoding of the Experi-
ences of Chronic Pain and Illness through the Lexicogrammar of Italian.
PhD Thesis, Queen Mary University of London.
Biber, D. and E. Finegan. 1994. Intra-textual Variation within Medical
Research Articles. In P. de Haan and N. Oostdijk, eds., Corpus-based
Research into Language. Amsterdam: Rodopi. 201–22.
Bloor, M. 2016. The Construal of Terminal Illness in Online Medical
Texts: Social Distance and Semantic Space. In S. Gardner and S. Alsop,
eds., Systemic Functional Linguistics in the Digital Age. Sheffield: Equinox.

120–33.
Brookes, G. and K. Harvey. 2015. Peddling a Semiotics of Fear: A Critical
Examination of Scare Tactics and Commercial Strategies in Public Health
Promotion. Social Semiotics 25(1): 57–80.
Brookes, G., K. Harvey, and L. Mullany. 2016. ‘Off to the Best Start’?
A Multimodal Critique of Breast and Formula Feeding Health Promo-
tional Discourse. Gender and Language 10(3): 340–63.
Brown, R. 2014. Physician–Patient Communication about Cancer Clinical
Trials. In H. Hamilton and S. Chou, eds., Routledge Handbook of Language
and Healthcare Communication. London: Routledge. 615–28.
Brown, R., P. Butow, D. Butt, A. R. Moore, and M. Tattersall. 2004.
Developing Ethical Strategies to Assist Oncologists in Seeking
Informed Consent to Cancer Clinical Trials. Social Science and Medicine
58: 379–90.
Brown, R., P. Butow, F. Boyle, and M. Tattersall. 2007. Seeking Informed
Consent to Cancer Clinical Trials: Evaluating the Efficacy of Doctor Com-
munication Skills Training. Psycho-Oncology 16: 507–16.
Bührig, K. 2004. On the Multimodality of Interpreting in Medical
Briefings for Informed Consent: Using Diagrams to Impart Knowledge.
In E. Ventola, C. Charles, and M. Kaltenbacher, eds., Perspectives on Multi-
modality. Amsterdam: John Benjamins. 227–41.
Butt, D. G. 1988. Randomness, Order and the Latent Patterning of Text. In
D. Birch and M. O’Toole, eds., Functions of Style. London: Pinter. 74–97.
Butt, D. G. 2000. Semantic Cycles: Structure Statements at the Level of
Meaning. In D. G. Butt and C. M. I. M. Matthiessen, eds., The Meaning
Potential of Language: Mapping Meaning Systemically. Unpublished mimeo.
Sydney: Department of Linguistics, Macquarie University. 228–39.
Butt, D. G. 2004. Parameters of Context: On Establishing the Similarities and
Differences between Social Processes. Unpublished mimeo. Sydney: Macquarie
University.
Butt, D. G. 2006. Multimodal Representations of Remote Past and Projected
Futures: From Human Origins to Genetic Counselling. In M.-C. Amano,
ed., Multimodality: Towards the Most Efficient Communications by Humans.
Nagoya: Graduate School of Letters, Nagoya University.
Butt, D. G. 2008 The Robustness of Realizational Systems. In J. J. Webster,
ed., Meaning in Context: Implementing Intelligent Applications of Language
Studies. Norfolk: Continuum. 59–83.
Butt, D. G., A. R. Moore, C. Henderson-Brooks, J. Haliburn, and R. Meares.
2010. Dissociation, Relatedness, and ‘Cohesive Harmony’: A Linguistic
Measure of Degrees of ‘Fragmentation’? Linguistics and the Human Sciences
3: 263–93.
Butt, D. G., A. R. Moore, and C. Henderson-Brooks. 2012. Discourse Correl-
ates of the Therapeutic Method and Patient Progress. In R. Meares,
N. Bendit, J. Haliburn, A. Korner, D. Mears, and D. G. Butt, eds., Borderline
Personality Disorder and the Conversational Model: A Clinician’s Manual.

New York: W. W. Norton. 267–90.
Butt, D. G., A. R. Moore, and K. Tuckwell. 2013. The Teleological Illusion
in Linguistic ‘Drift’: Choice and Purpose in Semantic Evolution. In
L. Fontaine, T. Bartlett, and G. O’Grady, eds., Choice: Critical Considerations
in Systemic Functional Linguistics. Cambridge: Cambridge University Press.
101–40.
Butt, D. G., A. R. Moore, and J. Cartmill. 2016. Transactions between
Matter and Meaning: Surgical Contexts and Symbolic Action. In S. White
and J. Cartmill, eds., Communication in Surgical Practice. Sheffield: Equinox.
181–205.
Caldwell, D., H. Tebble, and D. M. Clarke. 2006. The Language of Subjective
Wellbeing. In J. Blore, A. Gluskie, and Y. MacKay, eds., Proceedings of the
7th Australian Annual Conference on the Quality of Life. Melbourne: Deakin
University.
Candlin, C. N. and S. Candlin. 2002. Introduction: Discourse, Expertise and
the Management of Risk in Health Care Settings. Research on Language and
Social Interaction 35(2): 115–37.
Candlin, C. N., A. R. Moore, and G. Plum. 1998. From Compliance to
Concordance: Shifting Discourses in HIV Medicine. Paper presented at
the International Pragmatics Conference, Rheims, 19–24 July, 1998.
Candlin, S. 2000. New Dynamics in the Nurse–Patient Relationship? In
S. Sarangi and M. Coulthard, eds., Discourse and Social Life. Harlow: Pearson
Educational. 230–45.
Candlin, S. 2002. Taking Risks: A Measure of Expertise? Research on Language
and Social Interaction 35: 173–93.
Cartmill, J. and D. G. Butt. 2016. Towards a Language of Operative Surgery.
In S. White and J. Cartmill, eds., Communication in Surgical Practice.
Cassell, E. J. 1985. Talking with Patients. Cambridge, MA: MIT Press.
Chan, L. and M. Shum. 2011. Analysis of Students’ Reflective Essays
on Their First Human Dissection Experience. Korean Journal of Medical
Education 23(3): 209–19.
Chandler, E., D. Slade, J. Pun, G. Lock, C. M. I. M. Matthiessen, E. Espindola,
and C. Ng. 2015. Communication in Hong Kong Accident and Emergency
Departments: The Clinicians’ Perspectives. Global Qualitative Nursing
Journal 2: 1–11.
Chang, C. 2017. English-medium Instruction in a Medical School: Managing
Classroom Discourse. In J. Zhao and L. Dixon, eds., English-medium Instruc-
tion in Chinese Universities: Perspectives, Discourse and Evaluation. London:
Charon, R. 2007. Narrative Medicine: Honoring the Stories of Illness. Oxford:
Oxford University Press.
Clayton, J., K. M. Hancock, P. N. Butow, M. H. Tattersall, D. C. Currow, et al.
2007. Clinical Practice Guidelines for Communicating Prognosis and
End-of-life Issues with Adults in the Advanced Stages of a Life-limiting

Illness, and Their Caregivers. MJA 186: 76–108.
Clerehan, R. 2014. Quality and Usefulness of Written Communication for
Patients. In H. Hamilton and W.-Y. Chou, eds., The Routledge Handbook of
Language and Health Communication. London: Routledge. 212–27.
Clerehan, R. and R. Buchbinder. 2006. Toward a More Valid Account
of Functional Text Quality: The Case of the Patient Information Leaflet.
Text and Talk 26: 39–68.
Clerehan, R., R. Buchbinder, and J. Moodie. 2005. A Linguistic Framework
for Assessing the Quality of Written Patient Information: Its Use in
Assessing Methotrexate Information for Rheumatoid Arthritis. Health
Education Research 20: 334–44.
Relations of Context, Meaning and Grammar. Nottingham: Department of
English Studies, University of Nottingham.
Connor, U., M. F. Ruiz-Garrido, W. Rozycki, E. Goering, E. D. Kinney, and J. M.
Koehler. 2008 Patient-directed Medicine Labeling: Text Differences between
the United States and Spain. Communication and Medicine 5: 117–32.
Dahlgren, G. and M. Whitehead. 2007. European Strategies for Tackling Social
Inequalities in Health. Copenhagen: WHO Regional Office for Europe.
Denniston, C., E. Molloy, D. Nestel, R. Woodward-Kron, and J. Keating.
2017. Learning Outcomes for Communication Skills across the Health
Professions: A Systematic Literature Review and Qualitative Synthesis.
BMJ Open 7: e014570.
de Silva Joyce, H., D. Slade, D. Bateson, H. Scheeres, J. McGregor, and
E. Weisberg. 2015. Patient-centred Discourse in Sexual and Reproductive
Health Consultations. Discourse and Communication 9: 275–92.
Donovan, J. and D. Blake. 1992. Patient Non-compliance: Deviance or
Reasoned Decision-making? Social Science and Medicine 34(5): 507–13.
Driscoll, J. 2012. The Representation of Terminally Ill Cancer Patients:
A Transitivity Analysis of Advice and Interview Texts. PhD Thesis, University
of Liverpool.
Eckkrammer, E. M. 2004. Drawing on Theories of Inter-semiotic Layering to
Analyse Multimodality in Medical Self-Counselling Texts and Hypertexts.
In E. Ventola, C. Charles, and M. Kaltenbacher, eds., Perspectives on Multi-
modality. Amsterdam: John Benjamins. 221–6.
Eggins, S. 2014. Hospital Humor: Patient-initiated Humor as Resistance to
Clinical Discourse. In E. Stracke, ed., Intersections: Applied Linguistics as a
Meeting Place. Newcastle: Cambridge Scholars. 43–66.
Eggins, S. and D. Slade. 2012. Clinical Handover as an Interactive Event:
Informational and Interactional Communication Strategies in Effective
Shift-Change Handovers. Communication and Medicine 9(3): 215–27.
Eggins, S. and D. Slade. 2015. Communication in Clinical Handover:
Improving the Safety and Quality of the Patient Experience. Journal of
Public Health Research 4(3): 197–9.
Eggins, S. and D. Slade. 2016. Communication in Bedside Nursing Hand-

overs. In S. Eggins, D. Slade, and F. Geddes, eds., Effective Communication in
Clinical Handover: From Research to Practice. Berlin: Walter de Gruyter.
97–114.
Eggins, S., N. Cominos, and J. Walsh. 2016. Disciplinary Discourses:
Contrasting Representations of the Patient in Medical and Mental Health
Handover Interactions. In H. de Silva Joyce, ed., Language at Work: Analys-
ing Language Use in Work, Education, Medical and Museum Contexts. Newcas-
tle: Cambridge Scholars. 192–213.
Fleischman, S. 1999. ‘I Am . . ., I Have . . ., I Suffer from . . .’: A Linguist
Reflects on the Language of Illness and Disease. Journal of Medical Human-
ities and Cultural Studies 20: 3–32.
Frank, A. 1995. The Wounded Storyteller. Chicago: University of Chicago Press.
Fraser, H. and C. Gannon. 2003. Collaborative Learning: Using Functional
Analysis of Texts Written by Peers: Helping Medical Students to
Pass Examinations of Knowledge and Reasoning. CIEFL Bulletin 13(2):
71–101.
Fryer, D. L. 2012. Analysis of the Generic Discourse Features of the English-
language Medical Research Article. Functions of Language 19(1): 5–37.
Geddes, F., P. Della, E. Stewart-Wynne, and D. Jones. 2016. Interhospital
Transfer of Rural Patients: An Audit of ‘Patient Expect’ Documentation.
In S. Eggins, D. Slade, and F. Geddes, eds., Effective Communication in
Clinical Handover: From Research to Practice. Berlin: Walter de Gruyter. 157–96.
Goffman, E. 1963. Stigma: Notes on the Management of Spoiled Identity. Engle-
wood Cliffs: Prentice Hall.
Hage, G. 2017. On the Management of the Border between Necro- and
Bio-politics. Provocations Seminar, School of Humanities and Social
Inquiry, University of Wollongong. 13 April 2017.
Halliday, M. A. K. 2002. On Grammar and Grammatics. In J. J. Webster, ed.,
The Collected Works of M. A. K. Halliday, Volume 1: On Grammar. London:
Halliday, M. A. K. 2005. The Grammar of Pain. In J. J. Webster, ed.,
The Collected Works of M. A. K. Halliday, Volume 5: The Language of Science.
Halliday, M. A. K. and W. Greaves. 2008. Intonation in the Grammar of English.
Halliday, M. A. K. and C. M. I. M. Matthiessen. 1999. Construing Experience as
Meaning: A Language-based Approach to Cognition. London: Cassell.
Halliday, M. A. K., A. McIntosh, and P. Stevens. 2007. The Users and Uses of
Language. In J. J. Webster, ed., The Collected Works of M. A. K. Halliday,
Volume 10: Language and Society. London: Continuum. 5‒37.
Harvey, K. 2012. Disclosures of Depression: Using Corpus Linguistics
Methods to Interrogate Young People’s Online Health Concerns. Inter-
national Journal of Corpus Linguistics 17: 349–79.
Harvey, K. 2013. Medicalisation, Pharmaceutical Promotion and the Inter-

net: A Critical Multimodal Discourse Analysis of Hair Loss Websites.
Social Semiotics 23: 691–714.
Harvey, K. J., B. Brown, P. Crawford, A. Macfarlane, and A. McPherson.
2007. ‘Am I Normal?’: Teenagers, Sexual Health and the Internet. Social
Science and Medicine 65: 771–81.
ation. 181–219.
Hasan, R. 1985. Linguistics, Language and Verbal Art. Geelong: Deakin Univer-
sity Press.
Hasan, R. 1996. The Nursery Tale as Genre. In C. Cloran, D. G. Butt, and
G. William, eds., Ways of Saying, Ways of Meaning: Selected Papers of Ruqaiya
Hasan. London: Cassell. 51–72.
Hasan, R. 2009a. Semantic Variation: Meaning in Society and in Sociolinguistics.
Edited by J. J. Webster. Sheffield: Equinox.
Hasan, R. 2009b. The Place of Context in a Systemic Functional Model. In
Henderson-Brooks, C. 2006a. What Type of Person am I, Tess?: The Complex Tale
of Self in Psychotherapy. PhD Thesis, Macquarie University.
Henderson-Brooks, C. 2006b. ‘Words Being Its Marker’: A Linguistic Study
of Self as Shifting State in Three Types of Psychotherapeutic Conversa-
tion. In E. Swain, ed., Thresholds and Potentialities of Systemic Functional
Linguistics: Multilingual, Multimodal and Other Specialised Discourses. Gorizia:
University of Trieste. 229–67.
Herke, M., C. Matthiessen, J. McGregor, M. Manidis, H. Scheeres, D. Slade,
J. Stein-Parbury, and R. Iedema. 2008. Patient Safety: A Tri-stratal Inter-
pretation of Communicative Risk in the Emergency Departments of
Public Hospitals. In C. Wu et al., eds., Proceedings of the 35th International
Systemic Functional Linguistics Congress: Voices around the World. Sydney:
Macquarie University. 140–5.
Hirsh, D. 2001. Interpersonal Features of Talk in Interpreted Medical Consultations.
MA Thesis, Deakin University.
Hirsh, D., R. Clerehan, M. Staples, R. Osborne, and R. Buchbinder. 2009.
Patient Assessment of Medication Information Leaflets and Validation of
the Evaluative Linguistic Framework (ELF). Patient Education and Counseling
77: 248–54.
Hori, M. 2006. Pain Expressions in Japanese. In G. Thompson and S. Hunston,
eds., System and Corpus: Exploring Connections. Sheffield: Equinox. 206–25.
Hydén, L.-C. 1997. Illness and Narrative. Sociology of Health and Illness 19:
48–69.
Iedema, R. 2006. Medicine and Health: Intra- and Inter-professional Com-
munication. In K. Brown, ed., Encyclopaedia of Language and Linguistics.
Oxford: Elsevier. 745–51.
Iedema, R., ed. 2007. The Discourses of Hospital Communication: Tracing Complex-
ities in Contemporary Health Organisations. Basingstoke: Palgrave Macmillan.
Iedema, R., N. Mallock, R. Sorensen, E. Manias, A. Tuckett, A. Williams,
et al. 2008. Final Report: Evaluation of the National Open Disclosure Pilot
Program (Report for the Australian Commission on Safety and Quality in Health
Care). Sydney: University of Technology, Sydney.
Iedema, R., C. Jorm, J. Wakefield, C. Ryan, and S. Dunn. 2009. Practising
Open Disclosure: Clinical Communication and Systems Improvement.
Sociology of Health and Illness 31(2): 262–77.
Iedema, R., S. Allen, K. Britton, D. Piper, A. Baker, C. Grbich, A. Allan,
L. Jones, A. Tuckett, A. Williams, E. Manias, and T. Gallagher. 2011.
Patients’ and Family Members’ Views on How Clinicians Enact and How
They Should Enact Incident Disclosure: The ‘100 Patient Stories’ Qualita-
tive Study. BMJ 343: d4423.
Iedema, R., C. Ball, B. Daly, J. Young, T. Green, P. Middleton, C. Foster-Curry,
M. Jones, S. Hoy, and D. Comerford. 2012. Design and Evaluation of a New
Ambulance-to-Emergency Department Handover Protocol: ‘IMIST-
AMBO’. BMJ Quality and Safety 21(8): 627–33.
Jordens, C. 2002. Reading Spoken Stories for Values: A Discursive Study of Cancer
Survivors and Their Professional Carers. PhD Thesis, University of Sydney.
Jordens, C., J. Little, K. Paul, and E. Sayers. 2001. Life Disruption and Generic
Complexity: A Social Linguistic Analysis of Narratives of Cancer Illness.
Social Science and Medicine 53(9): 1227–36.
Jordens, C., B. Morrell, P. Harnett, et al. 2010. Cancergazing? CA125 and
Post-treatment Surveillance in Advanced Ovarian Cancer. Social Science
and Medicine 71(9): 1548–56.
Kaplan, S. H., S. Greenfield, and J. E. Ware. 1989. Assessing the Effects
of Physician–Patient Interactions on the Outcome of Chronic Disease.
Medical Care 27: 110–27.
Kappagoda, A. 2004. Semiosis as the Sixth Sense: Theorising the Unperceived in
Ancient Greek. PhD Thesis, Macquarie University.
Karimi, N., A. R. Moore, A. Lukin, A. Walczak, and P. Butow. 2018. Life
and Death: Consultations between Oncologists and Patients with
Advanced, Incurable Cancer. In A. Sellami-Baklouti and L. Fontaine,
eds., Perspectives from Systemic Functional Linguistics. London: Routledge.
315–37.
Kealley, D. J. 2007. ‘I Can’t Find a Pulse But That’s OK’: Nursing in Context:
A Systemic Functional Linguistic Examination of Nursing Practice. PhD Thesis,
University of South Australia.
Kealley, D. J., C. Smith, and B. Winser. 2004. Information Empowers
But Who Is Empowered? Communication and Medicine 1(2): 119–29.
Khoo, K. 2013. Cohesive Harmony: Exploring a Measure of Cohesion and Coherence
in Psychotherapy. BA Honours Thesis, Macquarie University.
Khoo, K. 2016. ‘Threads of Continuity’ and Interaction: Coherence, Texture
and Cohesive Harmony. In W. Bowcher and J. Liang, eds., Society in
Language, Language in Society: Essays in Honour of Ruqaiya Hasan. London:

Palgrave. 300–30.
Korner, A. 2015. Analogical Fit: Dynamic Relatedness in the Psychotherapeutic
Setting (with Reference to Language, Autonomic Response, and Change in Self-
state). PhD Thesis, Macquarie University.
Korner, A., N. Bendit, U. Ptok, K. Tuckwell, and D. G. Butt. 2010. Formula-
tion, Conversation and Therapeutic Engagement. Australasian Psychiatry
18(3): 214–20.
Körner, H. and C. Treloar. 2006. Representations of People with HIV and
Hepatitis C in Editorials of Medical Journals: Discourses and Interdiscur-
sive Relations. Communication and Medicine 3(1): 15–25.
Körner, H., O. Hendry, and S. Kippax. 2004. ‘I Didn’t Think I Was at Risk’:
Interdiscursive Relations in Narratives of Sexual Practices and Exposure
to HIV. Communication and Medicine 1: 131–43.
Lander, J. 2014. Conversations or Virtual IREs? Unpacking Asynchronous
Online Discussions Using Exchange Structure Analysis. Linguistics and
Lander, J., S. Reid, M. Walton, and G. Rubin. 2010. Learning about Quality
and Safety in an On-line Learning Environment. Focus on Health Professional
Lascaratou, C. 2007. The Language of Pain: Expression or Description? Amster-
Lassen, I. and J. Strunck. 2011. ‘I Think Danish Patients Would Feel the
Same’: Counter-discourses Emerging in the Danish Health Sector.
Communication and Medicine 8(3): 223–33.
Lobb, E., P. Butow, A. Barratt, D. G. Butt, T. Dudding, A. Ingrey, J. Kirk,
A. R. Moore, and K. Tucker. 2002. Discussing Cancer Risk with High
Risk Breast Cancer Families: A Communication Tool for Genetic Counselling.
Medical Psychology Unit, University of Sydney/Centre for Genetic Edu-
cation, NSW.
Lobb, E., P. Butow, A. R. Moore, A. Barratt, K. Tucker, C. Gaff, J. Kirk,
T. Dudding, and D. G. Butt. 2006. Development of a Communication
Aid to Facilitate Risk Communication in Consultations with Women
from High Risk Breast Cancer Families. Journal of Genetic Counselling 15:
393–405.
4(2): 187–243.
Mackenzie, F., A. R. Moore, and K. Tuckwell. 2007. Deconstructing DTCA:
Towards a Differentiated Policy Response to Direct-to-Consumer Adver-
tising in Australia. VELIM Annual Seminar, Centre for Values, Ethics and
the Law in Medicine, University of Sydney, 16 August 2007.
Mahboob, A. 2017. The NurD Project: Documenting and Teaching the
Genre of ‘Receiving/Admission Notes’. Research Seminars in TESOL and
Language Studies, University of Sydney, 15 May 2017.
Martin, J. R. 2004. Mourning: How We Get Aligned. Discourse and Society

15(2–3): 321–44.
Clause. Functional Linguistics 1: 1–24.
Martin, J. R. and G. Plum. 1997. Construing Experience: Some Story Genres.
Journal of Narrative and Life History 7(1–4): 299–308.
Martin, J. R. and P. White. 2005. The Language of Evaluation: Appraisal in
Martinez-Insua, A.-E. and J. Perez-Guerra. 2015. On Contentful/Light
Themes in English Historical Medical Texts: ‘As Shall Serue to Giue
Content to the Vnderstander’. Paper presented at the 42nd Inter-
national Systemic Functional Congress, Aachen, Germany, September
2015.
Matthiessen, C M. I. M. 2013. Applying Systemic Functional Linguistics in
Healthcare Contexts. Text and Talk 33(4–5): 437.
McDonald, D. 2016. Linguistic Change in an Online Support Group. PhD Thesis,
University of Melbourne.
McDonald, D. and R. Woodward-Kron. 2016. Member Roles and Identities
in Online Support Groups: Perspectives from Corpus and Systemic
Functional Linguistics. Discourse and Communication 10: 157–75.
McGinnis, J. M., P. Williams-Russo, and J. R. Knickman. 2002. The Case for
More Active Policy Attention to Health Promotion. Health Affairs 21(2):
78–93.
McGregor, J. and M. Lee. 2016. Strengthening Medical Handover Communi-
cation in Emergency Departments. In S. Eggins, D. Slade, and F. Geddes,
eds., Effective Communication in Clinical Handover: From Research to Practice.
Berlin: Walter de Gruyter. 69–90.
McGregor, J., M. Lee, D. Slade, and R. Dunston. 2011. Effective Clinical Hand-
over Communication: Improving Patient Safety, Experiences and Outcomes.
Sydney: University of Technology Sydney.
McInnes, D., K. Race, V. Kleinert, M. McMurchie, and M. Kidd. 2001.
Compliance Supportive Communication: Understanding Interactive
Sequence and Knowledge/Power Relations in HIV Treatment Negoti-
ations. In K. Race, ed., Adherence and Communication: Reports from a
Study of HIV General Practice. Sydney: National Centre in HIV Social
Research.
Meares, R., N. Bendit, J. Haliburn, A. Korner, D. Mears, and D. G. Butt, eds.
2013. Borderline Personality Disorder and the Conversational Model: A Clinician’s
Manual. New York: W. W. Norton.
Mishler, E. 1984. The Discourse of Medicine: Dialectics of Medical Interviews.
Norwood: Ablex.
Moore, A. R. 2004. The Discursive Construction of Treatment Decisions in the
Management of HIV Disease. PhD Thesis, Macquarie University.
Moore, A. R. 2005. Modelling Agency in HIV Decision-making. Australian

Review of Applied Linguistics 19: 103–22.
Moore, A. R. 2006. Interpersonal Engagement and Bodily Alignment in
Surgery. Paper presented at the European Systemic Functional Workshop
and Conference, Gorizia, Italy, 20–22 July 2006.
Moore, A. R. 2007. EBM™: Nominalising Evidence in Medicine. Workshop
on Evidence, The ARC Biohumanities Project, University of Queensland,
8–9 January 2007.
Moore, A. R. 2010. Text Analysis Meets the Reception Study: Using SFL to
Interrogate, Redesign and Test Health Communication. Paper presented
at the 37th International Systemic Functional Congress, University of
British Columbia, Vancouver, 19–23 July 2010.
Moore, A. R. 2015. Can Semantic Networks Capture Intra- and Inter-
Registerial Variation? Palliative Care Discourse Interrogates Hasan’s
Message Semantics. In W. Bowcher and J. Liang, eds., Society in Language,
Language in Society: Essays in Honour of Ruqaiya Hasan. London: Palgrave.
83–114.
Moore, A. R. 2016. Lovers, Wrestlers, Surgeons: A Contextually Sensitive
Approach to Modelling Body Alignment and Interpersonal Engagement
in Surgical Teams. In S. White and J. Cartmill, eds., Communication in
Surgical Practice. Sheffield: Equinox. 257–85.
Moore, A. R. 2017. Register Analysis and Message Semantics. In T. Bartlett
and G. O’Grady, eds., Routledge Handbook of Systemic Functional Linguistics.
Moore, A. R. forthcoming. Pills, Life, Agency: HIV Treatment Decisions as
Language in Social Context. Sheffield: Equinox.
Moore, A. R. and D. G. Butt. 2004. Risk and Meaning Potential: Why a
Network? International Systemic Functional Linguistics Conference,
Kyoto, 30 August–4 September 2004.
Moore, A. R. and J. Grossman. 2003. Exploring the Meaning of ‘Assignments
of Meaning in Epidemiology’. In I. Kerridge, C. Jordens, and E.-J. Sayers,
eds., Restoring Humane Values to Medicine: A Miles Little Reader. Annandale:
Desert Pea Press. 99–102.
Moore, A. R. and R. Wegener. 2010. Final Report on Functional Linguistic
Analysis and Recommendations for the I-CMI Project. May 2010.
Moore, A. R., C. N. Candlin, and G. A. Plum. 2001. Making Sense of HIV-related
Viral Load: One Expert or Two? Culture Health and Sexuality 3: 229–50.
Moore, A. R., D. G. Butt, J. Ellis-Clarke, and J. Cartmill. 2010. Linguistic
Analysis of Verbal and Non-verbal Communication in the Operating
Room. ANZ Journal of Surgery 80(12): 925–9.
Mullany, K., C. Smith, K. Harvey, and S. Adolphs. 2015. ‘Am I Anorexic?’
Weight, Eating and Discourses of the Body in Online Adolescent Health
Communication. Communication and Medicine 11(2): 1–14.
Muntigl, P. 2004. Narrative Counselling: Social and Linguistic Processes of Change.
Muntigl, P. 2006. Macrogenre: A Multiperspectival and Multifunctional

Approach to Social Interaction. Linguistics and the Human Sciences 2:
233–56.
Muntigl, P., N. Knight, A. Watkins, A. Horvath, and L. Angus. 2013. Active
Retreating: Person-centered Practices to Repair Disaffiliation in Therapy.
Journal of Pragmatics 53: 1–20.
Nagar, R. and J. Fine. 2013. Clause Complex Manifestation in Depression.
Text and Talk 33(4–5): 595–615.
Neil, D. 2017. Digital Drugs: The Marketing of Internet Addiction. Provoca-
tions Seminar, School of Humanities and Social Inquiry, University of
Wollongong, 6 March 2017.
Nwogu, K. N. 1990. Discourse Variation in Medical Texts: Schema, Theme and
Cohesion. PhD Thesis, Aston University.
Nwogu, K. N. and T. Bloor. 1991. Thematic Progression in Professional and
Popular Medical Texts. In E. Ventola, ed., Functional and Systemic Linguistics:
Approaches and Uses. Amsterdam: Mouton de Gruyter.
Overlach, F. 2008. Sprache des Schmerzes, Sprechen über Schmerzen: Eine
grammatisch-semantische und gesprächsanalytische Untersuchung von Schmerz-
ausdrücken im Deutschen. Berlin: De Gruyter.
Patterson, E. 2007. Structuring Flexibility: The Potential Good, Bad and Ugly
in Standardisation of Handovers. Quality and Safety in Health Care 17(1):
4–5.
Pounds, G. 2011. Empathy as ‘Appraisal’: A New Language-based Approach
to the Exploration of Clinical Empathy. Journal of Applied Linguistics and
Professional Practice 7: 139–62.
Predict. n.d. Online Mathematical Tool Designed for Patients and Doctors to
Help Them Decide on the Ideal Course of Treatment Following Breast Cancer
Surgery. Available online at: http://predict.nhs.uk. (Last accessed 13/05/
2017.)
Pryor, E. and R. Woodward-Kron. 2014. International Medical Graduate
Doctor to Doctor Telephone Communication: A Genre Perspective. English
for Specific Purposes 35: 41–53.
Reinsborough, P. and D. Canning. 2010. Re-Imagining Change. Oakland: PM
Press.
Rochester, S. and J. R. Martin. 1979. Crazy Talk: A Study of the Discourse of
Schizophrenic Speakers. New York: Plenum.
Rochester, S., J. R. Martin, and S. Thurston. 1977. Thought-process
Disorder in Schizophrenia: The Listener’s Task. Brain and Language 4:
95–114.
Santiago, N., L. Young, L. S. Baramy, R. Cabrera, E. May, R. Wegener, D. G.
Butt, and M. Parr. 2011. The Impact of the Medical Emergency Team on
the Resuscitation Practice of Critical Care Nurses. BMJ Quality and Safety
20: 115–20.
Sarangi, S. and C. Roberts. 1999. The Dynamics of Interactional and Insti-
tutional Orders in Work-related Settings. In S. Sarangi and C. Roberts,
eds., Talk, Work and the Institutional Order: Discourse in Medical, Mediation and
Management Settings. Berlin: Mouton de Gruyter. 1–57.
Sheehan, A. and W. Bowcher. 2017. Messages to New Mothers: An Analysis
of Breast Pump Advertisements. Maternal and Child Nutrition 13(2): 1–13.
Sinclair, J. and M. Coulthard. 1975. Towards an Analysis of Discourse. Oxford:
Slade, D. and S. Eggins. 2015. Communication in Clinical Handover:
Improving the Safety and Quality of the Patient Experience. Journal of
Public Health Research 4: 666.
Slade, D., H. Scheeres, M. Manidis, C. M. I. M. Matthiessen, R. Iedema,
M. Herke, J. McGregor, R. Dunston, and J. Stein-Parbury. 2008. Emer-
gency Communication: The Discursive Challenges Facing Emergency
Clinicians and Patients in Hospital Emergency Departments. Discourse
and Communication 2(3): 289–316.
Slade, D., H. Scheeres, H. de Silva-Joyce, J. McGregor, N. Stanton, M. Herke,
E. Weisberg, and D. Bateson. 2009. Developing Effective Communication
between Doctors and Clients: Sexual and Reproductive Health Consultations.
Research Report to Family Planning NSW.
Slade, D., E. Chandler, J. Pun, M. Lam, C. M. I. M. Matthiessen, G. Williams,
E. Espindola, F. Veloso, K. Tsui, S. Tang, and K. Tang. 2015a.
Effective Clinician–Patient Communication in Hong Kong Accident and
Emergency Departments. Hong Kong Journal of Emergency Medicine 22(2):
69–83.
Slade, D., M. Manidis, J. McGregor, H. Scheeres, E. Chandler, J. Stein-
Parbury, R. Dunston, M. Herke, and C. M. I. M. Matthiessen. 2015b.
Communication in Hospital Emergency Departments. Berlin: Springer.
Smith, S. K., L. Trevena, D. Nutbeam, A. Barratt, and K. J. McCaffery. 2008.
Information Needs and Preferences of Low and High Literacy Consumers
for Decisions about Colorectal Cancer Screening: Utilising a Linguistic
Model. Health Expectations 11(2): 123–36.
Sontag, S. 1978. Illness as a Metaphor. New York: Farrar, Straus and Giroux.
Sorensen, R. and R. Iedema. 2006. Integrating Patients’ Nonmedical Status
in End-of-life Decision Making: Structuring Communication through
‘Conferencing’. Communication and Medicine 3: 185–96.
Stenglin, M. and M. Foureur. 2013. Designing Out the Fear Cascade to
Increase the Likelihood of Normal Birth. Midwifery 29(8): 819–25.
Street, R. L., G. Makoul, N. K. Arora, and R. M. Epstein. 2009. How
Does Communication Heal? Pathways Linking Clinician–Patient
Communication to Health Outcomes. Patient Education and Counseling 74:
295–301.
Tebble, H. 1993. A Discourse Model for Dialogue Interpreting. In Proceedings
of the First Practitioners’ Seminar of AUSIT, November 1992. Canberra: NAATI.
1–26.
Tebble, H. 1996a. Research into Tenor in Medical Interpreting. Interpreting
Research 6: 33–45.
Tebble, H. 1996b. A Discourse-based Approach to Community Interpreter

Education. In Proceedings XIV World Congress of the Federation Internationale
des Traducteurs. Melbourne: Australian Institute of Interpreters and Trans-
lators. 385–94.
Tebble, H. 1998. Medical Interpreting: Improving Communication with Your
Patients. Geelong: Deakin University and Language Australia.
Tebble, H. 1999. The Tenor of Consultant Physicians: Implications for
Medical Interpreting. The Translator 5: 179–200.
Tebble, H. 2003. Training Doctors to Work Effectively with Interpreters. In
L. Brunette, G. Bastin, I. Hemlin, and H. Clarke, eds., The Critical Link 3:
Interpreters in the Community. Amsterdam: John Benjamins. 81–95.
Tebble, H. 2008. Using Systemic Functional Linguistics to Understand
and Practise Dialogue Interpreting. In C. Wu, C. M. I. M. Matthiessen,
and M. Herke, eds., Voices Around the World: Proceedings of the 35th Inter-
national Systemic Functional Conference. Sydney: Macquarie University.
Tebble, H. 2012. Subjectivity in the Discourse of Depressed Acute Care
Hospital Patients. In N. Baumgarten, I. du Bois, and J. House, eds.,
Subjectivity in Language and in Discourse. Bingley: Emerald. 115–35.
Tebble, H. 2014. A Genre-based Approach to Teaching Dialogue Interpret-
ing. The Interpreter and Translator Trainer 8(3): 418–36.
Thompson, G. 1999. Acting the Part: Lexicogrammatical Choices and Con-
textual Factors. In M. Ghadessy, ed., Text and Context in Functional Linguis-
tics. Amsterdam: Benjamins. 101–24.
Thwaite, A. 2015. Using the Multimodal Analysis Video Program for Regis-
ter Analysis: A Preliminary Study. TESOL International 10(1): 110–25.
Torsello, C. 1997. Linguistics, Discourse Analysis and Interpretation. In
Y. Gambier, D. Gile, and C. Taylor, eds., Conference Interpreting: Current
Trends in Research. Amsterdam: John Benjamins. 167–86.
Tuckwell, K. and A. R. Moore. in preparation. ‘Issues, Concerns, Worries’:
The Role of Graduation in Encouraging Discussion of End-of-life Issues in
Palliative Care.
Tversky, A. and D. Khaneman. 1981. The Framing of Decisions and the
Psychology of Choice. Science 211: 453–8.
van Leeuwen, T. 1995. Representing Social Action. Discourse and Society 6:
81–107.
van Moll, M. and M. O’Donnell. 2004. Automatic Recognition of Generic
Structure: Medical Discharge Notices. In D. Banks, ed., Text and Texture:
Systemic Functional Viewpoints on the Nature and Structure of Text. Paris:
L’Harmattan. 329–53.
Vincent, C. A. and A. Coulter. 2002. Patient Safety: What about the Patient?
Quality and Safety in Health Care 11: 76–80.
Walsh, J., N. Cominos, and J. Jureidini. 2016a. How Language Shapes Psy-
chiatric Case Formulation. Communication and Medicine 13(1): 99–114.
Walsh, J., N. Cominos, and J. Jureidini. 2016b. Maintaining and Generating
Knowledge in Multidisciplinary Mental Health Handovers. In S. Eggins,
D. Slade, and F. Geddes, eds., Effective Communication in Clinical Handover:

From Research to Practice. Berlin: Walter de Gruyter. 245–64.
Walsh, J., N. Cominos, and J. Jureidini. 2016c. Patient Voice: Including the
Patient in Mental Health Handovers. In S. Eggins, D. Slade, and F. Geddes,
eds., Effective Communication in Clinical Handover: From Research to Practice.
Berlin: Walter de Gruyter. 265–78.
Walton, M., R. Harrison, P. Kelly, J. Smith-Merry, E. Manias, C. Jorm, and
R. Iedema. 2017. Patients’ Reports of Adverse Events: A Data Linkage
Study of Australian Adults Aged 45 Years and Over. BMJ Quality and Safety
26(9): 743–50.
Watson, O. 2012. The Language of Clinical Empathy: Modelling Affiliation in
Doctor–Patient Communication. BA Honours Thesis, University of Sydney.
Williams, G. 2014. Medical Discourse and Semantic Variation. Sydney
University SFL Friday Seminar, 28 March 2014.
Willis, C. 2001. Linguistic Features of Rapport: An Investigation into Consultant
Physicians’ Linguistic Use of Rapport. BA Honours Thesis, Deakin University.
Woloschuk, W., P. H. Harasym, and W. Temple. 2004. Attitude Change
during Medical School: A Cohort Study. Medical Education 38(5): 522–34.
Woodward-Kron, R. 2016. International Medical Graduates and the Discur-
sive Patterns of Patient-centred Communication. Language Learning in
Higher Education 6(1): 253–73.
Woodward-Kron, R. and C. Elder. 2016. Authenticity in an English Lan-
guage Screening Test for Clinicians: A Comparative Discourse Study.
Language Testing 33(2): 251–70.
Woodward-Kron, R., M. Stevens, and E. Flynn. 2011. The Medical Educator,
the Discourse Analyst, and the Phonetician: A Collaborative Feedback
Methodology for Clinical Communication. Academic Medicine 86(5): 565–70.
Woodward-Kron, R., C. Fraser, J. Pill, and E. Flynn. 2014. How We
Developed Doctors Speak Up: An Evidence-based Language and Commu-
nication Skills Open Access Resource for International Medical Gradu-
ates. Medical Teacher 37(3): 31–3.
Woodward-Kron, R., J. Hughson, A. Parker, A. Bresin, J. Hajek, U. Knoch,
T. D. Phan, and D. Story. 2016. Culturally and Linguistically Diverse
Populations in Medical Research: Perceptions and Experiences of Older
Italians, Their Families, Ethics Administrators and Researchers. Journal of
Public Health Research 5: 43–51.
Wyer, M., R. Iedema, S.-Y. Hor, C. Jorm, C. Hooker, and G. Gilbert. 2017.
Patient Involvement Can Affect Clinicians’ Perspectives and Practices of
Infection Prevention and Control: A ‘Post-Qualitative’ Study Using Video-
Reflexive Ethnography. International Journal of Qualitative Methods 16: 1–10.
Zinn, J. and D. McDonald. 2016. Changing Discourses of Risk and Health
Risk: A Corpus Analysis of the Usage of Risk Language in the New York
Times. In M. Chamberlain, ed., Medicine, Risk, Discourse and Power.
New York: Routledge. 207–40.
26
Language and Literature
Donna R. Miller
For Ruqaiya Hasan (1931–2015) – in memoriam.
26.1 Introduction
This chapter is dedicated to a twofold expression that needs to be

problematized, not least because since the mid-twentieth century the
relationship between the components of the dyad ‘language and litera-
ture’ has been a – regrettably, but undeniably – unhappy one (McIntyre
2012:1). To make matters still more intricate, the ‘divide’ reduplicates
itself within each of the ‘camps’ as well. Peaceful coexistence, in brief,
cannot be said to characterize the state of the art(s).
Section 26.2 begins our excursus with a brief chronology of what’s
generally become known as ‘stylistics’, a discipline that initially aimed at
a rapprochement between the two fields of study. Rather than tracing a
comprehensive history of stylistics and its various stages, attention focuses
on major influences in the emergence of what is often dubbed ‘British’
stylistics (Selden 1989:83) and also, in brief, on how Systemic Functional
Linguistics (henceforth SFL) scholars, Halliday (acknowledged) and Hasan
(overlooked), were actively contributing to its development.
The task of Section 26.3 is to properly define/describe what the central
SFL take on literature and the analysis of the language in literature consists
in. Hasan provides us with the expression ‘verbal art’; with the name of the
practice of doing SFL stylistics, i.e. Systemic Socio-Semantic Stylistics
(henceforth SSS), as her last formulation styled it (Hasan 2015, personal
communication); and so also with the double-articulation descriptive and
analytical model presented here (1989, 2007). The rationale and key con-
cepts of this framework will be juxtaposed primarily with Simpson (2014),
as representative of current mainstream approaches to stylistics.
Language and Literature 691
A sample illustration of verbal art analysis in action will be offered in

Section 26.4, exemplifying the basic steps by means of the analysis of D. H.
Lawrence’s ‘Lonely, Lonesome, Loney – O!’, a poem abounding in linguistic
mechanisms that will be shown to ‘mean’ in the ways Hasan’s model
foresees – as well as by means of the ‘pervasive parallelism’ (Jakobson
1966:423) that I argue deserves to be slotted into the framework (see Miller
2010, 2012, 2013, 2016a).
Section 26.5 takes brief note of recent research within SFL specifically
using the framework, as well as indicating possible directions for future
work. One aspect of that future must be the space that a Corpus Linguistic
(henceforth CL) approach to the analysis of verbal art can be seen to have –
and not have (e.g. Miller and Luporini 2015; Miller 2016b). The final section
will offer some parting thoughts.
26.2 A Partial Recount of Stylistics as a Discipline
26.2.1 Going Back

Systemic Socio-Semantic Stylistics, as Hasan points out, is not ‘new’:
‘it actually predates the 1960s’ structural stylistics’ (Hasan 2007:21).
She maintains that the initial work that was incorporated into the perspec-
tive was done by the Russian Neo-Formalists and Prague Circle scholars,
especially Mukařovský (1977, 1978) in his 1928 discussion of ‘foreground-
ing’. But it is fitting that we make a short journey back in time to attempt to
trace the roots, not solely of SSS, but of stylistics tout court – its seeds,
germinations, and cross-fertilizations – and try to see how these entwine
with SSS, or not.
Forty years after Hatzfeld’s bibliography (1953) of ‘new stylistics’, Lecer-
cle (1993:14) made the claim that nobody has ever really known what the
term ‘stylistics’ means, and, given the assorted faces it has presented over
time, it is hard to rebut him. In diverse decades, diverse linguistic schools
had the ascendency, leaving their marks. Wales (2001:373) breaks these
down into formalist Generative Grammar in the 1960s, a move to function-
alist Discourse Analysis and Pragmatics in the 1970s and 1980s, and the
input of Critical Discourse Analysis and Cognitive Linguistics in the 1990s.
With the 2000s, different schools seemed to step up their competitive self-
promotion, and new factions even began to emerge, bringing us to what is
currently rather a textbook potpourri.
But I would start this overview at the point Hasan refers to and where
other scholars tend to (e.g. Busse and McIntyre 2010, on which I lean
considerably here, although not exclusively), with the advent of the twenti-
eth century and the dénouement of the tradition of purely author-centred
literary theory, i.e. with the work of the Moscow Linguistic Circle, and
specifically, with the most famous of the so-called Russian Formalists,
Roman Jakobson, well-known for his ‘poetic function’, the function whose
692 DONNA R. MILLER
focus is seen as being on the message for its own sake (Jakobson 1960).
The decade 1929 to 1939 marked the golden years of the Prague School.
Jakobson was collaborating with the Prague Structuralists, notably
Mukařovský, and the two were aiming at identifying the formal and func-
tional linguistic mechanisms responsible for articulating the aesthetic
motivation they both saw as the impetus of literature. They worked
together towards the theorization of what would become seminal and
durable stylistic concepts such as deviation – which Mukařovský saw as
fundamental to the creation of the defamiliarizing effect also central to
literary language – and grammatical parallelism (henceforth GP) – the
structural patterning in texts that Jakobson theorized as the empirical
linguistic evidence of his ‘poetic function’ and which he proposed also
generated defamiliarization. Mukařovský (1977, 1978) moved on to
hypothesizing the role of foregrounding, and Jakobson (1966) to that of
pervasive parallelism, but theoretical intersection endured, despite the
undoubtedly greater ‘success’ of foregrounding – not only in SSS, but in
stylistics tout court. In addition to Halliday and Hasan, mainstream stylisti-
cians in the last decades of the twentieth century, such as Leech and Short
(2007), were convincingly demonstrating the essential role of foreground-
ing (or in the Mukařovský-esque terms preferred by Halliday (2002a:131),
the de-automatization of grammar) in literary interpretation. And fore-
grounding has even been related directly to the concepts of figure and
ground in recent work in cognitive stylistics. Its mileage is impressive,
but perhaps Jakobson’s reach was just as significant, if not as wide-ranging,
as I hope to show.
The Second World War meant a close to Jakobson’s collaboration with
the Prague Structuralists. For a while he was an itinerant scholar, settling
in the USA in 1941, a move that proved decisive for the subsequent spread
of his ideas in America but also in Europe, where work on style in literature
was of course also being done – chiefly by the Austrian philologist, Spitzer,
on the literature of the Romance languages, privileging ‘objective’ over
‘impressionistic’ methods (see Wales 2001:296–7), as well as by scholars
like Auerbach, Bally, and Guiraud, whose work would then impact on the
development of the French analyse de texte.
26.2.2 Moving On
Jakobson’s move to the USA was also significant to the development, in
America and Britain respectively, of the New Criticism and Practical Criti-
cism movements, both of which employed techniques of the ‘close reading’
of literature, a practice which was widespread for decades (1930s–1960s).
The first of these focused on the description of the aesthetic qualities of a
literary text, while the latter, developed by I. A. Richards, engaged with the
psychological aspects of how readers comprehend texts – what the first
school called the ‘affective fallacy’ (Wimsatt and Beardsley 1949). Richard’s
approach displayed concerns that decidedly anticipated reader-response

theory in the 1960s and 1970s, and also indeed cognitive stylistics in the
1990s. Yahya Ali Bani Salameh (2010) offers a neat synopsis of the distinc-
tions between these, and other, movements.
These approaches – then avant-garde – have long been assessed by stylis-
ticians as out-dated analytical paradigms, because imprecise and inad-
equate. Halliday (2002a:128) sees their foremost deficiency as a failure to
associate the text to the linguistic system, i.e. to meaning potential.
In purely literary circles, this ‘new’ critical focus on the text itself began
in the 1970s to be seen by Marxist critics as fostering the decontextualized
and even dehumanized reification of the text – texts being treated
as (sacramental-like) objects, divorced from the social context of their
creation.1 One might hypothesize a link between this reified text critique
and the desire of Simpson – and other scholars, including Fowler, more on
whom below – to demystify the age-old veneration of the ‘lit crit’ for
literature, as well as with Simpson’s suspicion of the least degree of pre-
scriptivism being allowed to govern the definition, and interpretative tools,
of literature.
Be that as it may, non-literary stylistics, giving special attention to the
text–context connection, began to be explored in these same years by
Crystal and Davies (1969) and Enkvist (1973). Busse and McIntyre (2010)
note, however, that work in non-literary stylistics stalled at this point, and
only picked up again quite a bit later. Interestingly, as an explanation for
the temporary halt, they cite the lack of linguistic frameworks able to deal
with the new and important contextual issues emerging from these
scholars’ work. As we know, however, in the late 1960s and early 1970s,
both Halliday and Hasan were analyzing literature with an already substan-
tially tried and tested functional linguistic framework that was deeply
entrenched in context, both of situation and culture. In brief, an albeit
still-being-modelled linguistic framework was available.
So when do these two scholars start getting the attention they deserve?
Hasan never does, and this raises a question of what I have dubbed a
‘politics of exclusion’ (Miller 2010:48). Halliday instead has received ample
tribute as having been a major influence on the development of British
stylistics. It was Fowler (1966) who was responsible for initially publicizing
SFL’s transitivity system – the main mechanism investigated in the often
cited Halliday (2002b) – as a useful stylistic tool – a tool, along with that of
modality, that has enjoyed wide and frequent application in mainstream
stylistics (e.g. Simpson 2014:77, on transitivity). Butler (2003, as noted in
Lukin 2015; also see Lukin and Webster 2005; Butt and Lukin 2009) docu-
ments Halliday’s influence on a number of monographs on stylistics,
1
These are intricate Marxist-based critical theory issues that are brilliantly elucidated from a diachronic perspective by a
semi-anonymous author, a certain ‘Paul’, available online at: http://herrnaphta.wordpress.com/2011/01/09/
reification-and-american-literature. (Last accessed 14/09/2017.)
694 DONNA R. MILLER
including Leech and Short (2007) and Toolan (2001). But Hasan’s framework
is always, conspicuously, ‘missing’. Undoubtedly this has something to do
with Halliday being simply harder to ignore, but also, and primarily, with
the ease with which certain aspects of the SFL model can be integrated into
a non-SFL stylistics text book, without, that is, needing to deal with a more
unwieldy, holistic model of the kind Hasan proposes. But Fowler warrants
further attention.
26.2.3 Fowler and ‘Linguistic Criticism’

I have deliberately avoided use of the manifold labels that the linguistic
analysis of literature has gone and still goes under, but in this heading
I have appropriated the title of Fowler’s 1986 monograph, as it accentuates
the linguistic–literary rivalry he strenuously engaged in. The practice of
stylistics unquestionably owes much not only to the seminal work of
Fowler but also to his laudable struggles to reinstate literature, pace the
‘lit crits’, as a legitimate object of linguistic study, as the famously acerbic,
if also at times amusing, Fowler–Bateson debates (Fowler and Bateson 1967,
1968) testify.
Of the many aspects of Fowler’s subsequent thought that would merit
discussion, to be cited, because absolutely antithetical to Hasan’s, is his own
definition of what literature is:
No plausible essentialist or intrinsic definition of literature has been or is likely

to be devised. For my purpose, no such theory is necessary. What literature is
can be stated empirically, within the realm of socio-linguistic fact. It is: an open
set of texts, of great formal diversity, recognized by a culture as possessing
certain institutional values and performing certain functions.
(Fowler 1981:81)
In what lies the opposition? Surely not in their respective views re the
importance of culture and community. Indeed, Hasan says that ‘the litera-
ture text . . . embodies precisely the kind of “truths” that most communities
are deeply concerned with’ (Hasan 1989:100). She also recognizes that the
social impacts on verbal art: indeed ‘perhaps the most critical part it plays is
in the shaping of the ideological orientations of those who write and those
who read literature’ (Hasan 2007:25). And a text’s endurance as art will
always hinge on the value which is awarded it by successive generations of
readers: ‘The challenge for the creator of verbal art is that the symbolically
articulated Theme has to be capable of striking a chord in the reader over
substantial distances in time and space’ (Hasan 2007:25). But in spite of
these correspondences, the sum and substance is a clash in standpoints.
Hasan does not invest the power for ultimately deciding what is or is not
literature in the community, whether it be that of the time and place of the
text’s creation or at a further semiotic social distance (Hasan 2007:34). And,
ultimately, as we will see below, her definition is eminently ‘essentialist’.
In addition, to my knowledge, Fowler was also the first to draw proper

attention to ‘The Mukařovský-Jakobson Theory’, in astutely – and, I submit,
quite rightly – noting the theoretical convergence of Mukařovský’s fore-
grounding and Jakobson’s grammatical parallelism: ‘For both of these
writers, literary language draws the reader’s attention to its own artifices
of construction’ (Fowler 1986:73). But Fowler explicitly and energetically
distances himself from their ‘aesthetic’ positions on the functions and
motives of foregrounding and parallelism.
His aversion might in some part also be ascribed to the still ongoing
critique of the reifying New Criticism and can certainly be understood
against the backdrop of his non-essentialist definition, formulated five
years previously.
That said, Fowler’s keen perception of the unmistakable analogies
between Jakobson’s insights and those of Mukařovský is to be thanked for
the direction of my recent and ongoing arguments for recognizing – pace
Hasan’s misgivings – that Jakobson’s grammatical parallelism must be
likened, as foregrounding is, to Hasan’s symbolic articulation of verbal
art, what literature’s art resides in (Miller 2010, 2012, 2013, 2016a). The
relevance of Jakobson’s work for SSS is a vital ‘further direction’ that verbal
art studies need to take. To espouse Jakobson’s work, even in this postmod-
ern era, is not to advocate an obsessively ‘structural’ approach to language:
he continually expounded the inseparability of form and meaning. Rather
than a stringent, die-hard structuralist, he was foremost a linguist, atten-
tive to meaning and context as well (see Stankiewicz 1983:24; Caton
1987:223–4).
26.2.4 Kindred Stylistic Spirits

On the language of literature, Halliday and Hasan were unquestionably on
the same wavelength. Their ideas evolved in largely parallel fashion, as the
ideas of like minds in contact will.
In 1964, the same year as Hasan’s PhD thesis, Halliday wrote: ‘Literature
is language for its own sake: the only use of language, perhaps, where the aim
is to use language’ (Halliday et al. 1964:245, emphases added). And I would
note how this for its own sake echoes the focus of Jakobson’s poetic function,
cited above. In 1964, and again in 1971 and 1989, Hasan puts forth an
analogous proposition:
It is not that there is art [somewhere ‘out there’, so to speak], and the job of
language is simply to express it; rather it is that, if there is art, it is because of
how language functions in the text . . . in verbal art the role of language is
central. Here language is not as clothing to the body; it is the body.
(Hasan 1989:91; also see Hasan 1964, 1971)
In 1982, Halliday drolly makes the point for the somehow ‘different’ nature
of literature as text with the quip: ‘the paradox of “poetic” language [is] that
696 DONNA R. MILLER
there is no such thing . . . but we can all recognize it when we see it’ (Halliday
2002a:134, emphasis added).
In 2002, already engaging with Hasan’s work, Halliday (2002b) draws a
parallel between Mukařovský’s foregrounding and the motivated ‘promin-
ence’ of grammatical features. He suggests two levels of meaning, both
grammatically realized, but one ‘underlying’ and ‘deeper’ than the first,
‘immediate’ level, also glossed as ‘subject matter’ (2002b:118–20 et passim).
The deeper semantic meanings are said, significantly, ‘to serve a vision of
things . . . The vision provides the motivation for their prominence’. It will
be primarily Hasan’s task in the course of the 1970s to better delineate this
‘vision’.
In Halliday (2002a:131, emphasis added) the term foregrounding, as
noted above, is explicitly replaced by ‘de-automatization’, conceivably with
a view to highlighting the distinct if complementary roles of these two
semantic levels from an SFL perspective:
What is in question is not simply prominence but rather the partial freeing of
the lower-level systems from the control of the semantics so that they become domains
of choice in their own right. In terms of systemic theory the de-automatization
of the grammar means that grammatical choices are not simply determined
from above: there is selection and pre-selection. Hence the wording becomes
a quasi-independent semiotic mode through which the meanings of the
work can be projected.
I have presented this vital thought of Halliday’s here since the two distinct
semiotic modes of literature meaning-making he hypothesizes can with
impunity be said to be analogous to the scaffolding of the doubly articu-
lated framework for the analysis of verbal art which Hasan began to model
in 1971, and which will be properly described and illustrated below.
26.3 SSS vs. All the Rest
26.3.1 SSS and Mainstream Stylistics: Convergences and

Divergences
Hasan’s SSS model is the SFL framework I aim to describe here. To remove
any possible ambiguity immediately, although some SFL verbal art analysts
fully espouse the model, not all do, even though none has modelled or put
forth a rival SFL framework for the analysis of language in literature. Non-
SFL analysts keep their distance from the model, basically ignoring it,
though to what extent avoidance is deliberate is a moot point. The reason(s)
for the lack of attention it receives are equally hard to pinpoint, yet it must be
admitted that practising SSS is not ‘easy’. The first and most crucial obstacle is
that one needs to have a good knowledge of the SFL model, itself widely
perceived as being demanding. Systemic Socio-Semantic Stylistics is also fun-
damentally incompatible with other mainstream stylistic approaches to
literature, despite certain apparent intersections in how the discipline is

defined. I will address some of these overlaps here at the outset.
The reason Simpson (2014) has been chosen as the rival yardstick for this
chapter is that his ‘Resource book for students’ engages with theoretical
positions that help exhibit the dispute. The task of teasing out the diver-
gences from the convergences, however, is a slippery one, as the definitions
for key terms are not always transparent and the divergence ranges from
slight to radical.
At first sight, Simpson’s (2014:3) introductory definition of the discipline
does not strike one as particularly problematic or challenging to SSS:
‘Stylistics is a method of textual interpretation in which primacy of place
is assigned to language.’ Nothing wrong with that. Additional remarks are
also fully acceptable: ‘It is the full gamut of the system of language that
makes all aspects of the writer’s craft relevant in stylistic analysis’ and
‘stylistics is interested in language as a function of texts in context.’ Simp-
son even goes on to speak of patterns of language being an index to a text’s
function – no ground for grievance for the verbal art theory, ‘patterning’
being a central Hasanian concern. In fact, his definition of Mukařovskian
foregrounding – ‘a form of textual patterning which is motivated specific-
ally for literary-aesthetic purposes’ (Simpson 2014:52) – is just right. It is
noteworthy, however, that he changes his mind a mere three pages on in
affirming that ‘to separate off literature from other uses of language . . . is
not a desired outcome in stylistic analysis’ (Simpson 2014:55), thus effect-
ively overturning the definition – a point I will take up again at a more
fitting juncture. But with this about-face, the conflict is exposed. Apropos,
what foregrounding finally comes down to for Simpson is that well-known
dinosaur, ‘deviation’: either from a norm or as ‘more of the same’, through
reiteration (Simpson 2014:52). Both Halliday (2002b:100) and Hasan
(1989:92) argue, rightly I suggest, against deviation’s ultimate relevance to
stylistics.
So convergences turn out to be little more than skin-deep. But we need to
take up more systematically that ‘incompatibility’ between SSS and main-
stream approaches. The mismatch can perhaps be best encapsulated in
three of SSS’s convictions, the first, I suggest, being the terrain in which
the others have taken root:
(1) for SSS, literature is a ‘special’ kind of text, requiring a special type of
descriptive and analytical framework for its study, which we will take
a close look at below;
(2) the two-tiered approach should not, indeed cannot, be applied to the
investigation of other text types: rather than extending its brief to any
and all registers, as mainstream stylisticians (see Simpson 2014) tend
more and more to do, it sees literature as its sole object of study;
(3) the approach ought not to be promiscuous; i.e. it should remain
unadulterated; it demands exclusivity, philandering with incompatible
698 DONNA R. MILLER
stylistic approaches not being welcomed. It is not open to the hetero-

geneous methodological input of the proliferation of sub-disciplines,
such as feminist stylistics, cognitive stylistics, and discourse stylistics,
which mainstream stylisticians, again typified by Simpson (2014:2), see
as enriching the methods of stylistic practice – or even, in theory at
least, CL-based, or assisted, stylistics, about which more will also be
said below.
Systemic Socio-Semantic Stylistics is, in brief, ‘holistic’, to apply a conceiv-

ably positive epithet, or, pace Fowler (1981:81; see quote in Section 26.2.3
above), ‘essentialist’ and ‘intrinsic’; but it might also be labelled ‘uncom-
promising’, a far less flattering word.
Disentangling these convictions is too tricky. Thus we turn to ‘literature
as special’, the discussion of which will naturally and inevitably embrace
the second and third points above as well.
26.3.2 SSS: Literature as Special

So then, the conflict between SSS and mainstream approaches to stylistics is
rooted primarily in the SSS notion of literature as special, on just what
makes a text literature, or ‘verbal’ art. Let’s start with listening to Hasan
herself: ‘The framework I shall be presenting . . . commits one to taking
seriously both the art which is languaged and the language which is artis-
tic, to see literature as a variety of language, but the variety itself as
possessing attributes which are not matched by non-literature varieties’
(Hasan 1996:49). But what are these ‘attributes’ and in what way are they
‘not matched’?
To answer we can begin by recalling the celebrated quotes cited at the end
of Section 26.2 above. But to answer fully we also need to go back to the
years when Hasan’s work was actively shaping the SFL modelling of lan-
guage in use, or register theory – the systemic functional notion of each
instance of language use being rooted in the context of some social and
material situation which will tend to activate and be construed by mean-
ings that will tend to be realized in wordings (see Bowcher, this volume).
This research was proving fundamental to her critical deduction that verbal
art was not a register like any other. She concluded that, although literature
is indisputably instantiated language use in a particular social context, the
context–language connection in verbal art is different, because decidedly
more complex than for other functional varieties of text (Hasan 1975:54,
2007:22–3).
The contexts of a literature text include, to begin with, the ‘real’ and
‘fictional’ contexts, the first embracing writer and reader and the social act
of narrating/poet-‘izing’ and the second being what is dynamically created
by the text, its internal field and tenor in particular. The insight is based on
Malinowski (1923, 1935; also see Halliday 1978:146–7; Hasan 1996:50–4,
2007:35; Miller 1998:276, 282–3). This distinction between a ‘real’ and

‘fictional’ context is of course also valid for what Hasan calls ‘verbal art
manquée’ (Hasan 2011, personal communication), but it is the role these
play within the systematic analytical process of SSS that is missing in the
latter, and that makes the difference.
In addition, also posited is a ‘context of creation’ of the literature text,
comprising the language, worldview, and artistic conventions of the time/
place of writing, all of which the artist cannot help engaging with, either to
basically repropose, or to some degree distance him/herself from, them.
Finally, a ‘context of reception’ of the reader’s time and place of reception is
also theorized. Taken together, these contexts impact on the text and its
interpretation and require the analyst’s close and careful attention (Hasan
1989:101–3, 1996:50–4). And no other variety of language use can be seen
to display such context–language intricacy; hence, literature is different.
But what does Simpson have to say on the literature as special stance?
Unsurprisingly, he candidly, and categorically, contests it:
To argue for the existence of a distinct literary register is effectively to argue
for a kind of cliché, because it would involve reining stylistic expression into
a set of formulaic prescriptions. . . . To claim that literary language is special,
that it can somehow be bracketed off from the mundane or commonplace in
discourse, is ultimately to wrest it away from the practice of stylistics.
(Simpson 2014:106–7)
His monoglossically stated contentions here are fully consistent with his
position on the practice of stylistics needing to extend to any and all text
types. But do they hold up? One might suggest that to maintain the
distinctiveness of literature, as SSS does, is not to deal with clichés, nor to
apply automatic strictures to artistic expression, nor to necessarily argue
for a ‘literary’ register, not least because, as we will see below, the starting
point for the analysis of verbal art is identical to that for any other text
(Hasan 1989:92). Moreover, Hasan, in suggesting a way to distinguish
between what is verbal art and what, most likely, is not, actually pits
‘literary’ against ‘literature’:
If the patterning [foregrounding] of patterns is consistently utilised for a
second-order semiosis . . . then the text in question is a literature text. If,
however, such a role is not played by the patternings, then we have a literary
text. The recognition of this distinction is important, not least because the
techniques for the study and evaluation of the two are not identical.
(Hasan 1989:101)
So, ‘literariness’ has only a superficial likeness to verbal art. Recalling

Simpson’s change of heart or mind concerning the validity of the aesthetic
motivation of literature mentioned above, I now add its apparent inspir-
ation: ‘aesthetic motivation’ would appear to be at cross-purposes with his
otherwise admirable project of debunking the conventional reverence of
the ‘lit crit’ for literature (Simpson 2014:98–9). Now, I have no argument
700 DONNA R. MILLER
with the fittingness of demystifying such blind ‘veneration’ of literature, or

with fighting the ‘lit crit’ academy that would safeguard the artefacts of its
domain from what it sees as the intruding desecrations of the inept and
irreverent linguist – as Fowler openly did. But, it is wrong to conclude – as
Simpson does, but as also Fowler apparently did – that doing so means
needing to conclude that there is nothing meaningfully ‘special’ about
literature, or that a belief in its specialness means defeating a proper
stylistic agenda or unacceptably hemming in its territory.
There would be much more to say about the literature vs. literary text
divide as Hasan defines it, but the SSS metalanguage in the quote above still
wants disambiguating. Before doing so, however, a brief word needs to be
said regarding the third SSS ‘conviction’ – vetoing a model muddle as
it were.
In Hasan’s view, for producing any analysis that is ultimately going to be
worthwhile, a framework must be ‘maximally applicable to the genre [i.e.
to literature], irrespective of variations in time, sub-genre, and the critic’s
response’ (Hasan 1989:90). These qualities are crucial to the holistic nature
of her definition of literature and her model, but also to its impermeability
to eclecticism in approach. Be that as it may, Hasan is the only ‘stylistician’
I am aware of who has ever proposed – or had the temerity (+ve appreci-
ation) to propose – such a model – to which I finally turn.
26.3.3 The SSS Framework

As said, SSS hypothesizes that, because verbal art is different, it requires a
different theoretical and methodological take. Again we find significant
conflict with what mainstream stylisticians propose – i.e. that because
literature is essentially no different from any other text type, it needs none
other than the typical descriptive/analytical categories and tools used for
examining these (e.g. Simpson 2014:106–7). And yet, for the preliminary
steps of SSS analysis, the SFL categories and tools are exactly the same as for
any other text; meaning the starting point for its analysis of literature is
identical to that for any other text:
The starting point for the description of literature is identical to that

elsewhere . . . these descriptive categories are applicable to all uses of lan-
guage, irrespective of where they occur; the semantic values assigned to
them in the clauses in which they occurred is also a constant. There can be
no dispute that these patterns of language are not the prerogative of litera-
ture; wherever they occur, their initial analysis is the same.
(Hasan 1989:92, emphasis added)
The words I have italicized in the segment, however, quite clearly tell us
that there is more to the analysis of literature than there is to that of other
uses of language. And both the first level of analysis, along with that ‘more’,
can be visually grasped straightaway in Figure 26.1 below.
Figure 26.1 The double-articulation framework (adapted from Hasan 1989:99)
A full account of the figure can be had elsewhere (primarily in Hasan 1989,
with synopses in Miller 2010, 2012, 2013, 2016a), but we can easily recognize
the preliminary level, what is labelled the semiotic system of language, as the
multiple coding system which, from the SFL perspective, is the valid entry
point for the analysis of any text. The system itself is made up of strata, each
of which is realized, i.e. becomes accessible to us, through the one below it. In
short, semantics, or meanings, are realized in lexicogrammar, or wordings,
which become accessible to us in phonology, or soundings, or, in the case of a
written text, in graphology, in written symbols.
But SSS is not content with merely identifying isolated patterns of lan-
guage in verbal art and the meanings they reveal; rather, it insists that such
patternings be demonstrated, at a higher order, to be ‘significant’. This
involves applying Mukařovský’s (1964) notion of ‘foregrounding’ as ‘motiv-
ated’ contrast (also see Halliday’s (2002a:131) concept of the ‘de-automatiza-
tion of grammar’); it involves a second order of meaning, where first-order
meanings are repatterned; it involves this model of what Hasan has called
‘double-articulation’ (Hasan 2007).
The higher, additional level is the semiotic system of verbal art. Its first
stratum, Verbalization, includes all of the lower system and builds upon it –
hence the broken line. This it does by means of the ‘symbolic articulation’
of the ‘Theme’. Looking at it from ‘below’, symbolic articulation is the place
where the basic first-order meanings are expanded upon, or ‘enriched, or
‘deepened’, and are made ‘art’. How it accomplishes this for Hasan is
through foregrounding (or the patterning of patterns). This is the process
by which the meanings emerging through analysis of the semiotic system
of language are symbolically turned into signs, for the purpose, the aes-
thetic motive, or ‘artistic intention’ (Mukařovský 1977) of expressing a
Theme: ‘what the text is about when dissociated from the particularities
of that text. . . . very close to a generalisation, which can be viewed as a
702 DONNA R. MILLER
hypothesis about some aspect of the life of social [wo]man’ (Hasan 1989:97).
Perhaps, as Hasan (1989:97) also notes, this is akin to what Aristotle had ‘in
mind when he declared that art is truer than history’. Be that as it may,
without this ‘Theme’, this reflection on humanity, and its symbolic articu-
lation, there is, for Hasan, no verbal art. I now aim at illustrating the
application of the framework just sketched.
26.4 An SSS Analysis of a Poem
I will now offer a brief sample of SSS analysis of the functions of fore-
grounded patterning in a short poem by D.H. Lawrence (see Miller 2007).
The text is the following:
Lonely, Lonesome, Loney – O!

1 When I hear somebody complain of being lonely
2 or, in American, lonesome,
3 I really wonder and wonder what they mean.
4 Do they mean they are a great deal alone?
5 But what is lovelier than to be alone?
6 escaping the petrol fumes of human conversation
7 and the exhaust smell of people
8 and be alone!
9 Be alone, and feel the trees silently growing.
10 Be alone, and see the moonlight outside, white and busy and
11 . . . silent.
12 Be quite alone, and feel the living cosmos softly rocking,
13 soothing, restoring, healing.
14 Soothed, restored, healed
15 when I am alone with the silent great cosmos
16 and there is no grating of people with their presences gnawing
17 at the stillness of the air.
(in Lawrence 1932)
I begin with the semiotic system of language, and, in the stratum of lexico-
grammar – the wordings instantiated in the poem that make its semantics
accessible to us – with transitivity. Table 26.1 shows how the –er roles, or
Doers, in the poem are divided between animate somebody and poetic voice I,
on one side, and inanimate natural phenomena, trees, moonlight, and the living
cosmos, on the other. People/human conversation are present, but substantially
construed as embedded Qualifiers, and so downranked grammatically and
demoted as participants (see Berry, this volume; Fontaine and Schönthal, this
volume) – e.g. the petrol fumes (of human conversation).
But what do the Doers do? In Stanza 1, a no better identified somebody
complain[s] of being lonely or, in American, lonesome. Apparently it is a
Table 26.1 Taxis, –er roles, Verbal Groups and process types2
clause –er roles VG process type
1 I hear ment:perc.
embedded somebody complain verbal
2 I wonder and wonder mental:cog.
embedded they mean verbal
3 they mean verbal
3a they are a great deal alone relational:att.
4 what is lovelier than ”
embedded - to be alone ”
” - escaping material
” - be alone relational:att.
” - be alone ”
” - feel mental:perc.
” trees growing behavioural
” - be alone relational:att.
” - see mental:perc.
” moonlight [that is] outside, white and busy and relational:1 circ.
silent + 3 att.
” - be quite alone relational:att.
” - feel mental:perc.
” living [that is] softly rocking, soothing, behavioural
cosmos restoring, healing
5 elliptical [I] [am] Soothed, restored, healed relational:att.
5a I am alone ”
5b there is no grating of people… existential
embedded presences gnawing material
reiterative verbal process that I hear[s] now and again and which causes
him/her, somewhat surprisingly, to wonder and wonder what they mean, with I
as perplexed Senser of the queried embedded Phenomenon with they as
Sayers.
Line 4 stands alone graphically; it consists of a rather ingenuous rhet-
orical question, once again concerning they’s meaning, one that functions to
highlight the response the poetic voice proceeds to provide in the second
rhetorical question (Line 5) and then to elaborate on in the rest of the poem.
Somebody/they then exits from the poem, and the I too fades, to reappear
however, symmetrically, in the poem’s final stanza.
Relational processes enter the text with the second interrogative and
reign uninterruptedly through Stanza 3 Line 9, excepting the non-finite
material Process, escaping, in Line 6. What is being escaped affords the first
appearance of the demoted participants signalled above: the petrol fumes [of
human conversation] and the exhaust smell [of people] – more fittingly com-
mented in terms of evaluation, below.
2
I read VG mean as agnate to want to say, a VGC with a desiderative α verb and verbal β verb.
704 DONNA R. MILLER
Even more crucially, along with the interrogatives, the Attribute alone
materializes as well. Its seven instances dominate the poem from Line
4 through Line 15, four times as Complement to the non-finite base
form of be. Real time withdraws as the relationals continue and embedd-
edness multiplies: there are no finite verbs from Line 6 to Line 15 and
the present participle assumes command. However, even finites in the
poem (first three stanzas and last) are not typical simple presents or
present-in-presents but rather whenever presents – spanning past/present/
future.
The three mental processes of perception in Stanza 4, also in base form,
likewise work to construe a sense of timelessness: feel the trees silently
growing; see the moonlight outside, white and busy and silent; feel the living cosmos
softly rocking, soothing, restoring, healing. Trees are earthly, natural phenom-
ena, though not human, while moonlight, and even more so the living cosmos,
are preternatural, uncanny Doers. I will take up the special salience of this
fourth stanza again below.
The final stanza – an incomplete clause complex, unless one provides the
‘Subject + Finite’ in Line 14 – reproposes the doings of the cosmos, now
Attributes of the symmetrically reintroduced I. The enhancing temporal
clause also links back symmetrically to that in the first stanza. The existen-
tial clause concluding the poem eliminates the negative presence of the
already established antagonist, people, again grammatically downgraded to
Qualifier, the negated –er role assumed by their presences, injuriously
gnawing.
But it is time to take a closer look at the evaluation enacted in the poem,
from the perspective of appraisal systems (Martin and White 2005; also
see Hood, this volume). Briefly, and appropriating Thompson’s (2014:50)
neat summary:
The central system is affect: the set of choices to do with ‘emotional

responses’ – expressing reactions to, and feelings about, things, such as
liking or fearing. This is ‘institutionalized’, in Martin’s (2000: 147) term, in
two other systems, judgement and appreciation. judgement is the
realm of ethical and moral assessments of human behaviour, drawing on
and constructing ‘norms about how people should and shouldn’t behave’
(Martin, 2000: 155). appreciation, on the other hand, is the realm of
aesthetic assessments of ‘products, performances and naturally occurring
phenomena’. (2000: 159)
Such evaluation can be explicit (inscribed) or implicit (invoked) in a text and

carry positive (+ve) or negative (-ve) polarity. The categories are also mod-
elled into sub- and sub-sub-classifications, but for our purposes this account
will suffice.
I would note here at the start that a reader unacquainted with the writings
of D. H. Lawrence might well be excused for considering potential irony being
enacted in Line 5’s rhetorical question But what is lovelier than to be alone? – an
insincerity that those who know Lawrence’s work would never assume.
Regarding being alone, we can immediately posit +ve judgement with impun-
ity. Yet familiarity with the writer is hardly requisite, as the question itself,
taken seriously, implies such evaluation. In addition, the fast-accumulating
supporting evidence compounds and fully endorses this reading.
Non-Finites are of course ambiguous in their logico-semantic relations,
and the indistinctness here before ‘escaping’ is reinforced by the non-finite
base form be in Line 8. I hypothesize a Cause:Purpose enhancing relation
such as so that [you’re escaping/you can escape]; thus escaping is also being
evaluated in terms of +ve judgement and may also be seen as an indirect
expression of the +ve judgement of being alone, and an illustration of
what Thompson (2014) has called the ‘Russian doll dilemma’ in appraisal
analysis.3 As one typically expects, what is escaped is inherently negative in
assessment: the noxious car odours. However, the human Qualifiers con-
strue a paradox that enacts further -ve appreciation; this evaluation too
feeds back into the overriding +ve judgement of being alone.
The Cause:Purpose interpretation of escaping with reference to being
alone above is not unlike the meanings of Stanza 4’s looser paratactical
and linking the base forms of be with those of the Processes of percep-
tion. Being alone affords awareness of those uncanny personified par-
ticipants: +ve appreciation is enacted of moonlight outside, white and busy
and . . . silent; whereas I read +ve judgement of those more patent doings
of this inhuman but living cosmos softly rocking, soothing, restoring, healing.
And these assessments once again token the overall +ve judgement of
being alone.
In the first line of the fifth and final stanza these present participles
become past, and function as Attributes of (once again explicit) I, now
personally alone with the silent great cosmos, which as Thing is explicitly
evaluated with +ve appreciation – and which is a further token for the
global +ve judgement of being alone. The final Lines assert the now in-
Existent grating of people with their presences gnawing at the stillness of the air,
with negative soundings enhancing their meanings. Enacted here is
another -ve judgement on the behaviour of people and their presences, one
which again indirectly expresses that overarching +ve judgement of
being alone. The dolls might also be visually represented as in (my admit-
tedly over-simplified) Figure 26.2.
At this point a first formulation of the Theme of the poem can be
advanced. Solitude as bliss, diametrically opposed, or contratextual (Martin
3
As Thompson puts it (2014:49), ‘This relates to the way in which an expression of one category of attitude may
function as a token (an indirect expression) of a different category; and that token may itself function as an indirect
expression of yet another category, and so on. This raises the question of how many of these layers, one inside the
other, should be included in an analysis – and how to code the different layers.’ This ‘refinement’ of the systems is
remarked here, as the poem presents a noteworthy case in point but, given the not-necessarily expert-in-SFL nature of
our audience, is not analyzed in-depth.
706 DONNA R. MILLER
+ve +VE:moonlight outside, white and busy and silent/the

judgement: living cosmos soly rocking, soothing, restoring, healing/
BEING ALONE I -soothed, restored, healed -alone with the silent great
cosmos
-VE:the petrol fumes of human conversaon
and the exhaust smell of people/ grang of
people with their presences gnawing
Figure 26.2 The ‘Russian dolls’ of appraisal in the poem
1986), to the dominant Western cultural paradigm which sees ‘solitude as

sadness’ (as e.g. in Brontë’s Villette). But something more can be said.
To do so we need to move the analysis to the semiotic system of verbal art
and the symbolic articulation of Theme. As pre-announced, I will discuss
this in terms of GP as foregrounding, or the patterning of patterns. Jakob-
son (1960:358–9) sees this marked reiteration at the syntagmatic level of
(phonological, morphological, syntactic, and lexical) form as ‘the empirical
criterion of the poetic function’. The ultimate significance of grammatical
reiteration, however, is its capacity to call forth a corresponding recurrence
of ‘sense’, so that GP, according to Jakobson, following up on Hopkins’ 1865
insight, is seen to construe also, and at the same time, semantic parallelism
(SP). And GP/SP in this text unmistakeably qualify for the label of ‘pervasive
parallelism’ (Jakobson 1966:423). Leaving aside possibly relevant phonemic
reiteration, Table 26.2 describes and quantifies the main instances of GP
according to rank.
And here in this obviously pervasive grammatical and semantic parallel-
ism, we recognize the already commented strongly evaluated elements.
Clearly the ‘Theme’ as already hypothesized is being steadily hammered
home through the reiteration of key elements at all levels of the rank scale.
But we also need to focus on the semantic location of these.
Table 26.2 Overview of grammatical parallelism in the poem, by rank
morphemes -ing forms: escaping as action – foregrounded as material process and

against ACTs of grating/gnawing of people
-ed forms: past participles as Attributes: Soothed, restored, healed –
quasi-synonyms
words alone (7), quasi-synonymy with lonely/lonesome; silent (2) silently (1) –
lexical scatter – quasi-synonymy with softly/stillness;
feel (2); cosmos (2); people (2); conjunctive adjunct ‘and’ (9)
groups paratactic extension of: VG (Stanza 1); NG (Stanza 3); base forms/Attributes
(Stanza 4);
Predicator (of base forms)^Complement
phrases Qualifiers: of human conversation/of people (2)/ of the air; one
nominalized non-finite: of being lonely in circumstance: Matter of
complaint
clauses opening + closing hypotactic enhancing ‘when’ clauses;
7 non-finite base form clauses; 4 Be alone – one intensified (quite) – plus
feel (2) and see (1): extended with and once in Stanza 3 and three times
in Stanza 4;
15 embedded clauses, due in particular to base forms and embedded non-
finite clauses in Stanza 4: rocking, soothing, restoring, healing – last 3
quasi-synonyms, plus others: escaping, growing, gnawing
Stanzas 3 and 5 see-saw between alone-ness and what is escaped, thus

enacting the conflict. The fourth stanza, however, is in itself foregrounded
as a key semantic location of symbolic articulation, due to the recurrence of
various mechanisms, all severally pointed out above: unearthly, inhuman
participants; timelessness (non-finites, especially present participles – six of
the nine in the poem are here, and non-finite base forms – six of the seven
in the poem are here); softly and silently restorative doings; embeddedness,
but also notable paratactic extension (five of the nine occurrences of and in
the poem) are here – all working rhythmically, I submit, to mean in a way
that is quintessentially ‘Lawrentian’ and that Lawrence (1968:276) himself
assessed:
In point of style, fault is often found with the continual, slightly modified
repetition. The only answer is that it is natural to the author; and that every
natural crisis in emotion or passion or understanding comes from this
pulsing, frictional to-and-fro, which works up to culmination.
The final stanza slows things down again, and winds things up.
In order to test the validity of the Theme and complete its formulation,
the three ingredients of the context of creation of verbal art need to be
probed. We have just now glossed the first of these: the language of its
writer. His position re the artistic conventions of his time can be said to be
‘contra’ as well: beginning as a Georgian poet, he fast rebelled against
allowing any formal strictures to his art. With reference to the worldview
of Lawrence, we have seen his contratextual stance re solitude, but the
poem gives us more.
708 DONNA R. MILLER
If we recall the mechanized nature of what is being escaped: the petrol

fumes of human conversation and the exhaust smell of people – we get closer to
a fuller Theme of the poem. In Miller (2007), I wrote: ‘In Lawrence,
admittedly, there is also a rhetoric of ridicule, of invective, of railing
against the collapse of civilization, as Lawrence wanted it to be’, as he
idealized it should be. The sole solution for the individual in his estima-
tion was to draw apart in single and proud solitude – and to rail. This is
one of his ‘Last Poems’, by which time he had wholly given up on – and
cut himself off from – his fellow human beings, judged incapable of
receiving his prophetic, salvific message. The collapse of pre-
industrialized civilization and the model response of the single person
in the face of that collapse is indeed a staple Theme of Lawrence’s
copious works and integral to his ‘repertoire’ (e.g. Martin 2010). Beyond
the palpable ‘ode to solitude’ enacted, this may be said to be, ultimately,
the Theme of this poem as well – which might now be indeed reformu-
lated as Single, proud solitude as the ideal response of the individual to the
collapse of our civilization.
26.5 Recent Research and Future Directions
Space precludes lengthy treatment of recent studies, but the work of

several scholars within the SFL (if not always strictly ‘verbal art’) frame-
work should be cited here: foremost among these are Butt (1988), Lukin
and Webster (2005) in particular, on the Australian scene, Banks (2011) in
France, and, in Italy, Taylor Torsello (2007, 2016), Turci (2007, 2010),
Luporini (2016), and Miller and Turci (2007). Also deserving mention, for
their work on the translation of verbal art from the SFL perspective, the
publications of Manfredi (2012, focusing on its teaching, 2014) and Swain
(2014) in Italy and of Lukin and Pagano (2012). Other recent work can be
found in Wegener et al. (2018). Likely omissions are of course my own
responsibility.
In future, to begin with, my own efforts to champion the verbal art
framework – with Jakobson’s GP formally slotted into it – will remain a
priority. If Hasan has continuously acknowledged her debt to Mukařovský,
due to the incontestable correspondences between the two, the same atten-
tion is due to Jakobson, and for the same reason.
But the extent to which CL methods can/should be a part of the
ongoing development of a rigorous SSS is also a topic of future, and
indeed present, work by Miller and Luporini (2015, 2018) and Miller
(2016b). Quantitative stylistic studies are of course not new. In a search
for SFL-ers on the topic, one finds Bednarek (2008) offering a brief
introductory ‘Language+Literature+CL Methods 101’ from an SFL per-
spective, if not a Hasanian one, and Toolan (2009) making the case for
the analytic advantages of having corpus evidence to support qualitative
findings. Here too, incidentally, the name of Hasan only appears linked
to that of Halliday, and in terms of cohesion.
The essential query is whether foregrounding – at least at a preliminary
investigative stage – is in fact quantifiable. We recall Halliday’s (2002b:102–3)
own explicit caveat that prominence as motivated foregrounding is not
tantamount to mere statistical frequency. Also to be recollected is that auto-
matic analyses cannot give us the findings we need at the ‘higher’ levels of
semantics and context (Halliday and Matthiessen 2004:48–9) – no matter
what the text type. Yet also to be recalled is that Halliday did see ‘counting’
the linguistic choices of a writer as a step towards establishing the potential
prominence of patterns: for determining what features deserve further inves-
tigation re motivation. Indeed, as Miller and Luporini (2015, 2018) suggest,
the extent to which foregrounding is quantifiable would seem to be inscrut-
able without the assistance of corpus linguistic methods. But their findings
show that it is not enough; it plays but a supporting role, guaranteeing data
accuracy and statistical significance that cannot be manually achieved in
longer texts, but by no means supplanting the labour-intensive manual
analysis of co-textual logogenesis that SSS requires.4
26.6 In Closing
Our excursus ends here, though with no claim to comprehensiveness of all

that could or should have been said. Despite the aim of taking an objective
critical stance to the issues dealt with in this chapter, inevitably my own
partiality to the verbal art framework will have come to the fore. I too
believe that literature is a ‘special’ variety of language use and thus that it
only makes sense to work within a framework that is based on that
recognition and aims at investigating just what it is that makes such
language literature, or ‘verbal’ art, i.e. ‘language that is artistic and art that
is linguistic’ (Hasan 2014, personal communication). To my knowledge, SSS
is the only model available for doing this.
The SSS analytical model is an appliable linguistics, demandingly rigor-
ous in its method, but fully open to scrutiny, and so its steps are amply
retrievable and replicable (see Simpson 2014:4). It is enabling for the ana-
lyst, as it frees her from the authority of ‘lit crit’ and shifts her allegiance
solely to the literature text and its context(s). It provides a functional
metalanguage for talking about this special kind of languaging and rich
rewards to those willing to make the admittedly requisite effort to acquire
expertise in its practice.
4
Re the accuracy of CL findings, one needs to bear in mind that in relying excessively on frequency, one risks missing
significant non-frequent foregrounded items, and so vital parts of the full picture. There are those who imagine a brave
new digital world in which the future powers of algorithms would make Fillmore’s (1992) armchair linguist redundant.
I am not among them.
710 DONNA R. MILLER
With my contribution to this volume, I shall be content if I have managed

to ‘broadcast’ the SSS model of literature satisfactorily; if I have also man-
aged to impinge on the complacency and/or stimulate the curiosity of any
(actual or potential) stylisticians out there, I shall, of course, be delighted.
References
Banks, D. 2011. Comprendre l’incompréhensible: analyse d’un poème de

J. H. Prynne. In D. Banks, ed., Aspects linguistiques du texte poétique. Paris:
L’Harmattan. 219–29.
Bednarek, M. 2008. Teaching English Literature and Linguistics Using
Corpus Stylistic Methods. In Bridging Discourses: ASFLA 2007 Online Proceed-
ings. Available online at: www.asfla.org.au/category/asfla2007. (Last
accessed 06/05/2015.)
Busse, B. and D. McIntyre. 2010. Language, Literature and Stylistics. In
D. McIntyre and B. Busse, eds., Language and Style. Basingstoke: Palgrave. 3–14.
Butler, C. 2003. Structure and Function: A Guide to Three Major Structural-
functional Theories, Part 2: From Clause to Discourse and Beyond. Amsterdam:
John Benjamins.
Butt, D. 1988. Randomness, Order and the Latent Patterning of Text. In
D. Birch and M. O’Toole, eds., Functions of Style. London: Pinter. 74–97.
Butt, D. and A. Lukin. 2009. Stylistic Analysis and Arguments against
Randomness. In M. A. K. Halliday and J. J. Webster, eds., Continuum
Companion to Systemic Functional Linguistics. London: Continuum. 190–215.
Caton, S. C. 1987. Contributions of Roman Jakobson. Annual Review of
Anthropology 16: 223–60.
Crystal, D. and D. Davies. 1969. Investigating English Style. Bloomington:
Indiana University Press.
Enkvist, N. E. 1973. Linguistic Stylistics. The Hague: Mouton de Gruyter.
Fillmore, C. 1992. Corpus Linguistics or Computer-aided Armchair Linguis-
tics. In J. Svartvik, ed., Directions in Corpus Linguistics. Berlin: Mouton de
Gruyter. 13–38.
Fowler, R., ed. 1966. Essays on Style and Language. London: Routledge.
Fowler, R. 1981. Literature as Social Discourse: The Practice of Linguistic Criticism.
London: Batsford.
Fowler, R. 1986. Linguistic Criticism. Oxford: Oxford University Press.
Fowler, R. and F. W. Bateson. 1967. Argument II: Literature and Linguistics.
Essays in Criticism 17: 322–47.
Fowler, R. and F. W. Bateson. 1968. Argument II (continued): Language and
Literature. Essays in Criticism 18: 164–82.
Halliday, M. A. K. 1978. Language as Social Semiotic. London: Arnold.
Halliday, M. A. K. 2002a. The De-Automatization of Grammar: From
Priestley’s ‘An Inspector Calls’. In J. J. Webster, ed., The Collected Works
of M. A. K. Halliday, Volume 2: Linguistic Studies of Text and Discourse. London:

Halliday, M. A. K. 2002b. Linguistic Function and Literary Style: An Inquiry
into the Language of William Golding’s The Inheritors. In J. J. Webster, ed.,
The Collected Works of M. A. K. Halliday, Volume 2: Linguistic Studies of Text and
Discourse. London: Continuum. 88–125.
Halliday, M. A. K., A. MacIntosh, and P. Strevens. 1964. The Linguistic Sciences
and Language Teaching. London: Longman.
Halliday, M. A. K., and C. M. I. M. Matthiessen. 2004. An Introduction to
Functional Grammar. 3rd ed. London: Arnold.
Hasan, R. 1964. A Linguistic Study of Contrasting Linguistic Features in the Style of
Two Contemporary English Prose Writers. PhD Thesis, University of
Edinburgh.
Hasan, R. 1971. Rime and Reason in Literature. In S. Chatman, ed., Literary
Style: A Symposium. London: Oxford University Press. 299–329.
Hasan, R. 1975. The Place of Stylistics in the Study of Verbal Art. In
H. Ringbom, ed., Style and Text. Amsterdam: Skriptor. 49–62.
Hasan, R. 1989. Linguistics, Language and Verbal Art. Oxford: Oxford Univer-
sity Press.
Hasan, R. 1996. Teaching Literature across Cultures. In J. E. James, ed., The
Language–Culture Connection. Singapore: SEAMO Regional Language
Centre. 34–63.
Hasan, R. 2007. Private Pleasure, Public Discourse: Reflections on Engaging
with Literature. In D. R. Miller and M. Turci, eds., Language and Verbal Art
Revisited: Linguistic Approaches to the Study of Literature. Sheffield: Equinox.
41–67.
Hatzfeld, H. A. 1953. A Critical Bibliography of the New Stylistics Applied to the
Romance Literatures, 1900–1952. Chapel Hill: University of North Carolina
Press.
Jakobson, R. 1960. Closing Statement: Linguistics and Poetics. In T. A.
Sebeok, ed., Style in Language. Cambridge, MA: MIT Press. 350–77.
Jakobson, R. 1966. Grammatical Parallelism and Its Russian Facet. Language
42(2): 399–429.
Lawrence, D. H. 1932. Last Poems. Florence: Giuseppe Orioli.
Lawrence, D. H. 1968. Foreword to Women in Love. In W. Roberts and H. T.
Moore, eds., Phoenix II: Unpublished and other Prose Works by D.H. Lawrence.
London: Heinemann. 275–6.
Lecercle, J. 1993. The Current State of Stylistics. The European English Messen-
ger 2(1): 14–18.
Leech, G. N. and M. H. Short. 2007. Style in Fiction: A Linguistic Introduction to
English Fictional Prose. London: Longman.
Lukin, A. 2015. A Linguistics of Style: Halliday on Literature. In J. J. Web-
ster, ed., The Bloomsbury Companion to M. A. K. Halliday. London: Blooms-
bury. 348–69.
712 DONNA R. MILLER
Lukin, A. and A. Pagano. 2012. Context and Double Articulation in the

Translation of Verbal Art. In J. Knox, ed., To Boldly Proceed, Selected Proceed-
ings from the 39th International Systemic Functional Linguistics Congress.
Sydney: UTS. 123–8.
Lukin, A. and J. J. Webster. 2005. Systemic Functional Linguistics and the
Study of Literature. In R. Hasan, C. M. I. M. Matthiessen, and J. J. Webster,
eds., Continuing Discourse on Language: A Functional Perspective. Sheffield:
Equinox. 413–56.
Luporini, A. 2016. Spotlighting Fantasy Literature with the Tools of Frame
Semantics and Systemic Functional Linguistics: A Case Study. Quaderni del
CeSLiC: Occasional Papers. Bologna: Centro di Studi Linguistico-Culturali
(CeSLiC) e Alma Mater Studiorum, Università di Bologna. ALMADL – Area
Sistemi Dipartimentali e Documentali. DOI: 10.6092/unibo/amsacta/5162.
Malinowski, B. 1923. The Problem of Meaning in Primitive Languages. In C. K.
Ogden and I. A. Richards, eds., The Meaning of Meaning. London: Kegan Paul.
Malinowski, B. 1935. Coral Gardens and Their Magic. London: Allen and
Unwin.
Manfredi, M. 2012 Description vs Prescription in Translation Teaching:
A Bridgeable Gulf? In F. Dalziel, S. Gesuato, and M. T. Musacchio, eds.,
A Lifetime of English Studies. Essays in Honour of Carol Taylor Torsello. Padua: Il
Poligrafo. 545–53.
Manfredi, M. 2014. Translating Text and Context: Translation Studies and
Systemic Functional Linguistics, Volume 2: From Theory to Practice. 2nd ed.
Bologna: Asterisco.
Martin, J. R. 1986. Grammaticalising Ecology: The Politics of Baby Seals and
Kangaroos. In T. Threadgold, E. A. Grosz, G. Kress, and M. A. K. Halliday,
eds., Language, Semiotics, Ideology. Sydney: Pathfinder Press. 225–67.
S. Hunston and G. Thompson, eds., Evaluation in Text. Oxford: Oxford
English. Houndmills: Palgrave Macmillan.
eds., New Discourse on Language Functional Perspectives on Multimodality,
Identity, and Affiliation. London: Continuum. 1–34.
McIntyre, D. 2012. Linguistics and Literature: Stylistics as a Tool for the
Literary Critic. SRC Working Papers 1: 1–11.
Miller, D. R. 1998. Insegnando la lingua speciale del testo letterario:
l’approccio sociosemiotico. In M. Pavesi and G. Bernini, eds., L’apprendi-
mento linguistico all’Università: le lingue speciali. Roma: Bulzoni. 271–93.
Miller, D. R. 2007. Construing the ‘Primitive’ Primitively: Grammatical Par-
allelism as Patterning and Positioning Strategy in D. H. Lawrence. In D. R.
Miller and M. Turci, eds., Language and Verbal Art Revisited: Linguistic
Approaches to the Study of Literature. Sheffield: Equinox. 41–67.
Miller, D. R. 2010. The Hasanian Framework for the Study of ‘Verbal Art’
Revisited . . . and Reproposed. In J. Douthwaite and K. Wales, eds., Stylistics
and Co. (Unlimited): The Range, Methods and Applications of Stylistics. Textus 23
(1): 71–94.
Miller, D. R. 2012. Slotting Jakobson into the Social Semiotic Approach to
‘Verbal Art’: A Modest Proposal. In F. Dalziel, S. Gesuato, and M. T.
Musacchio, eds., A Lifetime of English Studies. Essays in Honour of Carol Taylor
Torsello. Padua: Il Poligrafo. 215–26.
Miller, D. R. 2013. Another Look at Social Semiotic Stylistics: Coupling
Hasan’s ‘Verbal Art’ Framework with ‘the Mukařovský-Jakobson Theory’.
In C. A. M. Gouveia and M. F. Alexandre, eds., Languages, Metalanguages,
Modalities, Cultures: Functional and Socio-discoursive Perspectives. Lisbon:
BonD. 121–40.
Miller, D. R. 2016a. Jakobson’s Place in Hasan’s Social Semiotic Stylistics:
‘Pervasive Parallelism’ as Symbolic Articulation of Theme. In W. L. Bow-
cher and J. Y. Liang, eds., Society in Language, Language in Society: Essays in
Honour of Ruqaiya Hasan. London: Palgrave Macmillan.
Miller, D. R. 2016b. On Negotiating the Hurdles of Corpus-assisted Appraisal
Analysis in Verbal Art. In S. Gardner and S. Alsop, eds., Systemic Functional
Linguistics in the Digital Age. Sheffield: Equinox. 211–28.
Miller, D. R. and A. Luporini. 2015. Social Semiotic Stylistics and the Corpus:
How Do-able is an Automated Analysis of Verbal Art? In A. Duguid,
A. Marchi, A. Partington, and C. Taylor, eds., Gentle Obsessions: Literature,
Linguistics and Learning: In Honour of John Morley. Roma: Artemide Edizioni.
235–50.
Miller, D. R. and A. Luporini. 2018. Software-assisted Systemic Socio-
Semantic Stylistics: Appraising tru* in J.M. Coetzee’s Foe. In R. Wegener,
A. Oesterle, and R. Neumann, eds., On Verbal Art: Essays in Honour of
Ruqaiya Hasan. Sheffield: Equinox.
Miller, D. R. and M. Turci, eds. 2007. Language and Verbal Art Revisited:
Linguistic Approaches to the Study of Literature. Sheffield: Equinox.
Mukařovský, J. 1964. Standard Language and Poetic Language. In P. L. Garvin,
ed. and trans., A Prague School Reader on Aesthetics, Literary Structure and Style.
Washington, DC: Georgetown University Press. 17–30.
Mukařovský, J. 1977. The Word and Verbal Art: Selected Essays by Jan Mukař-
ovský. Edited and translated by J. Burbank and P. Steiner. New Haven:
Yale University Press.
Mukařovský, J. 1978. Structure, Sign and Function: Selected Essays by Jan Mukař-
ovský. Edited and translated by J. Burbank and P. Steiner. New Haven:
Selden, R. 1989. The Cambridge History of Literary Criticism, Vol. 8: From Formal-
ism to Poststructuralism. Cambridge: Cambridge University Press.
Simpson, P. 2014. Stylistics. 2nd ed. London: Routledge.
Stankiewicz, E. 1983. Roman Jakobson, Teacher and Scholar. In A Tribute to
Roman Jakobson 1896–1982. Berlin: Walter de Gruyter. 17–26.
714 DONNA R. MILLER
Swain, E. 2014. Translating Metaphor in Literary Texts: An Intertextual

Approach. In D. R. Miller and E. Monti, eds., Tradurre Figure/Translating
Figurative Language. Bologna: Bononia University Press. 241–54.
Taylor Torsello, C. 2007. Projection in Literary and Non-literary Texts. In
D. R. Miller and M. Turci, eds., Language and Verbal Art Revisited: Linguistic
Approaches to the Study of Literature. Sheffield: Equinox. 115–48.
Taylor Torsello, C. 2016. Woolf’s Lecture/Novel/Essay A Room of One’s Own. In
D. R. Miller and P. Bayley, eds., Hybridity in Systemic Functional Linguistics:
Grammar, Text and Discursive Context. Sheffield: Equinox. 240–67.
Thompson, G. 2014. Affect and Emotion, Target-value Mismatches,
and Russian Dolls: Refining the Appraisal Model. In G. Thompson and
L. Alba-Juez, eds., Evaluation in Context. Amsterdam: John Benjamins.
47–66.
Toolan, M. 2001. Narrative: A Critical Linguistic Introduction. 2nd ed. Abingdon:
Routledge.
Toolan, M. 2009. Narrative Progression in the Short Story: A Corpus Stylistic
Approach. Amsterdam: John Benjamins.
Turci, M. 2007. The Meaning of Dark* in Joseph Conrad’s Heart of Darkness.
In D. R. Miller and M. Turci, eds., Language and Verbal Art Revisited:
Linguistic Approaches to the Study of Literature. Sheffield: Equinox. 97–114.
Turci, M. 2010. The Literary Text at the Borders of Linguistics and Culture:
A SF Analysis of Les Murray’s Migratory. In E. Swain, ed., Thresholds and
Potentialities of Systemic Functional Linguistics: Applications to Multilingual,
Multimodal and Other Specialised Discourses. Trieste: Edizioni Università di
Trieste. 334–46.
Wales, K. 2001. A Dictionary of Stylistics. 2nd ed. Harlow: Pearson Education.
Wegener, R., S. Neumann, and A. Oesterle, eds. 2018. On Verbal Art: Essays in
Honour of Ruqaiya Hasan. Sheffield: Equinox.
Wimsatt, W. K. and M. Beardsley. 1949. The Affective Fallacy. Sewanee
Review 57(1): 31–55.
Yahya Ali Bani Salameh, M. 2010. Stylistic Manipulations in Henry James’ Novel
The Portrait of a Lady. PhD Thesis, Department of Linguistics, Aligarh
Muslim University.
27
Language and Social
Media
Enacting Identity through Ambient Affiliation
Michele Zappavigna
27.1 Introduction
Social media services are web-based technologies that support online

networking amongst their users. This chapter explores how SFL can
illuminate the types of ‘ambient affiliation’ (Zappavigna 2011, 2012,
2018) that are central to this social networking in terms of the new
forms of sociality that are being enacted. Social networking sites (SNS),
one of the main forms of social media, allow relationships to be estab-
lished between user accounts in the form of ‘friendship’ or ‘following’
connections (Boyd and Ellison 2007). Different kinds of semiotic associ-
ations can also be generated in the discourse produced by social media
users through, for example, the affordances of ‘metadata’ (data about
data, e.g. location information), sometimes referred to as ‘conversational
tagging’ (Huang et al. 2010) or ‘social tagging’. Common examples of
SNS are Facebook, Twitter, Weibo, and Instagram. While platforms
constantly change with both social and technological imperatives, what
is interesting to the social semiotician is the types of multimodal
resources that are evolving as we find new ways to establish ‘commu-
nion of feeling’ (Firth 1957) with other language users in these online
environments.
This chapter will begin by reviewing some of the main communicative
features of social media discourse, before considering research into social
media identities across different disciplines. It will then explore the ‘user in
uses’ perspective for exploring identity, inspired by Firth (1957), that has
arisen out of SFL research in the areas of ‘individuation’ and ‘affiliation’
(Martin 2010). This research has been focused on understanding how iden-
tities are enacted and how social bonds are forged interactively in discourse.
The chapter concludes by surveying issues of collecting and analyzing social
media corpora as linguistic data sources for both qualitative and quantita-
tive research projects.
716 M I C H E L E Z A P PAVI G N A
27.2 Social Media Discourse
The main type of texts shared with ambient audiences using social net-
working services are ‘posts’: multimedia content typically arranged into
chronological ‘streams’ or ‘feeds’. In the case of microblogging services
such as Twitter, these posts are short character-constrained messages
that originally functioned predominately as status updates relating
to users’ activities. These have now taken on a broad range of communi-
cative functions, from broadcasting ideational ‘content’, to sharing feel-
ings. An example of a post, in this case a tweet, with a prominent
interpersonal function is the following comment on the 2014 Australian
federal budget in (1):
(1) Cruel callous cronyism #ThreeWordBudget #Budget2014
This is also an example of a Twitter meme, a form of linguistic and multi-

modal play whereby users are invited to contribute iterations on a theme,
often as a phrasal template (see further discussion of memes in Section
27.4.2 below). Social media posts of this kind are usually presented as a
temporally unfolding ‘social stream’, incorporating various types of multi-
media such as images (see Figure 27.1) and video.
27.2.1 Searchable Talk: Social Media Metadata

The drive to make our discourse searchable by using metadata such as
#Budget2014 has become a prominent social impetus, realized by a range
of online communicative practices. With the advent of social media
services this relation has intensified in a movement toward ‘searchable
talk’ (Zappavigna 2012, 2015, 2018), communication featuring collaborative
metadata embedded in social media texts and visible within the main
content of a post. Metadata is information that describes dimensions of a
Figure 27.1 An example of an Instagram feed (left) and an individual post (right)
Language and Social Media 717
data source, for instance, information about the author of a social media
post or the location where it was created. While metadata has a long history
in the domain of information management, this is the first historical period
where we see it so closely tied to enacting social relations, having extended
its semiotic reach as an information-organizing tool to a social resource for
building relationships and communities.
A popular form of ‘social metadata’, or ‘social tagging’, is the hashtag,
indicated with a # symbol followed by a keyword or concatenated phrase
or clause (for an analysis of the linguistic functions of hashtags, see
Zappavigna 2015, 2018). The tweet presented in the previous section
contains two instances of social tagging in the form of hashtags (#Three-
WordBudget and #Budget2014) used to indicate both the semantic domain
of the post, and to designate the post as part of a larger Twitter meme.
Hashtags of this kind are also able to realize a range of complex interper-
sonal and textual functions beyond such topic-marking. Hashtags as a
form of conversational tagging enable individuals to search social media
discourse to find out what people are saying about particular domains, or
to share feelings and opinions with like-minded users (or argue with those
who do not share your worldview). In this way, social metadata supports
forms of ambient communion that arise out of the ability to search for
and engage with other people’s posts in ‘real-time’ within the social
stream.
What makes social metadata particularly interesting to linguists is its
capacity to infiltrate the linguistic structure of the texts that it seeks to
annotate. While traditional metadata is typically hidden from the view of
users of an information system, or separated from the main body of a text
in some systematic way, social metadata is incorporated into social media
communication, and can perform a wide range of functional roles in the
discourse itself (Zappavigna 2015). While social media services collect
many forms of traditional metadata about the networks, users, and con-
tent that they manage, most of this information is not presented within
the main content of a post and certainly does not form part of the
linguistic function and structure of a post. In stark contrast, social meta-
data is user-generated and typically acts as a kind of ‘in-text’ tagging
with a range of novel communicative functions. For example, Zappavigna
(2014b) has explored how hashtags support attitudinal alignment around
iconized dimensions of experience, such as a positive attitude regarding
coffee:
(2) I do love #morningcoffee #coffee #coffeelover #blackcoffee
The tweet above presents a coupling of positive affect with #coffee, as well as
using related coffee tags to identify the social media user as a ‘coffeelover’.
It enacts alignment around coffee as a ‘bonding icon’ (Stenglin 2008) and
foregrounds coffee as part of this user’s identity performance, concepts that
we will explore in the second half of this chapter.
27.2.2 Social Streams and ‘Real-Time’: Streaming Data

and Temporality
Time is an important dimension in social media, and feeds of social media
content are sometimes referred to as ‘social streams’. For example, Twitter
describes itself as ‘a real-time information network’ (Twitter 2013) and
presents instant engagement with its content as a critical affordance.
‘Real-time’ as a concept ascribes value to an apparent increase in the pace
at which we engage with Internet-mediated resources, and with members
of our social networks. ‘Streaming data’ is information that unfolds in
sequence, typically over chronological time. For example, blogs and micro-
blogs are displayed in reverse chronological order and display timestamp
information indicating when the post was published. It is now common for
feeds of ‘live tweets’ to appear in news broadcasts, or during programmes
involving audience commentary, such as chat shows and panels. Because of
the continuous unfolding of social streams, they are associated with
listening practices where users ‘tune in and out’, consuming content in an
ad hoc manner, rather than tracking a feed exhaustively (Crawford 2009).
The term ‘real-time’ is often used uncritically as synonymous to ‘clock-
time’ or ‘synchronous time’, where social media users are posting about
events almost ‘as they happen’ via mobile media. The classifier ‘real’,
however, also has a legitimating function: web-based time is positioned as
being the equal to the ‘offline’ time of face-to-face social interaction. At the
same time, it is ‘hyper-real’, potentially affording superior access to what is
happening in our social networks in any given moment, though this pos-
ition is clearly available for critique.
27.2.3 Interdisciplinary Communication Research into Social Media

Research into social media is a broad interdisciplinary arena that has arisen
out of wider interest in using ‘big data’ as a lens through which to examine
the kinds social practices that are developing alongside digital technologies
(Boyd and Crawford 2012). Social media make available vast quantities of
naturally occurring discourse, together with complex metadata that can be
collected, stored, and analyzed. Communication-focused research into
social media forms part of a broader field known as Computer-Mediated
Communication (CMC) that brings together a range of communication
theories focused on understanding electronic discourse. However, because
of the rapid development of social media technologies, they can be some-
what of a ‘moving target’ for scholars (Hogan and Quan-Haase 2010).
Linguistic perspectives on social media have largely been developed within
sociolinguistics, pragmatics, and computational linguistics. Outside linguis-
tics, work has spanned disciplines such as sociology, informatics, and
computer science. Research employing SFL is an emergent area, with early
approaches focused on issues of identity and affiliation of the kind surveyed
in the following sections of this chapter.
27.3 Individuation: Enacting Identity with Social Media
27.3.1 Interdisciplinary Research into Social Media Identities

It is unsurprising that issues of identity are the focus of much social media
research, given the important function these channels have accrued in
terms of digital self-presentation. Identity is a difficult concept to pin down
as enacting identity, or doing ‘identity work’ (Benwell and Stokoe 2007),
means different things across disciplines. Aside from research into personal
uses of social media, there has also been great interest in professional
identities, branding, and micro-celebrity (Gilpin 2010; Marwick 2011; Page
2012a). One obvious way in which social media explicitly construes repre-
sentations of the self is through profile descriptions, which often include
the social media user’s ‘demographic information, interests and relation-
ship status along with a photograph and a self-description’ (Boyd and Heer
2006:2). More nuanced dimensions of identity that have been considered
within the broad interdisciplinary field of social media research include
ethno-racial representation (Grasmuck et al. 2009); sexual identity (Duguay
2014); gender, in particular, in relation to social media photographs (Albury
2015; Rose et al. 2012), and also lexical variation (Bamman et al. 2014); age,
in particular, the teenage years (Boyd 2014; Davis 2011); and location
variables (Schwartz and Halegoua 2014). An important theme is the issue
of privacy in terms of how much information about the self is revealed
when using an SNS (Madden 2012).
Identifying and predicating demographic variation from social media
sources is seen as valuable in areas from marketing to legal inquiry. Dimen-
sions that have been identified as important include contextual variables
such as location (e.g. location-based prediction using geotagged corpora)
and time (using the temporal metadata afforded by social media services).
In addition, variables relating to different orders of group membership
have been viewed as relevant. For example, studies have considered the
effect of age and gender on blogging (Argamon et al. 2007; Lu et al. 2010;
Rustagi et al. 2009; Schler et al. 2006).
27.3.2 Sociolinguistic Approaches

A body of sociolinguistic studies has begun to develop considering how
identity is construed in social media texts in terms of language variation.
Seargeant and Taggs’ (2014) edited volume brings together key work in this
area, including perspectives on the performance of the ‘ludic’ self (Deumert
2014), humour and impersonation (Page 2014), user-generated online
reviews (Vásquez 2014), self-presentation and social positioning (Lee
2014), and linguistic and discursive heterogeneity in terms of entextualiza-
tion and resemiotization (Leppänen et al. 2014). Some of this type of
research has focused on variables such as gender and stylistic variation in
emoticon use (Schnoebelen 2012), building on earlier approaches outside
the social media domain (e.g. Campoy and Espinosa 2012; Coupland 2007;
Eckert 2000). Other approaches adopt frameworks from related linguistic
areas such as Page’s (2012b) work applying narrative theory to social media
story-telling, considering dimensions such as the use of hashtags.
27.3.3 The ‘Users in Uses’ Approach to Identity

Linguistics is still coming to terms as a discipline with how to account for
what Firth referred to as the language of ‘persons’ and ‘personalities’ (Firth
1950). Hodge (2014:35) undertakes an interesting thought experiment,
contemplating what would have happened ‘if Whorf and Halliday’s work
already published by 1956 were taken as paradigm-forming works that
created identities in linguistics’. He suggests this would have initiated a
more productive engagement with the concept of identity in linguistics: an
engagement that factors in how identities are enacted through meaning-
making. It would have allowed for what Halliday (1978) refers to as a ‘social
semiotic’ perspective capable of accounting not only for the ways in which
identities are construed in discourse, but also for how such construal
affords different forms of sociality.
SFL has a history of exploring semantic variation that has informed
current work on how identities are enacted in discourse (Martin, this
volume; Hasan and Webster 2009; Martin 2010). Following Firth (1957),
Martin et al. (2013) conceptualize social semiotic concern with identity as a
‘users in uses’ approach, that is, an approach which considers how particu-
lar ‘uses’ of language are performed by particular language ‘users’:
Users of language perform their identity within uses of language. Identity, in

other words, is always already conditioned by register and genre, so that who
we are depends on the roles we play in a given situation. The identities we
enact with language at a particular point in time are influenced by the
particular stage of the particular genre in which we happen to be involved.
(Martin et al. 2013:468)
Martin (2009) is careful to distinguish this type of social semiotic approach

from more common-sense perspectives that tend to characterize or describe
identity in terms of individual people rather than semiotic personae, of the
kind envisaged by Firth:
One thing we have to guard against here as functional linguists is a neuro/

biological interpretation of individuals and communities instead of a social
semiotic one. As Firth warns, it is not psycho-biological entities we are
exploring, but rather the bundles of personae embodied in such entities
and how these personae engender speech fellowships. We’re not, in other
words, looking at individuals interacting in groups but rather at persons and
personalities communing in discourse.
(Martin 2009:563)
Martin (2009:576) argues that we have exhausted the synoptic modelling

potential afforded by the realization hierarchy, which ‘crystalises snapshots
of semiotic valeur at particular points in time’. For instance, he suggests
that the traditional distinction, grounded in the realization hierarchy,
between ‘dialects as different ways of saying the same thing, and coding
orientation as meaning different things’ does not accord with the general
principles of SFL, since this kind of modelling assumes that dialects do not
in themselves make meaning (Martin 2009:575–6). However, this problem
can be avoided if we acknowledge that meaning is made at all levels of
stratification, and if we develop ‘a third hierarchy alongside realization and
instantiation, focusing on the allocation of the meaning potential of culture
and its deployment for affiliation’ (Martin 2009:575–6).
Martin (2009, 2010) refers to this new hierarchy for fostering ‘users in
uses’ research as the ‘individuation’ hierarchy. The aim is to account for the
ways in which the meaning potential of culture is allocated amongst perso-
nae, in other words, the particular distribution and patterns of meaning
instantiated by different personae. This hierarchy is geared towards taking
account of ‘logogenesis – i.e. unfolding discourse at the instance end of the
instantiation cline, ontogenesis – i.e. individual development at the reper-
toire end of the individuation cline, and phylogenesis – i.e. the evolutionary
consequences of variation according to users (individuation) and uses
(instantiation)’ (Martin 2009:576). Figure 27.2 represents the individuation
Figure 27.2 The individuation cline (adapted from Martin et al. (2013:490))
hierarchy from the perspective of both allocation, that is, how semiotic
resources are distributed amongst users, and affiliation, that is, how semi-
otic resources are deployed to commune (Martin et al. 2013:490). The intent
is to model the relationship between personae, sub-cultures, master iden-
tities, and, at the most generalized end of the cline, culture as a system of
meaning potential. However, work on the exact nature of this hierarchy,
and what constitutes meaningful units of analysis, is still in its infancy (e.g.
see Zappavigna and Martin 2014).
Martin grounds his perspective on individuation in the Bernsteinian
notion of culture as a ‘reservoir’ of meanings from which an individual
can mobilize a particular ‘repertoire’. This repertoire arises out of the
registers and genres to which they have been exposed. He quotes the
following passage from Bernstein in order to illustrate this point:
I shall use the term repertoire to refer to the set of strategies and their
analogic potential possessed by any one individual and the term reservoir to
refer to the total of sets and its potential of the community as a whole. Thus
the repertoire of each member of the community will have both a common
nucleus but there will be differences between the repertoires. There will be
differences between the repertoires because of the differences between
members arising out of differences in members’ context and activities and
their associated issues
(Bernstein 2000:158)
According to this kind of approach, the interpersonal dimension of mean-

ing inflects the kind of personae that we can take up in social life, just as the
kinds of roles that these personae can adopt is modulated by the genres into
which we have been socialized (Martin and Rose 2008).
A collection of work on identity influenced by the ‘users in uses’ perspective
by mostly Sydney-based researchers was published as a volume edited by Bed-
narek and Martin (2010). Dimensions that have been considered in this volume
and in related work, include gender (Bakar 2014, 2015; Hamid and Bakar 2010;
Tian 2008), nationality (Tann 2010, 2013), emotionality (Bednarek 2010, 2013,
2015), multimodality (Caldwell 2010), and, most relevant to this chapter, ambi-
ence in relation to social media identities (Zappavigna 2014a). These approaches
vary from close qualitative approaches using multimodal discourse analysis, to
corpus-based and quantitative approaches (O’Donnell 2014). They also draw
extensively on the appraisal framework in order to explore the expression
of attitude and the ‘syndromes of evaluation which characterize an individ-
ual – their appraisal signature’ manifest as ‘the idiolectal reconfigurations
of meaning-making potential by which individual authors achieve a recog-
nisable personal style’ (Martin and White 2005:208).
27.3.4 Current SFL Work on Individuation in Social Media

Work from an SFL perspective on identity performance in social media is
still very much in its early stages. Some multimodal work on representation
of the self has been undertaken by Zappavigna and Zhao (2017) and Zhao
and Zappavigna (2017), and by Zappavigna (2016) in relation to the repre-
sentation of subjectivity in Instagram images and Tumblr posts. This work
has considered the different options for visually representing the self in
social media images, from the choice to explicitly represent the self in the
form of a self-portrait, or ‘selfie’, or the choice to imply or infer the
presence of the self through compositional choices (e.g. a cup of coffee near
the front of an image indicating the photographer outside the frame), or
through inclusion of part of the photographer’s body in the image (e.g. a
hand holding a coffee cup). Zappavigna (2016) models these semiotic
choices as a system of ‘subjectification’. The major choice in this system
is between ‘as photographer’ perspective, where the photographer’s sub-
jectivity is represented, inferred, or implied through various choices in
visual structure, and ‘with photographer’ subjectivity, where the photog-
rapher’s perspective is unmarked (Figure 27.3).
Drawing on this initial modelling, Zhao and Zappavigna’s work aims to
show that the function of the selfie as a multimodal genre is not solely to
Figure 27.3 Subjectification (adapted from Zappavigna 2016)
represent ‘the self’ but rather to enact intersubjectivity, that is, to generate
various possibilities of relations between perspectives on a particular topic,
issue, or experience and hence to open up potential for negotiating differ-
ent points of view.
Selfies, as performances of identity, have been widely criticized in news
and entertainment media as an inherently narcissistic practice. One of the
assumptions at the core of this criticism is the position that the naturalized
reading of a selfie is ‘Look at me’. It is a proposition difficult to refute, as the
self/subject is the primary representational object in selfies as a visual
media (Pham 2015). However, if we adopt a social semiotic approach that
factors in the importance of interpersonal meaning, we can propose a
complementary reading: that the visual structure seen in selfies fore-
grounds the perspective of the photographer on a particular object, phe-
nomenon, or issue. For instance, in the domain controversially known as
‘mommyblogging’, selfies taken by mothers can be read as ‘Look, it is my
perspective on motherhood’ or ‘Let’s look at motherhood through my per-
spective’ (Zappavigna and Zhao 2017). This subtle shift of focus from ‘me’ to
‘my perspective’ is a very important consideration, as it affords a shift in
analytical focus from the ideational to the interpersonal.
Zappavigna (2014b) has explored microblogging identities in relation to the
types of bonds that are enacted within different communities of users. For
example, this work focused on a cluster of bonds that occur in microblogging
discourse about motherhood, but which also appear to be generalizable across
different types of personae and communities. The first bond is a ‘self-depre-
cation bond’, where the social media user can be read as laughing off (Knight
2010), or lightly and irreverently mocking, a stereotype such as the concept of
the ‘perfect mother’ (Lopez 2009), an example of which is given in (3):
(3) Just spilt wine on son’s homework diary. #badmother
A co-occurring bond is the ‘addiction bond’, as in example (4), in which the
user rallies around everyday bonding icons such as coffee, wine, or technology:
(4) Need.Wine.
And finally a frazzle bond (example (5)), whereby the user relates fatigue or
exasperation resulting from engaging in the core activity of a particular
community of fellowship, for instance:
(5) What a spectacularly horrendous night. Max awake for 3 hours,
wanting ‘iPad! IPAD!!’ I think NOT baby. Then twins snuck into
bed @ 4am. Ugh
This interplay or ‘complex’ of bonds is also seen in personae enacted in
other domains, such as the world of computer coding: for instance,
example (6) is a parallel to the above frazzle bond:
(6) Getting ready to go to bed after a long night of coding up some
custom classes to handle xml parsing and database interaction.
Section 27.4 details how we may identify these types of bonds by the way in
which they are realized as particular couplings of attitude and experience
in discourse.
27.4 Affiliation: Communing with Social Media
27.4.1 Developing the Cline of Affiliation

As Figure 27.2 suggests, complementary to individuation is ‘affiliation’, that
is, how community is negotiated in discourse. Language is replete with
resources for enacting social relations and forming networks of semiotic
‘bonds’ (Knight 2010, 2013; Martin 2010). We can think of the formation of
community as involving ‘bonds’ realized in discourse as patterns of values.
These bonds are negotiated in interactions, producing different types of
membership or fellowship. As we will see in the next section, these connec-
tions may be ambient in the sense that they do not necessarily require direct
contact between users and may draw on mass forms of communion of feeling.
The discursive patterns of values associated with different bonds and bond
complexes may be instantiated across multiple dimensions of meaning
‘across metafunctions (ideational resources such as technical and specialized
lexis, interpersonal resources such as naming and vocatives) and across
strata (accents in phonology, grammatical variation and discourse semantic
style)’ (Martin 2010:25). They may also span different semiotic modes (Drey-
fus et al. 2011). For instance, Martin et al. (2013) have considered the role
that gesture plays in the interactive bonding process of face-to-face encoun-
ters. This work analyzed the way that particular gestures support the pro-
posal and response to different kinds of attitudinal alignments by realizing
particular social bonds in the discourse. In addition Hood has considered the
role of gesture in forging alignments in classroom discourse (Hood 2011).
A key technical concept deployed in this type of work on affiliation is the
notion of ‘coupling’ Martin (2000), a concept that has been used to model the
kinds of connections that can be forged across different kinds of linguistic and
multimodal systems. In its more general sense, coupling refers to textual
relations that involve ‘the temporal relation of “with”: variable x comes with
variable y’ (Zhao 2011:144), and these co-selections may be made ‘across ranks,
metafunctions, strata and modalities which are not specified by system/struc-
ture cycles’ (Martin et al. 2013:469). Most studies applying coupling have
focused on evaluative meaning, since evaluation of proposals and propositions
is a particularly important resource for construing solidarity:
Feelings are meanings we commune with, since we do not say what we feel
unless we expect the person we are talking with to sympathize or empathize
with us. We express feelings in order to share them . . . to build relationships;
where we misjudge the situation and get rebuffed, then a sense of alienation
sets in.
(Martin 2002:196)
We do not commune, however, around feelings disjointed from their idea-

tional targets, but around the connection of these feelings to dimensions
of our experience, as ‘couplings’ of attitude and ideation (Martin 2000).
For example, the following tweet in (7) instantiates a coupling of experien-
tial meaning with negative judgement (attitude in bold, ideation
underlined):
(7) She is such a hypocrite
Martin et al. (2013) represent this type of coupling with the following
notation:1 [ideation: she / evaluation: negative judgement]. In any given
encounter we are always proposing and reacting to bonds as we negotiate
couplings in discourse. It is in this sense that we can view ‘persons and
personalities as active participators in the creation and maintenance of
cultural values’ (Firth 1957:186). While the term negotiation characterizes
this as a very deliberate activity, the practices at work always involve a tacit
dimension: just as we do not usually consciously attend to our linguistic
choices (Zappavigna 2013), we are rarely directly aware of the patterns of
bonds that we propose and react to in an interaction.
Current work on affiliation is focused on how bonds pattern into higher-
order complexes, that is, how they ‘cluster as belongings of different orders
(including relatively “local” familial, collegial, professional and leisure/rec-
reational affiliations and more “general” fellowships reflecting “master
identities” including social class, gender, generation, ethnicity, and dis/
ability)’ (Martin et al. 2013:490).2 The aim is to use semiotic rather than
common-sense criteria to account for the kinds of memberships that
emerge when we study coupling patterns. According to this perspective,
identities are patterns of meaning inflected by membership in networks of
fellowship. In other words, they can be thought of as the disposition to
enact particular configurations of couplings that realize particular config-
urations of bonds. This disposition is informed by a persona’s particular
semiotic ‘repertoire’ that arises out of the potential semiotic ‘reservoir’
available via their membership in a given community (Bernstein 2000).
27.4.2 Ambient Affiliation in Online Media

Affiliation has largely been developed with a view to explaining dialogic
discourse, in the sense of interactants forging dynamic, interactive align-
ments through rallying around, condemning, or laughing off couplings
(Knight 2010). In terms of the modalities related to face-to-face interaction,
paralinguistic clues such as gesture and posture can assist in analyzing how
1
Other notation systems and ways of visually representing how couplings function in interactions are currently being
developed.
2
For a related approach which focuses instead on the tenor dimension, see Don (2012).
bonds are orchestrated.3 However, in communication via online media,

where we do not have some of these types of paralinguistic clues (depending
on the nature of the digital channel), markers of solidarity such as emoti-
cons and hashtags play a pivotal role in supporting bonding (Zappavigna
2012). In addition, some forms of social media discourse are not always
directly interactive: for example, a study of a corpus of thirty-eight billion
tweets in existence in 2014 found that only 25 per cent of the tweets in the
sample received replies (Liu et al. 2014). To add to the complexity, ‘commu-
nion of feeling’ in social media discourse can also incorporate mass-
communicative practices, such as social tagging and producing iterations
of Internet memes (media such as catchy phrases, image macros, or videos
that are widely shared through social networks). These practices do not
require direct interaction between particular users in order to forge attitu-
dinal alignments.
Zappavigna (2011) has employed the concept of ‘ambient affiliation’ to
explore how semiotic resources are deployed to commune, in particular,
when such communion operates beyond the negotiation of couplings
within explicit conversational exchange structures. For example, mass
performances of hashtagging are ambient in this sense. Consider the
following post, expressing negative judgement regarding the 2014 Austra-
lian federal budget with the meme, ‘ThreeWordBudget’. This meme
involved users posting three words, typically opinion or sentiment, targeted
at the budget, the government, the treasurer and related content in this
semantic domain. For example, the main function of the following post in
(8) taken from the #ThreeWordBudget corpus is not ideational (e.g. about
specific details of the budget) but interpersonal (sharing negative evaluation
of the budget):
(8) #ThreeWordBudget Cruel and Mean
The meme involved mass expression of negative judgement and negative

affect coupled with the budget, as is shown in examples (9) to (12):
(9) #ThreeWordBudget Lie, cheat, steal!

(10) Evil, malevolent shitheads. #threewordbudget
(11) I’m Absolutely Terrified #threewordbudget
(12) #ThreeWordBudget Unjust Unfair Unwanted
This kind of data pushes our social semiotic modelling of affiliation into
considering how bonds can function outside exchange structures.
Appending a hashtag to a post presupposes that there is an ambient audi-
ence who may share or contest the values construed.
3
For gesture analysis see Martin et al. (2013); for image analysis see Caple (2010).
Adopting a Firthian perspective on microblogging, we might think of

posts such as the above as proposing ‘bonds’ to an ambient audience in
order to invite communion of feeling. These bonds are realized as the
coupling of negative judgement/affect with the ideational label realized in
the hashtag also coordinating the meme. This function is related to the
notion of ‘phatic communion’ (Malinowski 1972), where the main commu-
nicative function is interpersonal. We might think of posts as always
having dual interpersonal affordances: construing a value at the same time
as entering the personae construed in the text into a relationship with other
personae in the social media stream, who in turn manifest repeated
coupling patterns.4
The repeated coupling coordinated by the #ThreeWordBudget hashtag is
related to Knight’s (2010) notion of ‘rallying affiliation’ in interactive dis-
course. The above posts are examples of users ‘communing’ around shared
negative assessment of the budget. We can also think of the tag as func-
tioning like a vocative, a call out to a putative ambient fellowship or
‘college’ of potential aligners who share this value. This is a process that
Zappavigna and Martin (in preparation) term ‘convocation’. Understanding
how communion and convocation function in social media is a current
critical challenge that is stretching our existent social semiotic tools for
understanding bonding to their limits.
27.5 Collecting and Analyzing Social Media Texts
27.5.1 Methodological and Technical Issues

The sections above have considered the theoretical details of how to begin
exploring identity and affiliation in social media texts. In this section we are
focused on data collection and analysis issues that inform how we can use
social media texts to address these theoretical interests. What is exciting
about high-volume social media data is not simply its size,5 but how
particular affordances of this type of data can illuminate different kinds
of social relations. For instance, the availability of metadata, allowing
search, aggregation, and cross-referencing, enables different viewpoints
on a data set (Boyd and Crawford 2012). These affordances also impact
how we can apply our SFL tools to these texts.
We can group the main issues that may arise during the data collection
phase of a social media research project into two areas: methodological
issues regarding the nature of the data to be collected to answer the
4
For an example of a Twitter study using the concept of ambient affiliation to explore attitudinal alignments during the
2011 Japan nuclear crisis, see Inako (2013).
5
Indeed Boyd and Crawford (2012:663) point out that ‘some of the data encompassed by big data (e.g. all Twitter
messages about a particular topic) are not nearly as large as earlier data sets that were not considered Big Data (e.g.
census data)’.
particular research question at hand, and technical issues about the prac-
tical mechanics of gathering the data, which in the case of social media
services, usually involves some proficiency in using APIs (Application
Programming Interfaces) for ‘scraping’ data from the service’s databases
(see Zappavigna 2012 for an overview of the key issues relating to API use).
An important methodological concern is sampling, that is, determining the
scope and criteria for collecting texts. Many of the kinds of difficulties that
may be encountered when sampling and collecting social media texts are
not unique to electronic discourse (think, for example, of the multitude of
considerations that must be accounted for when dealing with spoken
discourse that is to be transcribed); however, the relative novelty of social
media platforms means that standards for sampling are still being
developed, and we are yet to even clearly establish what the important
issues and pitfalls are. One popular form of sampling that has received
some criticism is hashtag studies where a tag is used to collect some form of
experience (Bruns 2013). However, as Crawford (2013) has noted in relation
to the general field of big data analysis, this type of perspective only
captures a slice of the potential voices involved in a dimension of social life
at any point in time. It may factor out people who are not using social
media technologies, possibly the most disenfranchised within a commu-
nity. As such it might create what she terms ‘algorithmic illusions’ that
pollute our data analysis.
A prerequisite for usefully accumulating social media data for linguistic
analysis is being able to specify both the unit of analysis (what we are
drawing conclusions about) and the unit of observation (the kinds of textual
data that need to be sampled in order to draw these conclusions). This is the
case for both qualitative and quantitative studies. For example, a study
might investigate the social practices involved in Facebook posts (unit of
analysis), realized in specific patterns of linguistic features in these posts
(unit of observation), identified by applying any of the multifaceted types of
analysis made available by SFL theory. Social media texts, and electronic
discourse more broadly, however, problematize some of the practices lin-
guists may have used in the past for denoting what constitutes a text to be
analyzed. While all texts enter into heteroglossic relations with other texts
(Bakhtin 1981), and can be approached from both dynamic (text as process)
and synoptic (text as artefact) perspectives, determining the ‘bounds’ of a
networked, electronic text is particularly challenging. For example, High-
field and Leaver (2014) have noted the instability of Instagram posts due to
the dynamic nature of this media, which allows users to go back to posts
and add or delete comments on the image. In this case the text is, from the
perspective of mode, never complete as a communicative unit since it can
always be modified and added to.
In addition, in terms of multimodal dimensions such as presentation (e.g.
font, colour choice, layout), the text is also not static since it may be
syndicated to different kinds of devices (e.g. a smart phone) where it may
appear visually different from its instantiation in a web browser running on

a desktop computer. Thus, a concern, although not unique to social media
texts, is how to collect, store, and manipulate multimodal semiotic
resources. In other words, how do we collect and track relations between
elements of texts that involve multimedia? Again, we need to think care-
fully about what matters to a study. For instance, do we need to collect
information about the presentational layout of a blog post, or are we
concerned with the verbiage of the blog post alone? Do we need to track
relationships between the verbiage and the layout? The seemingly simple
task of establishing the dimensions of a semiotic resource is often more
difficult than it might at first appear, and opens up a range of interesting
data collection issues that are intimately tied to the kinds of analysis that
will be able to be undertaken in a study. A fundamental question in most
social media studies will be ‘Do I need to use a database (and/or a script for
processing the data)?’6 or ‘Can I simply collect data in its native format or a
simple format such a plain text?’ Answering these questions means know-
ing which variables matter to the study and whether or not there are
multiple relationships between variables that need to be tracked.
If we are interested in tracking contextual variables alongside the ‘con-
tent’ of a social media text, the volume of data can quickly become
extremely large and unwieldy. For example, if we wanted to track social
relations based on metadata alongside the linguistic patterns in microblog-
ging, the data can rapidly expand in complexity. One early study, outside
linguistics, claiming to analyze the entire Twittersphere at the time, gener-
ated a collection of ‘41.7 million user profiles, 1.47 billion social relations,
4,262 trending topics, and 106 million tweets’ (Kwak et al. 2010:600). If
detailed metafunctional linguistic analyses were also to be added, the
number and kind of relationships to be quantitatively tracked would be
immense. We can partially overcome this type of problem if we ‘shunt’
between qualitative approaches and quantitative approaches to sampling
and analysis. For example, we might adopt what Bednarek (2009), develop-
ing Baker (2006), refers to as a ‘three-pronged’ method for corpus-based
discourse analysis. This approach incorporates close manual analysis of
single texts with manual, or partially automated, small-scale corpus-based
analysis, complementing quantitative work, often highly automated, using
large-scale million-word corpora. However, it needs to be noted again that,
because of the kinds of non-standard orthography and other features of
social media, many of the standard tools, such as POS taggers, available for
supporting automatic analysis, may not work – for an example of a POS
tagger that has been modified to account for Twitter data, see the GATE
Twitter POS Tagger (Derczynski et al. 2013).
6
Scripts are small programmes using a scripting language to automate text-processing tasks that would otherwise have
to be done manually by a linguist.
27.5.2 How Time Affects Data Collection and Analysis

The blend of asynchronous and synchronous time involved both in the
production and consumption of social media texts and streams can make
data collection and analysis problematic. On the one hand, social media
texts that involve streaming data (chronologically organized feeds of data)
might have a ‘real-time’ reference point traceable via a timestamp in the
metadata.7 On the other hand, it may also be asynchronously modified,
replied to, syndicated, and rebroadcast. When viewed as a ‘conversation-
like’ interaction, a social media post may be part of an interactional unit
evolving long after the original text was published.
In addition, the social processes involved in producing a social media
text, and the ‘scope’ of a text (e.g. the bounds of a conversation-like inter-
action), will also look different depending on the ‘timescale’ (Lemke 2000)
across which they are viewed (e.g. a week, a day, a decade). Different time
scales will allow the researcher to observe different relations between the
semiotic resources that they are studying (Zhao 2010). In addition, the rapid
speed at which technology changes means that our most refined tools for
social research, which have been developed for analyzing older semiotic
modes, are often unable to be rigorously applied:
Online behavior at Time X only predicts online behavior at Time X + 1 if (1)

the underlying population from which we are sampling remains the same
and (2) the medium itself remains the same. A changing media environment,
under early adopter conditions, violates both (1) and (2). Ceteris is not paribus.
All else cannot be assumed equal. One obvious consequence is that research
findings are rendered obsolete by the time they have been published. The
terrain we can explore with traditional social science techniques has
narrowed in scope.
(Karpf 2012:642)
In addition, time is important for working out the difficult issue of how we
know where a social media interaction begins and ends. Some of the
parallel parameters we find in other modes, such as ‘turn-taking’ (Sacks
et al. 1974), that might be used to characterize an exchange are not neces-
sarily present in social media interactions. An interaction will look differ-
ent depending on the ‘node’ in the social network from which it is
observed.8 Some ability to track which user is replying to, or rebroadcast-
ing, a particular user is typically available in social media metadata.
However, representing the discourse structure of non-linear, multicast
(many-to-many broadcast) interactions is a problem that has yet to be
solved. Indeed, linguists are still working on how to represent dynamic
exchanges in face-to-face interaction.
7
A timestamp is a record, usually encoded in a standard format, marking when an event occurred.
8
Study of the properties of social media networks is undertaken in an area known as Social Network Analysis (SNA).
27.6 Future Directions
As network capabilities advance, social media is becoming more image- and

video-intensive. Thus comprehensive descriptions of social media texts will
need to account for meanings made across multiple modes, from written to
spoken discourse. This will include meanings made via paralinguistic
systems such as gesture and image, for which we have a limited but
emerging multimodal SFL description. Social media research, particularly
in the areas of individuation and affiliation, will need to contribute to the
development of a multimodal metalanguage for theorizing and describing
these resources.
References
Albury, K. 2015. Selfies, Sexts and Sneaky Hats: Young People’s Understand-
ings of Gendered Practices of Self-Representation. International Journal of
Communication 9: 1734–45.
Argamon, S., M. Koppel, J. Pennebaker, and J. Schler. 2007. Mining the
Blogosphere: Age, Gender, and the Varieties of Self-expression. First
Monday 12(9). Available online at: http://firstmonday.org/ojs/index.php/
fm/article/view/2003/1878. (Last accessed 15/08/2018.)
Bakar, K. A. 2014. ‘A (Sensitive New Age Guy) with Difference’: Gendered
Performances in Online Personal Advertisements. Linguistics and the
Human Sciences 9(1): 5–34.
Bakar, K. A. 2015. Identity in Online Personal Ads: A Multimodal Investi-
gation. Asian Social Science 11(15): 313.
Baker, P. 2006. Using Corpora in Discourse Analysis. London: Continuum.
Bakhtin, M. 1981. The Dialogic Imagination: Four Essays. Edited by M. Holquist.
Austin: University of Texas Press.
Bamman, D., J. Eisenstein, and T. Schnoebelen. 2014. Gender Identity
and Lexical Variation in Social Media. Journal of Sociolinguistics 18(2):
135–60.
Bednarek, M. 2009. Corpora and Discourse: A Three-Pronged Approach
to Analyzing Linguistic Data. In M. Haugh, K. Burridge, J. Mulder,
and P. Peters, eds., Selected Proceedings of the 2008 HCSNet Workshop on
Designing the Australian National Corpus. Somerville: Cascadilla Proceedings
Project.
Bednarek, M. 2010. The Language of Fictional Television: Drama and Identity.
London: Bloomsbury.
Bednarek, M. 2013. ‘There’s No Harm, Is There, in Letting Your Emotions
Out’: A Multimodal Perspective on Language, Emotion and Identity in
MasterChef Australia. In N. Lorenzo-Dus and P. Blitvi, eds., Real Talk:
Reality Television and Discourse Analysis in Action. London: Palgrave Macmil-
lan. 88–114.
Bednarek, M. 2015. ‘What We Contrarians Already Know’: Individual and

Communal Aspects of Attitudinal Identity. In M. Charles, N. Groom, and
S. John, eds., Corpora, Grammar, Text and Discourse. Amsterdam: John
Bednarek, M. and J. R. Martin, eds. 2010. New Discourse on Language: Func-
tional Perspectives on Multimodality, Identity, and Affiliation. London:
Continuum.
Benwell, B. and E. Stokoe. 2007. Discourse and Identity. Edinburgh: Edinburgh
University Press.
Critique. Revised edition. London: Taylor & Francis.
Boyd, D. 2014. It’s Complicated: The Social Lives of Networked Teens. New Haven:
Boyd, D. and K. Crawford. 2012. Critical Questions for Big Data. Information,
Communication & Society 15(5): 662–79.
Boyd, D. and N. B. Ellison. 2007. Social Network Sites: Definition, History,
and Scholarship. Computer-Mediated Communication 13(1): 210–13.
Boyd, D. and J. Heer. 2006. Profiles as Conversation: Networked Identity
Performance on Friendster. Paper presented at the 39th Annual Hawaii
International Conference on System Sciences.
Bruns, A. 2013. Faster Than the Speed of Print: Reconciling ‘Big Data’ Social
Media Analysis and Academic Scholarship. First Monday 18(10). Available
online at: http://firstmonday.org/ojs/index.php/fm/article/view/4879. (Last
accessed 15/08/2018.)
Caldwell, D. 2010. Making Metre Mean: Rhythm and Identity in the Music
of Kanye West. In M. Bednarek and J. R. Martin, eds., New Discourse on
Language: Functional Perspectives on Multimodality, Identity and Affiliation.
Campoy, J. M. H. and J. A. C. Espinosa. 2012. Style-Shifting in Public: New
Perspectives on Stylistic Variation: Amsterdam: John Benjamins.
Caple, H. 2010. Doubling-Up: Allusion and Bonding in Multisemiotic News
Stories. In M. Bednarek and J. R. Martin, eds., New Discourse on Language:
Functional Perspectives on Multimodality, Identity and Affiliation. London: Con-
tinuum. 111–34.
Coupland, N. 2007. Style: Language Variation and Identity. London: Cambridge
University Press.
Crawford, K. 2009. Following You: Disciplines of Listening in Social Media.
Continuum: Journal of Media & Cultural Studies 23(4): 525–35.
Crawford, K. 2013. Algorithmic Illusions: Hidden Biases of Big Data. Paper
presented at the O’Reilly Strata Conference, Santa Clara, California.
Davis, K. 2011. Tensions of Identity in a Networked Era: Young People’s
Perspectives on the Risks and Rewards of Online Self-expression. New
Media & Society 14(4): 634–51.
Derczynski, L., A. Ritter, S. Clark, and K. Bontcheva. 2013. Twitter Part-of-
Speech Tagging for All: Overcoming Sparse and Noisy Data. Paper
presented at the Proceedings of the International Conference on Recent

Advances in Natural Language Processing.
Deumert, A. 2014. The Performance of a Ludic Self on Social Network(ing)
Sites. In P. Seargeant and C. Tagg, eds., The Language of Social Media: Identity
and Community on the Internet. London: Palgrave. 23–45.
Don, A. C. 2012. Legitimating Tenor Relationships: Affiliation and Align-
ment in Written Interaction. Linguistics and the Human Sciences 5(3):
303–27.
Dreyfus, S., S. Hood, and M. Stenglin. 2011. Semiotic Margins: Meaning in
Multimodalities. London: Continuum.
Duguay, S. 2014. ‘He Has a Way Gayer Facebook Than I Do’: Investigating
Sexual Identity Disclosure and Context Collapse on a Social Networking
Site. New Media & Society 18(6): 891–907.
Eckert, P. 2000. Language Variation as Social Practice: The Linguistic Construction
of Identity in Belten High. Malden: Blackwell.
Firth, J. R. 1950. Personality and Language in Society. The Sociological Review
42(1): 37–52.
Firth, J. R. 1957. Papers in Linguistics, 1934–1951. London: Oxford University
Press.
Gilpin, D. R. 2010. Working the Twittersphere: Microblogging as Profes-
sional Identity Construction. In Z. Papacharissi, ed., The Networked Self:
Identity, Community and Culture on Social Network Sites. New York: Routledge.
232–50.
Grasmuck, S., J. Martin, and S. Zhao. 2009. Ethno-Racial Identity
Displays on Facebook. Journal of Computer-Mediated Communication 15(1):
158–88.
Hamid, B. D. A. and K. A. Bakar. 2010. Articulating Male and Female
Adolescent Identities via the Language of Personal Advertisements:
A Malaysian Perspective. In A. Duszak and U. Okulska, eds., Language,
Culture and the Dynamics of Age. Berlin: Walter de Gruyter. 191–223.
Hasan, R. and J. Webster. 2009. Semantic Variation: Meaning in Society and in
Sociolinguistics, Volume 2. Sheffield: Equinox.
Highfield, T. and T. Leaver. 2014. A Methodology for Mapping Instagram
Hashtags. First Monday 20(1). Available online at: http://firstmonday.org/
article/view/5563/4195. (Last accessed 15/08/2018.)
Hodge, B. 2014. What if the Revolution Never Happened? A Thought Experi-
ment in Linguistics and Identity. Linguistics and the Human Sciences 9(1):
35–58.
Hogan, B. and A. Quan-Haase. 2010. Persistence and Change in Social
Media. Bulletin of Science, Technology & Society 30(5): 309–15.
Hood, S. 2011. Body Language in Face-to-face Teaching: A Focus on Textual
and Interpersonal Meaning. In S. Dreyfus, S. Hood, and M. Stenglin, eds.,
Semiotic Margins: Meaning in Multimodalities. London: Continuum. 31–52.
Huang, J., K. M. Hornton, and E. N. Efthimiadis. 2010. Conversational

Tagging in Twitter. Paper presented at the Proceedings of the 21st ACM
Conference on Hypertext and Hypermedia, Toronto, Ontario, Canada.
Inako, A. 2013. Affiliation over Crisis: Physicists’ Use of Twitter Mode on
Fukushima Daiichi NPP Accident. Proceedings of JASFL 7: 1–13.
Karpf, D. 2012. Social Science Research Methods in Internet Time. Infor-
mation, Communication & Society 15(5): 639–61.
Knight, N. K. 2010. Wrinkling Complexity: Concepts of Identity and Affili-
ation in Humour. In M. Bednarek and J. R. Martin, eds., New Discourse on
Language: Functional Perspectives on Multimodality, Identity, and Affiliation.
Knight, N. K. 2013. Evaluating Experience in Funny Ways: How Friends
Bond through Conversational Hum. Text & Talk 33(4–5): 553–74.
Kwak, H., C. Lee, H. Park, and S. Moon. 2010. What is Twitter, a Social
Network or a News Media? Paper presented at the 19th World-Wide Web
(WWW) Conference, Raleigh, North Carolina.
Lee, C. 2014. Language Choice and Self-presentation in Social Media: The
Case of University Students in Hong Kong. In P. Seargeant and C. Tagg,
eds., The Language of Social Media: Identity and Community on the Internet.
London: Palgrave. 91–111.
Lemke, J. L. 2000. Across the Scales of Time: Artifacts, Activities, and
Meanings in Ecosocial Systems. Mind, Culture, and Activity 7(4): 273–90.
Leppänen, S., S. Kytölä, H. Jousmäki, S. Peuronen, and E. Westinen. 2014.
Entextualization and Resemiotization as Resources for Identification in
Social Media. In P. Seargeant and C. Tagg, eds., The Language of Social
Media: Identity and Community on the Internet. London: Palgrave Macmillan.
112–36.
Liu, Y., C. Kliman-Silver, and A. Mislove. 2014. The Tweets They Are
A-changin’: Evolution of Twitter Users and Behavior. Paper presented at
the International AAAI Conference on Weblogs and Social Media
(ICWSM), Ann Arbor, June 2014.
Lopez, L. K. 2009. The Radical Act of ‘Mommy Blogging’: Redefining Mother-
hood through the Blogosphere. New Media & Society 11(5): 729–47.
Lu, H., J. C. Lin, K. Hsiao, and L. Cheng. 2010. Information Sharing Behav-
iour on Blogs in Taiwan: Effects of Interactivities and Gender Differences.
Journal of Information Science 36(3): 401–16.
Madden, M. 2012. Privacy Management on Social Media Sites. Pew Internet
Report. Available online at: www.pewinternet.org/2012/02/24/privacy-
management-on-social-media-sites. (Last accessed 16/09/2017.)
Malinowski, B. 1972. Phatic Communion. In J. Laver and S. Hutcheson, eds.,
Communication in Face-to-face Interaction. Harmondsworth: Penguin.
146–52.
Martin, J. R. 2000. Beyond Exchange: APPRAISAL Systems in English. In
S. Hunston and G. Thompson, eds., Evaluation in Text: Authorial Stance and
the Construction of Discourse. Oxford: Oxford University Press. 142–75.
Martin, J. R. 2002. Blessed are the Peacemakers: Reconciliation and Evalu-

ation. In C. Candlin, ed., Research and Practice in Professional Discourse. Hong
Kong: City University of Hong Kong Press. 187–227.
Martin, J. R. 2009. Realisation, Instantiation and Individuation: Some
Thoughts on Identity in Youth Justice Conferencing. DELTA: Documentação
de Estudos em Lingüística Teórica e Aplicada 25: 549–83.
eds., New Discourse on Language: Functional Perspectives on Multimodality,
Identity, and Affiliation. London: Continuum. 1–34.
Equinox.
Martin, J. R. and P. R. R. White. 2005. The Language of Evaluation. New York:
Palgrave Macmillan.
of Language: Embodied Identity in Youth Justice Conferencing. Text &
Talk 33(4–5): 467–96.
Marwick, A. 2011. To See and Be Seen: Celebrity Practice on Twitter. Conver-
gence: The International Journal of Research into New Media Technologies 17(2):
139–58.
O’Donnell, M. 2014. Exploring Identity Through Appraisal Analysis:
A Corpus Annotation Methodology. Linguistics and the Human Sciences
9(1): 95–116.
Page, R. 2012a. The Linguistics of Self-branding and Micro-celebrity
in Twitter: The Role of Hashtags. Discourse & Communication 6(2): 181–201.
Page, R. 2012b. Stories and Social Media: Identities and Interaction. London:
Routledge.
Page, R. 2014. Hoaxes, Hacking and Humour: Analysing Impersonated
Identity on Social Network Sites. In P. Seargeant and C. Tagg, eds., The
Language of Social Media: Identity and Community on the Internet. London:
Palgrave Macmillan. 46–64.
Pham, M.-H. T. 2015. ‘I Click and Post and Breathe, Waiting for Others to
See What I See’: On #FeministSelfies, Outfit Photos, and Networked
Vanity. Fashion Theory 19(2): 221–41.
Rose, J., S. Mackey-Kallis, L. Shyles, K. Barry, D. Biagini, C. Hart, and L. Jack.
2012. Face It: The Impact of Gender on Social Media Images. Communi-
cation Quarterly 60(5): 588–607.
Rustagi, M., R. Prasath, S. Goswami, and S. Sarkar. 2009. Learning Age and
Gender of Blogger from Stylistic Variation. In S. Chaudhury, S. Mitra,
C. A. Murthy, P. S. Sastry, and S. K. Pal, eds., Pattern Recognition and
Machine Intelligence. Berlin: Springer: 205–12.
Sacks, H., E. A. Schegloff, and G. Jefferson. 1974. A Simplest Systematics for
the Organization of Turn-taking for Conversation. Language 50(4):
696–735.
Schler, J., M. Koppel, S. Argamon, and J. Pennebaker. 2006. Effects of Age

and Gender on Blogging. Paper presented at the Proceedings the Spring
Symposium on Computational Approaches for Analyzing Weblogs.
Schnoebelen, T. 2012. Do You Smile with Your Nose? Stylistic Variation in
Twitter Emoticons. University of Pennsylvania Working Papers in Linguistics
18(2): 117–25.
Schwartz, R. and G. R. Halegoua. 2014. The Spatial Self: Location-based
Identity Performance on Social Media. New Media & Society 17(10):
1643–60.
Seargeant, P. and C. Tagg, eds., 2014. Language and Social Media: Communi-
cation and Community Online. London: Palgrave.
Stenglin, M. 2008. Olympism: How a Bonding Icon Gets Its ‘Charge’. In
L. Unsworth, ed., Multimodal Semiotics: Functional Analysis in the Contexts of
Tann, K. 2010. Imagining Communities: A Multifunctional Approach to
Identity Management in Text. In M. Bednarek and J. R. Martin, eds.,
New Discourse on Language: Functional Perspectives on Multimodality, Identity,
and Affiliation. London: Continuum. 163–94.
Tann, K. 2013. The Language of Identity Discourse: Introducing a Systemic
Functional Framework for Iconography. Linguistics and the Human Sciences
8(3): 361–91.
Tian, P. 2008. Playing with ‘Femininity’: Multi-modal Discourse Analysis of
Bilingual Children’s Picture Book. In C. Wu, C. M. I. M. Matthiessen, and
M. Herke, eds., The Ballad of Mulan. Proceedings of ISFC 35: Voices Around the
World. Sydney: The 35th ISFC Organizing Committee. 308–15.
Twitter. 2013. Twitter About Page. Available online at: https://about.twitter
.com. (Last accessed 16/09/2017.)
Vásquez, C. 2014. ‘Usually Not One to Complain But . . .’: Constructing
Identities in User-generated Online Reviews. In P. Seargeant and C. Tagg,
eds., The Language of Social Media: Identity and Community Online. London:
Palgrave Macmillan. 65–90.
Zappavigna, M. 2011. Ambient Affiliation: A Linguistic Perspective on Twit-
ter. New Media & Society 13(5): 788–806.
Zappavigna, M. 2012. Discourse of Twitter and Social Media. London:
Bloomsbury.
Zappavigna, M. 2013. Tacit Knowledge and Spoken Discourse. London:
Bloomsbury.
Zappavigna, M. 2014a. Enacting Identity in Microblogging. Discourse and
Communication 8(2): 209–28.
Zappavigna, M. 2014b. Enjoy Your Snags Australia . . . Oh and the Voting
Thing too #ausvotes #auspol: Iconisation and Affiliation in Electoral
Microblogging. Global Media Journal: Australian Edition 8(2). Available
online at: https://www.hca.westernsydney.edu.au/gmjau/?p=1139. (Last
accessed 15/08/2018.)
Zappavigna, M. 2015. Searchable Talk: The Linguistic Functions of

Hashtags. Social Semiotics 25(3): 274–91.
Zappavigna, M. 2016. Social Media Photography: Construing Subjectivity in
Instagram Images. Visual Communication 15(3): 271–92.
Zappavigna, M. 2018. Searchable Talk: Hashtags and Social Media Metadiscourse.
London: Bloomsbury.
Zappavigna, M. and J. R. Martin. 2014. Mater Dolorosa: Negotiating Support
in NSW Youth Justice Conferencing. International Journal for the Semiotics of
Law / Revue internationale de Sémiotique juridique 27(2): 263–75.
Zappavigna, M. and S. Zhao. 2017. Selfies in ‘Mommyblogging’: An
Emerging Visual Genre. Discourse, Context & Media 20: 239–47.
Zhao, S. 2010. Intersemiotic Relations as Logogenetic Patterns: Towards
the Restoration of the Time Dimension in Hypertext Description. In
M. Bednarek and J. R. Martin, ed., New Discourse on Language: Functional
Perspectives on Multimodality, Identity, and Affiliation. London: Continuum.
195–218.
Zhao, S. 2011. Learning through Multimedia Interaction: The Construal of Primary
Social Science Knowledge in Web-based Digital Learning Materials. PhD Thesis,
University of Sydney.
Zhao, S. and M. Zappavigna. 2017. Beyond the Self: Intersubjectivity and
the Social Semiotic Interpretation of the Selfie. New Media & Society 20(5):
1735–54.
28
Theorizing and Modelling
Translation
Erich Steiner
28.1 Historical Overview of the Development

of Translation within SFL
28.1.1 Early British Contextualism

The origins of a Systemic Functional approach to translation can be traced
back to the work of the anthropologist, Bronislaw Malinowski.1 Malinow-
ski’s important role in the development of Firth’s functional approach to
language and the subsequent development of Systemic Functional Linguis-
tics (SFL) is widely acknowledged, and this role also extends to the SFL
approach to translation. There are two major senses in which translation
as a phenomenon was critical for Malinowski’s understanding of language
and linguistics: translation, as he saw it, was a decisive way of explicating
the difference between a culture’s meanings and the anthropologist’s
intended readership. In other words, translation for him was less an instru-
ment in the service of assimilation of some foreign culture to a target
culture than a way of becoming aware of the differences between them
(Malinowski 1935:ix). This was an early foreshadowing of Venuti’s
(1995:148–86) advocacy of ‘foreignizing translation’.
The other sense is that translation was considered to be a process of
iterative contextualization of linguistic structures, e.g. words in phrases
in clauses in sentences in situational and ultimately cultural contexts. The
elucidation of ‘meaning’ was thus to proceed through a sequence of linguis-
tic to cultural levels until the full (difference and strangeness in) meaning of
the linguistic activity under study became understandable to its readership:
We see then that it is impossible to define a word by mere equation.

Translation in the sense of exact and exhaustive definition of meaning cannot
be done by affixing an English label . . . Translation in the sense of defining a
1
See Steiner (2005, 2015a) for more elaborate accounts of the history of SFL Translation.
740 ERICH STEINER
term by ethnographic analysis, that is, by placing it within its context of culture,
by putting it within the sets of kindred and cognate expressions, by contrast-
ing it with its opposites, by grammatical analysis and above all by a number
of well-chosen examples – such translation is feasible and is the only correct
way of defining the linguistic and cultural character of a word.
(Malinowski 1935:17, emphases in original)
Translation as a process of iterative contextualization became a prototyp-

ical methodological activity and orientation which Malinowski bequeathed
to functional and Firthian linguistics. And, although translation as such did
not figure prominently in Firth’s writing (but see Firth 1968a), what Firth
did inherit from Malinowski was a general programme of linguistic analysis
as a process of iterative contextualizations intra- and extra-linguistically
(e.g. Firth 1968b).
28.1.2 Scale and Category

Probably the most influential attempt at modelling translation against the
background of the scale-and-category version of Halliday’s linguistics (see
Halliday 1961; Halliday et al. 1964:111–34) is Catford’s A Linguistic Theory of
Translation (1965). Its lasting contribution is a demonstration of how the
scale-and-category-type architecture of linguistics can be used for an under-
standing of the relationship of translation between texts. This was a signifi-
cant step away from models of translation based on some version of the
‘container metaphor’, wherein translation was seen as the transfer of some
equivalent content from the ‘container’ of one language to that of another.
Catford showed how translation could be seen as relationships between
units in structures arranged in a hierarchy of ranks and levels. Apart from
these relationships, no separate level of ‘mental representation’ or ‘sense’
was postulated. Functional and early systemic linguistics were strongly
anti-mentalist in avoiding any separate levels of ‘mental representations’
in addition to whatever ‘meanings’ the contextualizations of units in their
structures and systems made available.
28.1.3 Systemic Functional Linguistics: Levels, Axes, and

Scales of Abstraction
From the early 1970s onwards Halliday’s SFL theory gave increasing space
to the system rather than the structure, to a semanticized grammar rather
than the more formal one of the scale-and-category version, and to context
in addition to text. This provided methodological refinement to discussions
of translation, and a significant extension of the range of phenomena
covered under the term of ‘translation’ became possible. At the same time,
SFL approaches maintained a firm linguistic and text-oriented base, at a
time when some translation studies disclaimed any grounding in linguistics
under the global methodological orientation of the ‘cultural turn’ in the
Theorizing and Modelling Translation 741
social sciences and humanities. Translation began to be modelled on a

variety of levels and of scales of abstraction.
28.1.3.1 Levels: Stratification

Halliday himself has not written very extensively on translation (but see
Halliday 1962; Halliday et al. 1964:111–34; Halliday 2001, 2009, 2012), but
he does name (machine-)translation as one of his own work’s major driving
contexts for theoretical development (Halliday 2010:128). The translation
studies community has often acknowledged SFL’s particular usefulness for
translation theorizing and modelling (Baker 1992:121–59, 180–214; House
2015; Munday 2012; Newmark 1988; Toury 1995:264–5), and a number of
systemicists and translation studies scholars have used the ongoing devel-
opment of SFL as a guideline for modelling aspects of translation.
One strand of work (Matthiessen and Bateman 1991; Steiner et al. 1988)
exploits the level of ‘local’ clause-based semantics, or semanticized net-
works, for explorations of a suitable level of transfer in transfer-based
architectures for (machine-)translation, sometimes and characteristically
related to work in multilingual text generation. Bell (1991) offers one of
the more comprehensive attempts at modelling translation and related
phenomena in an overall SFL-based framework.
Another strand of work (Baker 1992; Hatim and Mason 1990; House
2015; Taylor 1998) considers translation still at the stratum of semantics,
but no longer mainly as ‘local’ clause-based semantics, but as ‘register’.
These works use SFL ideas (register, genre, coherence, cohesion) to explore
questions of textuality in translation. House from the 1970s onwards pro-
vides an important attempt at translation quality assessment, intersecting
SFL ideas with ideas from pragmatics and comparative culture studies in a
wider sense. Some attempts at German–English translation additionally
focus on micro-level considerations, for example, in the exploitation of
the notion of ‘grammatical metaphor’ for translation (Steiner 2004).
The developments traced here move from clause-based to text-based
modelling within an overall SFL orientation, as well as exploring inter-
actions and shifts between metafunctions. The potential of modelling on
the different axes (system, structure) and along the scales of abstraction
offered by the theory is recognized, though in these studies not yet made
fully explicit. It is only in the work referred to in the next section that this
potential is more systematically explored.
28.1.3.2 Axes and Scales of Abstraction

Apart from work exploiting the notion of register, another important move
of theorizing translation along the instantiation dimension includes
corpus-based methodologies in investigations of translations and parallel
texts as instances (Ghadessy and Gao 2001; Hansen-Schirra et al. 2012;
Steiner 2004:125–81; Teich 2003). This development of SFL theorizing is
perhaps best represented in Matthiessen (2001).
742 ERICH STEINER
Matthiessen (2001:41–126) outlines an SFL model for theorizing translation

in a comprehensive statement. He locates translation within a typology of
systems, characterizing it as a semiotic process. This includes non-linguistic
semiotic systems and the possibility of translation between those and lan-
guage. The focus is on translation between languages, and within languages
between varieties. The latter possibility has been quite extensively discussed in
SFL work on registerial variation, especially between expert and non-expert
registers (Halliday and Martin 1993). Matthiessen (2001) makes full use of SFL
scales of abstraction in several dimensions. He gives due weight to the possi-
bility of ‘metafunctional shifts’ in translation, which may arise from different
metafunctional orientations of language systems and/or from constraints of
the intended register and genre of the target text. He also locates important
concepts of translation studies, such as ‘equivalence’, ‘shift’, and ‘free vs.
literal’, in this model. His instantiation-stratification matrix (Matthiessen
2001:92) defines a significant space for types of systemic investigations of
translation. Seen as a whole, his contribution charts the entire territory of
conceptualizations of translation opened up by SFL thinking in a comprehen-
sive and systematic way (cf. Halliday 2009 for application to a text).
The development traced here provides a significant extension of the range
of options for the translator, as well as a broadened range of phenomena to
investigate for the researcher. The translator is free to shift between ranks,
levels, and metafunctions when searching for translational equivalents. At
the same time, the meaning of the term ‘translation’ is extended to transla-
tions between diachronic and synchronic variants, on the one hand, and to
translations between different semiotic systems, on the other. The explicit
modelling of these highly complex relationships within one overall architec-
ture distinguishes SFL as a theory. At the same time, the firm grounding of
SFL models of translation in language and text helps to avoid the dangers of a
metadiscourse without operationalizations of terms, occasionally practised
in some more ‘anti-linguistic’ approaches to translation. These characteris-
tics of an SFL-based approach are illustrated in Sections 28.2 and 28.3.
28.2 Key Concepts of Models of Translation
Complementary to the ‘historical’ orientation of Section 1, this section

adopts more of a ‘systematic’ orientation. Our focus will be on the character-
istics of an SFL perspective in the wider field of translation studies: in
clarifying the specifics of translation in comparison to ‘interpreting’, on the
one hand, and to ‘multilingual text production’, on the other, we bring out
its characteristic properties. We then contrast ‘translation’, ‘paraphrase’, and
‘(textual) variation’ as types of relationship between texts. In a subsequent
step, SFL views are outlined on ‘equivalence’, the text type ‘translation’, and
the role of the ‘translator’, all of which constitute widely shared essential
concepts through which schools of translation studies can be compared.
28.2.1 Translation vs. Interpreting vs. Multilingual Text Production

The production of texts with a multilingual output and some sort of input
specification can be called ‘multilingual text production’. Examples include
multilingual advertising campaigns, the production of user handbooks
from a common content specification, software localization, or the writing
of some piece of EU-regulation in all EU-languages. If the input to a multi-
lingual text production is a linguistic text (source text), rather than some
other sort of meaning specification, and if input and output are off-line and
written rather than spoken, this is pre-theoretically called ‘translation’.
A theoretically motivated notion of translation recognizes the source text
as already linguistically encoded input, having the full intricacy of a lin-
guistic encoding. The re-encoding of the resulting intricate web of meaning
relations, together with the contrastive differences between language
systems involved, creates the uniquely challenging environment of transla-
tion. Each individual translation is text production under the specific
constraints of a source text. ‘Interpreting’, finally, is ‘translating in spoken
mode’, however, in very different ways depending on whether we are
concerned with simultaneous, consecutive, or community interpreting.
Consecutive and community interpreting share some characteristics with
multilingual text production, whereas simultaneous interpreting is closer
to translating – however, with the drastically different constraints on
production arising from the specific demands of on-line processing. Inter-
preting has so far not received any major input from SFL and will be
backgrounded in this survey.
28.2.2 Translation vs. Paraphrase vs. Variation

The differentiation between ‘variation’, ‘paraphrase’, and ‘translation’ is
less one in terms of production, than in terms of the relationships between
the different textual variants (Steiner 2001:179–80).
Variation, more specifically register-variation, is possible within sets of
intertextually related texts, both intra- and inter-lingually. Translated texts
may or may not differ in register from their sources, with the theoretically
motivated stipulation that the register differences should be small in the
case of semantically based translations. Whether or not the contextual
configurations in terms of field, tenor, and mode of discourse are different
between source and target is a different question.
Paraphrase is a truth-condition preserving relationship among sets of
propositions and the sentences expressing them. Paraphrases generally do
not preserve the textual semantics, and they have no very clear relationship
towards the interpersonal semantics. Paraphrases are possible within and
between languages, but the only relationship they have to a translation is
that they preserve important aspects of the experiential and logical seman-
tics of sentences, which under-determines what a translation usually has
to be.
744 ERICH STEINER
Translation, in the sense of literal translation, is only possible across

languages (including sub-languages and historical varieties), because a lit-
eral translation within one language would have to be a lexicogrammatical
copy of itself. A semantically based translation within one language is
possible because of the non-bi-uniqueness of lexicogrammar and semantics,
but yields a highly constrained set of possibilities. If we added explicitation,
we would additionally introduce de-metaphorized variants, approaching
paraphrases, into the set. Yet all the important criteria from language
comparison and typology would be missing, and at least in that sense,
translation is again only possible across languages. Examples (1) to (4) (see
Section 28.2.3) illustrate our point, where (3) and (4) are ideational and
interpersonal (yet not textual) translations of (1) within English, yet not
literal translations of it.
In the end, we may say that translation is an approximation of a multi-
functional paraphrase – rather than the mono-functional paraphrases of
logic-oriented semantics – under the constraints of the process of under-
standing and of the typology of the language systems involved. Finally, each
individual translation, i.e. situated language (instantiation), is text produc-
tion under the constraints of a source text.
28.2.3 Equivalence
‘Equivalence’ as a key notion of translation has an important place in SFL
modelling, though more diversified than in most non-SFL approaches, and
depending on the relationships between source context and target context
(see Yallop 2001; for a summary discussion across schools of translation
studies, see Halverson 1997). ‘Equivalence’ between some source and target
text can be sought and privileged at different linguistic levels (phonological,
lexicogrammatical, semantic, and contextual) and with different emphases
on ideational, interpersonal, or textual meaning. Any of these can be
privileged depending on the purpose of the translation at hand, even if a
sort of default case will usually be equivalence on the semantic level, and
here particularly with regard to ideational meaning. An optimized solution
in any translation task will be a compromise approximation, rather than
simply ‘correct’ or ‘incorrect’. The following illustrates how at the lexico-
grammatical level and between ideational, interpersonal, and textual func-
tions, originals and translations may diverge substantially in an SFL
analysis (Steiner 2004:150–61, example from Doherty 1991):
(1) The suspicion that volcanic eruptions are the primary source of
aerosols in the upper atmosphere has been around for many years.
(original, New Scientist 21 January 1982: 150)
(2) Seit vielen Jahren vermutet man schon, dass die Aerosole in den höheren
Schichten der Atmosphäre vor allem aus Vulkanausbrüchen stammen.
(German translation, Doherty 1991)
Literal back-translation of (2) into English:
(3) For many years one has been assuming already that the aerosols in the upper
atmosphere derive above all from volcanic eruptions.
Simplified Analysis
English Source Text:
• Experiential: Carrier (embedded: Token – Process:relational – Value) –
Process:relational:locational – Location – Circumstance of Time
• Logical: Carrier (Head ! [[projected fact]])
• Interpersonal: Indicative/Non-modalized
• Textual: Theme: unmarked, on Subject ‘The suspicion that . . .
atmosphere’
• Information: NEW strongly on ‘primary source . . . Atmosphere’ and
weakly on ‘for many years’
German Translation:
• Experiential: Circumstance – Process: mental: cognition – Processor Cir-
cumstance Idea (Carrier – Process:relational – Locational)
• Logical: α: Head-clause – projection – β: projected clause
• Interpersonal: Indicative + indicative/Non-modalized
• Textual: Theme unmarked, on ‚Seit vielen Jahren . . .‘
• Information: NEW on ‚vor allem aus Vulkanausbrüchen . . .‘
As we can see, there is no isomorphism and in this sense no equivalence
between experiential and logical structures at the lexicogrammatical level
between source and target. The configurations of clause functions vary, and
the mapping of semantic functions onto lexicogrammar varies too – with
the exception of the NEW-element.
In terms of a clause semantics (Halliday and Matthiessen 1999), however,
both source text and target text encode the following ideationally equiva-
lent ‘sequence of figures’ (states of affairs):
(4) Someone unspecified assumes/suspects that when volcanoes erupt,

this causes changes in the amount of aerosols in the upper
atmosphere.
In terms of grammatical transitivity categories, there is at best partial

equivalence between source text and target text. In particular, and very
significantly for translations, the target text is less grammatically meta-
phorical, and in that sense more explicit, than its source text. It is closer to
the sequence of figures in (4) above. We can also see that the primary loci of
equivalence at the lexicogrammatical level here seem to be information
distribution and mood/modality. Source text, target text, and the semantic
sequence in (4) are in a weak sense experiential paraphrases of each other:
truth conditions between source and target text also seem to be largely
746 ERICH STEINER
stable. The German (2) is thus truth-conditionally, interpersonally, and in

terms of information structure equivalent to its source, yet not in terms of
ideational grammar and thematic structure. In an analysis taking the
preceding co-text, and thus cohesion, into account, we would also see that
the German translation (2) represents a good language-specific solution in
terms of information-flow across clauses.
Moving on to the relationship between source context and target context, and
thus to the question of ‘equivalence’ between the contextual configurations
of source and target text, we are entering the area privileged as ‘translation
brief’ or ‘scopos’ in German-type ‘functional approaches’ (see Reiss and
Vermeer 1984). Here the default case for SFL would again be a maximally
close relationship between the contexts, but any of the parameters (field,
tenor, mode) can be required to be changed in principle. However, the
stronger these changes are, and the more register dimensions are involved
in such changes, the less the mode of text production will still be that of a
translation. A translation brief requiring strong changes in register, thus
downgrading the constraining influence of the source text encoding, will
become a case of text production under contextual constraints which are
not linguistically encoded – and thus simply a case of (multilingual) text
production in the more general sense. To give a realistic example: if a
translation brief requires the translation of a handbook which in the
original was meant as a text by experts for experts (high level of expertise
under tenor of discourse) into a text addressed by informed laypersons to
laypersons, this is a change of context and ‘scopos’ – yet in theoretical
terms only marginally a translation, and in practice often commissioned
as a multilingual text production.
28.2.4 Product: The Text Type ‘Translation’

Translation as a mode of text production is unique, in that it is the only one
happening under the constraints of a linguistic source text. Its input specifi-
cation not only has encodings (cues) about events and logico-semantic
relations between them, but also carries interpersonal meanings and
instructions about how to process the ideational and interpersonal mean-
ings. No other form of input specifications has these properties to the
same extent. The default-instruction deriving from the relationship of
‘translation’ between source and target texts triggers a very specific form
of text production and very specific evaluation criteria. This, and the
required parallel processing effort, gives the text type ‘translation’ some
specific properties. Translation properties have been variously postulated as
simplification, normalization, levelling-out, sanitization, disambiguation, conventio-
nalization, standardization, explicitation, and others. Some SFL approaches
would give a special status to ‘explicitation’, possibly together with
levelling-out and disambiguation. These are necessary properties of transla-
tions, because ‘understanding’ of the source text involves explicitation and
disambiguation, and the frequent pressure towards conforming with

target-culture norms makes sets of translated texts often more similar to
each other than to the originals. The other postulated properties of trans-
lated texts may be due to other factors, such as cultural dominance or
explicit norms formulated for translating, but they are not a unique conse-
quence of the production type ‘translation’ (see Hansen-Schirra et al.
2012:3–4, 255–80).
Understanding the source text involves relating given units of text to
more explicit and more ‘literal’ paraphrases. In contrast to the more con-
ventional notion of paraphrase, SFL adopts a notion of ‘multi-functional’
paraphrase (Steiner 2001:179–85). It is common knowledge in linguistic
theorizing that states and events represented in texts can be encoded in
different grammatical categories, such as clause complex, clause, phrase/
group, word, morpheme, and of nominal, verbal, etc., types. In each case, a
distinction can be made between ‘congruent’ (transparent, literal, direct,
non-metaphorical) variants, on the one hand, and ‘metaphorical’ ones, on
the other. Aspects of this general phenomenon are variously covered as the
mapping of a given semantic category onto different grammatical categor-
ies. It is also addressed in some of the translation procedures discussed in
Section 28.3.3 below. We are exploiting here Halliday’s (Halliday and Mat-
thiessen 1999, 2014: Chapter 10) comprehensive notion of ‘grammatical
metaphor’ and ‘fractal types’ for a modelling of understanding in
translation.
Any kind of linguistic meaning can be expressed on different lexicogram-
matical ranks, and within ranks by different types of unit, ranging from a
cohesive device, still outside grammar, through phrases headed by their
lexical heads, to words and morphemes. It is not the head alone which
expresses the meaning, but potentially the complete phrase. Furthermore,
change of phrasal type in the expression of some unit of information, say
from verb(group) to noun(group), usually involves lexicogrammatical
changes in other grammatical units within the same clause and sometimes
within entire clause complexes. According to Halliday and Matthiessen
(1999), metaphorization within a language seems to follow a certain hier-
archy, graphically represented in Figure 28.1.
Figure 28.1 visualizes the idea that with increasing ‘grammatical meta-
phorization’, qualities tend to be encoded as things, processes as qualities
and then things, circumstances as processes, then qualities, then things,
etc. This idea is embedded in a wider theory of packaging of information
which assumes that along the axes of ontogenetic, phylogenetic, and semo-
genetic (textual) development, linguistic meaning tends to be encoded more
and more ‘densely’. Within such a view, a central aspect of ‘understanding’
and ‘decoding’ can be modelled as a process of grammatical de-
metaphorization, and this is where it makes contact with understanding
source texts and re-producing target texts in translation. In principle,
grammatical metaphorization/de-metaphorization as basic processes apply
748 ERICH STEINER
Figure 28.1 Direction of metaphorization [note: numbers refer to examples in Figure 28.3]
(Halliday and Matthiessen 1999:264)
inter-lingually as much as they do intra-lingually in processes of under-

standing. Let us schematize an assumption about translation based on the
notion of grammatical metaphor in a very simple diagram (Figure 28.2).
Understanding as symbolized in Figure 28.2 crucially involves the
unpacking of grammatical metaphor, the downward arrow in the ‘under-
standing’ direction. At some point of depth of understanding (bottom end
of both arrows), re-production in the target language sets in, and here the
process of re-metaphorization is cut short below the degree to which it
might otherwise go (the two blocks halfway up the ‘production’ arrow). In
cases where the translation is even more densely packed than its source-
language unit, we speak of metaphorization. A higher degree of grammat-
ical metaphoricity may be due to a highly skilled translator, to the complete
avoidance of unpacking in cases of highly routinized stretches of text which
allow direct transfer with possibly ensuing further metaphorization, or to
constraints depending on the linguistic system or register involved. Our
notion of ‘explicitation’ is restricted to properties of linguistic encoding,
rather than to the full situationally instantiated interpretation of a commu-
nicative act. Full explicitation of all logical and pragmatic implicatures is a
different process. In terms of other translation properties mentioned at the
beginning of this section, de-metaphorization usually results in simplifica-
tion and explicitation. Re-metaphorization is an indicator of interference
from the source text, and metaphorization may go hand in hand with
implicitation and also with normalization.
The notion of ‘the translation process’ as a source of explanations for
observed properties of translated texts only makes sense if used in a con-
strained way. The translation process may have arbitrary contextual
Figure 28.2 Incomplete re-metaphorization in target language (from Steiner 2004:143)
conditions and cultural conditions impinging on it. These might include a

lack of effort or ability on the part of the translator; they might be a
consequence of the translator’s adopted translation strategy (‘overt vs.
covert’, i.e. ‘source-text oriented’ vs. ‘target-text oriented’, and thus
adapting a cultural filter in the sense of House (2015)); and they might have
to do with explicitation in the sense of Relevance Theory (Gutt 1991) – or,
indeed, with some as yet unknown factor. Such an uncontrolled blend of
factors cannot directly serve as a methodologically valid source of
explanation.
Our focus here is on the processing issues referred to in Figure 28.2 and
the influence which they exert on the density and explicitness (metaphori-
city) of information packaging (Alves et al. 2010). These might count as a
‘translation universal’ because translation seems to be the only type of text
production in which previously linguistically encoded information has to
be re-encoded in a different language system under some constraining
notion of ‘equivalence’. Assuming that linguistic encoding (in the source
text) already represents an ‘optimized’ solution to the complex problem of
communication, we would predict that any serious attempt at re-encoding
in a different language will pose unique challenges. These challenges would
be strongly subject to all the specifics of the language contrasts involved,
but their particular nature would arise out of the relationship (and process)
of translation. Any ‘weaker’ form of multilingual text production in the
sense of producing a target-context-adapted re-creation, would not show
these translation-specific properties – because it would not be facing the
specific challenges.
28.2.5 Process: The Role of the Translator

The translator is a text producer of a special kind, because of the constraints
of the source text and the requirements of ‘equivalence’. However, s/he has
an agentive role in terms of register theory (see Section 28.3.1), just like any
750 ERICH STEINER
text producer has. Ideally, translators have a strong responsibility towards

the source text, and hence are highly competent de-coders and ‘understand-
ers’, with a high awareness of the contrastive relationships of the language
and textual systems involved. They will be acutely aware of factors influ-
encing the evaluation of the translation product. Depending on the transla-
tion brief and on their decisions about value systems to be brought to the
evaluation of their translation, they may responsibly decide between
staying with translation proper, or moving, for the given task, closer to
other forms of multilingual text production. Apart from these ethically
based roles, the translator plays a significant role through level of expertise
and training, as well as capabilities of coping with technology and time
constraints – in short, all the factors impacting on the process of translat-
ing. Psycholinguistic experiments in Alves et al. (2010) have traced the role
of some of these factors in the choice of translation unit and intermediate
grammatical shifts before a final product is reached.
28.3 Tools for the Translation Process
This discussion of tools is focused on SFL-inspired tools, rather than on

more general-purpose tools for translators, such as translators’ work-
benches, including machine(-aided) translation systems, translation mem-
ories, post-editing systems, etc. without any particular theoretical
commitment.
28.3.1 Pre-translational Text Analysis

SFL has a rich tradition of developing detailed tools for pre-translational
text analysis (Baker 1992; Hatim and Mason 1990; House 2015; Munday
2012; Steiner 2004).2 They are centred around variants of register analysis
in terms of field, tenor, and mode of discourse (Halliday et al. 1964),
variously complemented by an underpinning of contrastive linguistics
and/or contrastive cultural studies. ‘Genre’ is either integrated into register
analysis or else a separate notion alongside register. Both pre-translational
analysis and evaluation are based on the assumption that the stability of the
register that is being translated can be regarded as a default case, but may
be changed due to the translation brief: translating between strongly dif-
ferent contexts in different cultures may involve deliberate changes in
situational/contextual variables and in register. Where these changes are
strong, translation as a process shades into multilingual text production.
Neither text analysis nor evaluation assumes that the default case is ‘one
text – one register’. In fact, registers usually change in different generic
2
Section 28.3.1 is a revised version of Steiner (2004:25–43).
Figure 28.3 Contextual configuration
stages of a text. Crucially, and this is different from other applications of

register analysis, the pre-translational text analysis as well as the evaluation
of the target text need to have a substantial component addressing lan-
guage-pair-specific problems (for examples, see Steiner 2004:37–40, 57–61).
Register is often discussed under the three major sub-headings of ‘field,
tenor, and mode of discourse’, each of which is then subdivided, depending
on the purpose and specificity of the task at hand, into a small number of
sub-variables (see Figure 28.3 for a version). For the sake of terminological
consistency, it must be remembered, though, that
[a] register is a semantic concept. It can be defined as a configuration of

meanings that are typically associated with a particular situational configur-
ation of field, mode and tenor. But since it is a configuration of meanings, a
register must also, of course, include the expressions, the lexicogrammatical
and phonological features that typically accompany or REALISE these meanings.
(Halliday and Hasan 1989:38–9, emphasis in the original)
28.3.1.1 Field of Discourse

‘Field of discourse’ is divided here into three sub-variables which collect-
ively specify patterns of variation: experiential domain, goal orientation,
and social activity. Experiential domain is what is informally referred to
elsewhere as subject field or subject matter. Its realizations are the struc-
ture of lexical fields set up in texts, terminologies, cohesive lexical and co-
referential chains, headings of various types, paragraphing, transitivity of
clauses, encodings of time, perspective, and Aktionsart. Whereas texts in
natural communication have at least one experiential domain, it has often
been in the form of language for special purposes/specialized domains that the
752 ERICH STEINER
phenomenon has attracted attention. Differences in the structuring of

domains across cultures, together with questions of the degree of termino-
logical professionalization and degree of specificity of some lexical field, are
standard problems of translation. Contrastive studies of lexis show lexical
fields of varying degrees of specificity across languages (on English and
German, see König and Gast 2012:25, 247). An example of different degrees
of professionalization and specificity can be seen where culture A classifies
some domain in terms of easily observable external properties of phenom-
ena and/or in relatively little detail, whereas culture B may classify that
same domain in terms of only scientifically/technologically accessible cri-
teria, usually with a high degree of specificity. Everyday instances of such
problems are provided by translating between cultures with different
degrees of terminology formation for some area of knowledge, or by having
to change the degree of specialization in translation. In the first case, the
translator may have to act as a creator of terminology, in the second, as a
technical writer with a well-developed reader-model. All too often the use of
terminology imported from English or some other dominant language has
taken the place of indigenous terminology formation with negative impli-
cations for identity, culture, and nation building (Drame 2015).
Goal orientation is another variable – sometimes subdivided into short-
term goal and long-term goal. The notion of goal can be applied in various
degrees of granularity, i.e. different ranks on which a text structure can be
discussed (e.g. text, paragraph, clause complex). Considering the goal(s) of a
text as a whole, the main options would be those of ‘interchange’, ‘expos-
ition’, ‘persuasion’, ‘argumentation’, ‘instruction’, ‘narration’, etc. (Martin
1992:560–70; Matthiessen 2015), or a closely related classification. Realiza-
tions are not restricted to ideational meanings: some of the more promin-
ent lexicogrammatical realizations of goal structure include clause mood,
modality, presence or absence of ellipsis, tense selection, transitivity,
agency patterns, patterns of identifiability, types, frequency and realization
of conjunctive relations, patterns of thematic progression, typical patterns
of topic construction, and paragraphing in written texts. These patterns run
across entire texts and paragraphs and define properties of entire dis-
courses.3 This has straightforward consequences for any conception of a
unit of translation, if, for example, someone has to translate a largely
expository paragraph, into a text(-fragment) which is, say, mainly narrative
(e.g. in advertising). The requirement may be due to constraints arising
from intercultural differences in the function of genres, or because the
target-language readership is different in age from that of the source-
language text. Even when goals remain constant in translation, the
3
Halliday and Hasan (1976:229) include these ‘text-type’ options under ‘mode of discourse’, realized through the
textual component of the lexicogrammar. We interpret them here as more directly deriving from the semiotic goals
under ‘field’, diverging from a strong version of the ‘context-metafunction-hook-up hypothesis’, also problematized in
Thompson (1999).
lexicogrammatical realizations of the goal orientation in the different cul-

tures concerned may differ.
Social activity refers to general types of activity recognized across a
culture as meaningful and directly relatable to the needs of text producers:
production, exchange, communication, reproduction, consumption, or
their subtypes. Any text, along with its lexicogrammatical properties, is
constrained by the particular activity of which it is a part, or a realization.
A text about a given topic will have very different properties, for example,
depending on whether it is an advertising text or whether it is an instruc-
tion for production of the same object or commodity.
28.3.1.2 Tenor of Discourse

The main sub-variables under ‘tenor’ are agentive roles, social roles, social
distance, and sometimes separately, affect. Problems may arise in transla-
tion whenever a tenor variable is expressed by different sets of lexicogram-
matical properties between languages, or where the variable itself has to be
changed because of different cultural practices.
Agentive roles (or ‘agent roles’ in Halliday and Hasan 1989:56–7) are
semiotic roles assigned through the text to author and reader/hearer, such
as ‘vendor vs. customer’; ‘giver vs. receiver’; ‘sayer vs. listener’; ‘teacher vs.
learner’; etc. Important lexicogrammatical realizations are mood selec-
tions, ellipsis, modality, the use of specialist language, options in key, etc.
For example, the vendor vs. customer roles in a sales interaction may be
textually encoded with very different degrees of directness or explicitness,
and the translator or interpreter needs a comprehensive awareness of these
roles and highly developed linguistic capacities to realize them textually in
order to make translations successful, at least in covert translations.
Social role has to do with how a text encodes social power relationships
between participants along dimensions such as social class, gender, level of
expertise, age, ethnicity, religious affiliation, and education. The basic
options here seem to be equal vs. unequal, and then subtypes of these. The
main lexicogrammatical realizations of such choices are the interpersonal
systems of the grammar, but, beyond these, all systems structuring texts and
dialogues in response to social hierarchies. In a case where a source-language
text strongly encodes gender specificity, the issue would be whether the
target-language text is meant to have that property, too, and if so, how it is
achieved lexicogrammatically. Linguistic realizations include everything
from topic selection through genre to lexical selection and non-verbal com-
munication. As with all the examples discussed here, the translation chal-
lenge can be viewed on several levels: to start with, gender may or may not be
exploited as a variable of social power in the wider contexts of culture
involved. Secondly, the corresponding power relationships may be expressed
in culture-specific ways of structuring contextual configurations. And
thirdly, irrespective of whether or not we find differences at the two levels
previously mentioned, the lexicogrammatical means of and proportionalities
754 ERICH STEINER
of the register will be language-specific. The translation relationship, then,

needs to be defined within this overall matrix of variation.
Social distance stands for the degree to which the contextual space is
shared by participants. This is determined, in the first instance, by the
frequency and range of previous interaction. Realizations include tagging,
modality, use of types of form of address, use of accents and dialects, and
sociolects, etc.
The variable of ‘affect’ is sometimes singled out for particular attention
(see Martin 1992:533–6; in translation, Munday 2012). Major options have to
do with positive vs. negative, permanent vs. transient, self vs. other. What is
semiotically encoded here is affect towards self, other, and subject matter, as
shown in lexical selections, but also in a wide variety of grammatical choices
and rhetorical devices, such as repetitions, parallelisms, etc. Any model of
translation will want to answer the question of whether, and if so, how, a
given degree of affective involvement should be maintained in translation.
28.3.1.3 Mode of Discourse

The ‘mode of discourse’ refers to the role language is playing, the channel
used for communication, and the medium of discourse. Languages, texts,
situations, and cultures are characterized by considerable heterogeneity
with respect to whether or not they have evolved specific codes for written
vs. spoken language, how they use these with different channels (printing,
face-to-face, electronic), and whether there are marked differences between
ancillary and constitutive language use.
In terms of ‘language role’, texts differ depending on whether they are
(part of ) a linguistic activity (constitutive), or part of a non-linguistic activ-
ity (ancillary). Lexicogrammatical and cohesive systems involved are ellip-
sis, mood, and reference. Consider, for instance, a source-language text
with a substantial amount of visual and graphic information (tables, fig-
ures, diagrams), as in many varieties of academic or scientific writing. The
target culture may have quite distinct and heavily prescribed ways of using
language in an ancillary role to refer to these other types of information
(pronouns, captions, ellipsis, etc.). Another example is the translation of
any case of language in action, e.g. in film dubbing. Here cultures have
developed a whole variety of ways of using language as an ancillary part of
some non-linguistic activity, realized in the way they rely on patterns of
turn-taking, boundary signals in dialogic exchanges, ellipsis, hesitation
signals, etc. The activities of film dubbing and subtitling are, in their
different ways, prime examples of language in an ancillary role under the
constraints of source-language versions.
‘Channel of discourse’ has to do with the physical channel employed for
the discourse under investigation. Major options are sound waves, elec-
tronic channels, paper, telephone lines, etc. The various types of channel
are of interest here to the extent to which they offer and constrain choices
in meanings and their realizations. A channel which does not allow face-to-
face interaction, for example, limits the use of gesture and body-language in
communication, but may support other (variants of ) semiotic systems
instead, such as parallel display of texts, icons, etc. A recognition of the
importance of channel constraints helps the translator give due weight to
conventions of punctuation and paragraphing when using any of the print
channels. Again, languages show considerable variation here. A historical
example of the pervasive influence of the channel (and medium) of dis-
course is the invention and spread of printing and its influence on standard
languages in particular. The spread of electronic channels may currently be
causing another wave of textual and cultural change. Relevant textual
realizations, possibly indicating such changes, are the increasing frequency
of multiple and de-personalized authorship with all its linguistic effects, the
use of hypertext techniques in addition to, or instead of, continuous linear
text, and consequently, new patterns of cohesion and coherence.
In ‘medium of discourse’, the major variables probably still are spoken
vs. written, although electronic channels might conceivably lead to a new
medium of discourse. Patterns of realization involved are the use of pro-
nouns vs. full words, exophoric vs. endophoric reference, types of cohesion
in general, certain types of clause complexity, grammatical metaphor, etc.
English and German, for example, differ in interesting ways as to how they
employ lexicogrammatical resources to structure information in written
and spoken language, for example, through types of clefting, extraposition,
inversion, and related phenomena – in general thematizing and focusing
devices. A model of translation needs to offer scope for deciding whether
the medium of discourse remains constant in the translation of a given text,
and what lexicogrammatical configurations are appropriate in the target
language for realizing the medium.
28.3.2 Evaluation and Criticism

Arguably the most widely used approach to the (human) evaluation of
translations has been Juliane House’s Translation Quality Assessment (1977,
2015) since the 1970s. She combines SFL-based register and genre analyses
with influences from contrastive pragmatics, intercultural communication,
and more recently, corpus-based and experimental studies of translation.
Steiner (2004) adds a strong influence from typologically based contrastive
linguistics to House’s approach to evaluation. In these SFL-based tools for
translation evaluation, a variant of register analysis as described in Section
28.3.1 is used and the target text evaluated against the default assumption
of equivalence in register. Register changes may be evaluated positively, yet
then require a conscious brief to re-set contextual variables. Munday (2012)
builds on Martin’s (1992:533) model of ‘affect’, later developing into
‘appraisal’, for the exploration of ‘attitude’, ‘graduation’, and ‘engagement’
in originals and for translation evaluation in target texts, thus focusing on
the ‘tenor of discourse’.
756 ERICH STEINER
28.3.3 Translation Methods and Procedures

Translation methods and procedures are among the key tools in most
approaches to teaching translation (Vinay and Darbelnet 1958; Newmark
1988; Fawcett 1997). In SFL modelling, such procedures and methods can be
located at the different levels and in the different functional components of
the theory’s architecture. At the level of lexicogrammar, the SFL notion of
‘grammatical metaphor’ offers a theoretically motivated and uniquely
comprehensive tool. Figure 28.4 interprets changes in lexicogrammatical
category with experientially largely constant meaning as translation pro-
cedures. The congruent variants are, when instantiated in their full gram-
matical unit, usually more explicit than their ‘metaphorical’ counterparts.
The assumed translation direction here is English into German. An SFL-
based account of ‘grammatical metaphor’ provides a theoretically
grounded model of grammatical translation procedures and the associated
processes of ‘explicitation/implicitation’, all of which are essential for
translation.
For the sake of illustration, sentences (1) and (2) are here repeated as (5)
and (6), identifying types of changes in grammatical metaphoricity and thus
translation procedures as in Figure 28.4:4
(5) The suspicion that volcanic eruptions are the primary source of
aerosols in the upper atmosphere has been around for many years.
(original, New Scientist 21 January 1982: 150)
(6) Seit vielen Jahren vermutet man schon, dass die Aerosole in den höheren
Schichten der Atmosphäre vor allem aus Vulkanausbrüchen stammen.
(German translation, Doherty 1991)
• The suspicion ! vermutet (2.1)

• Volcanic eruptions ! Vulkanausbrüchen (13.2)
• the primary source ! vor allem stammen aus (2.1 + 13.2)
• of aerosols ! die Aerosole (13.1.a)
• the upper atmosphere ! den höheren Schichten der Atmosphäre (13.2)
• has been ! schon (variant of 5.2 involving an adverb)
The translation procedures in Figure 28.4 find their more global counter-
parts in what Newmark (1988:45–53) calls ‘translation methods’, and what
SFL integrates into its overall model as ‘translation strategy’ (Teich
2001:212–13), globally applied to entire texts. They move between ‘free/
target-culture oriented’ translation privileging contextual equivalence, and
‘literal/source-culture oriented’ privileging grammatical equivalence.
Translation strategies at the level of cohesion are exemplified for English-
German in Steiner (2015b).
4
A fuller analysis of the grammatical relationships in (1) to (2) can be found in Steiner (2004:157–8).
Figure 28.4 Translation procedures: change in grammatical category in translation

(extended from Steiner 2004:141, based on a monolingual version in Halliday and
28.3.4 Workflow: Translation-oriented Language Technology

SFL has so far not made any direct contributions to the digital translation
workflow (translators’ workbenches, etc.). There are, however, promising
tools with a characteristic SFL input. On the one hand, we can identify an
interesting trend to develop electronic corpora of originals and translations
as sources for research, but also for teaching translation and potentially for
translating itself (Hansen-Schirra et al. 2012; Kunz et al. 2017; Teich 2003).
758 ERICH STEINER
Prerequisites for work with electronic corpora include tools for semi-
automatic annotation, such as the UAM corpus tool (O’Donnell 2008). On
the other hand, Bateman (2008), Hiippala (2012), and Taylor and Baldry
(2001) suggest approaches and initial implementations for the analysis and
translation of multimodal documents, strongly influenced by SFL
theorizing.
28.4 Future Contributions to Translation (Studies)
28.4.1 The Activity of Translating

The role of translation as an activity has drastically expanded and changed
over the past, roughly, 100 years. Newmark (1988:9) surveys the major
trends up to about the 1980s, noting an increasing emphasis on readership
and setting, an expansion of topics beyond the religious, literary, and
scientific towards technology, trade, current events, publicity, and to virtu-
ally every topic of writing. Along with that, he diagnoses an increase in
variety of text formats, standardization of terminology, the increasing
formation of translator teams, and the recognition of the reviser’s role.
In translator training, he emphasizes the impact of linguistics, sociolinguis-
tics, and translation theory. All these trends have continued into the pre-
sent. Here we characterize the more recent contexts of translation using
a non-technical version of the SFL register framework as illustrated in
Section 28.3.1.
In terms of the ‘field of discourse’, and more specifically the ‘experiential
domains’, translation has undergone increasing diversification. Legal, eco-
nomic, and in particular, technical domains have increased in importance,
with medicine being another growing domain. ‘Goal orientations’ and text
types have diversified with the domains, in some fields alongside increasing
standardization. In some technical areas, even the lexicogrammar has
become restricted, as can be seen in the introduction of ‘controlled lan-
guages’. The ‘social activity’ of translation has undergone modification
from one being mainly situated in the spheres of cultural reproduction
(religious and literary translation) up to the twentieth century, to being
situated very much in production directly (translation as a professional
activity, software localization, multilingual technical writing, consulting),
and of course in interaction (court interpreting, medical interpreting, com-
munity interpreting in general, etc.). Modern language service industries
cover a broad band of these activities, employing translation-oriented lan-
guage technologies in elaborate translators’ workbenches. Any approach to
translation which aims to cater for this wide range of activities needs a
broad and functionally oriented theoretical grounding.
As for the ‘tenor of discourse’ in translation, the ‘agentive roles’ of the
translator have become more diversified along with the goals and text
types. In addition to paying respect to the authors and recipients of the
source and target texts, the translator has to decide more consciously what
his/her own agentive role is. Should the translator become visible and create
something that is recognizable as a translation, or should s/he remain
invisible and strive for maximum adaptation to target-cultural norms?5
Likewise, ‘social roles’ have become more diversified, and so have the
means of realization. The possible degrees of levels of expertise have
become more differentiated. There are nowadays between the extremes of
the absolute layperson and the absolute expert all shades of semiotic
possibilities as encoded in texts. Social hierarchies are often expressed
textually. Along with a broader range of social roles, we find a broader
range of ‘social distance’. The possibility of styles from absolutely frozen to
intimate has to be mastered by the present-day translator, alongside cultur-
ally specific changes in these scales. Finally, the various ways of realizing
affect, both as positive or negative and as self-oriented or other-oriented (see
Martin 1992; Munday 2012) are of increasing importance, especially in fields
such as advertising and in direct interactions (such as in interpreting), but
also in the various forms of synchronization and subtitling of films.
It may be in the ‘mode of discourse’ of translation that some of the most
far-reaching changes are taking place. Considering ‘language role’, we find
all possible shades between the poles of ancillary and constitutive. At the
ancillary end, there are registers where language plays a rudimentary role
only, being very much subservient to visual communication (see Taylor and
Baldry 2001; Bateman 2008). Other instances of this include software local-
ization, with ‘canned language’ in menu-guided user interfaces used for
many types of current software. However, there are still numerous registers
in which language is absolutely constitutive, as in the translation of fiction,
(popular) science, and interactions within large international organizations
such as the UN and the EU. As for the ‘channel of discourse’, we have to be
aware that whereas until about 1980 the dominant channel for most types
of translation may have been paper, along with conventional sound waves
for interpreters, this traditional channel has become accompanied, and
occasionally overtaken, by the electronic channel. The translator nowadays
has to be aware of the possibilities and limitations of the major channels of
communication, and of the tools for working in these. Systems for termin-
ology management, machine-aided translation, and above all for straight
text handling and word processing are gaining ground in most areas of
translation and multilingual text production. With these processes, the
possibilities of using intertextual relationships between versions of texts
have expanded, and translation memories for the production, evaluation,
and modification of existing texts are being used in professional practice.
These developments change the notion of ‘a text’, and ‘a translation’, from
a static to a much more dynamic and changeable entity. Finally, the
5
See House’s (2015:63–70) distinction between ‘overt’ and ‘covert’ translation and Venuti’s (1995) discussion of the
visibility of the translator.
760 ERICH STEINER
‘medium of discourse’: the variation between spoken and written may still
be the primary one, and it remains to be seen whether the electronic
channels ultimately lead to the creation of a major new medium – however,
electronic channels are certainly creating new sub-varieties, transcending
the binary spoken–written divide. Hypertext, inside or outside of the Inter-
net, allows new ways of creating non-linear discourses, as well as new ways
of exploring language-specific systems of conjunctive relations and other
forms of cohesion. Translators have to work increasingly web-based, and
the potentials for multilingual text creation and management have to be
thoroughly familiar to them.
28.4.2 Researching Translation

One strength of SFL work in translation is its potential to make connections
to investigations of the semiotics of culture and language: ‘texts’ and
‘discourses’ are not simply reified products or structures of elements on a
single level, but networks of relations among configurations of meanings
across situations and cultures. The decisive units of meaning are not micro-
level structures such as individual words or units of clause structure, but
more macro-level units such as those realizing register, genre, and ideology
in a culture. As well as giving due consideration to detail, work on texts
crucially involves macro-levels, traditionally undertaken in literary studies,
anthropology, and studies of discourse in context. SFL concepts of register,
genre, contextual configuration, and culture can bridge the gap between
traditions of research examining macro- and micro-levels of textuality.
Important concepts such as ‘foreignizing vs. domesticating’ translation
(Venuti 1995:148–86), ‘translations as facts of culture’ (Toury
1995:23–39), and ‘translational laws’ (Toury 1995:259–80) require operatio-
nalization in terms of textual detail in order to become as convincing as
they deserve to be. The older literature on translation focuses on the
translation of highly valued texts, such as philosophical and literary texts.
One of the most significant objects of (the study of ) translation has been the
Bible (for an SFL contribution, see Gregory 2001). Questions of paramount
importance in this area seem to be those of the role of the translator in
an overall hermeneutics of translation, the role of ‘narrative stereotypes’
(see Iser 1979; O’Sullivan 2003), and of course questions of ‘value’
(Halliday 2001).
Key concepts of a hermeneutics of translation can be elucidated with an
SFL architecture: ‘hermeneutics’ postulates ‘understanding (Verstehen)’ and
‘explicating (Auslegung)/interpreting’ as preconditions of ‘translation’ (Cer-
cel 2013:224–37). ‘Explication’ is oriented towards the source text and the
(inferred?) intention of the author. ‘Interpretation’ is oriented towards the
reading of the translator, which may happen against a different ideological
background from that of the author (Stolze 2003). ‘Translation’ is then
situated somewhere between ‘explication’ and ‘interpretation’. ‘Exegesis’,
differently from the above, is ‘interpretation plus commentary’. In

engaging with these questions, SFL can bring its analytic instruments to
bear on the different readings possible of a text/discourse, acknowledging
that these different readings are a product of the interaction between
context and text. Each ‘reading’ of a text disambiguates, instantiates its
meaning – and a model of translation ultimately needs to be clear about
whether the object of a translation is the individual text or the reading of it
by the translator. Readings can be governed by reader’s/translator’s dispos-
itions and choices; they can be ‘tactical, resistant, compliant’ and thus
‘subjectified’ readings (see Martin and Rose 2003:269–72) of instantiated
texts, and so important for translators’ choices in decision-making (Munday
2012). Obviously, this is a core area where the active role of the translator
finds its place.
Narrative stereotypes, or the guiding semiotic roles of author and reader,
analyses of ‘agentive roles’, and social hierarchy in register analysis may
become a decisive methodological tool for bridging the gap between classical
narratology and SFL, not only for translation, but also for investigations of
textuality generally. If we assume the basic possibilities for configurations of
social roles between author and reader to be those of no hierarchy, superior-
ity of author, and superiority of reader, then we can ask whether there is,
and should be, any manipulation of this configuration in translation. This
becomes highly relevant in translational adaptations of a text to an assumed
stage of language development of the (implicit) reader (grammatical meta-
phoricity, rhythm, sound). It also plays a role in – controversial – interference
with properties of the original in the interest of ‘inclusive language use’, and
it is unavoidable in cases of assumed multiple readership (adults, children) in
the implicit reader. But note that arbitrary interference, with the meanings
encoded in the source text and thus violation of core aspects of equivalence,
runs counter to the notion of ‘translation’ as such. All of this can be modelled
with the help of SFL register analysis.
SFL holds the promise of closing a significant gap between topics and
methods from arguably the two most influential parent disciplines of
translation studies: literary studies and linguistics. Approaches based on
literary studies focus on discursive macro-structures and relations, for
example, the interpretation of entire works of literature, or even entire
configurations of works, in their socio-cultural-historic context. Methodo-
logically, this is usually attempted through a heavily interpretative discus-
sion of the individual work. Linguistics, on the other hand, has tended to
focus on the micro-level of linguistic realization, such as phonemes, mor-
phemes, words, and clauses, and traditionally not in the individual
instance, but in the linguistic system. Linguists have often worked with
formalizations, attempting to generalize grammatical patterns into rule
systems, not even using truly empirical data, but rather constructed
examples. Between the macro- and the micro-levels thus conceptualized,
there was, and is, so much of a gap that the two approaches hardly inform
762 ERICH STEINER
each other. Traditionally, stylistics and rhetoric, operating on something

like a meso-level, provided some degree of contact. SFL, through its extrava-
gant architecture involving levels of language and scales of abstraction,
holds a unique promise for bringing these levels together in one perspec-
tive. If a systemically based discourse semantics (Martin 1992; Munday
2012) can be exploited in studies of translation, and if its connections to
the more ‘micro-level’ strata of the model can be upheld and strengthened,
we should be able to link macro-, meso-, and micro-levels of semiotic
analyses of translations. It is only through such connections that we have
instruments for inquiring into what lies at the centre of intercultural
communication: texts in contexts across cultures (Munday and Zhang
2015).
There are other, though related, questions that systemically based studies
of translation should be able to address: among those I would rate promin-
ently ‘language contact’, ‘multilinguality’, and ‘language change’ (Thoma-
son and Kaufmann 1988; Hansen-Schirra et al. 2012:255–80). An SFL
perspective will give due consideration to systems alongside structures, to
the instance alongside the system, and to more abstract (and at the same
time, more empirical) types of contrast than have often been at the centre
of theorizing. Language contact along the channels of system and instance
may be shown to be the process through which ultimately sociocultural
motivations drive language change. The way languages influence each
other is often through relative frequencies of use, changes in markedness,
directness, and explicitness, rather than through directly changing sys-
temic categories themselves. Theoretically motivated studies of translations
and other forms of multilingual text production should be significant steps
towards addressing questions of this important nature – important in the
last resort not only for an understanding of language, but also for an
understanding of different human cultures.
References
Alves, F., A. Pagano, S. Neumann, E. Steiner, and S. Hansen-Schirra. 2010.

Translation Units and Grammatical Shifts: Towards an Integration of
Product- and Process-based Translation Research. In G. M. Shreve and
E. Angelone, eds., Translation and Cognition. Amsterdam: John Benjamins.
109–41.
Baker, M. 1992. In Other Words. A Course Book on Translation. London:
Routledge.
Bell, R. T. 1991. Translation and Translating. London: Longman.
Catford, J. C. 1965. A Linguistic Theory of Translation. Oxford: Oxford Univer-
sity Press.
Cercel, L. 2013. Übersetzungshermeneutik: Historische und systematische Grundle-

gung. St. Ingbert: Röhrig Universitätsverlag.
Doherty, M. 1991. Informationelle Holzwege: Ein Problem der Überset-
zungswissenschaft. Zeitschrift für Literaturwissenschaft und Linguistik
21(84): 30–49.
Drame, A. 2015. The Social and Organizational Context of Terminology Work:
Purpose, Environment and Stakeholders. In H. J. Kockaert and F. Steurs,
eds., Handbook of Terminology, Volume 1. Amsterdam: John Benjamins.
Fawcett, P. 1997. Translation and Language: Linguistic Theories Explained. Man-
chester: St. Jerome.
Firth, J. R. 1968a. Linguistic Analysis and Translation. In F. Palmer, ed.,
Selected Papers of J. R. Firth 1952–1959. London: Longman. 74–83.
Firth, J. R. 1968b. A Synopsis of Linguistic Theory 1930–1955. In F. Palmer,
ed., Selected Papers of J. R. Firth 1952–1959. London: Longman. 168–205.
Ghadessy, M. and Y. Gao. 2001. Small Corpora and Translation: Comparing
Thematic Organisation in Two Languages. In M. Ghadessy, A. Henry, and
R. L. Roseberry, eds., Small Corpus Studies and ELT: Theory and Practice.
Gregory, M. 2001. What Can Linguistics Learn from Translation? In
E. Steiner and C. Yallop, eds., Exploring Translation and Multilingual Text
Production: Beyond Content. Berlin: Mouton de Gruyter. 19–40.
Gutt, E. A. 1991. Translation and Relevance. Oxford: Basil Blackwell.
241–92.
Halliday, M. A. K. 1962. Linguistics and Machine Translation. Zeitschrift für
Phonetik, Sprachwissenschaft und Kommunikationsforschung 15(i/ii): 145–58.
Halliday, M. A. K. 2001. Towards a Theory of Good Translation. In E. Steiner
and C. Yallop, eds., Exploring Translation and Multilingual Text Production:
Beyond Content. Berlin: Mouton de Gruyter. 13–18.
Halliday, M. A. K. 2009. The Gloosy Ganoderm: Systemic Functional Linguis-
tics and Translation. In J. Webster, ed., The Collected Works of M. A.
K. Halliday, Vol. 11: Halliday in the 21st Century. London: Bloomsbury.
105–26.
Halliday, M. A. K. 2010. Putting Linguistic Theory to Work. In J. Webster,
ed., The Collected Works of M. A. K. Halliday, Vol. 11: Halliday in the 21st
Century. London: Bloomsbury. 127–42.
Halliday, M. A. K. 2012. Pinpointing the Choice: Meaning and the Search for
Equivalents in a Translated Text. In J. Webster, ed., The Collected Works of
M. A. K. Halliday, Vol. 11: Halliday in the 21st Century. London: Bloomsbury.
143–54.
Halliday, M. A. K. and R. Hasan. 1976. Cohesion in English. Harlow: Longman.
Language in a Social-semiotic Perspective. Oxford: Oxford University Press.
sive Power. London: Falmer Press.
764 ERICH STEINER

Halliday, M. A. K., A. McIntosh, and P. Strevens. 1964. The Linguistic Sciences
and Language Teaching. London: Longman.
Halverson, S. 1997. The Concept of Equivalence in Translation Studies:
Much Ado About Something. Target 9(2): 207–33.
Hansen-Schirra, S., S. Neumann, and E. Steiner. 2012. Cross-linguistic Corpora
for the Study of Translations: Insights from the Language Pair English–German.
Berlin: Mouton de Gruyter.
Hatim, B. and I. Mason. 1990. Discourse and the Translator. London: Longman.
Hiippala, T. 2012. The Localisation of Advertising Print Media as a Multi-
modal Process. In W. Bowcher, ed., Multimodal Texts from Around the World:
Cultural and Linguistic Insights. Basingstoke: Palgrave Macmillan. 97–122.
House, J. 1977. A Model for Translation Quality Assessment. Tübingen: Gunter
Narr Verlag.
House, J. 2015. A Model for Translation Quality Assessment: Past and Present.
London: Routledge.
Iser, W. 1979. Der implizite Leser. München: UTB.
König, E. and V. Gast. 2012. Understanding English-German Contrasts. Berlin:
Erich Schmidt Verlag.
Kunz, K., S. Degaetano-Ortlieb, E. Lapshinova-Koltunski, K. Menzel, and
E. Steiner. 2017. GECCo: An Empirically-based Comparison of English-
German Cohesion. In G. De Sutter, I. Delaere, and M. Lefer, eds., New Ways
of Analysing Translational Behaviour in Corpus-based Translation Studies.
Berlin: Mouton De Gruyter.
Malinowski, B. 1935. Coral Gardens and Their Magic: A Study of the Methods of
Tilling the Soil and of Agricultural Rites in the Trobriand Islands, Volume 2: The
Language of Magic and Gardening. New York: American Book Company.
Martin, J. R. 1992. English Text. Amsterdam: John Benjamins.
Matthiessen, C. M. I. M. 2001. The Environments of Translation. In
E. Steiner and C. Yallop, eds., Exploring Translation and Multilingual Text
raphy. Functional Linguistics. 2(9): 1–48.
Matthiessen, C. M. I. M. and J. A. Bateman. 1991. Text Generation and Systemic
Functional Linguistics. London: Frances Pinter.
Munday, J. 2012. Evaluation in Translation: Critical Points of Translator Decision-
Making. Abingdon: Routledge.
Munday, J. and M. Zhang, eds. 2015. Target (Special Issue: Discourse Analysis in
Translation Studies) 27(3).
Newmark, P. 1988. A Textbook of Translation. New York: Prentice Hall.
O’Donnell, M. 2008. Demonstration of the UAM CorpusTool for Text and

Image Annotation. In J. Lin, ed., Proceedings of the ACL-08: HLT Demo Session
(Companion Volume), Columbus, June 2008. 13–16.
O’Sullivan, E. 2003. Narratology Meets Translation Studies: The Voice of the
Translator in Children’s Literature. Meta: Translators’ Journal. 48(1–2):
197–207.
Reiss, K. and H. J. Vermeer. 1984. Grundlegung einer allgemeinen Transla-
tionstheorie. Tübingen: Niemeyer.
Steiner, E. 2001. Intralingual and Interlingual Versions of a Text: How
Specific is the Notion of Translation. In E. Steiner and C. Yallop, eds.,
Exploring Translation and Multilingual Text Production: Beyond Content. Berlin:
Mouton de Gruyter. 161–90.
Steiner, E. 2004. Translated Texts: Properties, Variants, Evaluations. Frankfurt
am Main: Peter Lang.
Steiner, E. 2005. Halliday and Translation Theory: Enhancing the Options,
Broadening the Range, and Keeping the Ground. In R. Hasan, C. M. I. M.
Steiner, E. 2015a. Halliday’s Contributions to a Theory of Translation. In J. J.
Webster, ed., The Bloomsbury Companion to M. A. K. Halliday. London:
Steiner, E. 2015b. Contrastive Studies of Cohesion and Their Impact on Our
Knowledge of Translation (English-German). In J. Munday and M. Zhang,
eds., Target (Special Issue: Discourse Analysis in Translation Studies) 27(3): 351–69.
Steiner, E., U. Eckert, B. Weck, and J. Winter. 1988. The Development of the
EUROTRA-D System of Semantic Relations. In E. Steiner, P. Schmidt, and
C. Zelinsky-Wibbelt, eds., From Syntax to Semantics: Insights from Machine
Translation. London: Frances Pinter. 40–104.
Stolze, R. 2003 Hermeneutik und Translation. Tübingen: Narr.
Taylor, C. 1998. Language to Language. Cambridge: Cambridge University
Press.
Taylor, C. and A. Baldry. 2001. Computer Assisted Text Analysis and Trans-
lation: A Functional Approach in the Analysis and Translation of Adver-
tising Texts. In E. Steiner and C. Yallop, eds., Exploring Translation and
Multilingual Text Production: Beyond Content. Berlin: Mouton de Gruyter.
277–306.
Teich, E. 2001. Towards a Model for the Description of Cross-linguistic
Divergence and Commonality in Translation. In E. Steiner and C. Yallop,
eds., Exploring Translation and Multilingual Text Production: Beyond Content.
Berlin: Mouton de Gruyter. 229–48.
Teich, E. 2003. Cross-linguistic Variation in System and Text: A Methodology for
the Investigation of Translations and Comparable Texts. Berlin: Mouton de
Gruyter.
Thomason, S. G. and T. Kaufman. 1988. Language Contact, Creolization, and
Genetic Linguistics. Berkeley: University of California Press.
766 ERICH STEINER
Thompson, G. 1999. Acting the Part: Lexicogrammatical Choices and Con-

textual Factors. In M. Ghadessy, ed., Text and Context in Functional Linguis-
tics. Amsterdam: John Benjamins. 101–24.
Toury, G. 1995. Descriptive Translation Studies and Beyond. Amsterdam: John
Benjamins.
Venuti, L. 1995. The Translator’s Invisibility: A History of Translation. London:
Routledge.
Vinay, J.-P. and J. Darbelnet. 1958. Stylistique comparée du francais et de
l’anglais. Paris: Les Éditions Didier.
Yallop, C. 2001. The Construction of Equivalence. In E. Steiner and
C. Yallop, eds., Exploring Translation and Multilingual Text Production: Beyond
Content. Berlin: Mouton de Gruyter. 229–48.
29
Language Typology
Abhishek Kumar Kashyap
29.1 Introduction1
Language typology evolved as a method of linguistic inquiry to explore cross-

linguistic diversity and genetic relationships among languages. With centur-
ies of research and teaching it graduated to be a fully-fledged sub-discipline
within general linguistics like other sub-disciplines in the field, e.g. phonetics/
phonology, morphology, and sociolinguistics. With the development of what
Nichols (2007:231) calls ‘hallmarks of a mature discipline’ (e.g. specialized
forums for publication and debate, dedicated journals, professional associ-
ations, conferences and symposia, classic works and textbooks, and research
and teaching programmes), recent years have seen renewed interest in this
field. This interest includes work which focuses on linguistic structures as a
primary concern of intellectual inquiry as well as work which is primarily
meaning-focused, such as that within the framework of Systemic Functional
Linguistics (SFL) (e.g. see Webster 2008). Typological insights are not just
confined to enriching our understanding of cross-linguistic diversity and
genetic relationships among languages, but they are also applied to
strengthen other fields of linguistics, with increasingly growing interests in
the application of typological insights in fields such as intercultural commu-
nication, translation, language acquisition, and language learning and teach-
ing (e.g. Matthiessen 2001; Filipovicˊ 2008; Bowerman 2010).
In this chapter I present a snapshot of the development of language
typology and discuss key strands of typological research in language, espe-
cially what the term ‘language typology’ refers to, how it has grown over
time and what difference typological insights make in our life. In Section 29.2,
I explain what we mean by language typology and what types of questions
typologists engage with. Sections 29.3 and 29.4 are devoted to the
1
Please see the end of the chapter for a list of abbreviations used in the examples in this chapter.
768 ABHISHEK KUMAR KASHYAP
development of this field. In Section 29.3, I sketch the historical growth of

this field through cross-linguistic and comparative works, and the essence
of different theoretical approaches are discussed in Section 29.4, with
Section 29.4.1 presenting a comparison of formal and functional
approaches to language typology and Section 29.4.2 focusing on SFL per-
spectives on language typology. Section 29.5 briefly discusses the applica-
tion of typological insights to other related fields of study. Section 29.6
summarizes the discussions of the chapter with a few concluding remarks.
29.2 What is Language Typology?
The term ‘typology’ derives from the Greek noun týpos and refers to the
study of types, or ‘a system for dividing things into different types’ (Hornby
2005:1656; Collins English Dictionary 2017). Retaining its meaning of classify-
ing things into categories according to their types, the term is used in
several disciplines such as anthropology (classification of cultures and races
and sociocultural norms), psychology (the classification of different human
personalities and personality traits), archaeology (the classification of
things according to physical characteristics), and theology (the relationship
of Old Testament to New Testament with respect to religious beliefs, events,
persons, or statements).
In linguistics, typology refers to the classification of the world’s lan-
guages according to similarities and differences in their linguistic struc-
tures and genetic relationships. Language typology, therefore, is essentially
comparative and cross-linguistic. That is, a typological analysis obligatorily
involves data from multiple languages, either of different language families
or of the same family, for comparison, and proposes generalizations on the
basis of the analysis.
Language typology has developed from language description, which has a
long history of work going back at least to the time of Pānini (fifth century
_
BC) and other Indian grammarians (such as Yāska, Kātyāyana, and Patañjali)
of ancient India and of which we have written record (see Shukla 2006;
Kiparsky 2009). Initially, the focus was on the description of just one
language and the primary purpose was teaching the language as well as
intellectual inquiry in the philosophy of language. Pānini’s Astādhyāyī, a
_ __
masterpiece of linguistic endeavour in the history of language study, was
motivated by both pedagogical purpose as well as his inquiry in the phil-
osophy of language. As the tradition of language study developed, however,
researchers began to ask more interesting questions in order to understand
language diversity, and genetic relationship among languages was a natural
extension of the intellectual inquiry into the philosophy of language that
led to comparisons of structures of different languages. This later came to
be known as language typology (see Section 29.3). Language typology was
thus introduced as a method to investigate genetic relationships among
Language Typology 769
languages and group them under different stocks on the basis of their
genetic origin and linguistic structures. Greenberg’s (1963) classification
of language stocks on the basis of the order of major constituents of the
clause, traditionally known as Subject, Object, and Verb, was based on this
principle. (With typological developments in SFL, SFL researchers have
observed that these traditional categories have inconsistencies at several
levels. SFL typologists therefore have begun to use the order of Subject,
Complement, and Predicator to categorize language (e.g. see Matthiessen
2004).2)
As can be expected, the findings of typological study help us to under-
stand the amazing cross-linguistic diversity of the world, and that can be
understood by no other means than describing, comparing, and analyzing
languages. For example, while the first-person pronouns in most, if not all,
languages of the world do not show gender and the first-person singular
pronouns generally lack the marking for honorification, we find interesting
variations in the second- and the third-person pronominal forms, as a
comparison of the pronominal systems of English (a Germanic language)
and Bajjika (an Eastern Indo-Aryan language (Kashyap 2014)) demonstrates,
even though both these languages have developed from the same root (i.e.
Indo-European). Bajjika has developed a four-level honorification system
(high-honorific, mid-honorific, plain honorific, and non-honorific) to refer
to addressees of different social statuses (Kashyap 2012; Kashyap and Yap
2017). English, in comparison, is limited to the use of only one second-
person pronoun you, regardless of the addressee’s social status.
This kind of linguistic diversity helps us to make sense of different
sociolinguistic situations in the world and often provides insights into the
range of linguistic and meaning-making systems in world languages. For
example, for a speaker of English, or even an expert of English with little
exposure to how different languages behave and how they develop struc-
tures to express meanings, it will be hard to make sense of the morpho-
logical complexity of the following example from Mundari (a Munda
language within the Austroasiatic family) given in (1) and whether it is a
word or a sentence.
(1) Omamtanain
om-am-tan-a-in
give-2sg-prs-1sg
‘I give you.’
Mundari (Verma 1991:130)
2
The traditional categories Subject, Object, and Verb are reinterpreted and renamed in some new twentieth-century
theories of linguistics. In SFL, while the category Subject retains its name, although with a fresh interpretation, Object is
reanalyzed as Complement, and Verb (at the clause rank) is reinterpreted as Predicator (see Halliday 1994; Halliday
and Matthiessen 2014; Matthiessen et al. 2010). In SFL there is no such term as Object. The category ‘verb’ refers to a
unit in word classes; a corresponding unit above word class is ‘verbal group’, which has a verb as the Head, e.g. walk in
is walking and target in has been targeted.
Even more difficult would be making sense of why in the Bajjika con-
struction in (2) the addressee (shown by bold) is encoded in the verbal form
if the addressee is not a functional constituent or referent in the clause:
(2) ek din ham dekh-l-i-aw je ...

one day 1.nom see-pst-1.nom-2hnnom that
‘One day I saw that . . .’
Bajjika (Kashyap 2012:1869)
The hierarchical social order of the Bajjika community demands that the
speaker acknowledges the presence of the addressee even if the addressee is
not a functional constituent in the clause, and therefore the language has
developed a system in the agreement paradigm (for details, see Kashyap
2012; and Kashyap and Yap 2017).
Example (3) demonstrates how languages develop ways to manifest dif-
ferent sociocultural phenomena and how a good understanding of linguis-
tic diversity helps to grapple with the linguistic and semantic variations.
This example, from Maithili, shows the complexity of agreement marking.
In this example, three referents are marked in its paradigm of verbal
agreement.
(3) ham to-rā kaniyā-kẽ dekh-au-l-i-au-nh.3

1.nom 2h/nh-dat bride-dat see-caus-pst-1.nom-2nh/h.
nnom-3h.nnom
‘I showed you the bride.’
Maithili (Bickel et al. 1999:482)
A comparison of agreement paradigms in Indo-Aryan languages shows a

great deal of diversity and demonstrates how languages of the same family
show variations in their linguistic structures and the semantic domain that
they relate to. For example, Maithili verbal agreement system allows three
discourse referents to be simultaneously encoded, Bajjika allows two referents
(example (2)), and Bhojpuri and Hindi allow only one referent, as shown in (4).
Even though these four languages belong to the Indo-Aryan language family,
the structural and semantic variations among them are remarkable.
(4) a. bajār jā-it bā-ni.

market go-prog aux.prs-1
‘(I) am going to the market.’ Bhojpuri (Kashyap 2017)
b. mai-ne us-e nahĩ dekh-ā

1-erg 2nh–acc neg see-1.erg
‘I did not see him.’ Hindi
3
The interlinear glosses in this example contain the modification by the author on the basis of his knowledge of Maithili.
As more and more languages are described and cross-linguistic data are
compared and typologized, we encounter unexpected linguistic facts that
demonstrate that some ‘exotic’ languages have unique ways of encoding
the same information. In several cases, we also find surprising similarities
that stimulate us to ask whether they had any historical relationship. The
kind of addressee-oriented agreement markers shown in Indo-Aryan Bajjika
and Maithili above have also been found in Basque languages (see example (12)),
although no genetic relationship has yet been established between the Indo-
Aryan and the Basque languages. Rather, Basque is classified as a language
isolate, which means that as yet we have no clue to its origin.
Language typology has a great deal of similarities with two related lin-
guistic sub-disciplines, contrastive linguistics and comparative historical
linguistics, although they each have a unique focus of attention. While
language typology often focuses on many languages, and preferably larger
language samples on which firmer generalizations can be made and that
can enrich our theoretical understanding of linguistic diversity and lan-
guage evolution, contrastive studies have traditionally focused on the study
of two languages. Moreover, while comparison of linguistic structures and
related phenomena is at the core of both language typology and contrastive
linguistics, they have a different preference of inclination: in typological
explorations equal weight is given to both similarities and differences,
while the primary focus of contrastive linguistics is on differences of
linguistic structures, and linguistic similarities take a back seat.
Whereas language typology seeks to explore cross-linguistic variations
and chart the language diversity that exists across languages, language
contrast, which is sometimes considered a branch of language typology,
aims to contribute to applied areas of linguistics such as language teach-
ing, error analysis, and translation. Language contrast developed for peda-
gogical reasons, to improve foreign language teaching, from Charles Fries’
(1945) contention that in foreign language teaching ‘the most effective
materials are those that are based upon a scientific description of the
language to be learned, carefully compared with a parallel description of
the native language of the learner’ (Fries 1945:9). His colleague Robert
Lado later advanced Fries’ suggestion and proposed the ‘Contrastive
Hypothesis’ (Lado 1957), which suggested that ‘the second-language
learner’s language was shaped solely by transfer from the native lan-
guage’ (Tarone 2006:134). Lado emphasized that the comparison of the
learner’s target language with the learner’s first language would accur-
ately predict the learner’s difficulty of learning a second/foreign language,
offer better insights for learnability, and help to improve language teach-
ing. It is worthwhile noting that the typological anchoring of SFL also
developed through the pedagogical concerns of Michael Halliday (e.g. see
Halliday 2007), and contrastive analysis was an area of interest particu-
larly in the regions where English had the status of a second or foreign
language (e.g. see Prakasam 1970).
Comparative historical linguistics, in contrast, seeks to investigate gen-

etic relationships among languages that help to discover the history of
languages and their respective language communities – their origin, dia-
chronic development, and structural change over the course of time. The
results of both the approaches help each other to develop a wider perspec-
tive and a firmer understanding of language diversity and language evolu-
tion. Historically, language typology evolved with the goals that today
influence both historical and typological linguistics, that is, to discover,
respectively, the language and corresponding speech community’s develop-
ment along time and cross-linguistic variations.
Even though the aims and scope of the field of language typology and what
typologists do are generally clear, there seem to prevail some ambiguities about
what is and what is not language typology; this calls for clarification, at least for
beginners in the field, e.g. students beginning to develop skills of language
study. I foreground here two basic ones. Firstly, language typology has not yet
developed into a fully-fledged discipline like sociology, anthropology, or any
other discipline for that matter; it is a sub-discipline within the discipline of
‘linguistics’. As a piece of evidence in support of typology as a sub-discipline,
one can see how universities treat this area of study: it is rare to see universities
with an entire department dedicated to language typology in the same way that
one can find independent departments of linguistics, sociology, and anthropol-
ogy, each of which is a discipline in its own right. Language typology makes its
presence in a department of linguistics, and within the department of linguis-
tics is sometimes found a centre dedicated to language typology, if the depart-
ment intends to focus on this area of study. Language typology, therefore,
should not be confused as a discipline; it is a branch of general linguistics and
has grown to be a sub-discipline within general linguistics.
Secondly, language typology is not a linguistic theory or framework; rather,
it is a method of linguistic analysis, which necessarily involves cross-linguistic
comparison of linguistic items, meanings, and related phenomena. Typo-
logical investigations can be carried out within any theoretical framework,
e.g. SFL, West-Coast functionalism, or generative theory (see Section 29.4).4
Collections of papers and typological generalizations proposed in Caffarel
et al.’s Language Typology: A Functional Perspective (2004), for instance, adopted
SFL as the theoretical framework. SFL is a theory of linguistics,5 which has
4
Although some works claim to be theory-free, ‘there is no such thing as atheoretical description’ (Dryer 2006:207; also
see Matthiessen and Nesbitt 1996) and the same is true with respect to typology, given that description is a
prerequisite of typological explorations.
5
Some scholars confusingly suggest that SFL is a sub-discipline of linguistics, which is clearly inaccurate. SFL is not a sub-
discipline; it is a theory of language, a comprehensive method of language study, or a ‘metalanguage’ (Matthiessen
2007), which is applied to study various subjects within a range of linguistic sub-disciplines, including (but not limited
to) language description, language typology, discourse analysis (e.g. Butt et al. 2004), language teaching/learning (e.g.
Rose and Martin 2012; Dreyfus et al. 2016; Hood 2016), World Englishes (e.g. Halliday 2003; Kashyap 2014),
stylistics (e.g. Halliday 1971), and verbal art (e.g. Hasan 1989; Miller and Turci 2007; Butt 2009). See Mwinlaaru and
Xuan (2016) for a detailed review and references of language description using SFL as the theoretical framework.
been applied to various types of linguistic investigations, e.g. language descrip-

tion and discourse analysis; language typology is just another area of study to
which the theory is applied. Likewise, numerous other typological explor-
ations have been conducted within other theoretical frameworks.
29.3 History of Language Typology
Language typology has a long history of investigations, long before theories

like SFL or other theories discussed in Section 29.4 emerged. SFL and other
theories have advanced this field by adding new insights to earlier achieve-
ments in the field. Here I briefly sketch how language typology has
developed to its present state.
Experts are of different opinions about when precisely this field of study
appeared. Shibatani and Bynon (1995:1) accept it is difficult ‘to ascertain the
first formulations of a research programme of language typology’ and
suggest that ‘the underlying assumptions that run throughout the history
of language typology can be gleaned from the older passages of the
nineteenth-century writings’. Greenberg (1974:13) reports that the first
known occurrence of this word in respect to linguistics was in 1928 in the
Prague linguists’ research theses, while Graffi (2010) considers Georg von
der Gabelentz (1840–1893), a German linguist, as ‘the originator of typology
of today’ because Gabelentz was the one who ‘coined the term “typology” to
refer to a branch of linguistics’ (Graffi 2010:25). Despite the array of uncer-
tainty and disagreement, Friedrich von Schlegel (1772–1829) is ‘tradition-
ally credited with the first use of the term ‘vergleichende Grammatik’’
which means ‘comparative grammar’ (Koerner 2006:18).
Published literature in the field, however, clearly reveals that typological
research had begun much earlier. Although Friedrich von Schlegel is often
credited for introducing typology as a method of investigation, as reflected
in the quotes from Koerner above, the seed of typological investigation was
planted well before Schlegel, in Sir William Jones’ (1746–1794) works on
the grammar of Sanskrit and other Indo-Iranian languages that ‘formed the
basis of scientific historical/comparative linguistics’ (Pierce 2006:134).
Jones’ famous statement in his lecture to the Asiatic Society in Calcutta6
in February 1786 deserves full quotation here:
The Sanscrit [Sanskrit] language, whatever be its antiquity, is of a wonderful

structure; more perfect than the Greek, more copious than the Latin; and
more exquisitely refined than either, yet bearing to both of them a stronger
affinity, both in the roots of verbs and the forms of grammar, than could
possibly have been produced by accident; so strong indeed, that no
6
Sir William Jones was the founder of this society. The society later was renamed as the Asiatic Society of Bengal and was
a primary forum for scholarly debate and publication.
philologer could examine them all three, without believing them to have
sprung from some common source, which, perhaps, no longer exists; there is
a similar reason, though not quite so forcible, for supposing that both the
Gothick [Gothic] and the Celtick [Celtic], though blended with a very differ-
ent idiom, had the same origin with the Sanscrit, and the old Persian might
be added to the same family, if this were the place for discussing any
question concerning the antiquities of Persia.
(Jones 1798:422–3; quoted in Teignmouth 1804:388; Pachori 1993:175)
This brilliant thesis of Jones is considered ‘a single event in the history of

comparative linguistics’ and ‘the first known printed statement of the
fundamental postulate of the Indo-European comparative grammar’
(Cannon 1990:246). It not only advanced the theory of the genetic affinity
of Sanskrit, the language that remained as the centrepiece of European
intellectual inquiry for a considerable period and a ‘norm for generations of
[European] comparativists’ (Rocher 2006:748), with a common source to
which Latin, Greek, and the Iranian languages belonged, but it also laid the
foundation of modern comparative linguistics and the typological classifi-
cation of the common source that is today known as Indo-European.
Exposure to William Jones’ works on Sanskrit and other languages of the
Indo-Iranian family guided Schlegel for his future linguistic works. Later,
Schlegel’s suggestion ‘to compare grammatical features in order to estab-
lish genetic relationships’ was significant for the development of historical
and typological linguistics in the nineteenth century (Koerner 2006:18; also
see Jankowsky 2006). Friedrich von Schlegel’s (1801) Über die Sprache und
Weisheit der Indier became a notable milestone in comparative and historical
linguistics, in which he discussed the languages of India and described
Sanskrit as ‘the actual source of all languages, of all thoughts and poetry
of the human spirit’ (quoted in Rocher 2006:748). Schlegel was ‘among the
first linguists to propose a typological framework on the basis of morpho-
logical characteristics’ (Shibatani and Bynon 1995:1). His linguistic classifica-
tion on the basis of linguistic structure initiated a novel research programme
which was further advanced by the linguistic investigations of his successors,
such as Franz Bopp and Jacob Grimm, by the inclusion of new insights.
The effect of William Jones’ work on Indo-Iranian languages was more
clearly visible a century later, in the last quarter of the nineteenth century,
when descriptive and typological studies began with a renewed vigour under
the leadership of Sir George A. Grierson (1851–1941). Grierson, an adminis-
trator in the Indian Civil Service by profession but a prodigious linguist by
disposition, began to study the languages and dialects of Bihar (where he was
deployed as a district magistrate) and analyzed the verbal forms of the Bihari
dialects with his typological observations with respect to the languages of
their neighbourhood. His detailed description of the Bihari dialects and the
comparison of their structures was published in his Seven Grammars of the
Dialects and Subdialects of the Bihárí Language (Grierson 1883–1887).
Later the British government of India began the survey of Indian ver-
naculars under his leadership, which culminated in the publication of the
monumental multi-volume Linguistic Survey of India (Grierson 1898–1928),

for which Grierson is best known. The purposes of the linguistic survey
were administrative, as they were used to develop training programmes for
British administrators and were ‘intended to enable [them] to communicate
directly with the local populace’ (Wright 2006:160). The project, however,
was designed ‘with a view to revealing the richness of modern India’s
vernacular cultures to a West hitherto mesmerized by India’s past’ (Wright
2006:160). The Linguistic Survey of India, as a result, was saturated with
descriptive, ethnographic, and typological radiations: it clearly described
the genetic affiliation and typological classification of the Indian languages,
and the description and comparison of their linguistic structures. Grier-
son’s works on the languages of India were essentially typological, although
he did not propose generalizations of the kind presented by Greenberg and
other Universalists. His primary focus was on the documentation and
description of the languages, their genetic classification, and the compari-
son of their morphological structures and sound patterns. Grierson’s works
are not mentioned in descriptive and typological literature as often as they
should be, and Grierson is not given credit as eloquently as he deserves, but
those who are familiar with the history of descriptive and comparative
linguistics know that Grierson’s contributions on the languages of India
were significant milestones in the field.
Throughout the nineteenth century, leading up to ‘the development of
fresh approaches to typological problems by structuralists in the second
quarter of the twentieth century, the scene was dominated by what is
here called morphological typology, the classic expression of which was
the threefold division of languages into isolating, agglutinative, and
inflective’ (Greenberg 1974:35) that emerged from the work of the Schle-
gel brothers.7 The morphological typological method proposed by Schle-
gel was further taken by the generations of Wilhelm von Humboldt,
Heymann Steinthel, Sapir, and Greenberg. The second half of the twenti-
eth century witnessed a shift of focus from structure to meaning with the
emergence of new approaches to language study, e.g. in Michael Halli-
day’s works (e.g. Halliday 1967–1968; also see Daneš 1974). Halliday’s
primary focus was on meaning and understanding what contributions
the social and anthropological traits make in understanding meaning
construed by linguistic forms.
29.4 Approaches to Language Typology
29.4.1 Formal vs. Functional

We have discussed above that the linguistic approaches of early years of
language study were essentially descriptive and that they led to the
7
Friedrich von Schlegel’s elder brother, August von Schlegel, was an equally influential typologist.
development of what is the field of ‘language typology’ today. The field was
greatly advanced by the works of Sir William Jones, Friedrich von Schlegel,
Franz Bopp, Georg von der Gabelentz, Sir George Grierson, and many other
philologists in the eighteenth and nineteenth centuries, and Franz Boas,
Sapir, Bloomfield, and other linguists in the early years of the twentieth
century, all under the rubric of general linguistics. Linguistic structure took
precedence over meaning in these approaches, except in new approaches
such as SFL that are primarily concerned with meaning and the analysis of
morphological and syntactic forms as a means of understanding how
meaning is made in natural languages, as we will see below.
In the second half of the twentieth century, we can trace the emergence
of two significant lines of approaches to language typology – functional and
formal. The functional typology included several distinct theoretical frame-
works and was in a sense led by Greenberg’s functional typology, until the
emergence of new approaches such as SFL (Section 29.4.2), West-Coast
Functional, and Dik’s Functional Grammar (Butler 2003).
Each of the functional and formal approaches emerged on the scene as a
reaction to different approaches. Generative grammar emerged as a reac-
tion against the behaviourist psychologists’ anti-universalist view that dis-
approved of the existence of innate and universal mental ability for
language learning and postulated that ‘linguistic competence is acquired
through learning of stimulus-response pattern’ (Croft 2017). The generative
linguists (or cognitive linguists, to put it more appropriately) led by
Chomsky, interestingly, assumed all languages to be ‘English-like but with
different sound systems and vocabularies’ (Evans and Levinson 2009:429),
and proposed the well-known Universal Grammar, which lacked the empir-
ical base of the kind of Greenbergian or Hallidayan functionalism.
Greenberg’s functional typology emerged in response to anthropological
relativism, which postulated that languages of the world vary arbitrarily, as
we see in this famous quote of Martin Joos: ‘Languages could differ from
each other without limit and in unpredictable ways’ (Joos 1957:96). Green-
berg, as we see in his research and subsequent publications (e.g. Greenberg
1978), advocated for more systematic sampling of a greater number of
languages, which ‘reveals not only range of variation but constraints on
that variation’. He strongly believed that those linguistic constraints would
‘demonstrate that languages do not vary infinitely, and the constraints
represent language universals’ (Croft 2017). Thus, although the approaches
of both Chomsky’s generative theory and Greenberg’s functional typology
had a fundamental theoretical juxtaposition, they laid a great emphasis on
language universals for different reasons.
While cognitive linguists of the generative tradition and Greenbergian
typologists emphasized language universals, later research by functional
typologists painted a strikingly different picture. Languages of the world
‘differ so fundamentally from one another at every level of description
(sound, grammar, lexicon, meaning) that it is very hard to find any single
structural property they share’ (Evans and Levinson 2009:429). Michael

Halliday held this view much earlier (as discussed below in this section).
There was a striking distinction between the generative and Greenberg’s
typologies in terms of empiricism and data sampling. Generative linguists
preferred in-depth study of one or two languages, while Greenberg advo-
cated for substantial language samples. His generalizations of language
universals presented in his seminal paper (Greenberg 1963) were based on
samples of thirty languages. An inherent problem with Greenberg’s
approach, however, was that the findings came from the analysis of indi-
vidual sentences examined in isolation. As a result, his universals, e.g. word
order classification, faced challenges from discourse-based studies. For
instance, (Modern Standard) Hindi, which is usually considered ‘an SOV
(Subject-Object-Verb) language’ (Shukla 2009:497) following Greenberg’s
classification, often violates the imposition of a rigid SOV word order rule
in natural text, and therefore linguists such as Michael Shapiro conclude
that ‘[t]here are no hard and fast rules governing the order of constituents
in sentences [of Hindi] as a whole’ (Shapiro 2003:271), as in the example (5)
taken from a spoken text. The Subject in this example is pushed towards
the end of the clause, culminating in an OVS order. In natural text of many
Indo-Aryan languages, the Subject does not appear until a new Subject is
introduced and when it does, there is no guarantee that it will appear at the
clause-initial position, often leading to an (S)OV order.
(5) assī hazār jīt ga-ye āp

eighty thousand victory asp-pfv.3h 3h
Traditional label Object Verb Subject
SFL label Complement Predicator Subject
‘You won eighty thousand (rupees).’
Systemic linguists, instead, consider the switch of constituents in a sen-

tence as being motivated by a range of semantic factors, including textual
ones, where each sequence of constituents construes special meanings
(Matthiessen 1995; Martin 1995). And what is considered to be a universal
category in Greenbergian typology is not necessarily universal across lan-
guages. According to Martin’s (2004a) analysis, for instance, Tagalog does
not have Subject. What is often treated as Subject is Theme in Martin’s
analysis (see Martin 2004b:284–95).
The typological research up to the 1970s was pervaded by language
classification on the basis of morphological features, beginning with Fried-
rich von Schlegel’s two-fold classification of affixing and inflectional lan-
guages, which was expanded by his brother August von Schlegel’s
formulation of ‘agglutinative’ to a three-fold classification (isolating,
inflective, and agglutinative languages), up to Greenberg’s typology. Mor-
phological classification was still a key component in Greenberg’s typology
(e.g. whether a language had preposition or postposition), but syntax was
equally significant for his typological classifications (for example, word
order universals). The typological research from the early 1980s, however,
shifted the focus from language evolution and genetic classification to
language diversity, and typological researchers began to focus more on
how different meanings are expressed in different languages, thus guiding
the focus on a combination of factors, e.g. morphosyntax and lexicogram-
mar, as well as the semantic and pragmatic inferences drawn from lexico-
grammatical patterns in language.
In the last forty years, language typology has seen tremendous advance-
ment, especially from functional perspectives, with the emergence of dis-
tinct functional theorists who conduct research in distinct ways and as an
inevitable consequence have opened up numerous theoretical frontiers.
The functional theories of linguistics that emerged in the second half of
the past century and have significantly contributed to typological research
include Systemic Functional Linguistics (SFL), also referred to as Systemic
Function Grammar (SFG), first developed by Michael Halliday (e.g. Halliday
1994; Halliday and Matthiessen 2014) and further expanded by other sys-
temic functional linguists (on the development of SFL, see Matthiessen
2005), West-Coast functionalism (e.g. Givón 1995), Simon Dik’s Functional
Grammar (e.g. Dik 1997a, 1997b), and Role Reference Grammar (Van Valin
and LaPolla 1997; Van Valin 2005). As a description and comparison of
these theoretical approaches is not within the scope of this chapter,
I narrow down my focus to SFL’s contribution to language typology to be
consistent with the aims and scope of this handbook. An appreciative and
comprehensive comparison of the functional theories can be found in
Butler (2003). For a snapshot of key functional theories readers are advised
to consult Butler (2005, 2006).
29.4.2 SFL Approach to Language Typology

While generative and other functional linguists focused on language uni-
versals, Michael Halliday envisaged a ‘need for a general linguistic theory of
description, as opposed to a universal scheme of descriptive categories’
(Halliday 2002:22), of the kind that emerged in the 1960s in the wake of
the famous Conference on Language Universals held at Dobbs Ferry (New
York) in 1963. Greenberg’s descriptive and typological generalizations are
classic examples of this. While Greenberg spearheaded the ‘universal
scheme’ of descriptive typological research, Halliday dedicated himself to
developing a general theory of language, treating theory as a problem-
solving enterprise that ‘can be brought to bear on everyday activities and
tasks’ (Halliday 2006:19) and as ‘appliable’ to various related areas, such as
language teaching and discourse analysis (Section 29.5). Halliday’s vision
led to the evolution of SFL through the works of Halliday himself in the
1960s onward and its further expansion through the works of other SFL
linguists (see Matthiessen 2005, 2007; Schleppegrell 2012; Bateman 2017).
Applications of SFL to language typology to explain language diversity and
variation have been consistent with Halliday’s vision of the ‘appliability’ of

a general theory of language.
The SFL approach to language study puts great emphasis on communi-
cation as the primary function of language and postulates that linguistic
structures are shaped by the sociocultural norms and needs of the respect-
ive speech community (see Halliday 1978; Halliday and Hasan 1985). This
view is very clearly expounded in the following quote from Halliday (1994:
xiii):
Language has evolved to satisfy human needs; and the way it is organized is
functional with respect to these needs.
Social and cultural factors, therefore, play significant roles in explaining

linguistic phenomena from SFL perspectives.
Halliday was of the opinion that language data should be kept ‘as close as
possible to real language, spoken or written’ (Halliday 1977:38), rather than
constructed sentences based on the linguist’s intuition. Consequently, SFL
linguists have relied primarily on language samples extracted from natural
spoken and/or written texts; often, typological generalizations are based on
the analyses of larger text samples (e.g. Matthiessen and Kashyap 2014;
Kashyap and Matthiessen 2017). The analyses presented in Caffarel et al.’s
Language Typology (2004), for example, are all based on the analysis of
natural texts from the respective languages. In many studies, larger
extracts are described with a view to explain how language works in real-
life contexts (e.g. Teruya et al. 2007).
Clearly, contrary to the formalists’ focus on sentence grammar, SFL
approaches typology from a discourse perspective: descriptive and typo-
logical studies from the framework of systemic grammar are text-based,
in which descriptions of grammatical structures are illuminated by refer-
ence to anthropological traits of the respective speech community (e.g.
Kashyap 2012; Kashyap and Yap 2017), seen through the contextual param-
eters of field, tenor, and mode (e.g. Matthiessen and Kashyap 2014; Kashyap
and Matthiessen 2017; see Bowcher, this volume).
As noted above, language in the SFL framework is understood as a
meaning-making resource, and each utterance is interpreted with reference
to context. For example, the Hindi example in (6b) will make no sense if it is
seen in isolation; but if it is read in context and taken as an answer to the
question in (6a), the utterance will make perfect sense.
(6) a. kitāb kahā̃ hai ?

book where cop(be).prs
‘Where is the book?’
b. sofā par.
sofa loc
‘On the sofa.’ Hindi
While other approaches to typology have continued their typological

investigations from the perspective of syntax, morphology, and/or phon-
ology, retaining these traditional categories, and thus showing their lineage
to traditional grammar,8 SFL typologists primarily concentrate on explor-
ing metafunctional diversity in language as their guiding principle, with a
fresh look at the descriptive categories and their interpretations (for
example, see footnote 1 and example (5) above). The descriptions of eight
languages (French, German, Japanese, Tagalog, Chinese, Vietnamese,
Telugu, and Pitjantjatjara) in Caffarel et al. (2004), for example, are based
on the descriptive schema of metafunctional diversity outlined in the first
chapter. The first chapter introduces systemic functional typology and
presents the key dimensions of SFL theory, with an emphasis on the meta-
functional organization of lexicogrammar. Each language covered in the
book is described in terms of its metafunctional schema, e.g. how the
language organizes the three modes of meaning (experiential, interper-
sonal, and textual) and how the meanings are realized by linguistic means
such as grammatical items, prosody, or a combination of the two. At the
end, based on the description of the eight languages, Matthiessen (2004)
presents a typological survey, using ‘the theoretical dimension of metafunc-
tion to map out lexicogrammatical systems’ (Matthiessen 2004:537).
The typological comparison can focus on the key aspects of a metafunc-
tion, as in Matthiessen (2004), or on a particular type of meaning, e.g. the
ways of encoding projection (e.g. quoting and reporting) in the six different
languages (Arabic, Hindi, Dagaare, Spanish, Japanese, and English) in Arús-
Hita et al. (2018). The focus is on meaning, and linguistic components such
as morphology and phonology are considered as a means of expressing
different kinds of meaning, e.g. how the logico-semantic relation of projec-
tion is achieved in these languages and how the status of quoting and
reporting is signalled by linguistic means.
The primary concern of SFL typologists is exploring how a particular
phenomenon is realized or how the particular meaning is expressed, and
grammatical items below the clause rank, e.g. specific affixes or groups/
phrases, are described and explained in the process of exploring the par-
ticular meaning. Teruya et al. (2007) in their typology of mood, for instance,
explore what linguistic diversity can be understood with respect to the
language of negotiation, with a special focus on the commodity and nature
of exchange in actual dialogue. Languages develop ways to express the
social need of exchanging information and goods and services, grammati-
calized across languages in the system of mood, e.g. interrogative
(demanding information) and declarative (providing information), and each
language is unique with respect to its ways of expressing specific meanings.
8
Greenberg’s (1963) classic paper on word order universals in language ‘contrasts sharply with most other work at the
time in assuming a set of descriptive notions that are to a large extent simply those of traditional grammar’ (Dryer
2006: 210).
Teruya et al. (2007) analyze six languages (Òkó, Spanish, French, Danish,
Thai, and Japanese), and find that all those languages ‘operate with the
prosodic mode of expression in the realization of options in the interper-
sonal system of mood, but the languages vary with respect to whether
these prosodies are manifested sequentially, segmentally or intonationally’
(Teruya et al. 2007:914). In English (a Germanic language), for example, the
polar interrogative and declarative moods are distinguished by the
‘the relative sequence of Subject and Finite’ (Teruya et al. 2007), as in the
examples of English in (7) and (8).
(7) a. Are you Alice? (Finite: are ^ Subject: you – interrogative:

demanding information)
b. Yes, I’m Alice. (Subject: I ^ Finite: ‘m – declarative: providing

information)
In Thai, in comparison, the polar interrogative and declarative contrast is

realized segmentally in terms of availability or absence of a Negotiator
(realized by interrogative particles), as in (8), and in Spanish this
contrast is realized intonationally by rising and falling intonation, respect-
ively, as in (9).
(8) a. indicative: polar interrogative

khun1 khǝ:j1 paj1 chiaə1maj2 maj5 khrap4
you asp:pfv go Chiangmai polar.nego polite.
nego
‘Have you ever been to Chiengmai?’
Thai (Teruya et al. 2007:901)
b. indicative: declarative
thǝ:1 maj3 daj3 ju:2 kruə1thep3:
she neg asp:pfv live Bangkok
‘She does not live in Bangkok.’
Thai (Teruya et al. 2007:901)
(9) a. indicative: polar interrogative
tú te acuerd- as de aquel señor?
2-sg 2.ref.sg remember 2.sg.prs.ind. of that man
‘Do you remember that man?’
Spanish (Teruya et al. 2007:892)
b. indicative: declarative
María se acord- ó de nosotros
Mary 3.ref. sg remember 3.sg.pst.ind of us
‘Mary remembered us.’
Spanish (Teruya et al. 2007:893)
The typological comparison thus reveals that languages differ in terms of

whether lexicogrammatical properties are deployed for the realization of
mood, as in English, or prosodic properties such as intonation are respon-
sible, as in Spanish, the Indo-Aryan languages Bajjika (Kashyap 2014) and
Hindi (Kachru 2006), and the Niger-Congo languages Òkó (Akerejola 2006)
and Dagaare (Mwinlaaru 2017). Note, however, that even in the languages
that have developed lexicogrammatical systems for distinguishing declara-
tive and interrogative moods, the system of intonation is an indispensable
property of the mood system. In such languages, the instances of (polar)
interrogative mood realized by virtue of rising intonation are not uncom-
mon, as in the example (10) from English:
(10) They used to have gas stoves, kerosene stoves, before?
In this example, the configuration of the utterance is identical with that

of the declarative mood (Subject: They ^ Finite/Predicator: used), but the
rising intonation in which the utterance was spoken (which is represented
by the question mark (?) in the written mode) realizes that it is asking for
information and is in the interrogative mood.
29.5 Application of Language Typology
The success of a theory or field of study lies in how it is used to solve

practical problems of life – its appliability (to use Halliday’s term). Language
typology is greatly successful in this respect, as its application helps us to
grapple with practical problems of day-to-day life, especially in the spheres
of human life that demand a grasp of the world’s multilingual and multi-
cultural diversity (Matthiessen et al. 2008). Closely related areas such as
translation, cross-cultural and intercultural communication, language
learning and teaching, and the documentation and description of languages
call for a sound understanding of multilingualism. Not surprisingly, typo-
logical insights contribute immensely to these fields of study (see Filipovicˊ
2017). Translation, for instance, by its very nature is bilingual, and an
understanding of meaning-making pathways in different languages, espe-
cially in the source and target languages, helps; and an understanding of
how specific linguistic features contribute to meaning-making is further
advantageous (see Filipovicˊ 2008; Matthiessen 2001; also see Steiner and
Yallop 2001).
It is well established that in the course of translation a lot of meaning
from the source text is lost. Slobin, for instance, in his research of motion
verbs in a range of languages has shown that in translated text much of the
meaning is lost, especially when translating a text from a language that
uses a wide range of verbs encoding delicate details about the manner of
motion (see, for example, Slobin’s analysis of translated texts from the
languages that are typologically characterized as satellite-framed (e.g. Eng-

lish) into verb-framed languages (e.g. Spanish), following Talmy’s (1985)
typology of motion verbs). Speakers of satellite-framed languages are
‘trained, by their language, to make more distinctions of motor pattern,
rate, effect, and evaluation of movement, in comparison with speakers of V-
languages [verb-framed languages]’ (Slobin 2000:113). Studies on verbs of
motion in translated texts have shown that the patterns of use in language
are ‘sensitive to the structural typological differences in the encoding of
motion events’ (Chen and Guo 2009:1753), and language typology plays a
significant role in developing a sense of structural and semantic differences
between those typologically divergent languages.
Typologically distinct languages vary remarkably in terms of
what semantic component they encode and how they encode; and exotic
languages in this respect vary unexpectedly. For example, many speech
communities are interpersonally elevated in that interpersonal role-
relationships take precedence over what is known in SFL as ‘ideational’
meaning, and the interactants’ social status and mutual relationship are
crucial for making sense of how the language is used in social life (see
Hasan 1984). Consequently, such languages have developed linguistic
systems which prioritize interpersonal meaning, which are difficult to
translate (compared to ideational meaning), and which pose difficulties
for inter- and cross-cultural communications. Typological insights in such
situations can help ‘explain why certain lexical and grammatical features
are harder to translate than others’ (Filipovicˊ 2017:4). For even an experi-
enced professional translator, for example, it is extremely difficult to trans-
late the meaning expressed by allocutive agreement suffixes in Bajjika (11)
or Basque (12), and therefore SFL’s perspective on understanding the envir-
onments of translation with respect to the three strands of meaning in text
(ideational, interpersonal, and textual) is crucial.
(11) sārhe tin baje pahũc-t-aw tren mujappharpur me.

_
half three time reach-fut- alloc:2h train Muzaffarpur loc
‘The train will reach Muzaffarpur at 3.30pm.’
Bajjika (Kashyap and Yap 2017:435)
(12) a. bilbo-ra n-oa-k.
Bilbo-all 1.s-go-alloc:m
‘I am going to Bilbao.’ [male addressee]
b. bilbo-ra n-oa-n.
Bilbo-all 1.s-go-alloc:f
‘I am going to Bilbao.’ [female addressee]
Standard Basque (Antonov 2015:57)
In examples (11) and (12), we can see that the addressee is not a constituent
in these clauses, and yet these examples host suffixes for a second-person
referent to reflect the presence of the addressee. The English translation
here shows the ideational meaning, but the interpersonal meaning encoded
by the allocutive suffixes for reference to the addressee’s social status is
missing because these suffixes have no role in the experiential segment of
the clause. In Bajjika, however, the meaning implied here is purely inter-
personal and indicates that the speaker is mindful of the presence and
social status of the addressee. Furthermore, such unfamiliar structural
and semantic differences also have relevance for inter- and cross-cultural
communications.
In the current scenario of globalization, when people more frequently
travel across territorial borders for various reasons such as business, immi-
gration, education, tourism, and diplomacy, language typology has a bigger
role to strengthen one’s intercultural and cross-cultural communication
skills. Communication styles significantly vary across cultures, and how
successfully one communicates largely depends on one’s social, linguistic,
and cultural backgrounds and training. For example, it is quite common to
express gratitude and apology by the use of, for instance, please and sorry, in
English culture, while Indian speech communities do not favour overt
verbalization of gratitude (Apte 1974) because most languages spoken in
India have developed verbal systems that register different strands of
politeness including gratitude and that are integrated into the speakers’
communicative styles (see Hasan’s 1984 work on Urdu). Moreover, in most
Indian communities, requests and expressions of gratitude are overtly
verbalized for strangers or those who have a weak level of solidarity with
the speaker but rarely among family members, close relatives, and friends.
In today’s multicultural globalized societies where colleagues and co-
workers of different linguistic and cultural backgrounds interact, the socio-
linguistic knowledge gained from language typology can help to empower
people with a better communication style and develop sensitivity toward
one’s colleagues and co-workers’ languages and cultures, and can facilitate
the creation of materials for training programmes in this area.
Researchers in language learning/teaching and acquisition have long
been interested in typological research, with a firm belief that a grasp of
linguistic diversity and how languages vary with respect to linguistic
structure and related meaning will help us to better understand the
cognitive processes of language learning by children as well as adults.
The evolution of contrastive linguistics was propelled by the impact of
typological differences in language learning, as we noted earlier. And the
famous Sapir-Whorf hypothesis reflecting this typology-acquisition/
learning duet is well known; according to the Sapir-Whorf hypothesis,
the structure of a language influences the cognitive processes of the
speaker of that language (see Whorf 1956; Wardaugh 1970:123). Cognitive
scientists have noted that typologically different languages significantly
vary with respect to which linguistic items children acquire first and
which items they acquire later (Berman and Slobin 1994; Slobin and
Bowerman 2007; Berman 2014).
Unarguably, typology has helped us to identify the linguistic items that

are easy and difficult to learn across languages and those whose acquisition/
learning vary significantly. Cross-linguistic comparison of children’s lan-
guage has allowed us to understand what is universal across languages and
what is particular to specific languages (see Slobin 1982). The understand-
ing of language structure provided by descriptive and comparative research
in linguistics has greatly contributed to the understanding of linguistic
principles of language learning by children and answers (or has the poten-
tial to answer) relevant questions in this respect, for example, why children
acquire certain linguistic items earlier and why they acquire other linguis-
tic items later.
Finally, as for the link between language typology and language description,
the relationship is complementary. Each field has benefited from the other. As
Mithun (2016:472) notes, ‘Documentation, description, and typology are sym-
biotic: each can provide tools important to progress in the others.’ Language
documentation and description have enriched our understanding of typo-
logical regularities and patterns of variation across languages. Nevertheless,
insights from typological research have been applied to the descriptions of
new languages and have greatly contributed to the field. The linguistic cat-
egories or patterns observed in other languages have helped us to identify
them in new exotic languages more quickly, and the understanding that a
particular linguistic feature is cross-linguistically rare ensures that those
features are comprehensively documented and remain prominent in descrip-
tions rather than simply logging typological checklists (Mithun 2016:467), e.g.
the identification of the allocutive agreement suffixes that are so uncommon,
found only in a handful of languages of the world, such as in some Bihari
languages and Basque. In fact, these markers had remained a mystery for a
long time, and linguists fumbled to describe them until it was recognized that
the allocutive agreement markers in Bajjika, a Bihari language, are similar to
those found in Basque (see Kashyap and Yap 2017).
29.6 Conclusion
The objective of this chapter was to present a snapshot of language

typology, and I have outlined here how this field has developed over
centuries of research and how it has contributed to our firmer under-
standing of cross-linguistic diversity in the world. This field has passed
through different developmental phases that have had a distinct
research focus; unarguably, language typology has always remained at
the epicentre of intellectual inquiry in language, even before the term
‘typology’ was coined. The field began as a method of linguistic inquiry
to explore genetic relationships among languages and cross-linguistic
diversity. The primary aim of the field remains intact – to chart linguis-
tic diversity and explore patterns of structural and semantic variations
in languages through cross-linguistic comparisons. While initial phases

of this field focused on structural forms of language, with the develop-
ment of the field of linguistics and the emergence of new approaches,
how different languages construe meaning has become significant. In
SFL, in particular, meaning lies at the centre of linguistics investigation,
and typological study is essential to developing a typology of meaning.
In this regard, the contributions of social and anthropological factors
are of immense significance.
In recent years linguistic insights gained from language typology have
stimulated several related domains that call for a sound understanding of
language diversity, e.g. translation, intercultural and cross-cultural com-
munications, language teaching/learning, and the description and docu-
mentation of languages. Other work focuses on the appliability of
research and how, for example, this research can help to address problems
of our social life, such as what it can offer in terms of developing training
programmes across a range of contexts.
Abbreviations: 1/2/3:First/second/third person; acc: accusative; all: alla-

tive; alloc: allocutive; asp: aspect marker; aux: auxiliary; caus: causa-
tive; cop: copula; dat: dative; erg: ergative; f: feminine; fut: future; h:
honorific; ind: indicative; loc: locative; m: masculine; neg: negative;
nego: negotiator; nh: non-honorific; nnom: non-nominative; nom: nom-
inative; pfv: perfective; prog: progressive; prs: present; pst: past; ref:
reflexive; sg: singular.
References
Akerejola, E. 2006. A Text-based Lexicogrammatical Study of Òkó: A Systemic

Functional Approach. PhD Thesis, Macquarie University.
Antonov, A. 2015. Verbal Allocutivity in a Crosslinguistic Perspective. Lin-
guistic Typology 19(1): 55–85.
Apte, M. A. 1974. ‘Thank You’ and South Asian Languages: A Comparative
Sociolinguistic Study. Linguistics 12: 67–89.
Arús-Hita, J., K. Teruya, M. A. Bardi, A. K. Kashyap, and I. N. Mwinlaaru.
2018. Quoting and Reporting across Languages: A Systemic Functional
Approach. Word 64(2): 69–102.
Bateman, J. A. 2017. The Place of Systemic Functional Linguistics as a
Linguistic Theory in the Twenty-First Century. In T. Bartlett and G.
O’Grady, eds., The Routledge Handbook of Systemic Functional Linguistics.
Berman, R. A. 2014. Cross-linguistic Comparisons in Child Language
Research. Journal of Child Language 41(S1): 26–37.
Berman, R. A. and D. I. Slobin. 1994. Relating Events in Narrative:
A Crosslinguistic Developmental Study. Hillsdale: Erlbaum.
Bickel, B., W. Bisang, and Y. P. Yādva. 1999. Face vs. Empathy: The Social
Foundation of Maithili Verb Agreement. Linguistics 37(3): 481–518.
Bowerman, M. 2010. Linguistic Typology and First Language Acquisition. In
J. J. Song, ed., The Oxford Handbook of Linguistic Typology. Oxford: Oxford
Functional Theories, Part 1: Approaches to the Simplex Clause. Amsterdam:
John Benjamins.
Butler, C. S. 2005. Functional Approaches to Language. In C. S. Butler, M. de
los Ángeles Gómez-González, and S. M. Doval-Suárez, eds., The Dynamics of
Language Use: Functional and Contrastive Perspectives. Amsterdam: John Ben-
jamins. 3–17.
Butler, C. S. 2006. Functionalist Theories of Language. In K. Brown, ed.,
Encyclopedia of Language and Linguistics. 2nd ed. Amsterdam: Elsevier.
696–704.
Butt, D. G. 2009. Technique and Empire in the Poetry of Ee Tiang Hong.
World Englishes 28(3): 365–93.
Butt, D. G., A. Lukin, and C. M. I. M. Matthiessen. 2004. Grammar: The First
Covert Operation of War. Discourse and Society 15(2–3): 267–90.
Caffarel, A., J. R. Martin, and C. M. I. M. Matthiessen, eds. 2004. Language
Typology: A Functional Perspective. Amsterdam: John Benjamins.
Cannon, G. 1990. The Life and Mind of Oriental Jones: Sir William Jones, the Father
of Modern Linguistics. Cambridge: Cambridge University Press.
Chen, L. and J. Guo. 2009. Motion Events in Chinese Novels: Evidence for an
Equipollently-framed Language. Journal of Pragmatics 41(9): 1749–66.
Collins English Dictionary. 2017. Collins English Dictionary. New York: Har-
perCollins Publishers. Available online at: www.collinsdictionary.com/
dictionary/english. (Last accessed 25/09/2017.)
Croft, W. 2017. Typology and Universals. In M. Aronoff and J. Rees-Miller,
eds., The Handbook of Linguistics. 2nd ed. Oxford: Wiley Blackwell. 39–56.
Daneš, F., ed. 1974. Papers on Functional Sentence Perspective. The Hague:
Academia.
Dik, S. C. 1997a. The Theory of Functional Grammar, Part 1: The Structure of the
Clause. 2nd ed. Edited by K. Hengeveld. Berlin: Mouton de Gruyter.
Dik, S. C. 1997b. The Theory of Functional Grammar, Part 2: Complex and Derived
Constructions. Edited by K. Hengeveld. Berlin: Mouton de Gruyter.
Dreyfus, S., S. Humphrey, A. Mahboob, and J. R. Martin. 2016. Genre Peda-
gogy in Higher Education: The SLATE Project. Basingstoke: Palgrave
Macmillan.
Dryer, M. 2006. Descriptive Theories, Explanatory Theories, and Basic Lin-
guistic Theory. In F. Ameka, A. Dench, and N. Evans, eds., Catching
Language: Issues in Grammar Writing. Berlin: Mouton de Gruyter. 207–34.
Evans, N. and S. C. Levinson. 2009. The Myth of Language Universals:
Language Diversity and Its Importance for Cognitive Science. Behavioural
and Brain Sciences 32: 429–92.
Filipovicˊ , L. 2008. Typology in Action: Applying Typological Insights in the

Study of Translation. International Journal of Applied Linguistics 18(1): 23–40.
Filipovicˊ , L. 2017. Applied Language Typology: Applying Typological
Insights in Professional Practice. Languages in Contrast 17(2): 255–78.
Fries, C. C. 1945. Teaching and Learning English as a Foreign Language. Ann
Arbor: University of Michigan Press.
Givón, T. 1995. Functionalism and Grammar. Amsterdam: John Benjamins.
Graffi, G. 2010. The Pioneers of Linguistic Typology: From Gabelentz to
Greenberg. In J. J. Song, ed., The Oxford Handbook of Linguistic Typology.
Oxford: Oxford University Press. 25–42.
Greenberg, J. 1963. Some Universals of Grammar with Particular Reference
to the Order of Meaningful Elements. In J. Greenberg, ed., Universals of
Language. Cambridge, MA: MIT Press. 58–90.
Greenberg, J. 1974. Language Typology: A Historical and Analytical Overview. The
Hague: Mouton.
Greenberg, J. H., ed. 1978. Universals of Human Language. 4 Volumes. Palo
Alto: Stanford University Press.
Grierson, G. A. 1883–1887. Seven Grammars of the Dialects and Sub-dialects of
the Bihárí Language. Parts I–VIII. Delhi: Bharatiya Publishing House.
Grierson, G. A. 1898–1928. Linguistic Survey of India. 11 volumes. Calcutta:
Office of the Superintendent of Government Printing.
Halliday, M. A. K. 1971. Linguistic Function and Literary Style: An Inquiry
into the Language of William Golding’s The Inheritors. In S. Chatman, ed.,
Literary Style: A Symposium. London: Oxford University Press. 330–65.
Halliday, M. A. K. 1977. Ideas about Language: Aims and Perspectives in Linguis-
tics. Canberra: Applied Linguistics Association of Australia.
Arnold.
Halliday, M. A. K. 2002. Some Aspects of Systematic Description and
Comparison in Grammatical Analysis. In J. J. Webster, ed., The Collected
Works of M. A. K. Halliday, Volume 1: On Grammar. London: Continuum.
21–36.
Halliday, M. A. K. 2003. Written Language, Standard Language, Global
Language. World Englishes 22(4): 405–13.
Halliday, M. A. K. 2006. Some Theoretical Considerations Underlying the
Teaching of English in China. The Journal of English Studies (Sichuan Inter-
national Studies University) 4: 7–20.
Halliday, M. A. K. 2007. General Linguistics and Its Application to Language
Teaching. In J. J. Webster, ed., The Collected Works of M. A. K. Halliday,
Volume 9: Language and Education. London: Continuum. 135–73.
Halliday, M. A. K. and R. Hasan. 1985. Language, Context and Text:
sity Press.

Hasan, R. 1984. Ways of Saying, Ways of Meaning. In R. P. Fawcett, M. A. K.
Halliday, S. M. Lamb, and A. Makkai, eds., The Semiotics of Culture and
Language. London: Frances Pinter. 105–62.
Hasan, R. 1989. Language, Linguistics, and Verbal Arts. Oxford: Oxford Univer-
sity Press.
Hood, S. 2016. Systemic Functional Linguistics and EAP. In K. Hyland and P.
Shaw, eds., The Routledge Handbook of English for Academic Purposes. London:
Hornby, A. S. 2005. Oxford Advanced Learner’s Dictionary. 7th ed. Oxford:
Jankowsky, K. R. 2006. Early Historical and Comparative Studies. In K.
Brown, ed., Encyclopedia of Language and Linguistics. 2nd ed. Amsterdam:
Elsevier. 43–9.
Jones, W., Sir. 1798. The Third Anniversary Discourse, Delivered 2nd Feb-
ruary, 1786: On the Hindus. Asiatick Researches 1: 415–31.
Joos, M., ed. 1957. Readings in Linguistics I. Chicago: University of Chicago
Press.
Kachru, Y. 2006. Hindi. Amsterdam: John Benjamins.
Kashyap, A. K. 2012. The Pragmatic Principles of Agreement in Bajjika
Verbs. Journal of Pragmatics 44(13): 1868–87.
Kashyap, A. K. 2014. Developments in the Linguistic Description of
Indian English: State of the Art. Linguistics and the Human Sciences 9(3):
249–75.
Kashyap, A. K. 2017. Lexicogrammar of the Bihari languages. In S. K. Singh,
A. K. Kashyap, B. War, S. A. Lyngdoh, and B. Khyriem, eds., Fabric of Indian
Linguistics. New Delhi: Lakshi Publishers. 25–52.
Kashyap, A. K. and C. M. I. M. Matthiessen. 2017. FIGURE and GROUND in
the Construal of Motion: A Registerial Perspective. Word 63(1): 62–91.
Kashyap, A. K. and F. H. Yap. 2017. Epistemicity, Social Identity and Polite-
ness Marking: A Pragmatic Analysis of Bajjika Verbal Inflections. Linguis-
tics 55(3): 413–50.
Kiparsky, P. 2009. On the Architecture of Pānini’s Grammar. In G. Huet, A.
_
Kulkarni, and P. Scharf, eds., Sanskrit Computational Linguistics: Lecture Notes
in Computer Science. Berlin and Heidelberg: Springer. 33–94.
Koerner, E. F. K. 2006. Schlegel, Friedrich von (1772–1829). In K. Brown, ed.,
Encyclopedia of Language and Linguistics. 2nd ed. Amsterdam: Elsevier.
Lado, R. 1957. Linguistics across Cultures: Applied Linguistics for Language
Teachers. Ann Arbor: University of Michigan Press.
Martin, J. R. 1995. More Than What the Message Is About. In M. Ghadessy,
ed., Thematic Development in English Texts. London: Pinter. 223–58.
Martin, J. R. 2004a. Metafunctional Profile of the Grammar of Tagalog. In
A. Caffarel, J. R. Martin, and C. M. I. M. Matthiessen, eds., Language
Typology: A Functional Perspective. Amsterdam: John Benjamins. 255–304.
Martin, J. R. 2004b. Mourning: How We Get Aligned. Discourse and Society

15(2–3): 321–44.
Matthiessen, C. M. I. M. 1995. Theme as an Enabling Resource in Ideational
‘Knowledge’ Construction. In M. Ghadessy, ed., Thematic Development in
English Texts. London: Pinter. 20–54.
Matthiessen, C. M. I. M. 2001. The Environments of Translation. In
E. Steiner and C. Yallop, eds., Exploring Translation and Multimodal Text
Matthiessen, C. M. I. M. 2004. Descriptive Motifs and Generalizations.
In A. Caffarel, J. R. Martin, and C. M. I. M. Matthiessen, eds. 2004.
Language Typology: A Functional Perspective. Amsterdam: John Benjamins.
537–673.
Matthiessen, C. M. I. M. 2005. Lexicogrammar in Systemic Functional
Linguistics: Descriptive and Theoretical Developments in the ‘IFG’ Trad-
ition since the 1970s. In R. Hasan, C. Matthiessen, and J.J. Webster, eds.,
Continuing Discourse on Language, Volume 2. Sheffield: Equinox. 765–858.
Matthiessen, C. M. I. M. 2007. The ‘Architecture’ of Language according to
Systemic Functional Theory: Developments since the 1970. In R. Hasan,
C. M. I. M. Matthiessen, and J. J. Webster, eds., Continuing Discourse on
Language, Volume 2. Sheffield: Equinox. 505–61.
Matthiessen, C. M. I. M. and A. K. Kashyap. 2014. The Construal of Space in
Different Registers: An Exploratory Study. Language Sciences 45: 1–27.
Matthiessen, C. M. I. M. and C. Nesbitt. 1996. On the Idea of Theory-neutral
Descriptions. In R. Hasan, C. Cloran, and D. G. Butt, eds., Functional
Descriptions: Theory in Practice. Amsterdam: John Benjamins. 39–83.
Matthiessen, C. M. I. M., K. Teruya, and C. Wu. 2008. Multilingual Studies
as a Multi-Dimensional Space of Interconnected Language Studies. In
J. J. Webster, ed., Meaning in Context. London: Continuum. 146–221.
Matthiessen, C. M. I. M., K. Teruya, and M. Lamb. 2010. Key Terms in Systemic
Functional Linguistics. London: Continuum.
Miller, D. R. and M. Turci, eds. 2007. Language and Verbal Art Revisited:
Linguistic Approaches to the Study of Literature. Sheffield: Equinox.
Mithun, M. 2016. Typology, Documentation, Description, and Typology.
Linguistic Typology 20(3): 467–72.
Mwinlaaru, I. N. 2017. A Systemic Functional Description of the Grammar of
Dagaare. PhD Thesis, The Hong Kong Polytechnic University.
Mwinlaaru, I. N. and W. W. Xuan. 2016. A Survey of Studies in Systemic
Functional Language Description and Typology. Functional Linguistics 3(8):
1–41.
Nichols, J. 2007. What, if Anything, Is Typology? Linguistic Typology 11(1):
231–38.
Pachori, S. A., ed. 1993. Sir William Jones: A Reader. Delhi: Oxford University
Press.
Pierce, M. 2006. Jones, William, Sir (1746–1794). In K. Brown, ed., Encyclo-
pedia of Language and Linguistics. 2nd ed. Amsterdam: Elsevier. 134–5.
Prakasam, V. 1970. The Syntactic Patterns of Telugu and English: A Study in

Contrastive Analysis. Hyderabad: Central Institute of English.
Rocher, R. 2006. Sanskrit: Discovery by Europeans. In K. Brown, ed., Encyclo-
pedia of Language and Linguistics. 2nd ed. Amsterdam: Elsevier. 746–9.
Rose, D. and J. R. Martin. 2012. Learning to Write, Reading to Learn: Genre,
Knowledge and Pedagogy in the Sydney School. Sheffield: Equinox.
Schleppegrell, M. J. 2012. Systemic Functional Linguistics. In J. P. Gee and
M. Handford, eds., The Routledge Handbook of Discourse Analysis. London:
Routledge. 21–34.
Shapiro, M. C. 2003. Hindi. In G. Cardona and D. Jain, eds., Indo-Aryan
Languages. London: Routledge. 250–85.
Shibatani, M. and T. Bynon. 1995. Approaches to Language Typology. Oxford:
Shukla, S. 2006. Pānini. In K. Brown, ed., Encyclopedia of Language and
_
Linguistics. 2nd ed. Amsterdam: Elsevier. 153–6.
Shukla, S. 2009. Hindi. In K. Brown and S. Ogilvie, eds., Concise Encyclopedia
of Languages of the World. Amsterdam: Elsevier. 494–7.
Slobin, D. I. 1982. Universal and Particular in the Acquisition of Language.
In E. Wanner and L. R. Gleitman, eds., Language Acquisition: The State of the
Art. Cambridge: Cambridge University Press. 128–72.
Slobin, D. I. 2000. Verbalized Events: A Dynamic Approach to Linguistic
Relativity and Determinism. In S. Niemeier and R. Dirven, eds., Evidence
for Linguistic Relativity. Amsterdam: John Benjamins. 107–38.
Slobin, D. I. and M. Bowerman. 2007. Interfaces between Linguistic Typ-
ology and Child Language Research. Linguistic Typology 11(1): 213–26.
Steiner, E. and C. Yallop. 2001. Exploring Translation and Multimodal Text
Production: Beyond Content. Berlin: Mouton de Gruyter.
Talmy, L. 1985. Lexicalization Patterns: Semantic Structure in Lexical
Forms. In T. Shopen, ed., Language Typology and Syntactic Description,
Volume 3: Grammatical Categories and the Lexicon. New York: Cambridge
Tarone, E. 2006. Interlanguage. In M. Berns, ed. Concise Encyclopedia of
Applied Linguistics. Amsterdam: Elsevier. 134–9.
Teignmouth, L. 1804. Memoirs of the Life, Writings, and Correspondence of Sir
William Jones. London: John Hachard.
Teruya, K., E. Akerejola, T. H. Andersen, A. Caffarel, J. Lavid, C. M. I. M.
Matthiessen, U. H. Petersen, P. Patpong, and F. Smedegaard. 2007. Typ-
ology of Mood: A Text-based and System-based Functional View. In
859–920.
Van Valin R. D., Jr. 2005. Exploring the Syntax–Semantics Interface. Cambridge:
Van Valin, R. D., Jr. and R. J. LaPolla. 1997. Syntax: Structure, Meaning and
Function. Cambridge: Cambridge University Press.
Verma, M. K. 1991. Exploring the Parameters of Agreement: The Case of

Magahi. Language Sciences 13(2): 125–43.
Wardaugh, R. 1970. The Contrastive Analysis Hypothesis. TESOL Quarterly
4(2): 123–30.
Webster, J. J., ed. 2008. Meaning in Context. London: Continuum.
Lee Whorf. Edited by J. B. Carrol. Cambridge, MA: MIT Press.
Wright, J. C. 2006. Grierson, George Abraham, Sir (1851–1941). In
K. Brown, ed., Encyclopedia of Language and Linguistics. 2nd ed. Amsterdam:
Elsevier. 160.
Index
abstract vs concrete 59, 98–9, 273–4, 289, 322, amplification 301, 340, 348, 350 (see also message
452, 528, 562–3, 565, 568, 572, 578, 623, semantics)
628, 631, 639–40, 645 anaphora, anaphoric reference (see under
abstraction 3, 14, 22, 26, 30, 46, 70, 77–80, 85–7, cohesion)
152, 156, 162, 205, 358, 384, 437–8, 441, ancillary 294, 754, 759 (see also Mode (contextual))
459, 488, 548, 565, 570–1, 575, 621, 636, annotation 347, 564, 568–9, 577–80, 758 (see
671, 741, 762 also corpora/corpus, automated tagging)
academic discourse 390, 397 aphasia 6, 589, 598–9, 601, 603–4
accent 644 appliable linguistics 1, 51–2, 92, 160, 276, 306–7,
and juncture 180 (see also juncture) 333, 657, 709
as in pronunciation 185 Application Programming Interfaces (APIs) 729
as in phonology appraisal and educational discourse 396–8
word accent 182 appraisal and legal discourse 399, 471, 480
acquired brain injury (ABI) 601 appraisal and media discourse 398–9
action literacy 542–3 (see also literacy) appraisal and multimodal discourse 400–1
action research 301–2, 537–8, 545, 551 appraisal (system of ) 4–5, 81, 301, 334, 360,
Action, Reflection, Contact (ARC) 152, 476, 494, 372, 375–6, 378, 382–4, 401, 592, 600, 660
503, 513–14, 543 (see also graduation , judgement ,
Activation (in context) 148–9, 151, 164, 221, 578 attitude , affect , appreciation ,
active voice 104, 337, 429 engagement )
activity sequence 216, 359 appreciation 360–1, 384–5, 388, 397 (see also
Actor 39, 94–6, 98–9, 337, 415, 465–6, 567 (see appraisal (system of ))
also lexicogrammar) apprize 343, 345–6, 349, 504–5, 658 (see also
addressee (Addressee) 41, 218, 235, 267, 344, semantic networks)
507, 570–2, 769–70 (see also Performer) Arabic 545, 564, 624, 626, 628
adjective group 130, 138, 246–7 architecture (of language) 3, 35, 60, 86, 151, 173,
Adjunct, adjunction 40, 42, 110, 113–15, 193, 260, 267, 274–5, 291, 302, 383, 549, 621,
240, 242, 244, 251, 253, 346, 349 623, 633, 638, 643–4, 740
circumstantial Adjunct 42 artificial intelligence 27, 562, 565, 570, 579, 643
comment Adjunct 378, 383 Asperger Syndrome 593
conjunctive adjunct 115, 314, 707 assumptive 344–7, 658 (see also semantic
Confirmation Seeker 238 networks)
adverbial group 122, 127, 131, 138, 246–7, 383 articulate, articulation, articulator 493–4, 622
Ælfric’s Homilies 413–14 (see also symbolic articulation, double
affect 39, 360, 378, 384, 386–7, 704, 717, 728, articulation)
754–5, 759 (see also appraisal (system of )) attention deficit hyperactivity disorder (ADHD)
Affected (as in lexicogrammar) 241, 250–1 589, 596 (see also neurodevelopmental
affiliation 5, 361, 399, 402–4, 655, 661, 725–7, disorders)
732 (see also ambient affiliation) attitude 5, 81, 301, 360–1, 384–6, 388–90,
affordance 361, 715, 728 392–4, 396–8, 401, 403, 545, 551, 726 (see
Agent 39, 241, 250–1, 498, 573 also appraisal (system of ))
Alzheimer’s Disease 6, 589, 598 audience 388, 402, 512, 528–9, 727
ambient affiliation 399, 726 (see also affiliation) autism spectrum disorder (ASD) 592
ambiguity 2, 160, 214, 601, 668 automated tagging 569
794 INDEX
autonomous wheelchair 572 (see also dialogue clause complex 59–60, 73, 75, 78, 82–3, 267,
systems) 301, 314, 321–2, 360–1, 365–6, 393, 415,
auxiliary verbs 40, 107, 110, 126–7, 135, 244, 445
252–3 CLIL 529
clinical linguistics 6, 277, 587–90, 608, 610
Bajjika 769 clinician–patient interaction 654 (see also
Basque 641, 771 doctor–patient talk)
belief system 234–5, 514 closed class (items) 315
Bhojpuri 770 COBUILD corpus 653
Bihar 774, 785 code 45, 463–4, 633, 656, 675 (see also
boundary marker 94, 175–6 Legitimation Code Theory (LCT))
elaborated 294, 463–4, 498
Cardiff model of language, Cardiff Grammar 4, restricted 294, 463
29, 58, 230–55, 260–1, 266, 268, 270, 275, cognitive development 497, 526
474 cognitive models 266, 278–9
Carrier 102, 250–1, 287, 445, 745 Cognitive Linguistics 232, 476–7
casual conversation 149, 180, 298, 305–6, 383, cohesion 4, 41, 60, 80, 186, 288–91, 300, 311–28,
403 (see also pragmatic conversation) 334, 358, 360–1, 377, 379, 417, 467, 525–6,
category (grammatical, linguistic) 13, 46, 78, 591–3, 595, 599, 654, 656, 746
102–3, 119, 206, 210, 231, 243, 266, 268, anaphora, anaphoric reference 314, 318–19,
279, 294, 301, 316–18, 320–1, 336, 384–7, 359, 361, 364, 378
422, 474, 521, 630, 756 cataphoric reference 319, 361
Channel 630, 638 (see under Mode) cohesive conjunction 4, 80, 83, 115, 311–28,
Chaucer 414–15, 623, 631 334, 358–60, 370–1, 378, 457
Child language 4–6, 21, 52, 63, 92, 286, 293, 396, cohesive devices 19, 291, 312–18, 323, 326–7,
489–94, 502, 504, 651 (see also 358, 599
protolanguage) cohesive relations 60
Childhood-onset fluency disorder 606 (see collocation 17, 20, 145, 275, 317–18, 360, 591
stuttering) comparative reference 315, 361, 365, 370
children’s story 179–80, 188, 190 ellipsis 114, 288, 312–13, 359–60, 366–7, 591,
Chinese 22, 29, 49, 232, 399, 402, 564, 653, 668 593, 599, 655
Choice 7, 17, 19, 23, 35–6, 44, 92, 116, 148–9, endophoric reference 314–15, 361, 593, 598,
151, 156, 158, 160, 163, 173, 177–83, 185, 755
187, 189–90, 195, 197, 220, 224, 238, exophoric reference 314–15, 361, 591, 593,
273–5, 279, 340, 343, 346, 348, 350–1, 602
375, 403, 433–4, 436, 439–41, 462, 464, expectancy relations 317–18 (see also lexical
467, 473, 479–80, 520, 562, 566, 643–4, cohesion under cohesion)
671 demonstrative reference 315
Circumstance (as in grammatical unit) 38, 47, lexical chain 326–8, 360, 363, 605 (see also
93, 104, 113, 131–2, 240–1, 267, 360, cohesive chain)
367, 371, 435, 442, 445, 504, 516, 600, lexical cohesion 80, 288, 312–13, 316–18, 359,
745 363, 378, 525, 591 (see also lexical
circumstantial Adjunct (see under Adjunct) organization under cohesion)
Circumstantial Role (CR) 241, 251 lexical reference, lexical identity 312, 316
circumvenience 287, 296, 303, 305 lexical organization 312–13, 316–18 (see also
class 38, 121–3, 130, 243, 318, 471, 496, 623, lexical cohesion under cohesion)
630, 634 (see also rank scale) personal reference 315
classification 158, 211, 213, 250, 295, 312, 322, reference 314–15, 361, 593, 598 (see also
363, 564, 569, 578, 640, 768 anaphora, cataphoric reference, lexical
classification 340, 348, 350 (see also message reference, personal reference)
semantics) substitution 288, 312, 315, 360, 591
classroom 295–6, 348–9, 377, 392, 395, 515, 520, taxonomic relations 317–18, 363 (see also
668 lexical cohesion under cohesion)
clause 3, 39, 41, 59, 64, 73, 92–116, 118, 120, visual 451
127, 129–30, 137, 177, 188, 193, 236, cohesion in clinical situations (see also Alzheimer’s
241–5, 251–5, 267, 290, 359, 365, 378–9, Disease (AD), aphasia, Asperger Syndrome (AS),
384, 393, 415–17, 429, 435–6, 438, 445, autism spectrum disorder (ASD), schizophrenia,
474, 500, 552, 567–8, 574, 580, 591, 595–7, traumatic brain injury (TBI))
600, 603, 607, 623, 628, 630, 632, 637, 769 cohesive chain 318–21, 602, 605
(see also syntactic unit) cohesive harmony 4, 290–1, 317, 320–1, 334,
co-ordinated clause 187 602, 605, 660, 663 (see also traumatic brain
declarative clause 42, 107, 187, 253, 368, 595, injury (TBI) and cohesion in clinical situations)
780 colligation 17, 20
embedded clause 42, 707 collocation (see under cohesion)
major 106, 339 communicating body 26–7
minor 106, 339 communication 8, 16, 173, 265, 336, 348, 361,
projecting clause, projecting verb 64, 83, 399, 411, 423–4, 434, 436, 587–9, 591–4,
393–4, 435, 501, 506 596, 598, 601, 603, 605, 608, 610
Index 795
communication disability 587 Field, Tenor, Mode, first-order context, second-

communication linguistics 302–4, 307 order context, pre-text context, via-text context)
COMMUNAL Project 231, 233, 569 contextualization 18–19, 86–7, 294, 383, 570 (see
comparative reference (see under cohesion) also serial contextualization)
competence (as in TG grammar) 16, 27, 270, 595, contextualization system networks 156, 163
666 contextual configuration (CC) 148, 151, 155–6,
Complement 42, 114, 241–2, 244, 251, 253, 346, 163, 291, 303, 436–7, 743 (see also context
414 of situation)
complementarity 48, 55, 72, 75–6, 85–6, 384, continuation 340, 348, 350, 597–8
440, 480, 640, 785 (see also intersemiotic continuative 322, 501, 606 (see also message
complementarity) semantics)
complex (unit complexing) 59–60, 73, 75, 78, 83, contrast (explicit) 411–12
120–1, 241, 267, 301, 314, 321–2, 666, 675, contrast (implicit) 410–11
747 contrastive linguistics 750, 771, 784
componence 243 Contrastive Newness (CN) (see under New)
composition (in appreciation network) 388, co-ordinated clause (see under clause)
436 (see also appreciation) corpora/corpus 269, 328, 386, 399, 402, 411,
compositional layout 441–2, 446, 451, 458 (see 418, 425–8, 473, 478–80, 577, 627, 727
also multimodal/multimodality) corpus linguistics 29, 479–80 (see also UAM
compositional meaning 442, 479 (see also Corpus Tool and Systemic Coder, COBUILD
compositional layout) corpus)
compound tone (see under tone) corpus-based discourse analysis 386, 722
computational approaches 4, 231, 277, 561, 563, correlation (in context) 140, 152, 164, 577, 602
565, 579 (see also computational linguistics) co-text 40, 346, 361, 384, 392, 398
computational complexity 563, 566, 568 coupling 60, 62, 71, 399, 403–4, 725
computational linguistics 6, 29, 277, 561–4, covert categories 22, 274, 278
567–9, 575, 667 (see also computational Critical Discourse Analysis (CDA) 5, 402, 436,
approaches) 462, 467, 473–4, 477–8
conceptual metaphor theory 476 (see under critical linguistics 464, 466–7 (see also Critical
metaphor) Discourse Analysis (CDA))
concessive contractor 391–2, 395 (see also Critical Period Hypothesis 515 (see also second
appraisal) language acquisition (SLA), second language
confirm 343, 345, 347, 349–50, 479, 505 (see also development/learning/teaching)
semantic networks) cross-cultural communication 782, 784 (see also
Confirmation Seeker (see under Adjunct, see also intercultural communication)
Cardiff Grammar) cryptotypes 22, 274, 278
conflation 107, 250–1 culmination 185, 188
congruent/non-congruent 47–8, 64, 73, 323, 378, curriculum 349, 397, 520, 528, 538–41, 544–6,
417, 521, 628, 634, 747, 756 548–9, 551–2, 675
conjunction (cohesive) (see under cohesion) curriculum genre 542
conjunctive Adjunct (see under Adjunct)
conjunction group 128–9, 132 Dagaare 780
conjunctive relations 360, 365–7 Danish 781
connotative/connotation 62, 69, 81–2, 84, 160–2 Darwin (Charles) 47, 417, 626, 639, 641
(see also denotative/denotation) dead languages 410
consciousness thinking 21, 50, 627 De-Anglicization 11
consonant 25 de-automatization 47, 63 (see also foregrounding)
syllable-initial consonant 179 declarative clause (see under clause)
consonant cluster 179 default options 64, 81, 155
construal 149, 266, 279, 346, 363, 365, 401, 518 deicticity, deixis 361, 365, 426
(see also realization) specific deixis 365
in context 151, 164, 628, 630, 632–3, 652–6, 659, delicacy 38, 145, 158, 182, 268, 340, 388, 400,
661, 664, 666 579, 671
constructionist models 259, 266 demonstrative reference (see under cohesion)
consultation (medical) 653–6 (see medical denotative/denotation 69, 82, 160–2 (see also
consultation) connotative/connotation)
context 334, 336, 351, 379, 388, 400, 426 (see deontic 108, 423 (see modality)
also context of situation, genre (as extra- descriptive linguistics 288
linguistic stratum), open context, register) desiderative process 100, 102
context dependent/independent 25, 48, 153, developmental disorders (see autism spectrum
336 disorder)
context-metafunction hook-up hypothesis 85, 291 diachronic approaches 5, 291, 693
context-metafunction resonance hypothesis 291 dialect 45, 49, 146, 291, 302, 342, 641, 644, 721,
context of creation 693, 699, 707 754
context of culture 48, 160, 304–5, 434, 518 dialogue systems 570–2, 574 (see also
context of situation 13, 22–5, 35, 40–1, 48, 52, computational linguistics)
140, 143–9, 151–2, 154–5, 158, 160, 164, diatypic variety 147 (see also register)
172, 291, 293, 338, 468, 571, 671 (see also digital humanities 576
796 INDEX
dimensions 38, 58, 72, 83, 86–7, 174, 218, Ender (E) 253 (see also Starter (St))
267, 569, 592, 594, 623, 719, 722, 729, endophoric reference (see under cohesion)
741 engagement 301, 360, 373, 375, 383–4, 390,
discourse 205, 225, 285–307, 488, 500, 503, 505, 394–6, 398–9, 401, 755 (see also appraisal
507–8, 522, 526–7, 529, 537–50, 601, (system of ))
662–3 English for Specific Purposes 525
academic discourse 397 episode 438, 442, 447, 458, 660 (see also
as social practice 467 multimodal/multimodality)
discourse marker 128, 314 epistemic 108, 423, 600 (see modality)
discourse semantics 4, 78, 80, 162, 300–3, 334, ergative 39, 412, 429, 495, 641, 786
358–9, 362, 375, 378–9, 382, 384, 393, 402, Ethnicity 726
438, 442, 651, 660, 762 etiology 609
electronic discourse 718, 729 evolution 45–7, 339, 387, 411, 427, 622, 641,
Field of discourse 751 (see under Field) 771–2
HIV discourse 7, 656 (see also clinician–patient evolutionary systems 28
interaction) evolutionary thinking 50, 627
media discourse 383, 398, 651, 664–5, 667, evolutionary theory 47, 645
715, 717, 727 exchange 16, 20, 36, 40, 295, 360, 435, 572–3,
medical discourse 7, 663 754
metadiscourse 370–1, 376, 378, 663, 742 exchange structure 81, 297, 360, 594, 601, 727
Mode of discourse 468, 743 (see under Mode) Exegesis 760
multimodal discourse 5, 333, 400–1 existential process, existential clause 104, 435,
spoken and written modes of 597 704
Tenor of discourse (see under Tenor) exophoric reference 591, 593, 602 (see under
discourse analysis 4–5, 283, 285–7, 298, 333, cohesion)
348, 436, 608, 664, 773, 778 expansion 59, 75, 82–3, 318, 360, 401, 415,
discourse semantic systems 5, 358–9, 375–6, 439–40, 459
378–9, 384, 445 (see also discourse semantics experiential (meta)function 38–9, 82, 93–105,
under discourse) 108, 116, 240, 249, 267–8, 301, 340, 436,
discursive practice 467 (see also discourse) 745
display stratum 438 (see also multimodal/ exploring 158, 218 (see also field of activity)
multimodality) exponence 20, 244, 246, 254
distance 389–90 (see also graduation) , expounding 158–9, 218, 625, 664 (see also field of
doing 39, 95, 98–9, 154, 158, 218, 267, 435, 537 activity)
(see also field of activity) extra-linguistic environment 62, 140, 164
doctor–patient talk 348 (see also clinician–patient extravagance 71–2, 75–6
interaction) evaluation 81–2, 360, 373, 375, 378,
double articulation (as symbolic articulation) 47, 382–4, 386–8, 393–4, 396, 400–4, 592,
63, 690, 701 704
drift (in grammatical metaphor) 429, 522, 621
(see also grammatical metaphor) Facebook 7, 715, 729 (see also social networking
dynamic context 224–5 sites)
factorial explanation 373, 375
echolalic responses 594 Field 48, 85, 147, 150, 152, 154, 157–8, 161,
eco-social environment 62–3, 66, 85, 87 163–4, 186, 218, 221–2, 291, 303, 359, 361,
education 29, 537, 783 363, 365, 367–8, 379, 383, 392, 396, 402,
primary 6, 51, 544–7 404, 415, 426, 436–7, 457, 468–9, 475, 480,
secondary 6, 51, 547–8 518, 539 (see also context of situation)
tertiary 6, 397, 537, 550 field of activity 159, 223, 537 (see also registerial
educational context 163, 327, 396 cartography)
educational discourse 396 (see also appraisal and figure 82, 359–60, 365, 367, 391, 438, 442, 446,
educational discourse) 458 (see also multimodal/multimodality)
educational linguistics 52, 277, 383 filling 242–4, 246, 254
elaborated code (see under code) film 313–16, 320–1, 577, 754
element (of syntax) 82–3, 105, 111, 113, 123–7, Finite 40–1, 106–7, 113, 126–7, 289, 366–7,
130, 133–4, 176, 188, 238, 241–55, 313, 412–13, 418, 422, 629, 704, 781
319, 338, 347, 376, 413, 416, 441, 443, 445, First-order context 164 (see also context of
451, 458, 745 situation, second-order context)
element (of intonation) 181, 183, 185, 290 Firthian influence 11–30, 728
ellipsis 591, 593, 599, 655, 752–3 (see under Flickr 577
cohesion) Focus
embedded clause (see under clause) information focus (system of ) (see under
emergency medicine 655, 657 (see also medical information)
discourse) in graduation network 5, 389, 391 (see also
emoticon 719, 727 graduation)
enabling 158, 218, 565 (see also field of activity) foot (in system of intonation) 176–7, 185, 196,
end of life care 658 (see also clinician–patient 199, 379
interaction, medical discourse) foot composition 177–8, 182
Index 797
Ictus, silent ictus 175, 177, 182–3, 185 hashtag 399, 717 (see also metadata, social
ictus state 177–8 tagging)
Remiss 178, 182–3 health/healthcare 348, 397–8, 404, 651 (see also
tonic foot 183 medical discourse, sexual health discourse)
foregrounding 41, 64, 600, 691–2, 695–6, 706, hedging 412, 427
709 heteroglossia 394–5, 397 (see also monoglossia)
foreign language teaching 4, 771 (see also second high autonomous professional (HAP) 464 (see
language acquisition, second language also code, low autonomous professional (LAP))
development/learning/teaching) high-functioning autism (HFA) 593
formalist models 260, 262, 264–5, 691 (see also Hindi 777
Russian formalist approaches) historical linguistics 49, 410, 771–2, 774 (see also
fractality 75, 77 old English)
French 411, 421, 427–8, 526, 564, 653 history (language of ) 46, 372, 397–8, 402, 428, 772
functional models 35–52, 230–1, 259–60, 262, history texts 163, 428, 548
264–5, 268, 664, 780 HIV discourse 656 (see under discourse)
functional syntax 4, 230–55 homophora 361, 365
functional-cognitive space 4, 259 human language 12, 628
functionality 35, 493, 574 (see also systematicity) human–wheelchair interaction (see autonomous
wheelchair)
gender 21, 719, 722, 753, 769 humour 423, 656, 719 (see also patient-initiated
generic stage 23, 217, 750–1, 753 humour)
generic structure 134, 150, 216–17, 219, 224–5, hybridity 627
248, 373, 601 hybrid voices 304
generic structure potential (GSP) 154, 223, Hypertext 441, 443, 755, 760
291–3, 601
genetic counselling 663, 674 (see also Ictus, silent ictus (see under foot)
clinician–patient interaction, medical ideational 35, 38–9, 42, 49, 65, 82–3, 149, 187,
discourse) 225, 267, 274, 287, 291, 339, 359, 382–3,
genitive cluster 248–9 393, 403, 435, 443–6, 448–51, 454–5,
genre agnation 160, 163 457–8, 604, 725, 783
genre-based literacy programs 362, 375 (see also ideational function in clinical linguistics 517 (see
genre pedagogy) aphasia)
genre (as text type) 160, 215, 217–22, 265, 293, ideology 20, 50, 398–9, 760
298, 300–2, 375–7, 518, 539–40, 588, 675, identity (social) 6, 35, 146, 398, 402
752 identity chain 319, 364, 366, 717–20, 722, 724,
genre (as extra-linguistic stratum) 161–2, 728 (see also cohesive chain)
216–17, 221, 358, 379, 382, 433, 436–7 identification 301, 334, 339, 358–9, 361, 365,
genre pedagogy 375, 545 370, 378, 588 (see also discourse semantics)
gestalt theory 439 IELTS 550
gesture 400, 434, 452, 575, 725 image–text combination 441, 443 (see also
Given, Givenness 41–2, 111–12, 185, 188, 267, intersemiosis, intersemiotic complementarity,
415, 417 (see also New) multimodal/multimodality)
given information 41, 594 (see Given, Givenness) imperative 40–1, 60, 64, 105–7, 187, 326, 453–4
globalization 784 indeterminacy 45, 69, 160, 268
Goal 336–7, 429, 445, 552 indicative 106, 345–6, 745, 781, 786
grammatical 39, 95–7, 114, 546 individuation 21, 715, 721–2
communicative 186, 296 Indo-Aryan languages 770
goal orientation 157, 301, 751 Indo-Iranian languages 773
graduation 5, 301, 340, 360, 384, 386, 388–93, ineffable 13, 26, 625
399–401 (see also appraisal (system of )) infographic 433, 450, 459
grammaticalization 110, 272, 412, 425, 427 Information 177, 181, 187, 189, 297, 343, 361,
grammars (types of ) 358, 740 367, 378, 436, 442, 745, 780 (see also Given,
notional 207, 212 New)
formal 15, 212, 236, 242, 248 information distribution (system of ) 174,
functional 206, 208–9, 212, 230, 232–3, 237, 181, 189, 267, 745
242, 312, 322, 776 information flow 42, 234, 346, 361, 417, 436,
grammatical intricacy 596, 607 451, 453, 746
grammatical metaphor 622–3, 741 (see under information focus (system of ) 41, 171, 181,
metaphor) 185, 189, 196, 240–1, 438, 442
grammatical structure 60, 74, 79, 122–8, 242–3, information giver 236, 241, 252
274, 292, 320, 379, 630, 779 information group 177
grammatics 37, 546 information grouping 177
graphology/graphological systems 38, 67, 148, information prominence 177, 368, 445–6
162, 174, 302, 363, 384, 438, 701 information status 240, 425
Gricean maxims 16 information structure 41, 190, 193, 222, 279,
group 3, 174–7, 179, 183, 185, 187, 267, 359, 361, 415, 746
365, 390–1, 393, 414–15, 420, 428, 438, information unit 41, 111, 175–6, 181, 185,
454, 707, 780 (see also phrase, rank scale) 188, 193, 197
798 INDEX
information thinking 627 Japanese 232, 399, 401, 524, 653, 781
Instagram 7, 577, 723, 729 (see also social journalistic discourse 4, 348, 350–1
networking sites) judgement 360, 373, 385, 387, 397–8, 623, 633,
instance, instantiation 35, 66, 70–2, 145, 149–51, 704 (see also appraisal (system of ))
162, 177, 231–2, 268, 270, 287, 338, 384–5, juncture 180 (see also accent)
388, 390, 393, 402–3, 519–20, 644, 703,
721, 730 key 187, 194 (intonational system of, choices in)
instantiation effect 44 KPML 563–4 (see also natural language generation)
instantiation hierarchy 162
institutional linguistics 277, 652 L2 6, 512–28, 530 (see second language acquisition,
institutionalization 152 second language development/learning/
intelligent wheelchair 573 (see autonomous teaching)
wheelchair, dialogue systems) language-based knowledge construction 526
interpreting 7, 661–2, 742–3, 758–9 (see also language-based linguistics 3, 12
translation) language change 45–8, 271, 286, 569, 762
intensity 389, 391–2, 626 language learning 4, 52, 58, 396, 487, 509, 512,
intercultural communication 755, 762, 767 (see 515, 519, 521, 523–4, 530, 537, 538, 541,
also cross-cultural communication) 546, 550, 553, 784 (see also child language,
interpersonal analyses in clinical linguistics (see education, educational linguistics, ontogenesis,
aphasia, traumatic brain injury (TBI)) second language acquisition, second language
interpersonal (meaning, metafunction) 12, 20, development/learning/teaching)
40, 48, 77, 80–2, 104–10, 116, 186, 225, language typology 402, 776, 778, 782, 785 (see
240, 244, 267–8, 287, 291, 301, 349, 358, under typology)
360, 379, 382–4, 393–4, 397, 400–3, 412, langue 1–2, 44–5, 152, 641
427–8, 435–6, 442–4, 453–4, 458, 517, latent patterns/patterning 28
630, 662, 675, 716–17, 724 Legitimation Code Theory (LGT) 464
interpersonal interactions 600–1 (see Alzheimer’s levels of analysis 2, 12, 19, 24, 173, 205, 700
Disease) lexical density 190
interpersonal metaphor (see under metaphor) lexical cohesion 525, 591 (see under cohesion)
interpersonal prosody 393 lexical metaphor (see metaphor)
interpersonal resonance 403 lexicogrammar 37, 92–116, 162, 174–5, 181,
intersemiosis 433, 440, 447, 455–6, 663, 674 (see 230, 235–6, 273–4, 289, 303, 324, 333, 339,
also multimodal/multimodality, 341, 351, 382, 384, 393, 438, 487, 570, 580,
resemiotization, semiotic) 632, 637, 702, 744, 756, 780
intersemiotic complementarity 440, 444 (see linguistic criticism 694 (see also literary criticism,
also intersemiotic relations, multimodal/ stylistics)
multimodality, resemiotization, semiotic) literacy 349, 542, 657, 665, 675 (see also
intersemiotic complementarity framework (see recognition literacy, action literacy, reflection
intersemiotic complementarity) literacy, learning to read/write)
intersemiotic relations 439, 444, 446–7, 450, 452 literary language, literary text 49, 218, 692, 700
(see also intersemiotic complementarity, literary theory 621, 691
multimodal/multimodality, resemiotization, literature 7, 690, 692–5, 698 (see also poem, poets/
semiotic) poetry, verbal art)
intersemiotic downranking 449 living language 410–11
intersemiotic package 450 logical metafunction 82, 122–5, 127, 133–4, 136,
interstratal relationship 148, 156, 220 (see also 149, 187, 267, 301, 340
intrastratal relationship) logical metaphor (see under metaphor)
intersubjectivity 27, 145, 490, 724 logogenesis 23, 152, 286, 709
interventionist medical linguistics 653 (see also London School 14, 145
medical discourse) low autonomous professional (LAP) 294–5, 494
intonation (in clinical linguistics) (see also (see also code, high autonomous professional
Asperger Syndrome, high-functioning autism (HAP))
(HFA))
intonation (features of, theory of ) 3, 25, 40–1, macrofunction 494–5 (see also microfunction, child
253, 297, 360, 438, 593–5 (see also tone, language)
tonality, tonicity) macrophenomenon 101–2
intonation (and literacy, learning to read) 655 Maithili 770
intonation systems 178–9 (see also intonational markedness 41–2, 44, 104, 368, 429, 595, 762
systems, tone, tonality, tonicity) marked Theme (see under THEME)
intonation unit 190 (see also tone group, tone) Marxism 49–51, 475
intonational systems 173, 177 (see also intonation material situational setting (MSS) 22, 152, 295
systems) mathematical discourse 425, 547
intrastratal relationship 148 (see also interstratal mathetic function 495 (see also macrofunction,
relationship) microfunction)
item (word, morpheme, or punctuation) 127, matrix 742
144, 242, 244–5, 290, 311, 313, 317–18, experiential matrix 17, 26
591, 605, 623, 772, 784–5 matrix RU 294
iterative contextualization 739–40 maxims (Gricean) (see Gricean maxims)
Index 799
meaning (definition of in SFL) 56–8, 143, 173, Mood 187, 267, 289–90, 567, 780–1 (see also
207–9, 232, 236–8, 240, 244, 254–5, 266, mood (system of ), lexicogrammar)
303, 384, 390, 520–1, 595, 601, 604 Declarative Mood 40–1, 105–7, 267, 289, 781
meaning potential 35–7, 43–5, 58, 68–70, 85, Imperative Mood 105–6, 267, 289, 453–4
148–9, 162, 242, 274, 289, 292, 333, 338, Interrogative Mood 105–7, 267, 289, 781
384, 398, 400, 434, 458, 621, 623, 639, 654, mood (system of ) 38, 40, 60, 64, 66, 80, 105–8,
693 149, 240–1, 255, 347 (see also Mood,
media discourse (see under discourse) (see also lexicogrammar)
radio broadcasts/commentary) mother–child interactions 4, 6, 80, 293, 334,
medical consultation 654 (see also clinician–patient 338, 341–8, 396, 544 (see also child
interaction, doctor–patient talk) language, primary carer)
medical discourse (see under discourse) mother tongue 487, 507
medical emergency teams 666 (see also surgical motion verbs 782–3
teams, medical discourse) move (in discourse) 81, 153, 160, 290, 296–9,
medical interpreting 661 360, 575
melody (in speech) 175–6 multifunctional 3, 250, 494, 744
meme 716, 727, 728 multilingual text production 7, 746, 749–50 (see
mental health 659–60, 664, 675 (see also health, also interpreting, translation)
medical discourse) multiliteracies 546
message 4, 290–1, 500 (see also message analysis Multimodal Analysis Image 441
and message semantics) Multimodal Analysis Lab 575
message analysis 500–1 (see also message multimodal corpora 578–9
semantics) multimodal/multimodality 293, 307, 319, 327,
message semantics 335, 338–9, 341, 348, 350, 400, 430, 433, 575
666 SF-MDA 433–40, 444, 458
metadata 715, 716, 725 (see also social networking SF-MDA Framework 439, 576–7
sites, social metadata, social tagging) multimodal semiosis (see multimodal/
metafunction 400, 403, 433, 435–6, 440, 503, multimodality)
517–18, 567, 587, 654, 658, 725 multiperspectivism 71–5
metafunctional shift 742 multiscalar model of context 164
metalanguage 13, 641, 700 multisemiotic 459, 541, 547–8 (see multimodal/
metaphor 621, 633, 669 multimodality)
conceptual metaphor theory 476 multi-stratal 174, 502, 644, 656, 660
grammatical metaphor 6, 46–7, 63–6, 72–4, multivariate structure 119, 124, 133–5 (see also
213–14, 266, 276, 371, 378, 397, 521, univariate structure)
622–3, 628, 637, 645, 741, 747 mutual expectancy 17–22
interpersonal metaphor 64–5, 360
lexical metaphor 46, 64–5, 81, 392, 401 narrative, narrating 43, 760
logical metaphor 360 native speaker, non-native speaker 180, 327, 642
metaphorization (in translation) 622–3, 747 natural language generation 235, 324, 562–3 (see
metaredundancy 3, 66–9, 72, 75, 86, 88 also KPML, Penman project)
metarepresentation 501 Negator (N) 244, 253
metaphenomenon 101–2 negotiation 81, 289, 301, 383–4, 654, 666 (see
fact 101 also discourse semantics)
idea 101 Neurocognitive disorders 598
metastability 66–9, 88, 149 Neurodegenerative disease 598 (see Alzheimer’s
microblogging 716, 724 (see also social networking Disease)
sites) neurodevelopmental disorders 592
microfunction 490 (see also macrofunction, New 111–12, 185, 188, 193, 267 (see also Given,
protolanguage) Theme)
microplanner 235 Contrastively New, Contrastive New (CN) 240–1
Middle English 410, 425, 623 Culmination of New 185, 188
modalization battery 600 (see also Alzheimer’s Hyper-New 361, 369, 394
Disease) macroNew 445, 448, 457
modality 40, 66, 80, 108–10, 267, 435–6, 654, New1 186
693, 745 New2 186
modalization 108–10, 391, 423, 592, 600 Unmarked New (UN) (unmarked newness)
modulation 108–9, 389, 423 188, 240
Mode (contextual) 17, 147, 149, 163, 165, 190, New Criticism (literary) 692, 695 (see also literary
219, 303, 399, 402, 468–70, 475, 480, 518, criticism)
539, 627, 633, 754, 758, 779 New information 445, 451, 454 (see New)
channel 17, 186, 436, 630, 638, 754, 759 Newton (Isaac) 415, 426, 624, 628, 631
medium 85, 399, 429, 573, 668, 754 nominal group 101, 103, 122–3, 129–30, 135–6,
role of language 217, 219, 469, 515, 620, 651, 177, 208, 238, 244–6, 248, 251, 292, 415,
754 454, 622, 637 (see also Epithet, Deictic,
rhetorical mode 218–20, 222, 291 Modifier)
Modifiers 124, 135, 246, 420, 428 Thing vs Head 125, 236–8, 245–6, 292, 420,
monoglossia 394 (see also heteroglossia) 452, 629
800 INDEX
nominalization 46, 266, 413–14, 417, 634, 636, phonetics 25, 174, 333, 571, 587–8, 604, 607
660 phonological paragraph 177, 190 (see also
non-discreteness of language 268 paraphone)
non-literary stylistics (see under stylistics) phonological rank scale 121, 172, 176
non-verbal action 144 (see also verbal action) phonological realization 37, 187
nursing 658 (see also medical discourse) phonological system 40, 378
phonological units (in English) 176
observation 642, 644 phonology 162, 171–2, 174–7, 179, 207, 273,
Òkó 782 333, 341, 359–60, 378, 520, 589, 604, 607
Old English 411–12, 427, 430 (see also segmental phonology)
oncology 656 (see also clinician–patient interaction, photograph (analysis of ) 346, 411, 426, 433, 436,
medical discourse) 441, 444–6, 575–6 (see also multimodal/
onset (in phonology) multimodality)
onset syllable 190 phrase 118–19, 121–3, 129, 131, 133, 135–6, 138,
rise–fall hook onset 192 140, 379, 393, 420, 438, 564, 567, 600, 655
fall–rise hook onset 192 (see also group, rank scale)
onset (in clinical linguistics) 595 phylogenesis 45, 152, 286
child onset fluency disorder 606 (see also picture books 349, 400, 546
stuttering) pitch 173, 182, 187, 189, 191–2, 196–8, 438
ontogenesis 63, 152, 286, 396, 487, 721 (see also pitch contour 174, 181, 183, 185, 191–2, 194–5,
child language, protolanguage) 197–9
open context 338, 351 pitch range 191–2, 196, 198
operator 127, 236–8, 244, 252, 254, 412–13 Pitjantjatjara 780
oral-literate continuum 525 planner 234–5, 238 (see also microplanner)
orthography 730 discourse planner 235
over-lexicalization 466 sentence planner 233, 236
poem 691, 702
paradigm 14, 26, 339, 610, 622 poets, poetry 625, 707
paradigmatic 18, 35, 43–5, 58, 156, 181–2, 223, polarity 240–1, 268, 345–6, 385, 507, 704
274, 278, 318, 351, 478, 520, 566, 568 (see polysystem/polysystemic 11–26, 172
also syntagmatic) population thinking 627, 640
paragraph, paragraphing 177, 190, 359, 363, posture 17, 19, 26, 400, 493, 642 (see also
375, 444, 751–2, 755 communicating body)
paralinguistic 726, 732 power (in language) 298, 305, 341, 348, 360, 436,
parallelism 629, 754 (see also foregrounding) 462, 467, 472, 543, 548, 753 (see also
grammatical parallelism 695 critical discourse analysis (CDA), critical
pervasive parallelism 691–2, 706 linguistics)
paraphone 177, 190 PRAAT 191
paraphrase 347, 742–3, 747 Practical Criticism (literary) 692 (see also literary
parole 1–2, 44–5, 152 criticism)
parser/parsing (in computational linguistics) 565 pragma 11–18
(see also computational complexity) Pragmatics 15, 23, 265, 274, 289, 473, 503,
automated parsing 569 587
Stanford Dependency Parser 567 pragmatic conversation 306 (see also casual
systemic parsing 567, 570 conversation)
Universal Dependency 564, 567 Prague School linguists 692
Participant Role (PR) 147, 214, 235, 241, 249–51, Predicator 106, 244, 412–13, 502, 707,
253, 435, 442 769
patient-initiated humour 656 (see also humour) prefaced 506–7 (see also semantic networks)
passive voice 104, 465, 525 prehension 18, 23
pattern 14, 161, 180, 189, 216, 220, 223, 265, prepositional group 123, 128, 132, 246,
275, 292–3, 297, 299, 305–6, 339, 479, 543, 251
550, 577, 590, 592, 600, 607, 610, 620, 627, prepositional phrase 119, 126, 129, 136, 138,
644, 654–5 244, 414, 521, 629–30
pedagogical discourse 348 pre-text context 165 (see also context of situation)
Penman project 231, 564 pretonic 184–5, 190–2, 194, 196–9 (see also
Performer 235 (see also addressee) tonic)
periodicity 86, 334, 358, 361, 371, 378 (see also pretonic prominence (see under prominence)
discourse semantics) pre-translational text analysis 750–1
permeability 155–6 preventive medicine 668 (see also medical
personal reference (see under cohesion) discourse)
pervasive parallelism (see under parallelism) primary carer 499 (see also child language)
phasal shift 224 primary progressive aphasias 609
phasal analysis 224–5, 334, 594 principal components analysis (PCA) 20
phase (of discourse), phase analysis 14, 224–5, probability 435, 591
298, 304 probability theory 568
phoneme 25, 176, 179, 379, 638, 761 probes 95–7
phonetic expectancies 19 process test 95, 99, 251–3
Index 801
process type 212–14, 238, 240, 250–1, 316, 422, realization rules or statements 56, 179, 209, 223,
435, 445, 579, 600, 604, 703 236–44, 255, 273–4
behavioural 104, 435, 447, 604, 703 same pass rules 238
existential 104, 422 preselection 238, 241
material 212–14, 238, 240, 250–1, 316, 415, realizational systems 644, 675
419–20, 422, 435, 445, 463, 495, 498, 553, recognition literacy 542–3 (see literacy)
626, 704 recommending 158, 218, 664 (see also field of
relational 102–4, 267, 422, 435, 449, 604–5, activity)
637, 703 reconstrue 39, 527
attributive 102, 445, 457 recoverability 41, 94, 106, 240
identifying 103, 435, 454, 457, 637 recreating 158, 218 (see also field of activity)
mental 99–102, 238, 378, 383, 386, 422, 516, reference chain 314, 327, 361 (see also cohesion)
604, 704 reflection literacy 542–3, 546–7 (see also action
verbal 104, 209–11, 214, 422–4, 478, 522–3, literacy, recognition literacy and literacy)
542, 546, 604, 703 register 29, 45, 71, 83–5, 145–6, 180,
progressive message 339 (see also message 214, 216–17, 219–24, 235, 265,
semantics) 290–1, 300, 303, 358, 379, 382, 426,
projecting verb (see under clause) 436, 518, 525–6, 538, 540, 542, 547, 552,
projection 39, 59–60, 75, 83, 160, 209–10, 394–5, 579, 666 (see also situation, context of
415, 507, 516, 661, 780 situation)
prominence 445–6, 696 (see also salience, register variation 17, 19, 98, 181, 190, 225,
tonicity) 296, 411, 457, 509, 552, 627, 659, 668,
tonic prominence 41, 178–9, 181–2, 185 742–3
textual prominence 193, 394 registerial cartography 157, 160, 552
prominent 177, 448, 458, 655 (see also Relevance theory 474, 749
prominence, information prominence) relevancy 145, 147, 149, 152, 164
prosodic analysis 24, 172 Remiss (see underfoot)
prosodic composition 180, 185 reporting 158, 218, 390, 394, 433, 444, 664–5,
prosody 372–3, 393–4, 465, 608 (see also semantic 675 (see also field of activity)
prosody) resemiotization 433, 439–40, 719 (see also
Proto-Indo-European 641 intersemiosis, multimodal/multimodality,
protolanguage 396, 489–94, 502, 504 (see also semiotic)
child language) resonance 403
protolinguistic sign 489–90, 492–3 (see interpersonal resonance 403
protolanguage) fractal resonance 75–6
protoscience 620 restricted code (see under code)
punctuation 179, 597 restricted language 16–19, 26 (see also register)
punctuative message 501–2 (see also message Rheme 42, 111–16, 188, 267, 294–5, 313, 413,
semantics) 415, 417, 425, 445, 524, 527 (see also
theme, lexicogrammar, New)
quality group 246–7 rhetorical mode (see mode)
quantity group 248–9 rhetorical relations 324–8 (see also Rhetorical
Structure Theory (RST))
radio broadcasts/commentary 180, 189 (see also nucleus–satellite relations 299–300, 325
media discourse) multi-nuclear relations 299, 326
Range 129 Rhetorical Structure Theory (RST) 4, 29, 83, 153,
prepositional phrase as 126, 130 159, 298–300, 324, 334
rank scale 38, 82, 120, 123, 131, 133, 137, 176–7, Rhetorical Unit (RU) 153, 155, 293–6, 334
242, 338, 706 rhythmic unit, unit of rhythm 177 (see also foot,
phonological 121, 177–8, 182, 341, 347, 438, rhythm, syllable)
495, 599, 655 rhythmic patterns 180
lexicogrammatical 120, 122, 125, 129–30, 132, rhythmic properties 27
135–6, 155, 336–7, 345, 378, 393, 491–2, role enactment 86, 340 (see also message
500, 502–5, 515, 517–18, 522, 539, 569, semantics)
572, 592, 596, 604, 668 role of language (see under Mode, context of
ranking clause 359, 367, 501 situation)
rankshift 38, 124, 134, 136, 138, 140, 426, 629 Rolland 573 (see autonomous wheelchair, dialogue
reactance (grammatical) 206, 208, 212 (see also systems)
realization)
reaction (in appreciation network) 378, 388 salience 185, 425
reading 473, 541, 604 (see appreciation) salient syllable (see under syllable) (see also
reading aloud 179–80, 189 Tonicity)
reading comprehension 540 Sanscrit/Sanskrit 773
reading to learn pedagogy 540 scale-and-category linguistics 119, 740
realization 37, 47, 56, 67–8, 148–9, 151, 155–6, scale of delicacy 145
161–2, 174–5, 187, 207, 219–20, 222, 226, schizophrenia 590
232, 248, 254, 299, 304, 401, 502, 644, 662, schizophrenic discourse/speech 591, 608
752 (see also realization rules) (analysis of, features of)
802 INDEX
science 383, 418, 475, 477, 479, 519, 524, 542, social activity 148, 154, 286, 514, 516, 751, 753,
545, 547–8, 551, 621 758
register 419, 475, 477, 479, 551, 579, 666 social class 21, 342
experimental science (discourse of ) 415, 628 social distance 165, 344, 347, 594, 659, 754
scientific developments 418, 420–1, 631 social media 399, 664, 715–16, 718–19, 722
scientific discourse 46, 52, 74, 414, 416, 428, social network 727
546, 548, 620, 623, 633 social networking sites 715
scientific English 414–15, 638 social processes 13, 158, 161, 277, 437, 440, 539,
scientific explanation 622, 631 731
scientific research article 418 social purpose 7, 292, 518, 539, 550, 627
science of meaning 15, 49, 644 social semiotic(s) 266, 269, 398, 401, 436, 440,
scientific themes 626 462, 474, 479, 487–8, 498, 514, 517–18,
scope 389–90 (see also graduation) 537–8, 542, 548, 550–1, 576, 720
Scope (grammatical element) 96–7, 248 language as 2, 488, 498, 514, 517–18, 537–8,
second language acquisition (SLA) 6, 512 542, 548, 550–1
second language development/learning/teaching theory of language 1–2, 233, 266, 279, 286,
349, 512 462, 474, 479, 487
second-order context 47, 164 (see also context of social metadata 716–17 (see also metadata)
situation, first-order context) social tagging 715, 717 (see also metadata)
segmental phonology 179 sound wave 754, 759
selfies 723 (see also identity) source text 744–7, 760–1, 782 (see also target text,
semantic cluster 19, 508 translation)
semantic component 59, 75 Spanish 113, 401, 645
semantic domain 388, 468, 470, 770 Specific Language Impairment (SLI) 593
transcategorical semantic domain 75, 82 speech function 267, 296, 334, 347, 360, 442,
semantic drift 595, 660 453, 594 (see also metaphor of mood)
semantic networks 44, 60, 77, 236, 333–41, 343, speech sound disorder 596
348–51, 502–8, 670 (see also message statements of meaning 12, 18
semantics) stratification 3, 56–8, 60–6, 220, 519–21, 632,
semantic prosody 13, 20, 28, 81 657, 721, 741
semantic stratum 44, 56, 58–63, 222, 289, 291, streaming 718 (see also social media, social
500, 550 networking sites)
semantic variation 21, 29, 334–5, 341–2, 347–8, stroke 589, 601 (see also acquired brain injury
464, 480, 496–7, 501, 503–5, 508–9, 538, (ABI))
544, 657, 770 structuralist theory 58, 171–2, 209
semantics 55–88, 251, 254, 265, 273–4, 289, 300, stuttering 606 (see also onset (in clinical linguistics))
333–4, 340–1, 351, 358, 473, 494, 508, stylistics 276, 400, 622, 690
520–2, 524, 542, 545, 741, 744 British stylistics 690
as discourse-structural meaning 60, 508 non-literary stylistics 693
as higher-level systemic meaning 59, 158, 174, structural stylistics 691
494, 520–2, 524, 542, 545 Systemic Socio-Semantic Stylistics 690
as topological meaning 60, 473 Subject 4, 41–2, 106–7, 113–15, 193, 238, 241,
semiotic 36, 285, 398, 462, 473–4, 478, 488, 244, 251, 253–4, 289, 316, 368, 412–13,
493–4, 508–9, 515–17, 520, 525–6, 529, 421, 429, 472, 539, 567, 642
537–8, 552, 726 supervenience 287, 305, 307
artefact 458, 729 Support Vector Machine (SVM) 577
mediation 497, 516, 544, 552 suprasegmental 171
processes 158, 223, 439, 474, 540, 547, 664 (see surgical teams 666 (see also medical emergency
also social processes) teams)
resources 37, 426, 433–4, 436, 450, 458–9, syllable 25, 176, 178–9, 181–3, 185, 193, 368, 379
540–1, 547, 576–7, 722, 727 (see also intonation)
systems 36–7, 49, 63, 67, 85–6, 153, 160–1, salient syllable 176–7, 182, 185
216, 221, 233, 271, 287, 293, 402–3, 434, weak syllable 177, 182
437, 495, 542, 636, 701, 706, 742 syllable-initial consonant (see under consonant)
semogenic 36, 634, 638 syllable margin 179 (see also syllable-initial
semogenic power 47, 68, 628 consonant, syllable-final consonant)
serial contextualization 18–19 symbolic articulation 7, 622, 695, 701–2, 707 (see
sexual health 669 (see also health, medical double articulation)
discourse) synchronic linguistics 410
SF-MDA 576–7 (see under multimodal/ syntagmatic 43, 58, 72, 177, 182–3, 186, 274,
multimodality) 318, 440, 479, 508, 520, 566, 706 (see also
signalling 323–4 paradigmatic)
silent ictus (see under foot) syntax 464, 487, 489, 523, 642–3, 777, 780
situation 468, 470, 472, 475, 503, 518, 571, 671, system network 43–4, 58, 69, 85, 94, 156–7, 163,
754 (see context of situation) 177–9, 182–3, 220, 231–2, 235–44, 254, 266,
situation type 146, 155, 518 (see context of situation) 268, 274, 279, 297–8, 334, 336, 338, 351,
SLA 512–16, 520, 523, 527–31 (see second 384, 395, 400, 434, 441, 492, 503, 569, 590,
language acquisition) 658 (see also contextualization system networks)
Index 803
systematicity 493 (see also functionality) tone unit 174–8, 181, 183–4, 189, 594
Systemic phonology 171, 177 (see also phonology) tonic 368 (see also tonic syllable, tonicity)
tonic foot (see under foot, see also tonicity)
tag phrases 107 tonic prominence (see under prominence, see also
Tagalog 777 tonicity)
target text 742 (see also source text, translation) tonic syllable (see under tonicity)
taxis 19, 38, 82, 435, 703 tonicity (system of, choices in) 174, 178–81,
Thai 781 186, 189, 368 (see also pre-tonic, pre-tonic
teacher education/teacher training 529, 537, prominence)
539, 550–1 unmarked 368
teaching–learning cycle 539 tonic syllable 368
Telugu 780 topology 375 (see also typology)
Tenor 48, 84–5, 147, 161, 164, 217–20, 225, 291, trace 164, 420, 440
293, 303, 360, 379, 426, 436–7, 443, 453, transdisciplinary 417, 462, 473, 488, 497, 499,
468, 643, 746, 751, 753 (see also context of 514, 528, 550, 644
situation) transformational generative grammar 29, 776
text analysis 359, 362, 379, 463–4, 541, 751 (see (see TG grammar)
also pre-translational text analysis) transformative/creative system 97–8
text-based transitive vs intransitive 95, 412, 425, 429
curriculum 540 Transitivity 95, 211, 213–14, 231, 240, 244,
pedagogy 541 249–51, 254, 255, 267, 317, 322, 429, 464,
syllabus 540 467, 567, 600 (see also lexicogrammar)
teaching 552 Transitivity analysis in clinical linguistics 702
text synthesis 362 (see Alzheimer’s Disease)
TextTiling 327 translation 276 (see also interpreting)
textual metafunction 358, 413, 436, 517, 593, equivalence 744–6
624, 744 hermeneutics of 760
textual function (in clinical linguistics) 503, 595 literal 744
texture 41, 60, 80, 154, 163, 288–9, 291, 293, machine translation 327, 561, 741
311–12, 319, 326, 334, 384, 437, 440, 577, process 748
638 quality assessment 741, 755
thematic structure 40–1, 222, 411, 415, 421, 425, translator 675, 749
429, 436, 746 traumatic brain injury (TBI) 601 (see also acquired
thematic progression 42–3, 328, 417–18, 668, brain injury (ABI))
752 trinocular perspective 62–3, 158, 207, 348, 518,
theme 41–3, 66, 111–16, 149, 188, 240–1, 538, 552
255, 267, 294, 346, 361, 367–9, 371, Twitter 7, 399, 576, 716 (see also social networking
394, 413, 415, 417–18, 420–2, 425, sites)
436, 442, 445, 448, 451, 458 (system of, tweeting 361
choices of see also lexicogrammar, type 151, 243, 567
New, Rheme) ‘typical actual’ 15–16, 18, 30
adjunct Theme 421–2 Typology 375, 402 (see also topology)
marked, unmarked Theme 41, 113–15, 240, Formal approach to language typology 776
595–6, 745 Functional approach to language typology 776
and New, and Information 42, 79 History of language typology 775–82
experiential Theme 42 Language typology 269, 402, 776
hyper-Theme 43, 361, 369, 371, 393, 445, 448, Typology of genres 160, 375 (see also genre)
453–4, 527, 549 Typology of grammatical metaphors 634 (see
interpersonal Theme 42 also grammatical metaphor)
macro-Theme 83, 361, 371, 373, 375, 445, 448, Typology of registers 157 (see also register)
453, 457
topical Theme 445, 448 UAM Corpus Tool 564, 758 (see also corpus)
textual Theme 42, 313 unit 566, 630 (see rank scale)
threat (semantic network category) 336–7 unit of analysis 4, 325, 338, 474
token 445, 449, 567 (see also cohesive harmony) universal grammar 12, 15, 27, 642–3, 776
central token 290, 321 univariate structure 119, 133–6, 140 (see also
relevant token 290, 321 multivariate structure)
peripheral token 290, 321 unmarked New (UN) (see under New)
tonality (system of, choices in) 178–81, 186–7, unmarked stress 593
189 unmarked Theme (see under THEME)
tone (system) 174, 178, 180–1, 184, 186–7, 194,
345, 347, 379, 428 valeur 14, 19, 25, 56–9, 69, 641, 721
compound tone 181, 183–4, 191, 193–4 valuation (in appreciation network) 373, 388
primary tone 181, 184, 191, 655 (see appreciation)
secondary tone 195, 197 verbal action 144, 210, 660
simple tone 193 verbal art 7, 46–7, 628, 633, 639, 690–1, 698, 701
tone group 111–12, 121, 174, 176, 178–9, verbal group 18, 123, 126–7, 130, 135–6, 138,
182–3, 185, 187, 190, 193, 593 244–5, 359, 391, 629, 637, 769, 772
804 INDEX
verbal science 7, 46, 621, 623, 638, 645 webpage 433, 435–7, 439–57
vernacular 620–2, 641 Weibo 715 (see also social networking sites)
via-text context 165 (see also context of situation) WHO 433, 435, 437, 439–41, 454
Vietnamese 401 word phonology 179
visual display 444, 447, 452, 457, 547 work 479 (see also multimodal/multimodality)
voice 104, 465, 524 (see also active voice, passive writing 361, 365, 376, 520, 604, 754
voice) written language 521, 525–6, 539, 544,
voice quality 172, 180, 182, 198–9 575–6
war (language of ) 291, 426 YESnography 302–5

weak syllable (see under syllable)
weak vowel 179, 183 Zone of Proximal Development 516

THOMPSON BOWCHER FONTAINE SCHONTHAL 2019 The Cambridge Handbook of Systemic Functional Linguistics PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

THOMPSON BOWCHER FONTAINE SCHONTHAL 2019 The Cambridge Handbook of Systemic Functional Linguistics PDF

Uploaded by

Copyright:

Available Formats

The Cambridge Handbook of Systemic Functional Linguistics

Part I. SFL: The Model:

Part II. Discourse Analysis within SFL:

Part III. SFL in Application:

theory and practice of Systemic Functional Linguistics evident throughout

I.1 The General Scope of This Volume

The Cambridge Handbook of Systemic Functional Linguistics presents a clear and

language as an already produced and an evolving system is predicated on the

These principle perspectives of SFL theory underpin the contributions in

I.2 The Structure and Contents of This Volume

The chapters in this volume cover a comprehensive range of theoretical

I.3 Part I – SFL: The Model

I.4 Part II – Discourse Analysis within SFL

I.5 Part III – SFL in Application

Fairclough, N. 1989. Language and Power. Harlow: Longman.

1.1 J. R. Firth (1890–1960): First Impressions and

‘[M]eaning’ is a property of the mutually relevant people, things,

commitment to being methodical or complete in his explanations. Thomas

My own approach in general linguistics and especially in the study of mean-

managed separately and autonomously, with linguists left to ‘mop up’

acoustic, or even social primes that are extrinsic to languages. Linguistic

1.2 Restricted Language

(a) whether linguists are accounting for a collective phenomenon or the

(c) whether linguistics should attend to the dominant motif of a unity

Firth’s approach assists us in reconciling these dichotomies in what Halli-

1.3 ‘Mutual Expectancies’: Pattern, Process, and Pragma

The spectrum of phonetic expectancies allows for ‘a closer approach to the

Conﬁning ourselves to English as the language of description, let us face the

So the same complex layering of continua in phonetics was to be seen at

morphemic exponents of the grammatical roles of the Latin word pedibus

A single word isolate such as Latin pedibus might have to be considered in a

The extension of meaningful relations through mutual expectancy also

different conﬁgurations, represent strategies within interaction that clearly

The concepts of ‘mutual expectancies’ and Hasan’s semantic variation are

(1956:83) in the notion: ‘conﬁgurational rapport’. Essentially, this conﬁgur-

1.4 So What Would Be a Firthian Approach in Situation,

1.4.1 Context of Situation

on traditional formulas, the speciﬁc locale (mountain and water for

I have suggested that language is systemic. . . . We may assume that any

England 1914–1945 (see Love 1988:151–62). After setting out Firth’s

1.5 Posture and the Communicating Body

In 1974, emphasising the rhythmic properties of expressions, I said, ‘when a

Firth’s concepts of ‘posture’ and ‘person’ suggest a general account of a

1.6 Prehension in Science and the Arts: The Zeitgeist

Prehension is a term central to Whitehead’s theory of natural, cosmic

Unlike most other inﬂuential theories of language and linguistics with

Allen, W. S. 1953. Relationship in Comparative Linguistics. Transactions of

Bateson, G. 1982. Difference, Double Description and the Interactive Desig-

Halliday, M. A. K. 1978. Language as Social Semiotic: The Social Interpretation of

Laver, J. 1980. The Phonetic Description of Voice Quality. Cambridge: Cambridge

Teich, E. 2003. Cross-linguistic Variation in System and Text. Berlin: Mouton de

2.1 Systemic Functional Linguistics

A Systemic Functional approach looks at how language functions to make

2.2 Meaning in SFL Theory

Language is the instantiation of an indeﬁnitely large meaning potential

in an act of interpersonal exchange, communicating my sense of my own

These sets of semantic options constitute a semantic system or system of

2.3 Grammar in SFL Theory

To the ‘architecture’ of grammar, Halliday and Matthiessen (2014:55)

2.4 Realizing Ideational Meaning

mention of who or what is participating, and possibly also of the circum-

2.5 Representing Interpersonal Meaning

Language functions to facilitate exchange, both of information and of

congruent realization congruent realization

literal meaning congruent realization