Professional Documents
Culture Documents
to Nominalization in English
WDE
G
Cognitive Linguistics Research
26
Editors
René Dirven
Ronald W. Langacker
John R. Taylor
Mouton de Gruyter
Berlin · New York
A Cognitive-Functional Approach
to Nominalization in English
by
Liesbet Heyvaert
Mouton de Gruyter
Berlin · New York 2003
M o u t o n de Gruyter (formerly Mouton, The Hague)
is a Division of Walter de Gruyter G m b H & Co. K G , Berlin
ISBN 3 11 017809 5
The present study reflects a research project to which many people, each in
their own way, have contributed. I would like to thank those people here.
First and foremost I would like to thank Kristin Davidse. As my Ph.D.
supervisor, she was involved in my research on nominalization patterns
from its earliest stages onwards. I am deeply grateful for the continual en-
couragement she has given me, for the openness with which she has shared
her own learning with me, and for the precision with which she has read my
texts, time and again supplying extensive commentary. This book, in which
any remaining shortcomings are my own, in the first place represents the
dialogue I have been able to conduct with her.
I would also like to thank Hubert Cuyckens, Ad Foolen, Ronald Lan-
gacker and Willy Van Langendonck for their rigorous comments on an ear-
lier version of this book. In addition, I am grateful to René Dirven and
Ronald Langacker who, as editors of the Cognitive Linguistics Research
Series, guided me through the proposal and writing process and whose
careful comments on various chapters of the book I have greatly appreci-
ated. I thank the managing editor of the series, Birgit Sievert, for her pa-
tience and for her professional assistance throughout the publication proc-
ess. I thank Jürgen Benteyn for his swift and always good-humoured help
with the formatting of the manuscript, and Hubert Cuyckens, Kristin
Davidse and Lieven Vandelanotte for their general support. Thanks also to
Lieven Vandelanotte for his help with the drawings and to An Laffut and
Lieven Vandelanotte for proof-reading parts of the manuscript.
My thanks also go to a number of people who have in the past years
commented on my work and helped me with their suggestions and criti-
cisms. In addition to Hubert Cuyckens, Kristin Davidse, Ad Foolen, Ronald
Langacker and Willy Van Langendonck, these are: Eirian Davies, Renaat
Declerck, Dieter Kastovsky, An Laffut, Odo Leys, Lachlan Mackenzie, Bill
McGregor, Hans-Jörg Schmid, Lieven Vandelanotte, Frederike van der
Leek, Jean-Christophe Verstraete and Emma Vorlat. The University of
Leuven has contributed to the publication of this book by allowing me to
work on it as a postdoctoral fellow of the Research Council (postdoctoral
grant PDM/02/034).
I am most grateful to my parents and my parents-in-law for their unfail-
ing support. Finally, I would like to thank Wouter, without whose uncondi-
tional support, practical help and love I would never have been able to
bring this book to a good end.
Contents
Chapter 1: Introduction 3
Chapter 3: Nominalization 41
1.
1.1. A usage-based account
Schematization of nominalization
and extension 42
43
1.2. Entrenchment 44
χ Contents
0. Introduction 99
1. -er nominalizations: Agent names or Subject names? 101
1.1. Agentivity and salience 102
Contents xi
0. Introduction 149
1. Representational semantics 151
1.1. Agentive-er nominalization 152
1.2. Non-agentive-er nominalization 153
2. A semantic typology of non-agentive -er nominalizations 154
3. The constructional properties of -er nominalization 158
3.1. Lexicalized -er nominalizations 159
3.1.1. Non-agentive -er nominalizations 159
3.1.2. Agentive -er nominalizations 161
3.1.3. A note on instrumental -er nominalization 163
3.2. Ad hoc nominalizations 165
3.2.1. Phorie ad hoc nominalizations 166
3.2.2. Non-phoric ad hoc nominalizations 168
3.2.3. Ad hoc-er nominalization: Conclusion 172
3.3. The constructional properties of -er nomináis: Summary 173
4. Deverbal -er suffixation as Subject-profiling 174
5. Conclusion 176
xii Contents
0. Introduction 181
1. Factivity as truth presupposition 183
2. Factivity as embedded projection 185
2.1. Embedding vs. taxis 186
2.2. Projection vs. expansion 187
2.3. Facts as embedded projections 188
2.4. Conclusion 194
3. Delineation of the fact category 195
3.1. Ambiguity and vagueness 195
3.2. Gerundive nominalizations: Acts or facts? 198
4. Towards an internal, nominal analysis 200
4.1. Halliday's dependency analysis 201
4.2. Type vs. instance nominalization 203
Conclusion 244
Appendix 255
Notes 257
References 267
Index 281
Parti
Towards a theoretical-descriptive approach
to nominalization
Chapter 1
Introduction
related structures and options in which both Cognitive Grammar and Sys-
temic-Functional Grammar situate a construction) (Section 5).
In Chapter 3,1 will show how these basic theoretical assumptions enable
us to identify some of the lacunae and weak points in the existing ap-
proaches to nominalization. Importantly, these principles also point out a
possible course of action in the description of nominalized constructions.
Going through the main theoretical divisions of Chapter 2 again, I will ar-
gue, firstly, that nominalization patterns have to be categorized both by
schematization and extension. It also has to be taken into account that
nominalizations may display varying degrees of entrenchment or automati-
zation: they may be lexicalized and function as fixed expressions in lan-
guage, or they may only be established as grammatical patterns or sche-
matic units, without themselves being included as conventional units in the
language system (Section 1). Secondly, I will stress the importance of
viewing nominalizations, like all constructions, as natural symbolic units of
which the meaning is encoded in the lexicogrammar (Section 2).
Thirdly, I will argue that nominalizations have to be viewed as func-
tional configurations which are situated on a particular level or rank of
functional organization (Section 3). As such, they find themselves in be-
tween a 'higher' rank in which the nominalization itself serves a function
and a 'lower' level that constitutes the nominalization's own internal or-
ganization. A full description of a nominalized structure requires an analy-
sis of its external, synthetic functioning, of its internal, analytic functional
outlook, and of the way in which the external and internal functions are
combined in the construction itself. Put differently, some nominalized con-
structions may represent an at first sight inexplicable mixture of external,
nominal behaviour and internal, clausal categories, but the nominalized
structure itself succeeds in combining these seemingly divergent functions
into one construction and therefore has to be accounted for. Moreover, for a
number of nominalizations, we need a concept of downranking, which al-
lows the possibility of functional configurations occurring at a level of or-
ganization which is not the one on which they normally function: clause-
like nominalizations such as íAaí-clauses can then be identified as having
been downranked from clausal to nominal level while preserving part of
their internal clausal outlook. By the same token, downranking entails its
own form of reclassification and involves external as well as internal re-
classification: by being reclassified, the clausal structure takes on the
grammatical and semantic properties inherent to the nominal unit. Failure
to recognize the reclassification involved in the rankshifted type of nomi-
nalization has been one of the main gaps in the description of nominaliza-
tion so far.
6 Introduction
Each case study will start with a chapter in which I look at the main in-
sights thus far offered in the literature and establish my own descriptive
position. In Chapter 5 an overview will be given of the literature on dever-
bal -er nominalization; Chapter 8 will discuss the most interesting analyses
that have thus far been proposed for the system of factive nominalization.
In my own descriptive analysis of -er derivation and factive nominalization
(developed in Chapter 7 and Chapter 9 respectively), I will attempt to show
that it is possible to move beyond existing descriptions and provide an an-
swer to some of the long-standing moot points in the analysis of specific
nominalization types if the functionally-oriented approach which I set out
in Part I is systematically applied. Importantly, my analyses will focus not
on the external functioning of the nominalized constructions, but on their
internal functional organization, i.e. on the clausal and nominal categories
which they integrate.
Apart from the general functional perspective which I will take and the
central role which I will assign to the functional categories of the nominal
and the clause, two main motifs can be said to run through both my descrip-
tive case studies. First, my analysis will be based on the assumption that the
semantics of a construction is encoded in its lexicogrammatical properties
and can therefore be revealed through a careful analysis of those properties.
I will, in other words, not only present a detailed analysis of the lexico-
grammar of -er nominalizations and factive nominalizations; I will also try
to shed light on the basic meanings which these nominalization types en-
code.
Secondly, in each of these studies, the description of the syntagmatic in-
tegration of clausal and nominal properties in the internal organization of
the nominalized constructions is supplemented with observations about the
paradigmatic relations which the nominalized structures hold. In other
words, the syntagmatic properties of nominalizations will be systematically
considered against the background of the larger system network which they
form part of and which they select from. Not surprisingly, the paradigmatic
relations that will turn out to be most valuable to the elucidation of nomi-
nalizations are those with clausal structures and with nomináis. First, due to
the clausal nature of their starting point and the clausal categories which
they embody, nominalizations tend to be related to clausal structures. The
identification of related or 'agnate' clause types will be shown to play a
crucial role in the analysis of nominalized constructions. In my description
of -er nominalizations (Chapter 7), I will thus elaborate on the systematic
relationship that exists between -er suffixation (e.g., bestseller) and the
clausal middle construction (e.g., That book sells well) and I will show how
an accurate analysis of the lexicogrammatical properties of middle clauses
8 Introduction
which associates the semantic unit [OPENER] with the phonological unit
[opener]).
Composition refers to the ability to integrate two or more (component)
structures to form a composite structure (Langacker 1999: 94). Composi-
tion gives rise to relationships on the syntagmatic plane of language (Lan-
gacker 1987a: 75). An example of a composite structure is can opener,
which combines the symbolic units [[CAN]/[can]] and
[[OPENER]/[opener]], which in turn consists of [[OPEN]/[open]] and
[[-ER]/[-er]].
Apart from symbolization and composition, the abilities of comparison
and schematization are also relevant to language use. According to Lan-
gacker (1987a), the language user solves the problem of coding a detailed,
context-dependent conceptualization in terms of a linguistic expression by
making comparisons or categorizing judgements (Langacker 1987a: 65-
73): he/she assesses whether a specific expression can be categorized as a
member of the category defined by a conventional linguistic unit or a unit
that is widely shared by the language users of a speech community. A us-
age event is therefore always systemically motivated or "structured and
evaluated with reference to the conventional units of a grammar" (Lan-
gacker 1987a: 426).
Because structures which at first sight appear very different, may be
"quite comparable in a coarse-grained view" (Langacker 1999: 93), catego-
rization necessarily involves the ability of schematization: to determine
whether a usage event satisfies the specifications of specific conventional
units, one may have to abstract away from certain points of difference and
portray the distinct structures with lesser precision and specificity (Lan-
gacker 1999: 93). A usage event can then either be identified as elaborat-
ing/instantiating a category, or it can be said to extend it. When a linguistic
category is elaborated or instantiated by a usage event, the latter conforms
to the specifications of the category but is, schematically speaking, "charac-
terized in finer-grained detail" (Langacker 1999: 93). An elaborative or
instantiating novel usage is conventional or - in more traditional terminol-
ogy - grammatical·, it is conforming to the linguistic convention embodied
by the linguistic unit, i.e. it is fully sanctioned by it. When, on the other
hand, a discrepancy exists between the category that is defined by the lin-
guistic unit and the usage event which is related to it, the category is ex-
tended and the usage event is either ill-formed or non-conventional.
In general, a novel usage, irrespective of whether it merely elaborates or
extends a conventional unit, can become conventionalized or acquire unit
status via entrenchment (Langacker 1987a: 59). When the unit [[MOUSE]/
[mouse]], for instance, was first used with reference to a piece of computer
Language: The relation between system and usage 13
equipment, its meaning was extended. This extension has by now achieved
unit status because of its frequent use and the entrenchment resulting from
it (Langacker 1999: 108). Entrenchment or the automatization of certain
structures is a matter of degree and the boundary between units and non-
units fluctuates continually because "every use of a structure reinforces it
and entrenches it more deeply, whereas non-use has the opposite effect"
(Langacker 1999: 100).
SCHEMA
THING PL
X -s
DOG PL TREE PL
dog -s tree -s
"most delicate grammar" (Halliday 1961: 267), its relation with grammar is
argued to take the form of a cline or a continuum (an idea which is also
found in Langacker's work) and the term lexicogrammar is used instead of
grammar.
Apart from the artificial distinction between lexis and grammar, the re-
lationship that has most often been misrepresented is the fundamental rela-
tion between lexicogrammar and semantics: grammar and semantics have
often been conceived as autonomous entities or separate 'components' of
language (Langacker 1987a: 12; see also Halliday 1988). A description of
language which is truly natural, however, can only be arrived at when the
relation between grammar and semantics is recognized to be natural or
non-arbitrary.
syntagmatic context of the linguistic sign and the relations that are estab-
lished through composition. In Section 5,1 will discuss the systemic context
of linguistic signs or their position "within the schematic networks that col-
lectively constitute the grammar of a language" (Langacker 1987a: 401).
to realize functions that are similar to those encoded in those three black
cats: it profiles a single instance of the schematically specified type 'human
female', is inherently definite, and its designatimi is identified as being dis-
tinct from the speech-act participants (Langacker 1991: 148). It is thus the
functional analysis of she which establishes it as a member of the class
'nomináis' and links it up with other nominal constructions. (From a
strictly structural perspective, she seems to have hardly anything in com-
mon with nomináis like those three black cats.)
The analysis of a construction thus essentially involves elucidating the
functions which it expresses (e.g., Haas 1954; Halliday 1994; McGregor
1997) and constructions can be interpreted as 'configurations of functions'
(Halliday 1994). In the following section, I will zoom in on the unique role
which functions play in the description and classification of constructions.
the noun class, then, the distinction between the subclasses of count and
mass nouns mainly depends on "whether the profiled region is construed as
being bounded within the scope of predication" (Langacker 1991: 18):
count nouns (e.g., lake) include the boundaries of what they designate in
their scope of predication, whereas mass nouns (such as water) do not.
The external, grammatical functioning of a symbolic unit and the highly
schematic semantic characterizations of some basic classes have weighed
heavily on discussions about grammatical classes. Interestingly, internal
properties have not: internally, the various instantiations of a particular
class are typically considered to be too diverse to be generalized across.
Nomináis, which vary from having a common noun as head, to being a
proper name or a pronoun, seem to be no exception to this. The internal
structural diversity that is characteristic of many classes has, in fact, explic-
itly been said to render classification on the basis of internal properties im-
possible. Halliday (1961: 261) thus states that a class is "not a grouping of
members of a given unit which are alike in their own structure": he (1961:
261) holds that, in terms of the rank scale, classes are derived 'from above'
and not 'from below'.
In what follows, an alternative approach to grammatical classes will be
presented which does build in the perspective from below. In a first part, I
will argue that a radically functional approach to classification includes the
internal functional organization of symbolic units, irrespective of whether
the internal functions are symbolized separately or not. A functional ap-
proach to grammatical class which includes the internal functions in its
analysis manages to identify symbolic units which are structurally widely
divergent as members of one and the same class. In a second part, I will
argue that classes which are functionally derived allow for a finer-grained
description of what can be said to constitute a prototypical instantiation of a
class.
(Haas 1954: 70). Until recently, it was difficult to see how such an analysis
could be envisaged of the nominal, in which, in view of the topic of this
study, we are particularly interested. While generalizations that deal with
the nominal's external functioning are fairly readily arrived at (as, for in-
stance, in Quirk et al. 1985), internal functional generalizations are not.
This is not only because of the apparent disparity of the units that can func-
tion in the nominal slots of a construction, but also, and more fundamen-
tally, because the question of the internal functioning of the nominal has
tended to be approached from an exclusively structure-based perspective. It
is only Langacker's (1991) analysis of the internal functional organization
of the nominal that makes clear that the various instantiations of the class
can be generalized across also internally. Langacker's analysis is pathfind-
ing, both descriptively and theoretically, and adopts a view of the nominal
which is very much 'radically functional' in that it recognizes that "seman-
tic function (rather than constituency) is the critical factor for understand-
ing their internal organization" (Langacker 1991: 51).
The semantic functions which, according to Langacker, are characteris-
tically realized by nomináis are those of type specification, instantiation,
quantification and grounding (Langacker 1991). The "universally valid
schematic characterization" which Langacker proposes for the nominal is
that it profiles "a thing construed as an instance of some type and further
incorporates some specification of quantity and grounding" (Langacker
1991: 54). Simple nouns provide nothing more than a type specification:
they specify "the basis for identifying various entities as being representa-
tives of the same class", but are "not tied to any particular instance of that
class" (Langacker 1991: 53). Full nomináis such as the site, an excellent
site and two convention sites in the Midwest, on the other hand, presuppose
instantiation of the type in question and designate one or more instances. In
full nomináis, "information is furnished concerning both the number of in-
stances and their status vis-à-vis the speech-act participants" (Langacker
1991: 53).
These semantic functions, Langacker (1991: 54) argues, are realized in
all nomináis, be it in many different ways. If they are coded separately in
the nominal's structure, then the head noun, together with its adjectives and
other modifiers that render it more precise, provides the type specification
(e.g., excellent convention site in the Midwest); the quantifier is added as a
separate layer (e.g., three black cats), and a grounding element is appended
at the outermost layer (e.g., those three black cats). Many nomináis, how-
ever, fail to display this type of assembly, in which case "the semantic
functions in question are not uniquely associated with distinct levels of
constituency" (Langacker 1991: 54). Apart from pronominal nomináis
26 Theoretical assumptions
(such as she, which was discussed earlier), proper names probably repre-
sent one of the most extreme cases of departure from the iconic type of
coding: in proper names, Langacker (1991: 59) proposes, "type, instantia-
tion, quantity, and grounding are conflated in a single expression". In the
proper name Stan, for instance, a type specification is incorporated (that of
'male human'), and because it characterizes a specific person, it presup-
poses instantiation and quantification. Grounding is subsumed as well, "for
the nominal is definite and portrays the profiled individual as being
uniquely apparent to the speaker and hearer on the basis of this name
alone" (Langacker 1991: 59). Because these functions are fulfilled, the pro-
per name qualifies as a nominal, and this in spite of its non-canonical cod-
ing (Langacker 1991: 59). As Langacker (1991: 53) concludes, the func-
tions of type specification, instantiation, quantification and grounding thus
allow us to "profitably examine the organization of nomináis from the
standpoint of semantic function, abstracting away from any details of struc-
tural implementation".
Classification by internal functional properties should be distinguished
from intrinsic or notional classification. Notional classification makes ref-
erence to abstract cognitive events and, to make sure that all members of
the class (including the less prototypical ones) instantiate the class schema,
it is necessarily highly schematic (Langacker 1987b: 54).7 As Langacker
(1991: 52) himself points out, the characterization 'thing or region in some
domain' applies to nominal predications "as a broadly defined class": it is
too abstract to distinguish simple nouns like site from full nomináis like an
excellent convention site, "each of which profiles a thing and qualifies as a
noun in this inclusive sense of the term" (Langacker 1991: 52). The intrin-
sic classification of nomináis is especially meant to distinguish them from
relational predications. The functional description in terms of type specifi-
cation, instantiation, quantification and grounding, in contrast, applies to
the internal organization of nomináis and helps to identify the differences
between nominal constructions and simple nouns (Langacker 1991: 53).
Classification in general, then, is intrinsic as well as functional, and the
functional criteria involve both the external functions served by the sym-
bolic unit and the internal functions of its components.
of some type', for instance, is applicable to all the members of the class
'nominal'. Put differently, all nomináis instantiate the same schema, which
"captures the pertinent generalization" (Langacker 1988: 130). All nomi-
náis are fully compatible with its specifications, but characterize them in
finer detail or "elaborate the schema in different ways along various pa-
rameters, to yield more precisely articulated notions" (Langacker 1987a:
68). By combining a schema with various instantiating structures, classes
form complex categories, or categories residing "in a family of structures
connected by categorizing relationships" (Langacker 1988: 149).
Among the categorizing relationships in a complex category, however,
Langacker distinguishes not only categorization by schema, but also cate-
gorization by prototype or extension. He (1987a: 371) views the two modes
of categorization as "inherently related and describable as aspects of a uni-
fied phenomenon", and his network model is a 'synthesis' of categorization
by prototypes and by schémas. Categorization by prototype offers a per-
spective on classification which is not so much alternative as complemen-
tary to that of categorization by schema. Both categorization by schema
and categorization by prototype can be said to involve comparing a usage
event or target structure with a conventionalized unit or standard. Instantia-
tion implies that all the specifications of the standard are satisfied by the
usage event. The standard is then viewed as a schema which only differs
from its instantiations in degree of schematicity or 'granularity'. This is
true of, for instance, the schematic characterization of nomináis as
'grounded instances of some type'. Extension of the standard, on the other
hand, implies that the target structure is only partially sanctioned by the
standard or deviates from it. The standard is then conceived as the proto-
type of the category.
It follows that a 'natural' approach to classification should not only look
for schematic generalizations, but should also take account of the occur-
rence of prototypical and less prototypical instances of a class (Langacker
1987a: 371). Instead of considering all members of a class as being fully
compatible with a schema, categorization by prototype judges class mem-
bership through a "perception of similarity" (Langacker 1987a: 69), which
is based on what is conceived to be a prototypical instance of the class. Pro-
totypicality thus results in "degrees of membership based on degrees of
similarity" (Langacker 1987a: 371; [italics mine]). The criteria used to set-
tle on the prototype of a particular class - and with them, the prototypes
themselves - are rather divergent. Semantic/intrinsic arguments can be dis-
tinguished from arguments that concern the structural implementation of
specific internal functions.
28 Theoretical assumptions
the element considered as Subject is the duke. Its role in the clause can be
characterized in various ways: the duke can be analyzed as the Agent of the
process with which it is construed (i.e. gave); it can be said to be "the con-
cern of the message" (Halliday 1994: 30), and it is that of which something
is predicated. Not all English Subjects, however, realize these three fea-
tures. Consider the following clause:
The Subject here is my aunt, but the Agent-role is realized in the fry-phrase
(i.e. by the duke) and the message concerns this teapot. The latter two fea-
tures - the role which a unit plays in the process designated by the main
verb and being the concern of the message - appear to be features which
can, but need not be, associated with the Subject: they can also be realized
by other components of the clause.9
What at first sight appear to be characteristic properties of a single unit
thus turn out to be basically functions on the clausal level. Rather than be-
ing properties or aspects of a specific unit (in this case, the Subject-unit),
they are distinct functions which are either mapped onto each other in a
30 Theoretical assumptions
It has repeatedly been pointed out by Halliday that this type of 'segmen-
tal' analysis of the clause does not "tell the whole story" (Halliday 1994:
36). For one thing, it captures the experiential metafunction more accu-
rately than it does the interpersonal and the textual metafunctions. In his
radically semiotic model of language, McGregor (1997) therefore argues
that syntagmatic relationships other than that of constituency underlie the
interpersonal and the textual metafunction. He proposes to view the inter-
personal metafunction, for instance, as a grammatical sign of which the
formal aspect is formed by the syntagmatic relationship of 'conjugation'.11
Conjugational relations are basically whole-whole relationships, which ei-
ther involve what McGregor (1997: 210) calls 'scoping' ("in which a unit
applies over a certain domain, leaving its mark on the entirety of this do-
main") or 'framing' ("in which a unit delineates the domain over which it
Constructions and the relations between them 33
applies, marking it off from everything else"). Μ the clause, the declarative
and interrogative moods are thus claimed to have 'scope' over the full
clause. An example of framing is formed by inter-clausal relationships
whereby one of the clauses expresses 'represented speech': like "a frame
around a picture, the representing clause delineates the represented clause
from its context... indicating that it is to be viewed or evaluated in a dif-
ferent way" (McGregor 1997: 253).
which accounts for part of the interpersonal choices made in the English
clause would be the one in Figure 2 (based on Halliday 1976: 93): it gives
an overview of the basic mood system of the clause and the various options
in it, each of which stands for a list of constructions. The option 'interroga-
tive, yes-no question', for instance, represents clauses such as Will you
close the door? and Did she climb that mountain?.
yes/no
• indicative
• interrogative
[
WH-
assertion
clause
• declarative
C
exclamation
jussive
• imperative
C optative
What do the notions of 'agnation' and 'enation' stand for? Gleason (1965)
proposes the terms 'agnation' and 'enation' to describe two basic kinds of
relationships which he distinguishes between constructions. The terms are
derived from Latin enatus 'related on the mother's side' and agnatus 're-
lated on the father's side' and they are deemed appropriate to designate
"two contrasting types of relations, neither of which can exist without the
other" (Gleason 1965: 199fn2). The relationship of enation links up con-
structions with "identical structures", that is, "the elements ... at equivalent
places ... are of the same classes, and ... the constructions in which they
occur are the same" (Gleason 1965: 199). The relationship between the
clauses in (3a) and (3b) is one of enation:
It is only because the clauses He heard it and He felt it, which are enates of
He saw it display a relation of agnation with It was heard by him and It was
felt by him, which are enates of It was seen by him, that the relationship
between He saw it and It was seen by him in (5) can be analyzed as being
one of agnation.
Likewise, relations cannot be identified as enate if they do not have
identical sets of agnates. Constructions such as The man saw a stranger and
The man seemed a stranger, for instance, are not enate because they belong
to different agnation networks: A stranger was seen by the man is possible,
while A stranger was seemed by the man is not. The man seemed to be a
stranger, on the other hand, is acceptable, while The man saw to be a
stranger is not (Gleason 1965: 203). The identification of agnate structures
is, in other words, an ideal heuristic tool for disambiguating constructions
which are seemingly identical (Davidse 1998a: 283). Because it is possible
that constructions share some agnates, but behave differently with respect
to others, Gleason also recognizes partial enation or enation between struc-
tures which are only partially structurally identical (which is elaborately
illustrated in Laffut 2000). As pointed out by Davidse (1998a: 283), it
seems consistent with Gleason's way of thinking to speak of a relationship
of non-enation when constructions are structurally non-identical. The rela-
tionship between The man saw a stranger and The man seemed a stranger
can then be labelled as being one of 'non-enation'.
In short, relationships of agnation and enation cannot be stated clearly
without one another: they are mutually defining notions that make up a
two-dimensional set to analyze constructions (Gleason 1965: 201). Most
importantly, they show that "formal evidence is not restricted to 'observ-
able characteristics' of the syntagm in question, but ... also includes the
syntagm's systematic relation to its agnates" (Davidse 1998a: 284). In more
Constructions and the relations between them 37
mainstream terminology, one would say that formal evidence includes the
syntagm's behaviour with regard to syntactic tests (Davidse 1998a: 284).
Gleason's claim that agnate and enate relationships form an integral part of
the analysis of a construction is theoretically significant in that it assigns a
central role to the paradigmatic or systemic aspect of language use. As
pointed out by Davidse (1998a), it has had a great impact on Halliday's
thinking, and even though Halliday has never theorized much about agna-
tion (Davidse 1998a: 287), his identification of system networks and,
linked to it, of the various metafimctions, is based on inter-constructional
relations. For one thing, the categories in his system networks essentially
represent the proportionality of relations between constructions. As Halli-
day puts it, a category's label is "no more than the name of a proportional
relation" between constructions (Halliday 1994: xxxii).
Proportional relations and, consequently, linguistic categories are de-
fined by agnation and enation (Davidse 1998a: 287): more precisely, they
are established as well as semantically interpreted by means of the two-
dimensional set of agnation/enation. To give an example, the categories in
the system network in Figure 2 represent constructions of the following
types:
Between constructions such as those in (7a) and (7c), situated on the hori-
zontal level of (8), there exists a proportional relation of agnation, which is
confirmed by the enate constructions on the vertical axis:
(8) You opened the door. :: Did you open the door?
She climbed the mountain. Did she climb the moun-
tain?
He washed the car. Did he wash the car?
digm on the right-hand side, it is the Finite that precedes the Subject. It is
this proportionality which is captured by the categories labelled as 'declara-
tive' and 'interrogative'. Importantly, these categories also derive their
'meaning' from it: linked to the declarative order of the mood element is
the interpersonal speech function of 'giving information', while the inter-
rogative involves the demand of information (Halliday 1994). On a inore
general level, the Subject-Finite unit can thus be said to serve a central role
in what is called the interpersonal function of language.
Particularly interesting, then, is that the inter-constructional relations
that underlie the categories in Halliday's system networks differ depending
on the metafunctional layer of language which they represent (Davidse
1998a): while interpersonal networks such as the one in Figure 2 map out
agnate structures (as shown in 7) and consist of categories "dealing with
clausal variants that are, by and large, applicable to any one clausal syn-
tagm" (Davidse 1998a: 290), agnation has a different use in system net-
works of the ideational domain. There, agnation is mainly used to distin-
guish non-enate structures: the categories of 'intransitive', 'transitive' and
'ditransitive' clauses, for instance, do not label proportionalities of agnate
structures, but of non-enate constructions (Davidse 1998a). Consider the
following ideational network (based on Halliday 1968):
clause
goal-directed (Actor - process - Goal)
e.g., John hit the ball
transitive
descriptive (Initiator - process - Actor)
e.g., John marched the prisoners
The non-enate status of the constructions John hit the ball and John
marched the prisoners, and, consequently, of the categories 'goal-directed'
and 'descriptive', is revealed through the different sets of agnates which
they can be related to (Davidse 1998a: 293). The construction John
marched the prisoners is part of the paradigm illustrated in (9):
As shown in (10), the construction John hit the ball does not belong to that
same paradigm:
and, ultimately, the system of language. When used as a heuristic for the
description of particular constructions, they seem to serve one of two pur-
poses: either they disambiguate structures which seem identical, but are
non-enate and they help to interpret the semantic differences between them
(e.g., John hit the ball vs. John marched the prisoners)·, or they reveal the
various construal types which a construction can occur in and indicate the
semantic impact of each construal type (e.g., He opened the door vs. Did he
open the door?). In any case, the relations which constructions hold with
other constructions seem to form an essential part of the 'family of struc-
tures' that constitutes a grammatical construction (Langacker 1988: 149).
In this chapter, I have set out the overall cognitive-functional view of the
language system in which my analysis of nominalization has to be situated.
Among the most prominent features of this cognitive-functional perspec-
tive, I have argued, are: its usage-based character (Section 1.1); the atten-
tion that is devoted in it to both schematization and extension or prototypi-
cality (Sections 1.2 and 1.3); its view of grammatical constructions as natu-
ral symbolic units which encode meaning (Section 2.2); its multifunctional
analysis of constructions (Section 4); and, finally, the interest which it
shows in the systemic or paradigmatic side of constructions and the impor-
tance which it attaches to inter-constructional relationships (Section 5). I
have also argued in favour of a radically functional approach to composi-
tion and to the construct of 'class' (Sections 3.1 and 3.3) and I have pointed
to the existence of a functional hierarchy or rank scale on which construc-
tions are situated and which identifies them as combinations of specific
external and internal functions (Section 3.2).
In the next chapter I will confront these theoretical insights with analy-
ses that have been suggested for nominalization. I will show how the cogni-
tive-functional framework set out in Chapter 2 not only enables us to iden-
tify the lacunae and weak points in the literature but also gives an indica-
tion as to how to move beyond them and set up a more coherent and sys-
tematic account of the constructional mechanisms that underlie nominaliza-
tion.
Chapter 3
Nominalization
1.2. Entrenchment
complexity is not relevant to lexicalization: lexical items can take the form
of morphemes, words, phrases and even longer expressions (Langacker
1991: 45; see also Halliday 1961: 274). Lexicalized constructions are to be
distinguished from grammatical patterns which are sufficiently entrenched
to be established as schematic units, but not as specific expressions. Gram-
matical patterns are either situated at word level (traditionally called 'mor-
phology') or at a level higher than the word (i.e. 'syntax').12 Lexicalizations
are thus not always simple words, and non-lexicalized items are not neces-
sarily larger than the word. Both for that reason, and because the degree of
entrenchment which determines the lexicalization of a unit fluctuates con-
tinually, the boundary between lexicon and grammar is not a sharp one, but
rather forms a continuum.
The distinction between lexicalized constructions and constructions
which are schematic rather than specific is highly relevant to the discussion
of nominalization patterns such as those in (5) and (6):
(5) Alex Gough, the Welsh destroyer of Del Harris's hopes on Tuesday
....(CB)
(6) ... a passenger steamer, which crossed the Channel escorted by a de-
stroyer. (CB)
The nominalizations in (5) and (6) show that, in a schematic network that
elaborates one and the same abstract schema (in this case, that of [[V/.. .]-
[[-ER]/[-er]]), some instantiations are lexicalized, while others are not.
Whereas destroyer in (6) forms a lexicalized nominal and has itself
achieved the status of a conventional unit, destroyer in (5) is what I will
call an 'ad hoc' nominalization: it has been formed according to the same
high-level schematic symbolic unit [[V/...]-[[-ER]/[-er]] as has the lexicali-
zation destroyer, but it is not included in the language system as a particu-
lar, fixed expression.13
As to the productivity of particular schematic units - i.e. the likelihood
of their being activated for the construal of novel units - Langacker (1991:
48) remarks that low-level schemata are more likely to be invoked than
more abstract schemata. He (1991: 48) argues that evidence in favour of
this claim comes from the fact that instantiations of a schema are normally
concentrated in certain regions which correspond to low-level schemata,
rather than being "distributed randomly through the space of possibilities
defined by the highest-level schema". Langacker's claim that it is low-level
rather than high-level schemata which determine the production of novel
expressions is comparable to what is traditionally referred to as analogy or
the "direct modelling of novel expressions on the basis of familiar ones"
46 Nominalization
(Langacker 1999: 144). It differs from analogy in its emphasis on the role
of schematization: "even the learning of specific expressions (required as
the basis for analogy) involves abstraction and schematization from actual
usage events" (Langacker 1999: 144).
The low-level nature of productivity which Langacker describes can be
clearly observed among the agentive and the non-agentive instances of -er
nominalization. Both are characterized by certain 'regions' of productivity
or low-level schemata which give rise to instantiations that are sufficiently
entrenched to be established as units. Agentive regions are, for instance,
those of agent-like instrumental -er nomináis (e.g., computer, printer,
parser), 'second-order' Agents (as inpageturner 'a book that almost forces
you to turn the pages' and yawner 'an event that makes you yawn') and
'professional' Agents (such as programmer and fire-fighter). Non-agentive
regions of productivity are especially those that designate non-agentive In-
struments (e.g., front-loader, stroller, stepper, walker) and items of food
(e.g., broiler 'a chicken suitable for broiling', cooker 'an apple for cook-
ing', sipper 'a drink that you have to sip', dipper 'vegetable/fruit/other
snack that has to be dipped before being eaten').
In spite of the fact that nominalizations are symbolic units which encode
their meaning, research on nominalization patterns often seems to consider
the analysis of lexicogrammatical properties as an end in itself and only
rarely attempts to link it up with the semantics of the construction. Certain
lexicogrammatical features moreover tend to be - erroneously - considered
as 'meaningless' and are therefore frequently left unanalyzed.
The tendency to leave things at a detailed account of a construction's
lexicogrammatical properties only is perhaps best illustrated by the follow-
ing nominalizations:
(7) And it's the third anniversary of the opening of East Germany's bor-
ders. (CB)
(8) Opening a book is like opening the door into a wonderful new world.
(CB)
(9) I still regret opening the door so suddenly, (personal example)
dive nominal referring to facts'. In spite of the fact that he gives an excel-
lent description of the lexicogrammatical characteristics of each of these
types of nominalizations, Lees does not succeed in tracing the origin of the
nominalizations' semantic differences in them. Since Lees (1960), numer-
ous analyses have been devoted to these nominalization types, some of
which significantly elucidate certain of their properties and the meanings
associated with them (e.g., Kiparsky and Kiparsky 1971; Schachter 1976;
Halliday 1985, 1994; Langacker 1991; Davidse 1991). Yet, it remains un-
clear why, for instance, the nominalization opening the door into a wonder-
ful new world in (8) designates what Lees calls an 'action', while opening
the door so suddenly in (9) is factive in meaning. Likewise - and in spite of
the fact that their lexicogrammatical differences have been extensively de-
scribed - the differences in meaning between the 'action' nominal in (7)
and the 'gerundive nominal' in (8) remain to be clarified.
One of the reasons why the semantic distinction between nominaliza-
tions like the ones in (7) and (8) has not yet been fully accounted for, is that
the semantic import of the grammatical marker of in (7) has tended to be
ignored (Langacker 1991: 35): along with the 's of Zelda 's in a nominaliza-
tion like Zelda's signing of the contract, o/has often been assumed to be
'empty' and inserted for purely grammatical purposes (Langacker 1991:
35). And even if grammatical markers like's and of are not explicitly ar-
gued to be meaningless, they are not taken into account in the analysis of
the nominalization in which they function. The analysis of nominalization
offered in Halliday (1994) and Halliday and Matthiessen (1999), for in-
stance, ignores the presence of periphrastic markers by relating nominaliza-
tions such as the writing of business programs and writing business pro-
grams to the same clausal agnate, i.e. to (people) write business programs.
However, grammatical markers such as the and of form part of the lexico-
grammatical properties of nominalizations and motivate a facet of their
composite structure. As Langacker (1991: 35) points out, they "invariably
have some kind of conceptual import, which may be quite abstract but is
nonetheless essential to their function". A natural account of nominaliza-
tions has to take account of all lexicogrammatical properties, including
those the import of which at first sight seems negligible.
Halliday speaks of downward rank shift, when "an item normally having
the function of (entering into the paradigmatic and syntagmatic relations
associated with) rank χ characteristically 'loses' these functions on taking
over those of rank y" (Halliday 1966b: 114). A clause which is rankshifted
to phrase level (Halliday 1961: 253 speaks of 'group' level) thus "cannot
enter into direct syntagmatic relations with clauses outside the structure of
that group" (Halliday 1966b: 114). As Halliday defines it, rank shift basi-
cally involves a change in the external functions of a symbolic unit: a
50 Nominalization
clause which is rankshifted to phrase level, for instance, acquires the exter-
nal functioning potential of phrases (as is shown, in, for instance, [That he
committed the crime] surprised us, where the that-clause has come to func-
tion as Subject of the verb surprised). Halliday (1961: 261) concludes that
"by reference to the rank scale, classes are derived 'from above' (or
'downwards') and not 'from below' (or 'upwards')".
Because rank shift involves "a change in part-of-speech membership"
(McGregor 1997: 128), McGregor defines downward rank shift as reclassi-
fication. Most importantly, and against Halliday, he holds that rank shift
involves external as well as internal reclassification. He (1997: 127) de-
fines rank shift as "the process whereby a unit of a given rank is as it were
demoted in size, and reclassified as a unit of lower rank, as a result of
which it takes on the grammatical and semantic properties inherent to the
lower ranking unit". He (1997: 128) further points out that, semantically
speaking, the reclassification of, for instance, a clause into a nominal unit
implies that the clausal meaning is 're-presented' "as a thing - a second-
order entity - rather than an ongoing phenomenon located at some point in
space and time". In addition, McGregor emphasizes that there must be
some formal registration of the rankshifted status of a unit and he argues
that rankshifted clauses are characterized by a reduced number of gram-
matical options that are available. Among the options that are no longer
available are "those which permit the anchoring of the situation to a spe-
cific occurrence in the referent world" (McGregor 1997: 130). From this
premise, McGregor (1997: 131) draws the conclusion - wrongly in my
opinion - that finite clauses cannot be rankshifted or nominalized, whereas
non-finite clauses can. (A similar view is expressed in Dik 1997.) Finally,
McGregor (1997: 130) states that, apart from having a reduced structural
potential, the embedded clause may also acquire morpho-syntactic proper-
ties that are normally associated with nomináis.
In my opinion, McGregor rightly points to the importance of including
the internal properties of rankshifted units into their description and classi-
fication. Yet, he fails to clarify their significance to the analysis of rank-
shifted units, because he only considers overtly realized structural proper-
ties, instead of taking in a functional perspective. By focusing on the
clausal structural properties of the downranked clause only, McGregor fails
to observe and account for its nominal functioning. This results in inexpli-
cable discrepancies between the external (nominal-like) functioning of par-
ticular rankshifted units (e.g., as Subject of a clause as in [That he was the
one that killed her] surprised us most) and the internal status of these units,
which McGregor describes as non-rankshifted. McGregor's analysis also
fails to clarify the correspondences in meaning and in external functioning
Nominalization as composition and reclassification 51
can relate "two constructions which are not on the same grammatical le-
vel", as is the case with the clause the boy runs and the nominalization the
boy's running (Gleason 1965: 212). In fact, relationships of agnation, ena-
tion and non-enation have always figured in descriptive accounts of nomi-
nalization patterns (and, indeed, of structural patterns in general: the central
idea underlying the construct of 'transformation' in Transformational
Grammar was that of capturing systematic relationships between different
structures). In Lees's (1960) Grammar of English Nominalizations, for in-
stance, it is thus relationships of agnation which underlie the transforma-
tional relation between nominalized structures and 'deep structure' clauses
(Lees [1960] 1968: 32, 65). Nominalizations and the clauses they are said
to be transformationally derived from are agnates in that they have the
same major lexical items, but differ in structure:
(12) his rapid drawing of the picture <=> He drew the picture rapidly.
(13) his bringing up the box » He brought up the box.
(14) John is the signer of the check. o John signs the check.
Notice that the 'tests' which Lees uses to identify the differences in struc-
tural behaviour of action nomináis (e.g., his drawing of the picture) and
gerundive nomináis (e.g., his drawing the picture) are likewise based on the
assumption that constructions can only be analyzed in more detail if their
systemic or paradigmatic relations are considered: in the examples in (15)
and (16), for instance, the paradigmatic relations of Aw and of drawing as
they function in his drawing of the picture and his drawing the picture are
considered (Lees 1968: 65-66):
As I pointed out in Chapter 2 (Section 5.1), Gleason does not view agnation
as "a unique relation between isolated pairs of sentences", but argues that it
is based on the 'pervading patterns' of language and that it is "therefore
always a recurrent thing, involving large numbers of sentences" (Gleason
1965: 202). Gleason has also pointed out that one agnate structure high-
lights a particular grammatico-semantic feature of a construction, but does
not uniquely identify the structure which it agnates with. A detailed picture
of the grammatico-semantic choices that are realized by a particular struc-
ture can thus only be provided by agnation networks, in which each agnate
structure sheds light on one or more of the structure's properties.
Gleason's understanding of agnation uncovers one of the central weak-
nesses of the transformational-generative and the systemic-functional ap-
proaches to nominalization, viz. their being focused on one - necessarily
clausal - agnate.15 Clausal agnates are very important to the analysis of
nominalized structures, but other, non-clausal agnates may contribute to the
identification of the lexicogrammatical properties of a nominalization as
well. The focus on one agnate structure which is clausal in nature tends to
lead to incomplete or even inaccurate analyses. For instance, in Halliday
and Matthiessen's (1999) account of nominalization, both a structure such
as Tom's cleaning the kitchen and Tom's cleaning of the kitchen would be
related to the clause Tom cleans the kitchen. This, however, fails to identify
the relevant features of the nominalizations it is related to and needs to be
accompanied by a set of other agnates to distinguish the gerundive nominal
Tom's cleaning the kitchen from the action nominalization Tom's cleaning
of the kitchen}6
Agnation and nominalization 55
As noted in Chapter 2 (section 5.1.), Gleason (1965: 203) shows that ag-
nates can crucially be used to disambiguate apparently identical structures.
This also applies to the disambiguation of nominalizations. Consider, for
instance, the nominalized structures in (17) and (18):
At first sight, the structure not going to school in (17a) is identical to that in
(18a). Yet, if one considers the set of agnates that is related to not going to
school in (17a) and in (18a) (illustrated in 17d, 17e, 18d and 18e), as well
as the paradigmatic options of their component functions (illustrated in 17b,
17c, 18b and 18c), it becomes clear that they are different structures:
I have pointed out in Section 3.1 that Langacker's analysis of the order of
assembly of this nominalization is essential to distinguish it from the action
nominalization Sam's washing of the windows. Langacker (1991: 32) has
pointed out that the gerundive nominalization in (19) nominalizes a
"higher-level structure" or "a processual expression that has all the ingredi-
ents of a finite clause except an explicit subject and a grounding predica-
tion". The fact that this 'higher-level' structure combines washing and the
windows into one component unit of the assembly Sam's washing the win-
dows, which is then premodified by Sam's, is confirmed by agnation. Was-
hing the windows agnates, for instance, with wash the windows as it is used
in the constructions in (20), (21) and (22):
The clauses in (20), (21) and (22) illustrate that, also within clauses, a verb
(minus the grounding elements of tense and modality) and its Object(s)
combine into a separate component unit (Halliday 1994). The clause-level
agnates thus provide evidence in favour of the claim that the order of as-
sembly described in Langacker (1991) is the correct one and that washing
the windows constitutes a component unit in the assembly of Sam's was-
hing the windows.
In this final section, I put forward my view that agnation is best thought of
as involving degrees of schematicity·. relationships of agnation between
specific examples of constructions always involve relations between the
schemata which they instantiate. I also argue that agnate relationships bet-
ween nominalizations and clauses reveal the importance of clausal catego-
ries to the analysis of nominalized constructions.
Agnation and nominalization 57
The overall function of a nominal is to refer to a thing and make it "a mo-
mentary focus of attention" within the speech event (Langacker 1991: 53).
To single out the thing one wants to talk about - to the exclusion of other
things - one can basically employ two strategies. Either one provides the
thing with its own, unique label and uses a proper name·, or one follows a
strategy based on type specifications which one then specifies in terms of
62 The functional organization of nominal and clause
quantity, as well as in terms of their relation to the speech event and its par-
ticipants (the ground) (Langacker 1991: 53). It is by being grounded in the
speech event that the instance of the type specification that one wants to
talk about becomes uniquely apparent to the speech participants. However
different the latter, common noun strategy may seem from that based on
proper names, both strategies give rise to nomináis that conform to the
schematic characterization of the nominal as a 'grounded, quantified in-
stance of some type'. They only differ in the "structural implementation of
these semantic functions" (Langacker 1991: 148): unlike proper names,
nomináis with a common noun as head tend to reflect the semantic func-
tions which they realize iconically, with the type specification as the in-
nermost layer of organization, the grounding predication as the outermost
layer and intermediate between the type specification (or simple noun) and
the grounded type (or nominal), the functions of instantiation and quantifi-
cation (Langacker 1991: 54).
Which is the contribution of each of the semantic functions which Lan-
gacker (1991) distinguishes to the ultimate grounded status of the thing des-
ignated by the nominal? And how are these functions encoded in the struc-
ture of the nominal? Both formally and semantically, the type specification
constitutes the innermost functional layer. Semantically speaking, it makes
"an initial delimitation among the potential objects of thought" and "speci-
fies the basis for identifying various entities as being representatives of the
same class" (Langacker 1991: 53). In the nominal my office, for instance,
the noun office specifies the type or class of entities that is being talked
about and thus forms the representational core of the nominal (Halliday
1994: 189). Structurally speaking, it is the head noun which provides the
type specification, along with any adjectives or other modifiers which ren-
der the type specification more precise and add certain refinements: in the
nominal an excellent convention site, the head noun convention site is thus
rendered more specific by the inclusion of excellent. Postmodifying struc-
tures can also add to the type specification, as illustrated by in the Midwest
in the nominal an excellent convention site in the Midwest (Langacker
1991: 52). Together with the head noun, these modifiers create a 'higher-
order type specification' which forms a kind of nucleus within the nominal
and which typically "supplies a vast amount of detailed conceptual content"
(Langacker 1991: 143).
A nominal type specification is instantiated when it is thought of as
"having a particular location in the domain of instantiation" (Langacker
1991: 57). For nomináis, the typical domain of instantiation is space. The
instantiation of a type specification implies that one conceives the domain
of instantiation as being able to support the simultaneous manifestation of
From noun type specification to nominal 63
guishes between absolute or true quantifiers, on the one hand, and relative
quantifiers, on the other. Absolute quantifiers "offer a direct description of
magnitude" (Langacker 1991: 83). Examples are seven, many, few, little.
Relative quantifiers specify a quantity in relation to a 'reference mass',
which in the default case consists of the "maximal instantiation of the per-
tinent category" (Langacker 1991: 82). A relative quantifier such as all in
all dogs, for instance, profiles a quantity that is identical to the reference
mass. In most dogs, in contrast, the profiled mass only constitutes a subpart
of the reference mass, be it one "that comes reasonably close to exhausting
it" (Langacker 1991: 82). Because they situate the instances which they
quantify relative to a definite reference mass, relative quantifiers also serve
a grounding function. This explains why they cannot be used with other
grounding predications (e.g., *the all dogs, *those most dogs). Absolute
quantifiers do not imply a reference mass: the nomináis many dogs, several
dogs and seven dogs give a (vague or precise) indication as to the quantity
of dogs involved, but they do not situate this quantity relative to a reference
mass. Absolute quantifiers do not therefore serve a grounding function and
can be preceded by a grounding predication (e.g., those three dogs, the
many cats I have owned) (Langacker 1991: 82-83).
Langacker argues that every nominal contains some indication of quan-
tity, either in absolute terms (e.g., three cats) or proportionally (e.g., most
cats). He does not, however, elaborate on the quantificational information
provided by those nomináis which leave the size of the profiled instance
implicit or characterize it only 'schematically' (Langacker 1991: 89), such
as proper names and nomináis with articles, zero-determiners, demonstra-
tives and possessives. It is interesting in this respect to consider the analysis
which Milsark (1976, 1977) gives of determiners according to the type of
quantification that they can express (Davidse 1999: 216-217). Milsark's
analysis is to be situated in the context of a systematic distributional analy-
sis which he makes of the use of determiners in the existent NP of existen-
tial constructions, more precisely in the unmarked 'cardinal' type of exis-
tential such as There are five weak spots in the human body (for an exten-
sive discussion of existentials, see Davidse 1999). Milsark distinguishes,
firstly, 'cardinality words' which "express the size of the set of entities de-
noted by the nominal with which they are construed" (Milsark 1977: 23).
Cardinal determiners can be used in the existent NP of cardinal existentials
and they include, apart from the quantifiers which Langacker (1991) classi-
fies as absolute, also the indefinite article a, the zero-determiner with plural
and mass nouns in non-generic reading and, finally, cardinal numbers. If
determiners do not give a cardinal value themselves, Milsark argues, they
"must always be understood with reference to a set of some cardinality"
From noun type specification to nominal 65
(Milsark 1977: 23), i.e. they are what have been called 'relative' quantifi-
ers. The relative group of determiners comprises the definite article, bare
plural and mass nouns with generic reading, pronouns, proper names, pos-
sessive and demonstrative determiners, genitives and various relative quan-
tifiers, such as all, every, both, most (Davidse 1999: 217). The latter group
of determiners is excluded from the cardinal existential construction.
Davidse (1999) argues that Milsark's analysis of the quantifying proper-
ties of determiners can be integrated with Langacker's analysis of the
nominal as 'quantified instance of some type'. The distinction between ab-
solute (or cardinal) and relative or proportional quantification can, in other
words, be extended to all nomináis, irrespective of whether they contain an
explicit quantifier. Some of the analyses which are proposed in Davidse
(1999) are the following. With regard to non-generically used bare mass
nouns and plural count nouns (as in 1 and 2), Davidse argues that they real-
ize absolute quantification (Davidse 1999: 222), because they alternate sys-
tematically with nomináis introduced by absolute quantifiers such as some,
much and many (illustrated in 3 and 4):
Generic bare plurals and mass nouns (as in 5 and 6), in contrast, are
claimed to receive a relative quantificational reading (Davidse 1999). They
resemble proper names in that they designate unique instances. Or, as Carl-
son (1978) puts it, they are proper names of kinds conceived of as individu-
als. Unlike proper names, however, the instances which they designate are
instances in type space and not in physical space: the nomináis koalas and
laurel do not refer to specific, spatially-instantiated animals or plants.
Rather, they have to be understood with reference to a type universe, in
which they constitute unique individuals (for a more detailed account, see
Davidse 1999: 209-215). They realize relative quantification because they
"quantify universally, as it were, over the one instance contained by the
reference mass in type space" (Davidse 1999: 225).
With the size of the profiled instance specified, it remains for the
speech-act participants to establish contact with the instance: the objective
of the speaker in using a nominal is to refer to an instance of a type which
the hearer can identify. The speaker has to make the profiled instance into
something that can be exchanged in communication, something that can be
'talked about'. As Langacker (1991: 91) puts it, "the speaker (S) and hearer
(H), who jointly form the ground (G), face the task of coordinating their
mental reference to some instance t¡ of type Τ drawn from the reference
mass RT". Within the system of identification, the basic options are definite
vs. indefinite identification. Indefinite identification presents instances
which are not presumed known to the hearer; definite identification marks
an instance as 'presumed known'. Definite identification occurs when "full
coordination of reference" between speaker and hearer is achieved (Lan-
gacker 1991: 91).
As pointed out in Davidse (2001: 8), the functions of identification and
quantification have not always been properly distinguished: it is thus un-
clear how the definite-indefinite distinction applies to the various determin-
ers that are found in the nominal, and how this distinction relates to that
between relative and absolute quantifiers. Davidse (2001) suggests viewing
all definite reference as implying relative quantification. She (2001: 9) ar-
gues that "speaker and hearer can be thought of as coordinating their men-
tal reference to P, the actually designated mass, which, through the opera-
tion of definite reference, is delineated as an identifiable set or mass".
Unlike indefinite reference, definite reference delineates or delimits the
portion of the reference mass which the hearer should make mental contact
with. Note that this definition of definite reference is compatible with that
by Langacker (1991) in terms of 'full coordination of reference to some
instance t, of type Τ drawn from the reference mass Rf '; it only differs from
it in that the mechanism of relative quantification is brought into the defini-
tion and is identified as a prerequisite for definite identification. Davidse
(1991: 9) thus takes over the schematic representation of definite reference
in Langacker (1991: 92) and only changes the profiled mass t¡ into a Ρ re-
sulting from a presupposed operation of relative quantification:
From noun type specification to nominal 67
Domain of instantiation
enee puts an instantiation on the scene but is "not sufficient to put the
hearer in mental contact with a uniquely determined instance of the cate-
gory". In the example Hand me a wrench (Langacker 1991: 104), for in-
stance, the speaker tells the hearer 'what sort of thing' he wants, but a
wrench does not delimit a set of one specific wrench. In that the hearer has
to be able to recognize an instance corresponding to the type specification
used by the speaker, indefinite reference presupposes 'type-identifiability'
(Gundel et al. 1993: 275, as cited in Davidse 2001: 13). Besides represent-
ing the instantiation relation between t¡ and T, Figure 5 also indicates that
speaker and hearer both have to make mental contact with the type T.
It is thus not only determiners that contribute to the reference act: the
type specification makes an essential contribution to it as well: as Davidse
(2001: 12) points out, "successful reference is dependent as much on a suf-
ficiently informative and contextually well-positioned 'categorization' as
on the appropriate choice of determiner". In fact, the type specification
turns out to play a different role with definite and indefinite identification.
In nomináis which are introduced by a definite determiner, Davidse (2000b,
2001) argues, the type description defines a 'reference mass' of 'all the in-
stances corresponding to that type in the discourse context'. The following
examples are given to illustrate this point:
(9) When God asked why they were wearing clothes, it was the man who
spoke first.
(10) Here are the exams that still need correcting.
(11) If you take care of your exams that still need correcting, I'll take care
of mine.
In (10) and (11), the contextually relevant reference mass is 'all exams that
still need correcting'. While the definite article in (10) refers to all of those,
the possessive your in (11) picks out only some instances. In (9), the con-
textually relevant reference mass (in the context of the Garden of Eden)
consists of only one instance of the type 'man'. The use of the definite arti-
cle to refer to 'man' is therefore correct.
Indefinite nomináis, then, merely designate "'some' instantiation of a
type, without indicating whether there are - or aren't - other instances of
the type in the discourse context" (Davidse 2000b: 1113). The type specifi-
cation has a purely "classificational" function (Davidse 2000b: 1113). The
designated instance may be presented as indefinite, but specific (as in 12):
the instance is then being introduced into the discourse for the hearer's
benefit. Or no specific instance may correspond to the indefinite nominal
for either speaker or hearer (as in 13) (Davidse 2000b: 1113):
From noun type specification to nominal 69
(12) I had only just arrived when a man suddenly entered the restaurant
and started shooting.
(13) A tie is always an appropriate present for a man who is rather con-
servative.
The type specification, Langacker (1991: 149) argues, is in this case pro-
vided by the proposition conveyed by the nominalized clause. Because the
clause is finite and uniquely related to the time and the participants of the
70 The functional organization of nominal and clause
speech event, the type specification which it implies has only a single in-
stance which is definitely grounded.
Other nomináis consist of two structures which are themselves nomináis
and which are in apposition: the two components have "equal claim to the
status of profile determinant" and the resulting, composite structure is a
"doubly-characterized, doubly-grounded thing" (Langacker 1991: 149).
Examples of such appositional structures are the fact that whales are mam-
mals·, all those kittens and my daughter the lawyer (Langacker 1991: 149).
Still other nomináis consist of a grounding predication only (e.g., that,
these, some, any, all). They can be said to incorporate a type specification,
in spite of the fact that they lack a head noun. A nominal such as these thus
schematically characterizes the entity which it grounds as being a replicate
mass or plural; each, on the other hand, incorporates the type specification
of a discrete entity (Langacker 1991: 151).
In short, the semantic functions of type specification, instantiation,
quantification and grounding can be encoded in many different ways and
they do not always receive individual symbolization. Yet, they are realized
in every nominal and together, they serve to turn a 'thing' into something
which is related to the speech event and its participants, something which
one can zoom in on and talk about. Crucial for the analysis of nominaliza-
tions is Langacker's claim that the "relation between a simple noun and a
nominal is directly analogous to that between a simple verb and a finite
clause" (Langacker 1991: 191). In the next section, I will go more deeply
into the functional organization of the clause. Again, the analysis that is
presented is based on Langacker (1991). More than in the description of the
nominal, however, a number of complementary or alternative analyses that
have been proposed for certain aspects of clausal assembly are included as
well. The latter are especially based on Halliday (1994), Davidse (1992b,
1997), Heyvaert (2000) and Verstraete (2002).
The pivotal element in the clause is the verb. It may stand on its own (as in
15a), or it may form part of a more elaborate kind of verb group with a
varying number of auxiliaries (as in 15b and 15c):
Langacker (1991) argues that, on the basis of semantic function and gram-
matical behaviour, the verb group falls apart into two parts, viz. into a
grounding predication and a type specification. Importantly, this distinction
does not coincide with that between the main verb and its auxiliaries.
Rather, Langacker suggests "that the specification of tense and modality be
analyzed as the grounding predication, with the remainder of the verb
group (other auxiliaries and the main verb) regarded as a complex clausal
head analogous to a head noun" (Langacker 1991: 195). For the verb
groups in (15), this leads to the following analysis (with the grounding
predications in italics):
Semantically speaking, tense and modality seem to stand apart from the
rest of the verb group in that they are "the only auxiliary elements that spe-
cifically invoke the ground as a reference point" (Langacker 1991: 195):
tense relates the process to the time of the speech event, while the presence
of a modal explicitly situates the profiled relationship with respect to the
judgement of the speaker. A structural argument in favour of the proposed
division comes from the fact that an infinitive (marked by to) can be
formed on the main verb together with any combination of auxiliaries, mi-
nus the markers of tense and modality. It is thus possible to have to clean,
to be cleaned, to have cleaned, to have been cleaning. It is not possible to
do the same with a verb group which is marked for tense or modality, e.g.,
*to should wash, *to had been washing (Langacker 1991: 195). There is, in
other words, a distinction that cuts right through the auxiliaries of the verb
group and distinguishes those auxiliaries that serve to ground the process
type from all the others. Together with the main verb, the latter, non-
grounding auxiliaries make up the clausal head. In what follows, I will suc-
cessively discuss the properties of the clausal head (Section 2.1), the way in
which it is instantiated and quantified (Section 2.2), and the means that ex-
ist of grounding it (Section 2.3). Both in the instantiation and in the ground-
ing of a process type specification, the Subject will be argued to play an
important role. A third function of the clausal Subject, viz. that of 'speech
functionally responsible' element in the clause (Halliday 1994), will be de-
scribed in Section 2.4.
72 The functional organization of nominal and clause
Langacker argues that the head-part of the verb group, which provides the
process type specification of the clause, may be quite complex internally:
like the head in nomináis, the essential semantic content provided by the
main verb (e.g., clean) may be elaborated with any desired degree of preci-
sion by the non-grounding auxiliaries with which it is combined (Lan-
gacker 1991: 196). Each auxiliary element is said to impose its profile on
the structure it combines with, "thereby deriving a distinct higher-order
type specification" (Langacker 1991: 197). It is, however, only "the left-
most verb in the sequence" (Langacker 1991: 196) which determines the
profile of the entire verb group. If a verb group is to function as clausal
head and provide a clausal type specification, Langacker (1991: 199) ar-
gues, its leftmost verb must profile a 'process', or a series of component
states which are distributed continuously through time and scanned in a
sequential way (Langacker 1991: 21, 246). The function of the auxiliaries
have (present perfect), be (progressive aspect) and be (passive voice) is
precisely to impose a processual profile on the atemporal relation that is
formed by the affixal/inflectional pair members -ed (perfect aspect), -ing
(progressive aspect) and -ed (passive voice) (Langacker 1991: 199). More
particularly, the "members of a given pair must co-occur if the resulting
expression is to be processual and thus able to serve as clausal head" (Lan-
gacker 1991: 200). Examples of clausal heads which illustrate the use of
specific auxiliary pairs are the following:
(17) a. They must have known it all the time, (perfect aspect)
b. She will be singing in several operas this year, (progressive as-
pect)
Similar to grounding predications, the -ing suffix and the to-infinitive can
be used with both so-called 'perfective' and 'imperfective' verbs. Imperfec-
tive verbs profile processes with identical component states (e.g., know —>
knowing/ to know). Perfective verbs profile some kind of change due to the
differences that exist among the component states (e.g., hit -» hitting/ to
hit) (Langacker 1987a: 254-267).
I propose that the to-infinitive and the -ing forms in (19) represent
clausal heads that have been atemporalized. They constitute an alternative
to grounding the clausal head by tense or modality. Significantly, ground-
ing a clausal head and atemporalizing it are mutually exclusive choices.
Atemporal markers cannot therefore be added to a grounded structure, nor
can an atemporalized structure be grounded:
However, if the atemporal -ing suffix is elaborated into a structure with be,
thus forming the auxiliary pair be V-ing, a new processual structure or
clausal head is formed and grounding becomes possible again: when, for
instance, being dropped is turned into be being dropped, it can be grounded
in the speech event, e.g., is being dropped.
Can atemporalized clausal heads effectively figure as head of the clause,
and which is their semantic import? Like clausal heads that have been
grounded in the speech event, atemporalized clausal heads can serve the
function of head of the clause. As with finite clausal heads, a varying num-
ber of other clausal components (e.g., Subject, Object, Complement) tends
to accompany them. In the non-finite clause in (21), for instance, the apples
functions as Subject, and the ¿y-phrase realizes the Agent of the process. In
(22), then, the atemporal clausal head is accompanied by an adverbial {con-
scientiously) and an Object (the woods he loved).
From process type specification to clause 75
(21) Those who ate the apples ran off wild and mad into the forest - the
apples having been poisoned by a woman who had loved Merlin ....
(CB)
(22) There his father ... lived the life of a recluse, conscientiously tending
the woods he loved. (CB)
Observe that the clauses in which atemporal clausal heads function are nec-
essarily subordinate, i.e. they cannot stand alone as independent clauses. I
will come back to this briefly in Section 2.3.3.
As to their semantic import, I hold that atemporal clausal heads basi-
cally represent a different way of looking at the process type specification.
They do not ground it or instantiate it, but preserve its status as type speci-
fication. The change of processual structures such as play or have been
playing into the atemporal structures playing or having been playing, for
instance, does not instantiate them.21 It does, however, change the way in
which the component states of the verb type are scanned: rather than the
step-by-step, serial way of scanning which is typical of processual struc-
tures (such as kill, have killed, be killing) and finite clauses, atemporalized
clausal heads offer a holistic view and activate the component states of a
verb cumulatively, through summary scanning (as defined in Langacker
1991: 554). They thus present the component states of a verb as "a single
gestalt" (Langacker 1991: 21). This is true for both atemporal clausal heads
in -ing and for those that take a io-infinitive.
It is beyond the scope of the present study to examine the precise se-
mantic differences that exist between -ing and to-infinitives as they func-
tion in clausal heads. It is important, though, to point out that the -ing of an
atemporal clausal head should not be confused with progressive -ing. Pro-
gressive -ing has been characterized as "restricting its profile to a series of
component states that does not include the initial and final states" and as
rendering a verb "imperfective" in that its component states "are construed
at a level of schematicity that neutralizes their differences" (Langacker
1991: 209). Verbs that are clearly imperfective cannot therefore take pro-
gressive -ing. As pointed out before, however, the -ing of atemporal clausal
heads can be used with perfective as well as imperfective verbs (e.g., know-
ing the truth, having been there) and the clausal head which it forms can
itself contain a progressive (e.g., having been washing; having been playing
in the garden). As Langacker has moreover pointed out to me, an atemporal
clausal head such as eating that whole chocolate cake in [Eating that whole
chocolate cake] was a big mistake does not take an internal perspective on
the atemporalized process (i.e. one that excludes its endpoints). Rather, the
atemporal clausal head designates the entire event of eating the cake. In
76 The functional organization of nominal and clause
that respect, the -ing of atemporal clausal heads also differs from construc-
tions in -ing that follow verbs of perception (e.g., We saw [the ship sink-
ing], We heard [him yelling]) in which -ing can be attributed the same
value that it has in the progressive (i.e. it excludes the endpoints of the
event and offers an 'inside perspective') (Langacker 1991: 443).
To conclude, then, the process type specification which forms the core
of the English clause is either grounded in the speech event, or it is ex-
cluded from grounding by being explicitly atemporalized. In the latter case,
the process type specification remains ungrounded. Atemporal or un-
grounded clausal heads need not, however, remain uninstantiated: in what
follows, I will argue that atemporal clausal heads, like all clausal heads, are
instantiated by being combined with a Subject. Crucially, the role of the
clausal Subject should be distinguished from that of the Object, which is
primarily type-specifying: the clausal Object forms part of the clausal head,
which the Subject serves to instantiate. I will first elaborate on the type-
specifying role of the clausal Object (Section 2.1.2), and will only in Sec-
tion 2.2. turn to the instantiating properties of the Subject.
We have thus far identified the semantic function of the Residue as be-
ing that of clausal head, providing the type specification within the clause.
In the next section, I will go more deeply into the question of how instan-
tiation and quantification of the clausal head are realized.
having written which is marked on the clausal type itself. This type of sec-
ondary tense in the non-finite verb group is not deictic, i.e. it does not itself
anchor the clausal type in the here-and-now of the speech event (Halliday
1994: 204). It does, however, imply grounding (or primary tense), which, in
turn, presupposes that the clausal type specification is viewed as being in-
stantiated.
It can be observed here that the claim that the Subject has special status
with respect to the instantiation of the clause fits in with the Cognitive
Grammar analysis of the Subject as 'primary focal participant' (Langacker
1991: 321) and that of 'initial reference point accessed in arriving at the
processual conception' (Langacker p.c.). As pointed out to me by Lan-
gacker, it also meshes nicely with other ideas that have emerged from re-
search in Cognitive Grammar. It reveals, for instance, an interesting parallel
between the instantiation of the clausal process type and that of a nominal
type with possessor: in both of them instantiation is effected by the element
that is characterized as the primary reference point (i.e. the Subject and the
possessor respectively) (see also Langacker 1993). In my opinion, the Sub-
ject in that respect differs fundamentally from Objects or 'secondary' focal
participants, which, I argued before (Section 2.1.2), cannot themselves in-
stantiate a type specification (though they can serve to specify in more de-
tail the instance that has been established by the clausal type specification
and the Subject).
How, then, can the size of the single clausal instance which the Subject
establishes through instantiation be further specified, i.e. how are clauses
quantified!? Langacker (1991: 197, 421) proposes that it is the non-
grounding auxiliaries of aspect (perfect and progressive) that serve to quan-
tify the verb's processual profile: they "indicate, with respect to some tem-
poral reference point, whether the process designated by the content verb
has been completed or is still in progress (hence only partially accom-
plished)" (Langacker 1991: 197). At the same time, however, Langacker
argues that these non-grounding auxiliaries serve to derive a higher-order
type specification, and he compares their function to that of number in the
nominal, where the plural of a noun like pebbles creates a type specification
that is distinct from that of pebble·, while pebbles designates a replicate
mass type (which belongs to the category of mass nouns), pebble desig-
nates a discrete entity and represents a count noun (Langacker 1991: 78).
The claim that the non-grounding auxiliaries of aspect derive a higher-order
type specification, however, seems to conflict with the idea that they ex-
press quantification, which presupposes instantiation (Davidse 1997: 423).
On this point, I will thus follow Davidse (1997), who argues that it is
not the non-grounding auxiliaries which quantify over a clause's processual
80 The functional organization of nominal and clause
(27) You guys must / not have closed the door all the way. (CB)
(28) Your friend must/ not have very good eyes. (CB)
(29) You may/ not have read in the Constitution about Election Day for the
Supreme Court.... (CB)
(30) The researchers say the findings suggest women may/ not have
equivalent access to these procedures (CB)
In this view, the clause closely resembles nomináis with common noun
head, realizing the functions which it encompasses iconically, with the
process type specification as the innermost layer of organization and the
functions of instantiation and quantification added to it. The clause can
leave it at that and profile a quantified instance of a type which is not
grounded in the speech event, witness the occurrence of non-finite clauses
with atemporal clausal heads. In Section 2.3,1 will look into quantified in-
stances of a process type that are grounded and I will give an overview of
the types of grounding which they incorporate.
in the speech event" in that the speaker "associates with the thesis an indi-
cation of its status and validity in his own judgment" (Halliday 1970a:
335). This type of modality is called 'interpersonal' modality or 'modaliza-
tion'. It comprises the epistemic modal assessments of probability/ possibil-
ity/ certainty, as well as those of usuality/ typicality. A second type of mo-
dality is called 'modulation' and is argued to be "part of the thesis", i.e.
"part of the ideational meaning of the clause" (Halliday 1970a: 336). Under
modulation, Halliday includes the modals of volition and ability (grouped
together as 'dynamic' modals by Palmer 1990: 2, as cited in Verstraete
2002: 20), as well as instances of deontic modality to do with obligation
and permission. In Lyons (1977), the terms 'subjective' and 'objective'
were introduced to refer to the interpersonal and ideational types of modal-
ity respectively: subjective or interpersonal modality thus encodes the
speaker's position with regard to the propositional content of the clause.
Objective or ideational modality, in contrast, is internal to the proposition
(Verstraete 2001: 1507).24
Halliday (1970a) and Verstraete (2001, 2002) suggest that a subjective
type of modality exists in addition to that of epistemic modality: they argue
in favour of recognizing a subjective type of deontic modality as well. Ex-
amples of subjective deontic modality are given in (31) and (32). The mo-
dals in (33) and (34) illustrate objective deontic modality (based on Ver-
straete 2002: 38):
(31) Almost daily, I was told No, Clarkson, you are a fool and you may not
wear training shoes for school. (CB)
(32) For taking my sister's life, and all the other lives, too, she must be
locked up forever. (CB)
(33) But Ramadan means more than just physical deprivation. It has spiri-
tual and moral obligations, too. Followers must refrain from bad
thoughts, words and actions .... (CB)
(34) As you know, Mr. Karmal, a king must always return a gift with
something of similar value. (CB)
In the clauses in (31) and (32), the modal serves to encode a particular posi-
tion of the speaker with regard to the propositional content. The only dif-
ference between this type of deontic modality and epistemic modality con-
cerns "the relevant domain for the position", i.e. "deontic modality encodes
positions concerning the desirability of actions, whereas epistemic modality
encodes positions about the plausibility of propositions (Verstraete 2002:
40). The examples in (33) and (34), in contrast, do not encode a particular
position of the speaker, but they "report on the existence of a particular ob-
84 The functional organization of nominal and clause
When the various types of subjective and objective modals are analyzed
in terms of whether or not they establish epistemic grounding, another pic-
ture emerges and the main distinction turns out to be that between subjec-
tive deontic modals (and imperatives), on the one hand, and all the other
types of modality and the indicative, on the other:
tive deontic modals, objective deontic modality does not encode the
speaker's commitment to the obligation, but merely describes the existence
of such an obligation. And this can be located in past, present or future, just
like any other proposition (Verstraete 2002: 63):
(35) In Bangalore, computer companies have to/ had to/ will have to do
privately what they cannot depend on the government to do.
The same counts for clauses with dynamic modality, which are also tensed
(Verstraete 2002: 65):
Significantly, the presence of tense in clauses such as those in (35) and (36)
distinguishes them from clauses with subjective deontic modality: subjec-
tive deontic modals apply to tenseless or virtual domains. The state of af-
fairs that combines with subjective deontic modals is desired and, conse-
quently, purely virtual.
To conclude, the finite element does not simply ground the clause either
by tense or by means of modality: at a schematic level, all clauses that real-
ize tense can be analyzed as being modal in nature. They establish epis-
temic modality and express the speaker's position with regard to the plausi-
bility of the prepositional content of the clause. As Davies (2001) argues,
the semantics of indicative mood and of epistemic modality can thus be
treated as an integrated continuum. Interestingly, the objective type of de-
ontic modality and the dynamic modalities of ability and volition also seem
to establish epistemic modal grounding and thus form part of the epistemic
continuum as well. It is only the subjective type of deontic modality and the
imperative which do not imply any epistemic position on the part of the
speaker about the plausibility of the proposition. Instead, they express the
speaker's desire or wish and therefore deal with states of affairs which are
purely virtual and tenseless.
third person (Davidse 1997: 425). It is obvious that first and second person
refer to the ground: they refer to the speech participants who are primary
constituents of the ground. With regard to third person, note the unique ref-
erential status of the Subject which is pointed out by Keenan as a 'truly
universal' property of Subjects: the Subject's reference "must be determin-
able by the addressee at the moment of utterance. It cannot be made to de-
pend on the reference of other NPs which follow it" (Keenan 1976: 313).
The structure * He-self admires John cannot therefore replace John admires
himself. A third-person Subject is thus the non-speaker/hearer referent that
is most immediately accessible for identification by speaker and hearer.
How is the system of person deixis encoded in the clause? Davidse ar-
gues that it is coded subjectively or implicitly (Langacker 1991: 93) by the
Finite, i.e. through person and number marking, and objectively or explic-
itly by the Subject.28 The Subject thus makes explicit the grammatical per-
son which is marked subjectively on the Finite.29 Notice that there is no
person and number marking on non-finite or atemporal clausal heads: it
seems logical that atemporal clausal heads, which do not ground the proc-
ess type temporally or modally, also fail to realize the person deixis that is
inextricably linked to it. As to the Subject function in non-finite clauses,
then, I suggest that it is reduced to its instantiating role. It is only when a
Subject functions in a finite clause that it co-operates in the grounding of
the clause by establishing objective person deixis. Whereas the nominal
may be grounded either in terms of spatial proximity (by means of demon-
stratives) or in terms of person deixis (via possessives), the clause is thus
grounded both in terms of temporal/modal proximity and person deixis
(Davidse 1997: 424).
can now be completed, with grounding located both in the Subject and in
the finite element of the clause:
Figure 8. The role of the Subject and the Finite in clausal grounding
I have thus far shown that a clause is always grounded by person deixis and
in terms of modality: when the clause realizes epistemic modality, it en-
88 Thefiinctionalorganization of nominal and clause
codes the speaker's position with regard to the plausibility of the proposi-
tion; subjective deontic modality, in contrast, expresses the speaker's posi-
tion with respect to the desirability of the proposition. For a clause to func-
tion as a fixll speech act, however, it has to provide an indication as to who
takes responsibility for the modal position that is expressed in it. This indi-
cation is provided by the basic clause types declarative, interrogative and
imperative: together with the modal positions that can be taken in the
clause, these clause types serve to define the speech function of the clause
(Verstraete 2002: 83). In what follows, I will first briefly discuss these
speech functions and show how they indicate whether it is the speaker or
the hearer that is responsible for the modal position that is implied in the
clause. My discussion of the speech functions is based on Davies (1979)
and, in particular, on Verstraete (2002: 83-106). In a second part, I will
zoom in on the speech functional role of the Subject. My analysis of the
role of the Subject in the clause as speech act builds on Halliday (1994) and
Davidse (1997).
Declarative and interrogative structures serve to encode alternative as-
signments of responsibility for the modal position which they imply: while
the declarative assigns responsibility to the speaker in his or her own turn,
the interrogative transfers responsibility to the hearer in the next turn. As
Davies (1979: 51) puts it, in an interrogative, the speaker assigns to the ad-
dressee "the role of subsequent teller in so far as the speaker's use of this
construction incorporates a demand for reply". In the declarative examples
in (37) and (39), the speaker thus takes the responsibility in his or her own
turn and is committed to the assessment of possibility realized by might and
the assessment of certainty implied in the indicative in (39). The interroga-
tives in (38) and (40), on the other hand, signal that the speaker transfers
the responsibility to the hearer in the next turn. Verstraete (2002: 85) ob-
serves that this does not imply that the hearer will take this position:
"rather, transfer of responsibility implies that it is up to the interlocutor to
accept, reject or modify the modal position which the speaker has chosen as
the starting point for the exchange" (the examples are based on Verstraete
2002: 85):
(37) Gene therapy might very well become a major new revolution in
medicine. (CB)
(38) Might it be of some value? (CB)
(39) Then he got us involved in Iran-Contra. (CB)
(40) Were the police involved in that? (CB)
From process type specification to clause 89
Subject is not so much held responsible for the existence of the obligation,
but for carrying out the desired process: it depends on the Subject whether
the command will actually be carried out and will thus be successful. In
active clauses, the Subject is held responsible for carrying out the desired
process (as shown in 47a and 47b):
Observe the difference between (48b) and its objective deontic alternative
She has to be locked up forever·, in the latter, the sAe-Subject is held re-
sponsible for the existence of the obligation: she has committed a crime,
and therefore the obligation of locking her up exists.
Like the analysis of the Subject as clausal 'instantiator', its characteriza-
tion as 'modally responsible' element in the clause is reconcilable with the
Cognitive Grammar analysis of Subjects as 'primary focal participants'
(Langacker p.c.). Modal responsibility is assigned to the Subject-entity by
the speaker and therefore resides in the speaker's conceptualization of the
scene (Langacker 1990b). Like the notion of 'primary focal participant',
that of 'modally responsible element' is thus essentially a subjective cate-
gory. It manages to specify the precise nature of the speaker's conceptuali-
zation in more detail by linking it to the modal position (of plausibility,
truth or desirability) which the speaker takes in the clause. Importantly, the
subjective characterization of Subjects in terms of 'responsibility' is sche-
matic and generalizes across the range of Subjects that can be found in the
clause. Subjects can be said to differ, however, as to whether they are also
'objectively responsible', i.e. as to whether there is an objective basis for
92 The functional organization of nominal and clause
that is realized throughout the clause (Halliday 1994: 179). The representa-
tional functions in the clause are thus not located in the clausal head only
(even though the clausal head does provide the basic categorization of ex-
perience in the clause): representational functions are mapped onto the Sub-
ject as well.
Halliday (1994: 35) argues that the textual metafunction in the clause,
which relates the clause to the larger discourse context, is expressed by
means of "peaks of prominence located at beginnings and endings": the
textual component in the clause expresses the textual status of parts of the
discourse by assigning them to the boundaries of the clause. There is thus a
special significance attached to 'coming first' and 'coming last' (Halliday
1979: 69). In the structure of the nominal (which is less flexible in terms of
word order), by contrast, textual functions are primarily located in the de-
terminer: determiners can thus relate the nominal to the discourse context
or to an immediate referent "within the text" (Halliday and Hasan 1976:
33). Determiners, in other words, establish a direct link with the speech
event and its participants or they have an immediate referent in the dis-
course or text.
Finally, as to the interpersonal metafunction, it was shown that Halli-
day's (interpersonal) Mood-Residue analysis of the clause matches Lan-
gacker's functional description of the clause quite closely. While the Resi-
due expresses the clausal type specification, the Mood-element turned out
to realize instantiation, quantification and grounding. Not surprisingly,
therefore, the interpersonal differences between the clause and the nominal
- which also realizes a 'grounded instance of a type' - are smaller than are
those in terms of the representational and the textual metafunctions. Inter-
personally speaking, both the nominal and the clause consist of a compo-
nent that is ungrounded and combines with grounding elements that relate it
to the speech event, thus turning it into something that can be exchanged in
communication. In the clause, it is the Mood-element consisting of the Sub-
ject-Finite link that is most essential to grounding and realizes instantiation
and quantification; in the nominal, it is determiners (in the broad sense) that
assume the functions of instantiation, quantification and grounding.
Now that we have gained insight into the functional parallels and differ-
ences that exist between nomináis and clauses, we can turn to nominaliza-
tion again and try to establish the impact of the functional organization of
the nominal and the clause on the process of nominalization.
94 The functional organization of nominal and clause
I argued in Chapter 3 (Section 3.2) that one of the main problems related to
the traditional analyses of nominalization is that they fail to see nominaliza-
tions as constructions in their own right, i.e. as constructions which com-
bine clausal and nominal categories in a unique way but which are essen-
tially nominal in status and function as such in larger configurations. The
nominal character of nominalizations derived from a verb stem is not con-
troversial. That of nominalizations containing clause-like units, as in [Being
able to play the piano] has always been a dream of mine has eluded lin-
guists more and has not been accurately described yet. The nominal fea-
tures of the latter type of nominalization have often failed to be recognized
because they tend to be encoded non-prototypically, i.e. as non-discernible
functional components. I argued in Chapter 3 that for the internal analysis
of nominalizations, a functional account of the nominal class is needed
which is sufficiently schematic to cover all nominal constructions, proto-
typical as well as non-prototypical. As I have shown in Section 1 of this
chapter, the functional analysis of the nominal in terms of type specifica-
tion, instantiation, quantification and grounding can provide such an ac-
count and thus makes an analysis of the nominal status of the various types
of nominalized constructions - including those that contain clause-like
units - possible.
In addition to the nominal characteristics of nominalized constructions,
the clausal categories which accompany the clausal level of assembly that
forms the starting point of nominalizations have to be accounted for as
well. The schematic characterization of the clause which I have set out in
this chapter (Section 2) allows us to define the distinct levels of assembly
which nominalizations can start from more accurately (as first suggested in
Langacker 1991): nominalization either applies to a simple process type
specification or verb stem, to what I have defined as an atemporal clausal
head or to a fully grounded (or finite) clause. I also pointed out in Chapter 3
(Section 5) that nominalizations should not be reduced to being clause-like
constructions, but should be analyzed as being systematically related to
clausal constructions through agnation. More specifically, I argued that re-
lationships of agnation between nominalizations and clausal constructions
can shed light on the clausal categories that play a role in nominalization.
The functional model of the clause and, in particular, the description of the
types of grounding that are found in it and of the various functions served
by the Subject, I will show, form an excellent tool for the description of the
clausal agnates of nominalized structures.
Nominalization as functional reclassification 95
don 7 regret [having waited for him]), and nominalizations that take a full
finite clause as starting point (as in [That he left her] surprised me). In spite
of the fact that they derive from different levels in the functional assembly
of the clause, all of these structures are nominalizations, i.e. they have been
reclassified and have come to function as nomináis. My analysis is aimed at
elucidating the reclassification process that leads from processual starting
point to nominalization.
In Part Π, I will look at deverbal -er nominalizations. In Part ΙΠ, I will
turn to the system of factive nominalization, which includes three lexico-
grammatical subtypes: gerundive nomináis (e.g., his being fired), that-
nominalizations (e.g., that he was fired) and the fact /Aai-nominalizations
(e.g., the fact that he was fired). When discussing gerundive nominaliza-
tions, I will also mention gerundive nomináis with an 'action' meaning
(e.g., [Firing someone] is a terrible thing to do) and action nominalizations
(e.g., the writing of that book).
Part II
Deverbal -er nominalization
Chapter 5
Deverbal -er suffixation: Towards a descriptive
position
0. Introduction
6; Lees 1960; Marchand 1969; Quirk et al. 1985). Apart from human
Agents, -er nomináis also designate animate Agents or animals (e.g., re-
triever, wood-pecker) and immaterial Agents (e.g., pointer, reminder). In-
strumental nomináis like scraper, scratcher, toaster, transmitter have been
argued to come close to being material Agents (see, for instance, the -er
entry in the Oxford English Dictionary and Marchand 1969). However,
Modern English has also seen the emergence of many different types of
non-agentive deverbal -er nominalizations. The non-agentives that are most
often identified in the literature are those based on verbs of cooking (as in
1), those denoting items of clothing (as in 2), locative nouns (as in 3), and,
finally, nomináis designating a variety of other non-agentive entities (as in
4):30
The issue that needs to be tackled here is whether the Modern English
-er suffix can still justifiably be analyzed as agentive, or whether another
characterization has to be looked for which describes the present distribu-
tional properties of the suffix more accurately. Put differently, is agentivity
still the highest-level schema which deverbal -er nomináis instantiate, or do
we have to look for another schema? And if so, what does this alternative
high-level schema look like?
Not only the occurrence of non-agentive formations has to be accounted
for. Deverbal -er nominalizations also differ according to the degree to
which they are entrenched·. I referred earlier to the occurrence of 'ad hoc'
nominalizations (as in 5) and lexicalized -er nomináis (as in 6):31
-er nominalizations: Agent names or Subject names ? 101
(5) Alex Gough, the Welsh destroyer of Del Harris's hopes on Tuesday
....(CB)
(6) destroyer 'a small, heavily armed warship'
Research into the semantic profile of deverbal -er nominalization has al-
ways struggled with the system's tolerance of non-agentive -er nomináis.
102 Deverbal -er suffixation: Towards a descriptive position
Some of the early accounts of the suffix simply ignore the existence of non-
agentive -er nomináis and refer to -er as an agentive suffix (e.g., Oxford
English Dictionary; Lees 1960; Quirk et al. 1985). Others acknowledge that
-er can also derive non-agentive nomináis, but do not attempt to integrate
the non-agentive cases in the system of -er suffixation in general: their
account of -er nominalization is restricted to a mere list of possible forma-
tions (e.g., Jespersen 1914-1929, 6; Quirk et al. 1972; Kastovsky 1985).
Still others point to the disparity between agentive and non-agentive -er and
conclude that the non-agentive type is not part of the system of -er nomi-
nalization, that it is idiosyncratic or non-systematic in nature (e.g., Keyser
andRoeper 1984; for Dutch: Booij 1986; Taeldeman 1990).
Linguists who have tried to come to terms with non-agentive -er nomi-
nalization are, firstly, Ryder (1991, 1999a, 1999b), who offers an analysis
of the system of -er nominalization which focuses on the prototypically
agentive status of the designated entity and its salience within the event in
which it figures; secondly, Panther and Thornburg (2002), who suggest that
-er nominalization forms a complex conceptual category which is centred
on an agentive prototype to which a large number of other meanings are
linked through metaphor and metonymy; and, thirdly, Levin and Rappaport
(1988) and Rappaport Hovav and Levin (1992), who analyze -er nomináis
from the perspective of argument structure configurations. In the following
sections, I will have a closer look at the descriptive analyses which they
propose.
The analysis which Ryder (1991, 1999a, 1999b) proposes to account for the
variety of profiles in deverbal -er nominalizations is "semantic/pragmatic"
in focus (Ryder 1999b: 291). It analyzes -er nominalization "in terms of
semantic case, event structure and prototype reanalysis" (Ryder 1991: 299)
and builds on the assumption that semantic roles "are better viewed as pro-
totypes rather than as absolute categories" (Ryder 1991: 300). While some
participants will thus "be considered agents by everyone, other, less proto-
typical ones will be considered agents to the extent that they can be con-
strued as matching or approximating the prototype" (Ryder 1991: 300).
Ryder argues that the non-agentive cases of -er can all be regarded as ex-
tensions or reanalyses of the agentive -er prototype, reanalysis being possi-
ble when a non-agentive entity is conceived as being salient or particularly
noticeable within the designated event (Ryder 1991: 309; 1999b: 288-289).
Salience is what relates deverbal -er nominalizations to denominal ones
-er nominalizations: Agent names or Subject names ? 103
(e.g., prisoner, weekender) in which "the referent of the -er noun ... should
be higher in salience than that of the nominal base" (Ryder 1999b: 290).
The schema which is instantiated by all -er nominalizations - deverbal and
non-deverbal - can thus be summed up as "a person, animal or thing sali-
ently connected with X" (Ryder 1999a: 307).
How did in Ryder's opinion the distinct extensions of the agentive pro-
totype in deverbal -er nominalizations come about? She (1999b: 287) ar-
gues that two factors are important here, viz. an entity's salience and identi-
fiability in the event, and, secondly, its location on the causal chain of the
event in which it figures (Ryder 1999b: 287). The first and most obvious
extensions of Agent -er nomináis are those nomináis that designate Instru-
ments. Instruments can easily be foregrounded as salient, agent-like enti-
ties: "because they are independent of the Agent in many ways, the event
can be construed as having them as the head of the part of the causal chain
that is in focus, and so, like Agents, they are highly salient" (Ryder 1999b:
288). Especially the great proliferation of implements and machines from
the 16th century onwards explains the extension of the almost exclusively
agentive -er suffix to instrumentais in the early Modern English period (as
attested in Dalton-Puffer 1994).
Once -er nomináis came to designate Instruments, "clothing intended to
be worn while performing central actions in an episode" was construed as
similar to an Instrument used in that episode (Ryder 1991: 305). It is the
'identifiability' of clothing items, or the fact that they are "specifically de-
signed for certain activities, and so can be readily identified by those activi-
ties" (Ryder 1999b: 290) which explains why various -er nomináis denot-
ing items of clothing have since been coined: e.g., loafers, sneakers, wad-
ers, loungers, jumper. Patientive -er nomináis referring to food, then, pro-
file structures in which "the agent is outside the boundaries of the episode,
leaving the original patient as the most agent-like element remaining" (Ry-
der 1991: 309): the nominal steamer 'edible clam', for instance, profiles a
clam that steams. Nomináis like broiler, baker and frier are analyzed simi-
larly. Ryder (1991: 309) says they all designate "agent-like active patients".
Comparable to the instrumental and clothing types of -er nomináis, "the
food is sufficiently salient and ... sufficiently identifiable by the verb" to
become the referent of an -er nominalization (Ryder 1999b: 289). The sali-
ence of these items of food is attributed to two features in particular: firstly,
the processes which they are related to refer to a "specialized type of prepa-
ration" (Ryder 1999b: 289), which the animals, vegetables and fruit are
bred or selected for; and, secondly, "the food that is cooking, at least in
modern kitchens, is often the most prominent participant actually present in
the event" (Ryder 1999b: 289). Ryder refers here to the fact that the "per-
104 Deverbal -er suffixation: Towards a descriptive position
son in charge of the meal performs few if any overt actions during the
cooking process, and neither does the device in which or by which the food
is cooked" (Ryder 1999b: 289).
Although originally the referents of the food-related nomináis may have
been interpreted as active Patients, Ryder (1991: 309) claims, they were
reanalyzed as truly patientive (e.g., steamer as a clam that steams was re-
analyzed as a clam that is steamed) and thus paved the way for -er forma-
tions "that have few if any agent-like qualities", e.g., scratcher 'a lottery
ticket that is scratched to reveal the potentially winning patterns' (Ryder
1991: 310) and dipper 'something that is dipped before being eaten' (Ryder
1991: 309). That the latter, in spite of their lack of agent-like properties,
can be profiled by an -er nominalization, Ryder (1999b: 289) argues, is
only due to their being sufficiently "identifiable in their event".
Ryder's analysis of -er nominalization has the merit of trying to offer an
integrated analysis of agentive and non-agentive -er nominalization on the
basis of a wealth of data. It includes interesting observations, for instance
on the relation between instrumental -er nomináis and the agentive proto-
type of -er nominalization. Ryder's approach, however, in my opinion, also
has its weak points and in the next section, I will point to some of them. In
Section 1.1.2 I will discuss Panther and Thornburg's (2002) alternative
cognitive proposal for the analysis of -er nominalization and show that it
significantly improves on the analysis proposed by Ryder and contains
several valuable descriptive observations concerning the semantics of -er
formation. In Section 1.1.3, then, I will argue that, as shown in Lemmens
(1998), a detailed analysis of the nature of the base verbs of deverbal -er
nominalizations is necessary to arrive at an accurate description of their
profile.
seems to have taken place within the system of deverbal -er suffixation is,
in other words, left unaccounted for. hi Chapter 7 I will show that it is only
when the compositional value of instrumental and patientive -er nomináis
is analyzed more closely that the reason behind their semantic resem-
blances becomes clear.
Summarizing, Panther and Thornburg show that denominal -er nomináis
can be profitably examined from the perspective of metonymy and that
their semantics resembles that of deverbal -er nomináis to a certain degree.
They also make interesting observations concerning the semantics of in-
strumental and patientive -er nomináis. In my own analysis of -er nomi-
nalization, I will restrict myself to deverbal -er nomináis because their se-
mantics is encoded in the process-participant configuration which they
designate through the constructional link which they establish between a
verb and the -er suffix. I will, more particularly, argue that it is only the
analysis of the constructional properties of the various types of -er suffixa-
tion that can shed a new light on their semantics, and, ultimately, on the
high-level schema which all deverbal -er nomináis instantiate.
instance, dipper, sipper and scratcher can now be argued to be due to the
ergative nature of the process which they nominalize and, related to it, to
the Medium which they profile.34 More in general, Lemmens's account of
how the transitive and ergative paradigms operate in -er derivation points to
the importance of considering the contribution of the verb from which -er
nomináis are derived: the compositional value of a deverbal -er nominaliza-
tion largely depends on its verbal base.
To conclude, then, both Ryder's analysis and Lemmens's refinements of
it are indicative of the rich representational or ideational semantics of
deverbal -er nominalization. The analysis of the representational functions
which are realized in a nominalization system forms an essential part of a
multifunctional approach to nominalization (as discussed in Chapter 3,
Section 4). Representational categories are important for the description of
deverbal -er nominalizations in that they reveal the variety that exists
among them and allow us to identify the category's prototype, i.e. the
Agent-profiling type of -er nominal. However, representational categories
seem unable to generalize across all instances of deverbal -er suffixation.
The category of Agent, for instance, fails to account for the non-agentive
instances of -er nominalization. In representational terms, moreover, the
non-agentive type of -er nominalization is necessarily viewed as a periph-
eral, non-prototypical instance of -er nominalization. A schematic view on
-er nominalization which, I argued in Chapter 3 (Section 1.1), necessarily
complements an analysis in terms of prototype cannot therefore be based on
the representational categories of agency and process type. In the following
section, I will examine an attempt at schematization suggested in Levin and
Rappaport (1988) and Rappaport Hovav and Levin (1992).
Levin and Rappaport (1988) and Rappaport Hovav and Levin (1992) claim
that the derivation of -er nomináis "refers not to particular semantic-role
labels but rather to argument-structure configurations" (Levin and Rappa-
port 1988: 1067). -Er nomináis may bear a wide range of semantic relations
to the verbs from which they are derived, but "they most systematically
correspond to the external argument of the base verbs" (Rappaport Hovav
and Levin 1992: 145) and therefore only derive from verbs which have
external arguments (Levin and Rappaport 1988: 1070). What does the term
'external argument' stand for?
The notion of 'external argument' was introduced in Williams (1980,
1981) to refer to a particular, "distinguished argument, not a syntactic posi-
-er nominalizations: Agent names or Subject names ? I l l
tion, a case, or something else" (Williams 1981: 83). The external argument
is the argument "that corresponds to the NP in a sentence of which a [verb]
phrase with that item as its head is predicated" (Williams 1981: 84). The
example which Williams gives to illustrate the notion is that of the verb hit.
Hit has two arguments, viz. one which serves the semantic role of Actor
and another one which is Theme. The Actor is the external argument be-
cause in a clause it "must be specified in a position external to the verb
phrase of which hit is the head" (Williams 1981: 84). The Theme, on the
other hand, "must be specified within the verb phrase" (Williams 1981: 84).
This seems to correspond to what in Levin and Rappaport (1988: 1074) is
understood by 'external argument': the external argument of a verb is de-
fined there as its "underlying (D-structure) subject", while an internal ar-
gument is an "underlying (D-structure) direct object".
Passives, middles and ergative one-participant structures illustrate that
external arguments do not necessarily correspond to the so-called "surface
structure" Subject. In passive clauses, for instance, the arguments of the
verb are reorganized and "the internal argument of the verb occupies the
canonical subject position" (Haegeman and Guéron 1999: 123). In the pas-
sive clause Louise was invited (by Thelma), the internal argument Louise
thus occupies the Subject position, while the external argument Thelma
"may appear optionally in a PP adjunct" (Haegeman and Guéron 1999:
123). Ergative, one-participant structures such as The ball rolls (called 'un-
accusative' in Perlmutter 1978) have only one argument, which functions
as Subject and is not an external, but an internal argument (Levin and Rap-
paport 1988: 1074). In the corresponding two-participant structures, the
internal argument functions as Object (e.g., I roll the ball). Middle clauses
such as This book reads well, finally, constitute yet another type of clause
in which the Subject is the internal rather than the external argument of the
verb (Williams 1981; Keyser and Roeper 1984).
In their analysis of -er nominalizations, Levin and Rappaport (1992:
145) discuss four groups of nomináis which, they claim, corroborate their
"external argument generalization", viz. instrumental -er nomináis, nomi-
náis derived from the spray/load-type of verbs, nomináis based on intransi-
tive verbs, and, finally, nomináis which are related to the middle form of
the verb.
As far as instrumental -er nomináis are concerned, Levin and Rappaport
claim that they must designate "intermediary" instruments or instruments
which are able "to perform the action in some sense autonomously" (Levin
and Rappaport 1988: 1071). The verb open, for instance, can take -er suf-
fixation to refer to an intermediary instrument involved in the opening of
something, as in
112 Deverbal -er suffixation: Towards a descriptive position
With respect to -er nomináis derived from intransitive verbs, Levin and
Rappaport claim that they "should be based on a unergative rather than an
unaccusative verb, since only the single argument of a unergative verb is an
external argument" (Levin and Rappaport 1988: 1074). The distinction
between unergative and unaccusative intransitives goes back to Perlmutter
(1978) and Burzio (1986, as cited in Levin and Rappaport 1988: 1074).
Unergatives largely coincide with what in Davidse (1991, 1992b) are called
'intransitives' (e.g., run, climb, work and sin). The unaccusative category
includes all the ergative verbs of Davidse's model in their 'non-caused'
sense, as in the door opened, the ball rolled, the water froze, as well as
processes such as disappear, happen and die. The single argument of unac-
cusatives "is not an external argument but rather a direct internal argument,
that is, an underlying (D-structure) direct object" (Levin and Rappaport
1988: 1074). It cannot therefore undergo -er nominalization:
1077), because they are generally assumed to profile an internal rather than
an external argument. Levin and Rappaport (1988: 1078) point out, how-
ever, that these nomináis "do not receive the interpretation that would be
expected if they were derived directly from the transitive or unaccusative
uses of the related verb". Rather, their meaning seems closer to the inter-
pretation that the related verb receives in the middle structure (Levin and
Rappaport 1988: 1078). A nominal like broiler, for instance, designates a
chicken which is bred or intended to be broiled, which comes closer to the
interpretation of a middle construction such as This chicken broils well than
to This chicken broils. Apart from semantic correspondences, middles and
nominalizations in -er also show the same constraints with respect to the
verbs which they can be based on: both of them, it is argued, can only be
derived from verbs with 'affected objects'. A structure like *this planet
sees well and the corresponding -er nominal *seer are, for that reason, im-
possible (Levin and Rappaport 1988: 1079).
Does the relationship with middle formation imply that cooking -er
nomináis, like all -er nominalizations, profile the external and not the inter-
nal argument of the base verb? Levin and Rappaport (1988: 1078) admit
that most accounts of the middle construction consider the single argument
of middles as internal (e.g., Keyser and Roeper 1984; Roberts 1987), but
they argue that the occurrence of -er nomináis like broiler, baker and
steamer might be taken as support for an "alternative analysis of the middle
construction" (Levin and Rappaport 1988: 1078). The analysis which they
propose involves "the externalization of the direct internal argument of the
verb in the lexicon" (Rappaport Hovav and Levin 1992: 149). An -er nomi-
nalization which is related to middle formation therefore profiles an exter-
nal rather than an internal argument.
Levin and Rappaport's analysis is meant to show that -er nominaliza-
tions cannot be generalized across in terms of semantic roles but only in
terms of the notion of 'external argument'. The latter, however, proves to
be an ill-defined notion: what it seems to stand for is 'being the most likely
candidate for serving as Subject', and this, by and large, corresponds to
being the Agent of the process or being an entity that can be conceived as
agent-like. They thus fail to distinguish the representational function of
Agent from the 'interpersonal' or instantiating/grounding function of Sub-
ject (see Chapter 4, Sections 2.2 and 2.3). Levin and Rappaport also bring
in the Subject of middle constructions, but without investigating the true
extent of this category in English. This leaves a number of -er types unac-
counted for. In what follows, I will briefly run through those types of -er
nominalization which Levin and Rappaport fail to elucidate. The main ones
are: -er nomináis designating what I will call 'oblique participants' (Laffut
-er nominalizations: Agent names or Subject names? 115
Following Laffut (2000: 139), I will use the term "oblique participant" to
refer to participants which are clausally realized in the form of a preposi-
tional phrase (as in 15). Laffut (2000: 137-156) lists a number of semantic
and formal arguments which show that oblique participants differ from
circumstantials as in (16) in that they are basically "complements" and
elaborate a salient e-site in the verb with which they are used (Langacker
1987: 305) (examples based on Laffut 2000):
The existence of -er nomináis that profile oblique participants counters part
of Levin and Rappaport's analysis of instrumental -er nomináis, of nomi-
náis based on spray/load, verbs and of nomináis derived from unaccusative
verbs. Nomináis like stroller, walker and viewer fail to profile the agent-
like, external argument of the processes of strolling, walking and viewing
(this is also pointed out by Ryder 1999b: 272-273): instrumentais such as
stroller and viewer cannot be analyzed as doing the strolling and viewing
themselves, i.e. a stroller does not stroll and a viewer does not itself view.
Instead, they seem to profile oblique participants.
Apart from oblique participants which indicate the means whereby a
process can be carried out, -er nomináis can also designate the place of a
process, as in:
These nomináis illustrate that -er nomináis derived from intransitive verbs
(in Levin and Rappaport's terminology, unergative verbs) do not necessar-
ily designate the external argument of the base verb: rather than profiling
the entity which does the stepping/kneeling/sleeping, they designate the
surface that you step on, the chair that you kneel on and the bed that you
sleep in.
Finally, Levin and Rappaport only focus on the Locatum argument of
spray/load verbs (e.g., Bill loaded the truck with cartons, Bill loaded car-
tons onto the truck) and they ignore the Location of the process, as if that
category is not relevant to -er suffixation (e.g., Bill loaded the truck with
cartons, Bill loaded cartons onto the truck). Yet, I have found several -er
nomináis which designate the Location of the verb load:
(20) [about a boat:] She came up the Thames to Chelsea Harbour, and was
craned out on to a low-loader. (CB)
(21) The most popular type of washing machine is the front-loader but
there are still alternatives. (CB)
(22) Top-loaders are useful for a kitchen short on space as they can be
placed at the end of a run of units. (CB)
The Location cannot be analyzed as the external argument of load. The fact
that it can be profiled by an -er nominalization thus forms yet another
counterargument to the external argument generalization which Levin and
Rappaport propose for -er nominalization.35
Levin and Rappaport are right in pointing out that (unaccusative) verbs of
existence and disappearance do not normally take -er suffixation (e.g.,
*exister, *happener, *disappearer). Yet, in some cases they do, witness the
following (attested) examples of -er nomináis with a one-participant unac-
cusative as basis:
(23) When I get busy, I tend to forget to water my plants, so it's a good
thing I've got a plant that's a wilter. The minute it gets a little low on
-er nominalizations: Agent names or Subject names ? 117
The label of 'Subject' has thus far been assigned to -er nominalizations
only sporadically, and then particularly as an empty label representing a
purely syntactic category. The term 'Subject' is in fact often used in one
and the same breath with the semantic role that is typically mapped onto it,
viz. that of Agent. In the same way, patientive -er nomináis are then classi-
fied as belonging to the 'Object' type of -er nomináis.
Marchand's (1969) account o f - e r nominalization nicely illustrates this
specific use of the concept of Subject in the description of -er suffixation.
Marchand (1969) distinguishes between a Subject type of deverbal -er
nominalization, an Adverbial Complement type and an Object type. In the
Subject group, he situates all those nomináis that are agentive, ranging from
-er nominalizations: Agent names or Subject names ? 119
which can likewise be used anaphorically once they come to function in the
structure of the nominal. Still, it cannot be denied that ad hoc formations
are frequently used to rephrase a previous clause by nominalizing it, as
happens in the following examples:
(26) One guy jumped right into the fight, but his friend immediately van-
ished. Hie police came and hauled off the fighter, after which the van-
isher promptly reappeared, laughing, (cited in Ryder 1999b: 283)
(27) "Not him," says Mrs. Cook. "This one, though. A bit of a poseur, but
he moves very nicely." I will not describe the nice mover, to spare his
blushes, but let us say that he had dressed for the part. (CB)
(30) Recent research has established that merely breathing smoke-laden air
- passive or 'second-hand' smoking - puts the breather at risk. (CB)
(31) Trouble sleeping is the most common complaint, including an inabil-
ity to get to sleep, terrible, frightening dreams that wake the sleeper in
a state of anxiety, and waking early feeling unrefreshed but unable to
get back to sleep. (CB)
The nomináis breather and sleeper refer to the one who breathes and the
one who is sleeping respectively, but they do not designate specific, actual
occurrences of the processes of breathing or sleeping. As to the "ability to
inherit complement structure" (Rappaport Hovav and Levin 1992: 132),
then, this does not seem to be possible in ad hoc nominalizations only: the
nominalizations in (32) and (33) are both lexicalizations and they are fol-
lowed by o^phrases which periphrastically realize an Object which speci-
fies the range or scope of the process (in 32) or the Beneficiary of the
teaching (in 33):
The system of deverbal -er nominalization may have started out as an agen-
tive system, it can no longer be characterized as such: non-agentive -er
nominalizations are numerous and too productive to be described as non-
systematic. In the next two chapters, I will defend the claim that deverbal
-er suffixation has evolved from profiling Agents to profiling Subjects. I
will argue that an analysis of -er nominalization in terms of the Subject
function provides a more accurate generalization for the present distribu-
tional properties of -er nominalization than one in terms of agency. In par-
ticular, I will show that the various types of non-agentive -er nominaliza-
tions are systematically equivalent to the Subject of middle constructions
and of so-called "Setting-Subject" constructions (Langacker 1991: 346;
e.g., This tank does not squirt water.). I will moreover argue that deverbal
-er suffixation construes equivalents to the basic grounding or deictic func-
tion of the clausal finite verb. The correspondences between the clausal
Subject-Finite unit and -er nominalization elucidate the semantics of the -er
suffix significantly.
The evolution in the system of deverbal -er nominalization may run par-
allel to certain changes of predominance in the various functional layers of
organization in the clause. Within the Prague School (which has always
integrated the comparative and the diachronic in its functional approach), it
has been hypothesized that the English clause has developed from an
Agent-oriented to a predominantly Subject-oriented form of organization.
Or, as Vachek (1966: 92) puts it, "the main function of the grammatical
subject in an English sentence is no longer to express an agent". Unproto-
typical or non-agentive Subjects as in middle constructions are on the rise.
Significantly, it is this clause type, which combines a non-agentive Subject
with an active finite verb, which turns out to be particularly important for
the analysis of non-agentive -er nominalization. It will therefore constitute
the focus of the next chapter. In Chapter 7, then, I will present my own
124 Deverbal -er suffixation: Towards a descriptive position
(1) More than 50 years later, the book still sells well. (CB)
(2) The new Holden Berlina handles like a junior sports sedan.... (CB)
(3) And it's latex paint, so it cleans up quickly and easily with soap and
water. (CB)
(4) If they wanted to do Eliot, why didn't they do The Magi? It's narrati-
ve, it reads well.... (CB)
Middles typically combine an active verb form with a Subject that is pa-
tientive in nature and thus reminiscent of passive constructions.39 Semanti-
cally speaking, the middle is often associated with a facilitative meaning
(Fawcett 1980; Kemmer 1993): it is said to express how easy or difficult it
is to realize a particular process with respect to the entity construed as Sub-
ject. Prototypical middle constructions are construed with the adverb easily,
as in
Rather than account for the unique constructional link which middles
realize between a non-agentive Subject and an active verb form, most
analyses of middle formation describe middle constructions by reference to
other constructional paradigms and, as a result, leave many questions unan-
swered. In this chapter, I will start with a brief sketch of the two main exist-
ing approaches to middle formation (Section 1). I will show that they tend
to reduce the representational potential of middle formation and focus on
only part of the middle system. Among the middle constructions that have
been systematically ignored in the literature are, for instance, those based
on intransitive processes, with an oblique participant as Subject. They will
be the main topic of Section 2.1 will then zoom in on two aspects of middle
formation which, in my opinion, are crucial to identify the high-level
schema which middle constructions instantiate: I will first discuss the vari-
126 The middle construction
ous semantic profiles of middle formation, and argue that some middles,
rather than profile the facility with which a process can be carried out, fo-
cus on the destiny of the Subject or on the result of carrying out a particular
process on it (Section 3). In a final section, then, I will elaborate on the
essentially modal nature of the relationship which middles establish be-
tween a Subject and a Finite (Section 4).
In the literature on middle formation, some have stressed the active, one-
participant nature of the construction and have claimed agentive status for
its Subject, while others have ascribed passive value to the construction,
emphasizing that its Subject is affected and that an Agent is implied. I will
refer to the former as the ergative and the latter as the passive approach to
middles. Those who situate middle constructions close to ergative one-
participant structures are, among others, Hale and Keyser (1987), Sinclair
et al. (1990), Francis et al. (1996) and Sohn (1998). They argue that mid-
dles resemble ergative one-participant structures in that, firstly, they allow
for the addition of an "event of causation" (Hale and Keyser 1987: 6),
which is generally assumed to be characteristic of ergative one-participant
structures such as the one in (7):
which, Van Oosten claims, represent the linguistic raison d'être of middle
constructions: "the purpose of the patient-subject construction is precisely
to enable the speaker to assert that ... properties of the patient bear the re-
sponsibility for the occurrence of the action of the verb" (Van Oosten 1986:
93). In Lakoffs view as well, "the point of using the patient-subject con-
struction is to say that properties of the patient are more responsible for
what happens than the agent is" (Lakoff 1977: 248).
An alternative analysis of middle formation is found in, among others,
Sweet (1891), Jespersen (1914-1929, 3), Halliday (1967), Smith (1978),
Keyser and Roeper (1984), Fellbaum (1986) and Fellbaum and Zribi-Hertz
(1989). All of them start from the assumption that the middle realizes one
participant overtly, but contains two. The two-participant approach to mid-
dle formation identifies the Subject of the middle as patientive. The Agent
of the process is argued to be necessarily implied. Evidence in favour of the
patientive nature of the middle Subject is said to come from the fact that
middles cannot be used in the imperative (Fellbaum and Zribi-Hertz 1989:
19) (illustrated in 9), and from the fact that the middle Subject, unlike the
Subject of an ergative one-participant construction, cannot be presented as
doing things by itself (see 11) (Keyser and Roeper 1984: 405):40
As Keyser and Roeper (1984: 405) point out, a contradiction exists be-
tween the notion all by itself and the Agent which is implied in middles
(see also Smith 1978): the middle Subject cannot be perceived as agent-like
because the Agent of the process is implied.41 The Subject has moreover
been claimed to be necessarily affected by the process (Levin 1993: 26):
Finally, following Sweet (1891), the middle has frequently been compared
to and sometimes even been located within the passive paradigm (Halliday
1967; Keyser and Roeper 1984; Declerck 1991b). At the same time, how-
ever, it is pointed out that middles also differ from ordinary passives. Most
importantly, middle constructions do not allow the implied Agent to be
realized overtly, while passives do (Keyser and Roeper 1984: 406):
128 The middle construction
Apart from transitive and ergative verbs, however, also intransitive verbs
can figure in middle constructions, a possibility which, while recognized in
descriptions of the Dutch middle construction, has thus far been almost
unanimously denied in the literature about English.43 The Dutch middles in
(20), (21) and (22) qualify as intransitive. The example in (20) is based on
Haeseryn et al. (1997: 52); that in (21) was extracted from the DDL
corpus44; the middle in (22) is taken from De Vries (1910: 132):
(23) This music dances better than the other one. (Van Oosten 1986: 8)
(24) The Strathayr track is racing consistently nowadays. (Barry Blake
p.c.)
(25) The pitch is playing truly/well. (Barry Blake p.c.)
(26) [about a tennis court:] It is slightly coarser, so it plays a bit slower.
(CB)
(27) "They rolled the green just before the match and it ran three seconds
faster," said Curtis.... (CB)
(28) You wouldn't give my tip a chance of landing Aintree's Martell Red
Rum Chase ... but with the ground riding slower, he should improve
dramatically. (CB)
130 The middle construction
The Subjects of these middle constructions designate either the means that
is used to carry out the process (music) or they specify its location
(examples 24 to 28). They are normally realized as prepositional phrases:
you dance to music, run on a track, bowl on a pitch, play on a court, run on
a green and ride on a ground. Yet another type of intransitive middle has
an Instrument for Subject (Smith 1978: 103):
sive' view on middle formation) and it does not always undergo a change
(as claimed by the adherents of the ergative view on middle formation),
witness the following examples:
(31) Girls don't make it just because they are skinny. It's about how they
photograph and move. (CB)
(32) His earlier short stories don't read so well. (CB)
The girls in (31) are not affected or changed by being photographed. Like-
wise, the short stories in (32) are not affected or changed by being read.
The middle Subject may not be agentive, but is the Agent of the process
then obligatorily implied? As pointed out by Langacker, a middle like This
ice cream scoops out quite easily cannot but be interpreted as involving an
Agent which does the scooping out: we do not picture the ice cream scoop-
ing itself out (Langacker 1991: 334). It is, in fact, the semantics of the tran-
sitive verb scooping out which invokes the Agent: if for This ice cream
scoops out quite easily we do not imagine the ice cream scooping itself out,
it is because we follow the semantic instructions given by the active transi-
tive verb scoop out, which necessarily implies an Agent. What is presented
as most prominent by the middle construction, however, is the relationship
between the process and the non-agentive entity construed as its 'focal par-
ticipant' or Subject (Langacker 1991: 335). Intransitive middles can be
given an analogous semantic analysis: they most prominently present an
intransitive process in its relation to a Location or Means, but they imply an
Agent which actually executes that intransitive process. In, for instance,
The court plays a bit slower, an Agent is invoked which does the playing.
Likewise, in This music dances better than the other one, an Agent is im-
plied which dances to the music.
Van Oosten and Lakoff s emphasis on the agent-like responsibility of
middle Subjects should therefore be toned down: it is widely agreed that
the middle attributes the fact that a particular process can be carried out to
the inherent properties of the entity construed as Subject (apart from Van
Oosten 1977, 1986 and Lakoff 1977, also Fellbaum 1985, 1986; Langacker
1991; Massam 1992; Rosta 1995; Lemmens 1998), but the fact that the
Subject has properties that make a certain process possible should not be
confused with agentivity. Or, as Rosta (1995: 128) puts it:
In The book sold well... the properties of the book are a necessary condi-
tion for the book to sell well, but need not be (and typically are not) a suffi-
cient condition; if it were not for the properties of the book, it would not
sell well, but even though the book has the properties it has it will not sell
132 The middle construction
well unless some other participant (an agent) gets involved in getting the
book to sell well.
Rather than assign agent-like value to its Subject, the middle, in other
words, foregrounds the fact that the Subject-entity has properties which
influence the occurrence of a particular process. It has been pointed out that
when non-human entities function as middle Subjects, these inherent prop-
erties are often part of their design (Levin and Rappaport 1988; Lemmens
1998). As Lemmens (1998: 80) puts it, "the properties emphasized in the
middle construction are often also those for which the entity has been de-
signed in the first place". In the case of the latex paint in (33) and the lip-
stick in (34), for instance, it is obvious that cleaning up easily with soap
and water and slipping easily into your handbag or pocket is part of what
they are designed for:
(33) And it's latex paint, so it cleans up quickly and easily with soap and
water. (CB)
(34) Compact lipstick size design, slips easily into your handbag or pocket.
(CB)
Animals, then again, are often bred for a particular process: a chicken that
broils well has probably been bred to be broiled (Levin and Rappaport
1988: 1078).
(35) The window opened only with great difficulty. (Langacker 1991: 334)
(36) [about conditioning milk:] ... rinses easily away and really works!
(CB)
(37) That is easily done because the car handles superbly. (CB)
(38) If they wanted to do Eliot, why didn't they do The Magi? It's narrati-
ve, it reads well.... (CB)
(39) The new Holden Berlina handles like a junior sports sedan.... (CB)
(40) Her life story reads like the Hollywood movies she's starred in. (CB)
(41) "They rolled the green just before the match and it ran three seconds
faster," said Curtis .... (CB)
(42) [about a cosy car seat protector:] Quickly attaches/removes with ela-
stic straps and velcro tabs. (CB)
(43) This title usually ships within 2-3 days, (www.amazon.com)
134 The middle construction
The middle constructions discussed so far all zero in on the effect of the
properties of the Subject-entity on how the process can be carried out.
However, the nucleus of non-agentive Subject and active verb phrase in the
middle can also focus on various other stages. I distinguish between three
other possible foci of middle constructions: middles can simply highlight
the feasibility of the process (i.e. whether or not the process can be carried
out); they can zero in on what the process is destined for; or they can focus
on what the result is like when the process is carried out.
First, middles of the feasibility type (Davidse 1991: 44) merely focus on
whether the properties of the entity construed as Subject make a process
possible. They do not specify how this can be done or whether or not it can
be done easily. Examples are:
(44) [about a cook book holder:] Folds up and packs away for convenient
storage. (CB)
(45) This umbrella folds up. (Fellbaum 1986: 9)
(46) I thought we were out of gas, but the car DRIVES! (Fellbaum 1986:
9)
(47) This dress buttons. (Fagan 1992: 57)
Examples like these illustrate that the most prototypical cases of middle
formation may be those with an adverb (such as easily, welt), but that the
presence of an adverb is not an absolute prerequisite: middles which focus
on whether it is feasible to carry out a certain process on a particular entity
do not seem to be characteristically accompanied by specific adverbs.
Some middle constructions contain a locative oblique participant (Fagan
1992: 80). Interestingly, the occurrence of such Locatives appears to ac-
count for the destiny-oriented focus which characterizes these middles
(Davidse p.c.):
(48) [about a children's coat:] Outer flap wraps around little hands and
secures with Velcro.... (CB)
(49) The purpose of the device is to alert deer of your approach, not the
other way around. It is a little whistle that attaches to your car with
self-adhesive tape. (CB)
(50) The ultimate travel pillow. Resteaz fixes to the headrest providing
comfortable head and neck support. (CB)
(51) Playset folds up into a storage case with handle for easy carrying.
(CB)
Towards a semantic typology of middle constructions 135
These middles do not mean that the result of selling the book or driving the
car is good. Rather, the meaning of well in (54) comes close to fast or in
large quantities, whereas in (55), well can be paraphrased as easily,
smoothly. The adverb well, in other words, has various different contextual
entailments, which in the middle either trigger off a facility- or quality-
oriented reading (as in 54 and 55) or a result-oriented one (as in 52 and 53).
Other result-oriented middles are:
The middle in (56) emphasizes that the result offolding up the bag is likely
to be neat. As in middle constructions with an inanimate Subject in general,
it is implied that the Subject-entity has been designed in such a way that
this result is possible. The middle in (57) emphasizes that photographing
the Subject-entity will probably not lead to nice pictures.
The facility-, quality-, feasibility-, destiny- and result-oriented middle
structures which I have thus far discussed all 'target' (Langacker p.c.) one
particular facet of the complex event that is profiled by the middle con-
struction: feasibility-oriented middles highlight the beginning of the event,
facility- and quality-oriented ones its middle stage, while destiny- and re-
sult-oriented ones zoom in on the end stage of the event. Corpus data reveal
that middle constructions often combine two or more of these foci, and this
by integrating adverbs and oblique participants which are characteristic of
different semantic types. Consider the following middle constructions:
The middle in (58), for instance, highlights the destiny of the cards and
shapes (i.e. in the desk) and it specifies that the result of storing them in it
is likely to be neat. The middle construction in (59), then, focuses on the
time span in which the process can be carried out (in seconds), the way in
which it can be carried out (magnetically) and the destiny of the wood val-
ance (to a standard curtain rod).
To conclude, middle constructions may be prototypically patientive and
facility-oriented, but they allow for other processes and Subject-entities as
well and they can highlight a wide range of facets of the interaction be-
tween the non-agentive entity construed as Subject and the process. What
seems to characterize all middle constructions, irrespective of the process
which they are centred on, the entity which they construe as Subject and the
facet(s) which they select for highlighting, is that the properties of the Sub-
ject-entity are presented as making (or not making) it conducive to a par-
ticular process. This conduciveness, I will argue in the following section, is
basically a modal meaning which is realized through the unique construc-
tional link which middles establish between a non-agentive Subject and an
active finite verb: it is the specific way in which middle constructions code
the relation between a process and a non-agentive entity - be it Patient,
Location, Means or Instrument - which typifies the construction and pro-
vides the right generalization.
It has been argued before that middles imply a modal meaning (Fellbaum
1985; Fagan 1992; Massam 1992; Iwata 1999). In Massam (1992: 122), it
is posited that all middle constructions contain what is called a null element
"which is in essence equivalent to the modal can". Like the auxiliary can,
middles are argued to realize the modality of possibility/ ability (Massam
1992: 124; Fagan 1992: 54). Fagan (1992) thus suggests the following
paraphrases for the middle constructions in (64a):
However, no evidence has thus far been adduced in favour of a modal in-
terpretation of the middle: the claim that middles realize the modality of
possibility/ability is only based on the observation that middles seem to be
paraphrasable by clauses with can followed by a passive verb. In this sec-
tion, I will try to provide arguments for the modal nature of middle con-
138 The middle construction
The meaning of this middle deviates from that of This pan cleans easily, in
which no adverbial modal is present: rather than create "double modality"
(Halliday 1970a: 330fnl2) and reinforce the meaning of the middle This
pan cleans easily, the epistemic modality of possibility that is expressed by
the adverbial possibly introduces a new scope and forms a middle which
comes closer to the structure
(68) Plunge under cold water: the thin elastic membrane and any pieces of
fat should peel off easily. (CB)
(69) This book could sell. (Roberts 1987: 233)
The modal should in (68) realizes the modality of virtual certainty, and, as
Roberts points out, the modal could in (69) "can only be epistemic here",
i.e. it can only mean it is possible for the book to sell and not the book was
able to sell (Roberts 1987: 233). Including the speaker's comment on the
possibility or probability of the process is, in short, an option in the system
of middle formation that requires the explicit realization of a modal auxil-
Middle constructions and modality 139
iary or adverbial. It does not shed light on the relation between the Subject
and the finite verb in middle formation in general.
Do middle constructions then implicitly realize the deontic or root mo-
dality of ability? Research on middle formation has in general never got
beyond the observation that middles can be paraphrased by passive clauses
with the modal auxiliary can. As also hinted at in Roberts (1987), the mid-
dle itself is unable to contain the modal can in its ability meaning:
In my view, passive clauses with the modal auxiliary of ability can do not
agnate with middle constructions because they cannot be systematically
related to them: as I argued earlier, not all middle constructions can be re-
lated to a passive construction:
clauses which express the modality of ability, it is clear that the potency is
located in the Subject as well (e.g., He can play the piano).
I propose that the modality of middle constructions comes close to the
kind of modality realized in the Modern English dynamic modals. In the
middle construction as well, it is the Subject which forms the locus of po-
tency directed towards the realization of the process. However, the potency
of the middle Subject is not the one which is typically associated with cau-
sation, i.e. it does not itself bring about the process. Rather, the Subject has
properties that make it conducive to a particular process: by being con-
strued as Subject of a middle construction, a non-agentive entity is pre-
sented as having the necessary properties or potential to let a process be
carried out. Interestingly, 'letting' has in Talmy's force-dynamic frame-
work (Talmy 2000: 443) been characterized as being part of the "greater
modal system" of language. The letting-foTce which is unleashed by the
middle Subject is, unlike the locus of potency that unleashes it, typically
realized maximally subjective, i.e. it is not explicitly realized in the clause.
I will show that one of the arguments in favour of the presence of an objec-
tive, dynamic type of modality in middle constructions comes from the
occurrence of a subtype of middle construction in which it is explicitly
realized: structures such as The door would not open express "negation of
willingness" (Quirk et al. 1985: 229) or the absence of inclination on the
part of the Subject. They thus explicitly deny the letting modality that is
implied in the middle construction.
Like the dynamic modals of willingness and ability, middle construc-
tions realize a kind of modality that is objective·, the modality of letting is
internal to the proposition and specifies the representational meaning of the
relationship that is established in it. Like dynamic modals, middle construc-
tions therefore also have schematic epistemic modality (as described in
Chapter 4, Section 2.3.1, based on Verstraete 2002). This may either be
realized as indicative mood, or it takes the form of an explicit epistemic
modal, expressing the modalities of possibility, probability or certainty and
situating "the process at varying distances from the speaker's position at
immediate known reality" (Langacker 1991: 246), as illustrated in (74) and
(75):
(74) Plunge under cold water: the thin elastic membrane and any pieces of
fat should peel off easily. (CB)
(75) This book may sell, (personal example)
Because it can take explicit epistemic modals, the modality that is realized
in the middle comes closest to the periphrastic type of dynamic modals
142 The middle construction
such as be able to and be willing to, which can also be used with explicit
epistemic modals (e.g., He may be able to come, He may not be willing to
do it).
When middles are used in the indicative (which expresses the speaker's
commitment to the truth of the proposition, see Verstraete 2002; Davies,
2001), they can be part of immediate or 'proximal' reality, which typically
coincides with the present tense; or they can be situated in non-immediate
reality, i.e. in the past (Langacker 1991: 245).47 Examples of past middles
are:
(76) The lap portion of the car seat belt slotted into place easily, but the
belt clip at the back was rather difficult to thread. (CB)
(77) "They rolled the green just before the match and it ran three seconds
faster," said Curtis .... (CB)
tion (discussed in Section 1 of this chapter as the ergative and the passive
interpretations of middle formation). The non-agentive entity construed as
Subject of a middle structure is a letting-entity or a conducive entity, i.e.
one whose properties either let through the energy flow instigated by the
Agent, or blocks it.
Conduciveness, and consequently also the notion of 'letting', can be ar-
gued to constitute general properties of middle constructions in that they
schematize over the various process types and Subject-entities which the
middle can take, as well as over the wide range of facets of the interaction
between them which middles can select for highlighting. The schematic
characterization in terms of letting and conduciveness counterbalances the
traditional focus on the prototypical middle construction with a facility-
oriented profile and a patientive Subject. It constitutes a schema which is
instantiated by all middle constructions, prototypical as well as non-
prototypical. In, for instance, the green ran faster in (80), the properties of
the green are claimed to have let one run faster on it; the middle in (81)
conveys that the properties of the ladder are such that it lets one fix it
quickly and easily under the window ledge. Likewise, the middle in (82)
expresses that the properties of the Subject-entity are such that it lets the
process of washing be carried out with a good result. The umbrella in (83),
finally, is characterized as letting one fold it up.
(80) "They rolled the green just before the match and it ran three seconds
faster," said Curtis .... (CB)
(81) ... the ladder fixes quickly and easily under the window ledge. (CB)
(82) It washed well and didn't shrink much .... (CB)
(83) This umbrella folds up. (Fellbaum 1986: 9)
(84) ... Mr. Fabius chose to exit by a false door. ... the scenery began to
shake in response to the ex-prime minister's desperate efforts, but still
the door would not open. (CB)
(85) ... Mrs. Dambar was having some difficulty with the new dress she
had bought .... For some reason, it wouldn't zip all the way up in the
back. (CB)
Middle constructions and modality 145
(86) I tried to open the door, but the key wouldn 't turn. (Quirk et al. 1985:
229)
(87) I am at α sentence that will not write. (Jespersen 1914-1929, 3: 349)
It has been pointed out in the literature that the negation of will/would in
these clauses seems to "have something of the personificatory force of 're-
fusal'" (Quirk et al. 1985: 229; see also Ehrman 1966; Huddleston 1969):
mrs. Dambar's new dress in (87), for instance, as it were refuses to be
zipped. The modal auxiliary won 't/wouldn't in this use denies the typical,
letting properties of the Subject and shifts the focus to the fact that the Sub-
ject instead hinders the implied Agent in carrying out the process. When
the modal will in middle constructions is followed by internal negation, its
meaning seems to move more towards the epistemic, even though one can
also argue that there is a notion of preventing or resisting something that is
not desired, as in (88) and (89):48
(88) Just press and personalize clothes in minutes with clear, legible im-
prints that will not wash out. (CB)
(89) Place gifts in cardboard boxes that won't crush easily. (CB)
(90) ... although the key turns, the door will refuse to lock (or to unlock).
(CB)
(91) ... when the matches refused to strike. (Jespersen 1914-1929, 3: 348)
In Fagan (1992: 211), the German variant of this type of 'let itself -
construction is elaborately discussed and classified as a type of middle con-
struction:
5. Conclusion
I have suggested that the existing analyses of middle formation can be re-
fined in three ways. First, the construction's representational potential can
be mapped out more accurately and turns out to include also intransitive
processes. Secondly, middle constructions vary according to which facet of
the designated event that they highlight: in addition to facility-oriented
middles, also feasibility-, destiny- and result-oriented middles can be dis-
tinguished and linked up with the occurrence of specific adverbs or oblique
participants. And, finally, the link between a non-agentive Subject and an
active Finite, rather than restriction to a specific verb class, typifies the
middle construction. The essentially modal value of the relationship which
middles establish, and, more particularly, the letting force-dynamic status
of it, allows us to siring all the characteristics of middle formation together.
The middle's emphasis on the properties of the Subject-entity ties in with
the Subject being the locus of the potency of the letting modality; the fact
that the middle profiles a letting relationship explains the meaning of con-
duciveness which all middles have in common, as well as the non-agentive
nature of the Subject and the need for an implied Agent to actually carry
out the process. By being construed as Subject of a middle construction, a
non-agentive entity thus acquires speech functional responsibility and its
properties and the fact that they let one carry out certain processes (in a
particular way) are highlighted. In Chapter 7,1 will show how these proper-
ties of the clausal middle construction shed light on the category of non-
agentive -er nominalization and, ultimately, on the category of deverbal -er
derivation itself.
Chapter 7
A multifunctional approach to -er nominalization
Within a schematic network ..., certain nodes and relationships are far
more prominent and important than others, both cognitively and linguisti-
cally. In particular, special significance attaches to the nodes that function
as the category prototype and as the highest-level schema. The prototype is
significant because of its developmental priority and notable cognitive sali-
ence. As the primary basis for extension, it defines the center of gravity for
the category. The highest-order schema is significant because it embodies
the maximal generalization that can be extracted as a characterization of
the category membership. (Langacker 1987a: 380-381)
0. Introduction
have run parallel with the increased productivity of clausal middle forma-
tion, and even with the Subject-centred nature of contemporary English as
such. Possibly, the rise in -er nomináis based on nouns and other, more
complex expressions (such as hardliner, down-and-outer, lefthander) will
constitute another factor of instability within the system of -er nominaliza-
tion in general, which will again refashion the system and may have reper-
cussions for deverbal -er nominalization as well. But for now, the Subject
generalization seems to be the most accurate generalization for deverbal -er
nominalizations.
I claim that non-agentive -er nominalizations can be accounted for in
terms of subjecthood, provided that the link between non-agentive -er
nominalization and middle formation is recognized for what it is, viz. as a
relationship of agnation which is based on a fundamental correspondence
between the speech event-related choices made within the Subject-Finite
unit of the middle and those made in non-agentive -er nominalization.
Middle constructions are thus much more fundamental to the analysis of
non-agentive -er derivation than has been suggested thus far (in Levin and
Rappaport 1988; Rappaport Hovav and Levin 1992; Lemmens 1998). To
elucidate their role, however, it is essential that their own Subject-Finite
structure is interpreted correctly, i.e. as being centred on an active finite
verb and allowing for a wide range of verb types, including intransitives. If
non-agentive -er nomináis are related to middle constructions, then, this
would solve two things. Firstly, non-agentive -er nomináis construing a
Patient are not linked to a verb in the passive voice. Secondly, oblique par-
ticipants such as Locatives also fall within the generalization and can be
said to relate to the Subject function of the middle. The relationship of ag-
nation between middle formation and non-agentive -er nominalization
should, moreover, be seen as a basically schematic relation between sys-
tems. Non-agentive -er nomináis instantiate schemata which are partly
identical to those instantiated by middle formation. Each system, however,
also has its own schemata, which account for the unique character of its
instantiations. In short, if analyzed precisely, the relationship of agnation
between the middle construction and non-agentive -er nominalization can
refute all arguments raised against a Subject generalization for deverbal -er
nominalization.
To map out the overall schematic network of deverbal -er suffixation
and unveil the schematic choices which it has in common with the Subject
and the Finite in clausal constructions, a multifunctional analysis is re-
quired: the representational semantics of -er nominalization has to be inte-
grated with the constructional (or, in Halliday's terms, the interpersonal or
speech-event related) status of the relationship which -er nomináis imply
Representational semantics 151
and with their textual functioning (see Chapter 4, Section 3). It is only
through a multifunctional analysis that we can string together the form and
meaning aspects of -er nominalization and come to an account of its prop-
erties which is truly natural.
The outline of my analysis is as follows: first (in Section 1), I will give a
brief overview of the representational options which -er nominalization
offers and list some of the subschemata which can be said to instantiate the
schematic categories of agentive and non-agentive deverbal -er suffixation.
In a second part, I propose a semantic typology of non-agentive -er nomi-
nalization, modelled after that of middle clauses (Section 2). In the third -
and most substantial - section, I present an elaborate account of the con-
structional relationship which -er nomináis establish between an agentive
or non-agentive entity and a process type. I go more deeply into the distinc-
tion between lexicalizations and ad hoc nominalizations and point to the
different kind of interpersonal relation which they imply between an entity
and a process and the differences in textual functioning that seem to corre-
late with it. In a fourth and final section, then, I elaborate on the Subject-
generalization which I propose for the system of deverbal -er nominaliza-
tion.
1. Representational semantics
Agent) to turn the pages. Similarly, a weeper is a movie that makes you
weep and may, in fact, have been designed to make you weep (Panther and
Thomburg 2002: 181).
These constitute the main subtypes of agentive -er nominalization.
Cross-cutting this classification and further refining it are various other
distinctions. One can, for instance, distinguish between human Agents (e.g.,
baker, preacher), animals (e.g., wood-pecker, retriever), inanimate agent-
like entities or Instruments which designate material processes (e.g.,
toaster, stapler, scraper) and inanimate Agents based on immaterial proc-
esses (e.g., pointer 'a piece of advice or information which helps you to
understand something or find a way of making progress', reminder,
thriller) (based on Marchand 1969; see also Panther and Thornburg 2002).
Somewhat more peripheral in terms of agentivity, but nonetheless closer to
the agentive than to the non-agentive group of -er nomináis are those
nomináis that derive from mental processes or processes of sensing (e.g.,
admirer, lover, believer) (Halliday 1994). They designate the Senser of the
mental process which they are based on. Nomináis like container and
holder are derived from relational processes expressing possession and
part-whole relations and belong to the agentive rather than to the non-
agentive schematic category because they do not designate the things that
are in the container/holder, but the entity which contains/holds them. The
schematic category of agentive -er nominalization, in short, spans a wide
range of participant roles.
Non-agentive -er nomináis fall apart into two main categories: there are
those that designate a patientive participant, and there are -er nomináis that
profile oblique participants such as Locatives and Instruments (for the no-
tion of oblique participant, see Chapter 5, Section 1.2.1). Within the first
group, a further distinction can be made between nomináis that profile pro-
totypical Patients or inert, affected entities (e.g., scratcher, squeezer),
nomináis that designate entities which merely specify the Range of the
process (Halliday 1994) (e.g., reader, leaner), and nomináis which profile
the Medium or central participant of an ergative process, which is not
purely patientive but coparticipates in the process (e.g., broiler, cooker,
steamer).50
The other group of -er nomináis can be further divided into, firstly,
nomináis profiling oblique locative participants (e.g., kneeler 'a kind of
chair which you have to kneel on', stepper 'a raised surface on which you
154 A multifunctional approach to -er nominalization
have to step as a way of working out' Jotter 'something which you can jot
things on or in'); and, secondly, nomináis designating instruments which
lack agent-like properties, such as walker, stroller and viewer.
Without doubt, more and finer-grained representational categories can
be distinguished among agentive and non-agentive -er nominalizations.
This is, however, as far as I will take the representational description: the
main aim of my analysis is to integrate the overall representational seman-
tics of deverbal -er nominalization with the system's constructional proper-
ties. In the following section, I will elaborate on the semantics which all
non-agentive -er nominalizations have in common, and I will propose a
semantic typology of non-agentive -er derivation much along the lines of
that which I set out for middle constructions in Chapter 6. The semantic
correspondences between non-agentive -er nominalization and middle for-
mation will then in Section 3 be shown to derive from a more fundamental,
lexicogrammatical equivalence between the Subject-Finite unit in middle
constructions and the relationship which deverbal -er suffixation estab-
lishes in non-agentive nomináis.
more than middle clauses, -er nomináis focus on this process, which forms
the ultimate "destination" of the profiled entity (Lemmens 1998: 138).
Think, for instance, of Levin and Rappaport's (1988: 1078) analysis of
broiler as 'a chicken bred to be broiled'. Another example is killers, desig-
nating 'cattle ready for slaughter' (Oxford English Dictionary on CD-rom,
as cited in Lemmens 1998: 138):
(1) As killing cattle or killers (cattle ready for killing) they are inferior to
corn-fed stock, (attested in 1937, Supplement to the Oxford English
Dictionary 1972)
what is being referred to is that it sells well/ easily: the -er nominalization
picks up on the facility- or quality-oriented meaning which is prototypically
associated with middle formation and specifies how easy it is to carry out
the process on the entity construed as Subject or how fast it is sold. As
pointed out by Lemmens (1998: 139), even a nominal like seller, in which
the modal adverbial has been left out, is oriented towards the facility/speed
with which the action of selling is carried out: if you describe a particular
thing as a seller, it is implied that it sells well. Two other examples of non-
agentive -er formations that highlight the way in which a process can be
carried out are easy-rider and easy-walker. An easy-walker is the latest
type of stroller, which has properties that facilitate walking with it because
it has only three wheels which are highly manoeuvrable. An easy-rider is a
kind of motorbike which rides easily or smoothly, as in
(2) The 1.41 rides smoothly while the 1.6 is much firmer. (CB)
(3) It is a little whistle that attaches to your car with self-adhesive tape.
(CB)
can sleep (well); a train that is referred to as a sleeper is a train which has
all the facilities to let travellers sleep.51 The various non-agentive -er nomi-
náis that designate items of clothing also seem to profile especially if {fea-
sibility) and how (facility) they allow a certain process to be carried out:
waders are boots with which you can wade; slippers are heelless shoes
which one can slip into easily.52 Other non-agentive -er nominalizations
also express feasibility, but especially foreground the fact that undergoing
the process on which the nominalization is based is what the profiled entity
has been designed for (Lemmens 1998). It seems to me that nomináis such
as kneeler, squeezer, jotter and stepper especially foreground the processes
which the entities that they profile have been designed for. A kneeler thus
designates a type of chair which is designed in such a way that you can
kneel on it; a squeezer is a kind of bottle that one has to squeeze to extract
something from it; a jotter is meant to jot things down in; a stepper is a
raised surface that one has to step on as a way of working out.
A final group of non-agentives, then, can be argued to resemble the re-
sult-oriented type of middle formation. I am thinking here of the -er nomi-
nalizations that profile items of food, such as cooker, steamer, broiler and
fryer. They designate animals and vegetables that have, in many cases,
been bred or raised to be prepared in a certain way.53 But they also imply
that, if a particular process is carried out on the item of food in question, the
result will be good. Middle structures such as those in (4), which Levin and
Rappaport (1988: 1078) associate with non-agentive -er nomináis referring
to food, primarily hint at how to prepare the food to achieve the most deli-
cious or the best possible result: the adverbial well is used here in its result-
oriented meaning. A boiler, for instance, designates a type of chicken that
is best boiled and used for soup; a cooker is an apple which is best cooked
and used for, for instance, applesauce.
In short, the semantic distinctions that were introduced for the analysis
of middle formation seem pertinent to the description of non-agentive -er
nominalization as well. Importantly, they elucidate what all non-agentive
-er nominalizations have in common, viz. the conduciveness of the profiled
entity towards a specific process and the fact that the properties of the pro-
filed entity are responsible for it. It is the properties of the easy-walker that
make it easy to walk with it; it is because it has an opening at the front, that
158 A multifunctional approach to -er nominalization
a front-loader can be loaded via the front; a viewer has properties that allow
you to view photographic transparencies with it; a squeezer is squeezable,
i.e. it has properties which enable you to squeeze it when you have to ex-
tract something from it; a broiler has certain properties which turn broiling
into a good way of preparing it. Like middle clauses, non-agentive -er
nominalizations may even foreground specific facets of the interaction be-
tween the process and the non-agentive entity which they profile, ranging
from the feasibility of the process, the facility or speed with which the
process can be carried out, and the place it is oriented towards, to the result
of carrying it out. However, since few non-agentive -er nominalizations
explicitly indicate the facet which they want to highlight, vagueness with
respect to the distinction between facility-, destiny- and result-orientation
seems to be part of the system of non-agentive -er nominalizations.
In this section, I want to put forward the claim that the semantic correspon-
dences between non-agentive -er nominalization and middle formation
which I described in Section 2 reflect a more fundamental level of corre-
spondence between both construction types. I argue that non-agentive -er
nominalizations establish a relation between a non-agentive entity and a
process which is equivalent to that of the Subject-Finite unit in middle
clauses: like middle clauses, non-agentive -er nomináis designate a rela-
tionship that is characterized in terms of objective, dynamic modality, and,
like middle constructions, non-agentive -er nominalizations have a sche-
matic, epistemic modal claim mapped onto this relationship:
tions that are designated and serves to derive a more elaborate meaning: a
cooker is therefore more than an apple that cooks and a broiler is more
than a chicken that broils. In particular, the objective type of modality that
is realized in non-agentive -er nomináis comes close to the dynamic mo-
dalities of volition and ability, because it relates a process to an entity
which is conceived as the locus of potency of the modality: the profiled
entity in non-agentive -er nominalizations may be non-agentive, but it is
the locus of potency of the process because it lets a process be carried out,
i.e. it is conducive to a specific process. A kneeler is thus conducive to
kneeling, i.e. it enables you to kneel on it; a sleeper is a train that has beds
that let you sleep; a walker is a frame that makes it possible for you to walk,
it lets you walk. Importantly, these correspondences with the middle's Sub-
ject-Finite complex suggest that non-agentive -er nominalizations can be
analyzed as profiling what is clausally construed as the locus of potency of
a dynamic modal relationship, i.e. they profile a subject-like entity.
Like middle constructions, non-agentive -er nomináis moreover imply a
schematic, epistemic modal claim or position in addition to the objective
modality of their propositional content (see Chapter 4, Section 2.3.1). As in
middle structures, the epistemic modal position of the speaker in non-
agentive -er nominalizations relates to the 'truth' of the proposition: the
relationship which is designated by non-agentive -er nomináis is presented
as 'known to be true' and it is the profiled entity that is held responsible for
the truth of it (Chapter 4, Section 2.4). In particular, non-agentives imply
that the entity which they focus on has properties that make it 'modally' or
'speech functionally' responsible for the truth of the proposition: the fact
that the profiled entity is conducive to a particular process and lets it be
carried out is, as in middle constructions, attributed to the properties of the
profiled (Subject) entity. Kneelers, walkers and sleepers thus have proper-
ties that make them conducive to the processes of kneeling, walking and
sleeping. The fact that non-agentives imply that the modal responsibility of
the relationship which they designate is assigned to the entity which they
profile is reminiscent of the modal responsibility of the Subject in the
clause. It therefore provides additional support for the claim that non-
agentives relate to the clausal Subject profile.
It is the modal value of the relationship which non-agentive -er nomi-
nalizations imply which explains why they are so-called 'non-event' nomi-
nalizations, i.e. why they do not presuppose that the process which they
nominalize has actually taken place (Levin and Rappaport 1988; Rappaport
Hovav and Levin 1992). Something can be called a squeezer before anyone
has actually squeezed in it precisely because the -er nominal makes a mo-
dal claim about the entity which it profiles: the entity is presented as having
The constructional properties of -er nominalization 161
pages'. Rather than being a direct causer, the profile of an agentive -er
nominalization may also be an 'instigative setting' (Vandenberghe 2001):
nominalizations such as gusher 'a flowing oilwell' and squirter 'a container
that ejects liquid in a jet from a narrow opening' relate to what have been
termed "Setting-Subject" clauses by Langacker (1991: 346), as in
(8) A supertanker continues to gush oil off the coast of Spain. (CB)
(9) Tank is non-functional and does not squirt water. (CB)
(11) All Purpose Peeler - Peels fruit and vegetables quickly and safely.
Simply the best. (CB)
(12) ... a clever stainless steel double grater that grates ingredients finely
or coarsely.... (CB)
In the next section, I will show that the analysis of ad hoc nominaliza-
tions helps to complete the picture of the Subject-Finite-like relationship
which deverbal -er nominalizations encode. While lexicalizations were
shown to prefer modalized Subject-Finite relationships, it will be argued
that ad hoc nominalizations exploit the other option which the Subject-
Finite unit in clauses offers and designate primarily temporally grounded
instances of a process type.
I pointed out earlier that the distinction between lexicalized -er nominaliza-
tions and ad hoc formations is centred on differences in the degree to which
-er nominalizations are 'entrenched': while lexicalizations are sufficiently
automatized to be themselves included as units or fixed expressions in the
language system, ad hoc formations are derived 'ad hoc', on the basis of a
schematic unit or structural pattern (Chapter 3, Section 1.2). Especially
interesting for my analysis of deverbal -er suffixation is that lexicalizations
and ad hoc formations also differ in their semantic and formal properties.
As to the meaning of ad hoc nominalizations, I referred earlier to Levin and
Rappaport's (1988) 'event' interpretation, according to which ad hoc nomi-
nalizations designate events or processes that have actually occurred, in
contrast with the so-called 'non-event' value of lexicalizations. Regarding
the formal properties of ad hoc nominalizations, I have in Chapter 5 (Sec-
tion 2) pointed to the frequent occurrence of adverbial-like adjectives (e.g.,
He is a hard worker, Mackenzie 1990: 137), the impossibility to change the
accent position of the verb (e.g., He is a PHOtographer of snow-scenes,
Mackenzie 1990: 137), and the frequent use of postmodifying structures
(e.g., the Welsh destroyer of Del Harris's hopes, CB).
I will argue that, in spite of their semantic and structural differences,
lexicalized and ad hoc -er nominalizations instantiate the same high-level
constructional schema of deverbal -er nominalization. Like lexicalized -er
nominalizations, ad hoc ones can be shown to establish a relationship
which is similar to that which in the clause is realized between the Subject
and the Finite. Of the options that are available within the Subject-Finite
constellation, however, ad hoc nominalizations tend to choose options
which are partly different from those realized in lexicalizations. Ad hoc
nominalizations, in other words, realize a low-level schema which is differ-
ent from that of lexicalizations. More precisely, unlike lexicalizations, ad
hoc nominalizations tend to establish a relationship between an entity and a
process which is primarily temporal (see also Strang 1968, 1969).
166 A multifunctional approach to -er nominalization
(13) One guy jumped right into the fight, but his friend immediately van-
ished. The police came and hauled off the fighter, after which the van-
isher promptly reappeared, laughing. (Ryder 1999b: 283)
(14) I'm not a frequent resigner. (CB)
In (13), the -er nominalizations vanisher and fighter have definite, specific
reference and refer back to participants which were introduced in the pre-
ceding discourse context. I will call them 'phoric ad hoc nominalizations'.
Phorie ad hoc nominalizations differ from ad hoc nominalizations such as
resigner in (14), which is indefinite and part of a non-referential nominal:
rather than establish the presence of a referent in the universe of the dis-
course, it ascribes a quality to the entity construed as Subject of the copular
clause (Kuno 1970). In what follows, I will consider the properties of both
types of ad hoc nominalizations more closely.
(15) One of them was faking. ... Could the faker keep up free association
...? The faker, whichever he was, had practised or had natural talents.
(Kastovsky 1986b: 410)
The constructional properties of -er nominalization 167
(16) One guy jumped right into the fight, but his friend immediately van-
ished. The police came and hauled off the fighter, after which the van-
isher promptly reappeared, laughing. (Ryder 1999b: 283)
(17) "Not him," says Mrs. Cook. "This one, though. A bit of a poseur, but
he moves very nicely." I will not describe the nice mover, to spare his
blushes, but let us say that he had dressed for the part. (CB)
(18) Recent research has established that merely breathing smoke-laden air
- passive or 'second-hand' smoking - puts the breather at risk.
(CB)
(19) I was a teacher at an approved school and I saw a staff member beat
up a boy. I was asked to turn a blind eye but I couldn't and because of
that I was shunned by colleagues who expected me to support the
teacher. (CB)
The nominalization the teacher has definite reference and refers back to the
staff member mentioned earlier. The contextually defined reference mass of
teacher thus consists of only one instance, over which the teacher quanti-
fies universally. Lexicalizations can also have indefinite reference. As
pointed out in Chapter 4, when it is used indefinitely, the nominal type
specification does not define a reference mass, but serves a purely "classifi-
cational" function (based on Davidse 2000b: 1113). Indefinite lexicalized
nomináis may present the designated entity as indefinite, but specific (and
therefore paraprasable as 'a certain X', Dirven p.c.), as in:
(20) But it was when a teacher realized the lead singer of the group,
Freddie Mercury, was a bisexual who had died of AIDS last fall, that
the controversy began. (CB)
In this section, I will show that nomináis such as resigner in I'm not a fre-
quent resigner are situated in between phoric ad hoc -er nominalizations
and true lexicalizations. Unlike the former, they have indefinite grounding,
The constructional properties of -er nominalization 169
but in contrast with the latter, they tend to make their type specification
more informative, by elaborating it with pre- and postmodifiers.
As argued in Davidse (2000b, 2001), the type specification in indefinite
constructions has a purely classificational function and it "has to enable the
hearer to conceptualize instances corresponding to the categorization used
by the speaker" (Davidse 2001: 13). For this reason, the speaker will take
care to use a type specification which is sufficiently identifiable or clear. In
the case of ad hoc formations, I would suggest, this tends to lead to more
elaborate type specifications: because these -er formations are not them-
selves entrenched or lexicalized units, they tend to be turned into more
elaborate, higher-order type specifications. The higher-order type specifica-
tion in ad hoc formations is typically realized by means of explicit markers:
as in ordinary nomináis, a higher-order type specification is derived when
pre- and/or postmodifying structures are added to the head noun, i.e. the -er
derivation as such:
(22) ... he could swing the ball and was a precocious taker of wickets.
(CB)
(23) She was, in that relatively brief time, a girlfriend, a pregnant woman,
... the carrier of an unwanted second child,.... (CB)
(24) I was too self-absorbed to be a giver of pleasure. (CB)
(25) To get herself back out, she befriends a myriad of oddballs including
... a professional loser of boxing matches. (CB)
(28) Legal secretary about her boss: I swear, the moment I need to talk to
Max, he's suddenly gone. I'm beginning to think he's a professional
vanisher. (Ryder 1999b: 282)
(29) Ask a frequent flyer like Margie Crase .... (CB)
(30) I'm not & frequent resigner. (CB)
Unlike what has been suggested in, for instance, Levin and Rappaport
(1988) and Mackenzie (1990), postmodification is also possible in nomi-
náis with lexicalized -er nouns. However, because the head of such nomi-
náis is a lexicalized -er noun, the postmodifier may in that case serve vari-
ous functions: it may be restrictive and derive a higher-order type specifica-
The constructional properties of -er nominalization 171
tion which is still more elaborate than the type specification of the lexicali-
zation (as in 34). Or it may be non-restrictive, as in the nominalization in
(35):
(34) ... Allen has told his lover of 12 years that when the custody hearing
begins next week there would be nothing of her left standing. (CB)
(35) The teacher, of West London, was led away to the cells .... (CB)
In his oldfriend was 'a giver, not a taker ', however, they are used as if they
are lexicalizations and designate the type specification 'someone who has
the habit of giving/taking'.
172 A multifunctional approach to -er nominalization
(38) ... he could swing the ball and was a precocious taker of wickets.
(CB)
(43) One of them was faking. ... Could the faker keep up free association
...? The faker, whichever he was, had practised or had natural talents.
(Kastovsky 1986b:410)
speech event). It is therefore in the Subject profile that the semantic com-
monality of all -er nominalizations lies.
I pointed out in Chapter 5 that the notion of Subject has figured in the
analysis of -er nominalization before, but then as a more or less empty no-
tion, the meaning of which is primarily linked to the representational role
that is prototypically mapped onto it, viz. that of Agent (e.g., Marchand
1969; Bauer 1983; Keyser and Roeper 1984; Levin and Rappaport 1988;
Mackenzie 1990; Rappaport Hovav and Levin 1992). Halliday's multifunc-
tional analysis of the clause, however, makes clear that the functions of
Subject and Agent form part of two distinct layers of organization (see
Chapter 2, Section 4). The Subject function clusters together with the vari-
ous options which are available to the language user to ground the proposi-
tion and assume a modal position with regard to it. As such, the Subject
serves an 'interpersonal' function. The Agent, on the other hand, expresses
how the speaker experiences the world which he describes and forms part
of the 'representational' metafunction of language. If we are to come to an
accurate account of the Subject function, and, related to it, of the profile of
-er nominalizations, therefore, the Subject function has to be described in
terms of the interpersonal functions which it serves.
I argued in Chapter 4 that for the description of the function of Subject,
a syntagmatic or configurational analysis is required which considers the
different steps in the organization of the clause as a construction related to
the speech event, viz. the functions of process type specification, quantifi-
cation and grounding. I thus identified three distinct functions of the clausal
Subject: it was argued to function as instantiator of the clausal process type
specification; it was shown to co-operate with the finite element of
grounded clauses by establishing objective or explicit person deixis; and,
finally, it was identified as the modally responsible element of the clause
(Halliday 1994).
I suggest that what deverbal -er nominalizations profile is comparable to
the function of the Subject in clauses. Firstly, -er nominalizations charac-
terize people and things in terms of a particular process type which they
instantiate. -Er suffixation applies to a process type specification and es-
sentially represents a mechanism whereby a process type (given by the base
verb) is instantiated by tying it to an entity (marked by -er). As in the
clause the instantation is not necessarily causal in nature but may also in-
volve other types of force, such as as that of 'letting'. Secondly, -er nomi-
nalizations imply categories of grounding and thus also profile the 'gram-
matical person' to which the process is tied. More specifically, the person
deixis which they establish is that of the 'third person' (Davidse p.c.). Fi-
nally, like the clausal Subject, the entity which comes to be profiled
176 A multifunctional approach to -er nominalization
through -er suffixation is held 'responsible' for the proposition that is im-
plied in -er nominalization: in it is vested the truth of the proposition. The
entity either has a certain potential for carrying out a process or for letting
others carry it out (in lexicalizations); or it is an entity that has done/ is
doing/ will do something (in phoric ad hoc nominalizations). Lexicalized
-er nominalizations in addition typically profile entities in which, as in the
Subject of the dynamic modals of ability, volition and letting, a certain
potency is located: the profiled entity has the necessary skills/ knowledge/
physical appearance/ design/ inherent properties to carry out or let others
carry out a process. In phoric ad hoc nominalizations that establish a rela-
tionship of which the precise temporal status has to be retrieved from the
context no such potency is implied.
5. Conclusion
causing). Finally, the modal value of lexicalized -er nomináis was shown to
contrast with the predominantly temporal focus of ad hoc nominalizations.
To conclude, a multifunctional approach to deverbal -er nominalization,
or an approach which is not restricted to the representational properties of
the system, but includes the constructional categories which it realizes and
the textual functioning of the nouns that result from it, manages to offer a
new perspective on the mechanisms of -er derivation. It unveils the interac-
tion between the nominal and clausal interpersonal categories of type speci-
fication, instantiation and grounding and it shows that the high-level se-
mantics of -er nominalization is identifiable only in terms of these catego-
ries.
Part III
Factîve nominalization
Chapter 8
Factive nominalization: Towards a descriptive
position
0. Introduction
The gerundive, that- and the fact i/zaZ-constructions that I want to focus on
in Chapter 9 are of the following types:
(1) To this day he regrets that one of the unanticipated side effects of the
Welfare State was the disappearance of granny and grandpa from the
family hearth into the old people's homes. (CB)
(2) I don't like his having to do so much .... (CB)
(3) One characteristic regarding farm holidays is the fact that the different
offers are so varied. (CB)
Structures like the ones in (1) and (2) have traditionally been described as
complement clauses, or as clauses functioning as Subject or Object in an-
other clause (e.g., Kiparsky and Kiparsky 1971; Menzel 1975; Noonan
1985; Wierzbicka 1988).55 All three construction types are typically charac-
terized as being 'factive', a feature generally thought of as the "presupposi-
tion by the speaker that the complement of the sentence expresses a true
proposition" (Kiparsky and Kiparsky 1971: 345).
Strikingly, the system of factive nominalization tends to be primarily as-
sociated with ¿Aaí-structures, which are conceived as prototypical instances
of factive nominalization. As to the position of gerundive constructions and
the fact /Aai-structures, opinions diverge. Lees (1968: 59-60), for instance,
includes only ίΑαί-structures in his category of 'factive nomináis':
The Kiparskys (1971), on the other hand, seem to view gerundives (as in
6b) primarily as 'tests' for identifying factive predicates and their that-
182 F active nominalizatioti: Towards a descriptive position
(13) a. That there are porcupines in our basement makes sense to me.
[factive]
b. It makes sense to me that there are porcupines in our basement.
But c. *That there are porcupines in our basement seems to me. [non-
factive]
d. It seems to me that there are porcupines in our basement.
Finally, Kiparsky and Kiparsky (1971: 362) show that both factive and
non-factive clauses can be replaced by it, which is the "pro-form of noun
phrases" (see 14a and 14b), while only non-factive clauses can take so, or
the "pro-form of sentences", as illustrated in (15):
(14) a. John supposed that Bill had done it, and Mary supposed it, too.
[non-factive]
b. John regretted that Bill had done it, and Mary regretted it, too.
[factive]
(15) a. John supposed that Bill had done it, and Mary supposed so, too.
[non-factive]
b. *John regretted that Bill had done it, and Mary regretted so, too.
[factive]
What is interesting about the Kiparskys' approach is not only that they
situate factivity within a specific formal paradigm. Important is also that
they do not restrict the notion of factivity to one structural element only: in
Factivity as embedded projection 185
their discussion of the factive constructions in (16) and (17), for instance,
the Kiparskys first state that the "factive sentence ... carries with it the pre-
supposition 'it is raining'" (Kiparsky and Kiparsky 1971: 348). Then they
continue by saying that the "speaker presupposes that the embedded clause
expresses a true proposition" (Kiparsky and Kiparsky 1971: 348). Their
conclusion, finally, is that "predicates which behave syntactically as fac-
tives have this semantic property" (Kiparsky and Kiparsky 1971: 348).
As Delacruz (1976) points out, the Kiparskys thus use the notion of the fac-
tive presupposition in three different senses: to refer to the sentence as a
whole, to the main verb that functions in it, and, thirdly, to the speaker ut-
tering the clause (Delacruz 1976: 180). And, of course, the complement or
nominalization itself is also called 'factive' (Kiparsky and Kiparsky 1971:
356). There is no reason why the four interpretations of factivity should be
incompatible.56 In fact, they illustrate that, rather than being traceable to
one structural unit only, factivity results from afusion of several factors.
The Kiparskys may point to a number of interesting lexicogrammatical
differences between clauses which are factive in meaning and clauses
which are not, but the precise relationship between the formal characteris-
tics of the factive category and the notion of factivity is never made clear.
As Davidse (1994: 260) puts it, "tracing systematically different paradig-
matic affiliations to establish distinct constructions is α heuristic technique,
not α grammatical explanation". The notion of factivity should be linked
up with the category's grammatical behaviour, and because they refrain
from doing so, the Kiparskys ultimately fail to shed light on it. The Kipar-
skys' characterization of facts as propositions that are 'presupposed true'
thus remains vague and requires further elaboration.
First, the term 'embedding' is not new: it is also found in mainstream lin-
guistics, where it is used in a meaning which is closely associated with that
of 'complementation': embedded clauses are defined as complements or
subordinated clauses that function as arguments of another clause (Noonan
1985: 42). Traditionally, the /^-clauses in (18) and (19), for instance, are
both analyzed as embedded clauses or complements (examples taken from
Noonan 1985: 90):
ever the horse stopped, he fell o f f , for instance, both clauses are direct rep-
resentations of experience. When, on the other hand, one of the clauses in
the relationship represents experience not directly but instead designates
"linguistically processed phenomena" (Halliday 1968: 195), Halliday pre-
fers to speak of projection. Projections are metaphenomena (Halliday 1994:
249): they are representations "of a (linguistic) representation" (Halliday
1994: 250). They represent "phenomena already encoded in language"
which then participate in other linguistic structures (Halliday and Hasan
1976: 131). Among the most important subtypes of projection which Halli-
day distinguishes are the following (Halliday 1994):
These clause complexes differ in terms of the process type that functions in
their main clause and in terms of the relationship that exists between the
two clauses which they consist of: quotes are paratactically related to
clauses which typically contain a verbal process (i.e. asked in 21). Reports
are hypotactically related to the 'dominant' clause: in (22), that she wanted
to go home thus depends on the main clause Mary said/thought. Reports
can follow mental processes (e.g., thought), as well as verbal processes
(e.g., said). When they are construed with a mental process, they are tradi-
tionally referred to as 'indirect speech' representations.
Halliday (1968, 1994) analyzes facts as projections which are not tactically
related to the clause with which they are used, but are embedded in it as a
constituent. Facts are said to typically combine with clauses that contain
mental and relational processes, and, less frequently, verbal processes. An
example of a factive nominalization that functions in a relational clause or a
clause that sets up a relation between two entities (Halliday 1994: 119) is
given in (23):
(23) Mr. Archer said it was significant that Queensland, a growing sector
in the Australian property industry, had been granted this important
recognition. (CB)
(24) I regret the role of mother isn't generally more highly regarded by
everyone. (CB)
(25) Its owners cheerfully admitted that the bird was mad, and seemed
genuinely offended when I suggested it needed to be put down. (CB)
(26) David thinks that all championships would be better run on Grand
Prix lines. (CB)
(27) King pointed out that Meredith's admission was merely a crumb
thrown black Americans in lieu of real equality for all. (CB)
(28) Probably you resent that peace demonstrations include kids who can
wait out the war in college. (CB)
(29) I regret that it all happened. (CB)
come from and what do they stand for? In his (1976) article, Delacruz ob-
serves that the sentence Bill regrets that John resigned can receive a read-
ing in which both the speaker and Bill regard the proposition John resigned
as a fact, as well as a reading in which John resigned is a fact to Bill (i.e.
the processer of the mental process regret) only. Building on Delacruz's
observation and on the basis of a corpus study of factive constructions,
Davidse (2003) suggests that there are three basic interpretations of the no-
tion of factivity and that these interpretations can be linked to a number of
general semantic schemata. Firstly, facts may be facts for the speaker only
when they function in discourse contexts that imply no other consciousness
than that of the speaker (e.g., in relations of 'addition' or 'causality',
Davidse 2003; e.g., Add to that the fact that over half the menus the NRA
analyzed include sandwiches, and you 've got a trend worth noting, CB). In
environments which either represent or imply a 'second consciousness'
distinct from the speaker's, by contrast, facts may be facts either to both the
speaker and the processer that is interacting mentally with the proposition
or to the processer only. In The council have woke [sic] up to the fact that
Glasgow and a lot of areas are slums (CB), for instance, the use of the verb
wake up conveys that the speaker was subscribing to the factivity of the
proposition that Glasgow and a lot of areas are slums before the council
came to accept it (Davidse 2003). A pure 'processer fact', on the other
hand, can be found in What the 37 year-old finds most soul destroying
about Haiti is the fact that things have not changed since his childhood
(CB): the processer the 37year-old is depicted here in his or her emotional
reaction to a proposition which s/he perceives as a fact but which is not
necessarily a fact to the speaker (Davidse 2003). Note that even facts that
are facts to the processer only represent propositions which are not pro-
jected by the clause with which they combine - as are reports and quotes -
but which are 'pre-existent to' (Davidse 2003) and 'manipulated by' the
relation in which they participate.
By characterizing facts as projections, Halliday manages to elucidate
their commonality with clause complexes that have typically figured in the
discussion of factives, i.e. those involving direct and indirect speech and
thought. It is the status of factive nominalizations as embedded constituents
which turns out to be crucial to distinguish them from the other types of
projection. In what follows, I will go more deeply into the arguments that
have been given in the literature to support the analysis of factives as em-
bedded in rather than hypotactically related to the clause with which they
combine. The discussion is based on Halliday (1968, 1994), Halliday and
Hasan (1976) and Davidse (1991, 1994).
F activity as embedded projection 191
Halliday's claim that factive clauses are embedded entails that they are
viewed as downranked constituents which serve a function in another
clause. Which arguments can be given in favour of the downranked status
of factive structures? Evidence that ί/ζαί-factives fill a nominal slot in the
clause comes from the observation that, unlike their hypotactic counterparts
in (31), the nominalized ίΑαί-constructions in (30) can be coded as Subject
of the passive (as in 30a), they can be given thematic prominence by being
fronted (e.g., 30b), and they can be extraposed as Objects, while the Object
slot is filled by it (see 30c):
In spite of the fact that they do not take over all of the functions which an
ordinary nominal can fulfill in the clause, however, factive /Λα/-structures
clearly function as nomináis and their embedded or nominal constituent
status shows up in many of their lexicogrammatical properties.
F activity as embedded projection 193
It also becomes clear now that the 'tests' which the Kiparskys use to
distinguish factive from non-factive clauses primarily pick up on the dis-
tinction between hypotaxis (non-factive) and embedding (factive). That
factive clauses cannot be substituted by so, for instance, has to do with their
embedded, nominalized status: as shown by Halliday and Hasan (1976:
131-134), substitution can only be used with the same function and class as
the wording which it substitutes for. In the case of hypotactic clauses, so is
possible because the clause which it replaces serves the function of clause.
By contrast, embedded or rankshifted clauses have lost their clausal status
and function first and foremost as nomináis in another clause (Davidse
1994: 274). Reference, in contrast, does not depend on whether the refer-
ence item has the same function or class as the clause which it refers to
(Halliday and Hasan 1976; Davidse 1994). It is therefore possible to have
substitution in (45), but not in (46). In (46) only reference is acceptable:
(45) Do we really have to wait in line? It seems so. Drew says so.
(46) I regret that you had to wait in line. I regret it/*so too.
Finally, the fact that factive nominalizations cannot have their Subject
turned into the Subject of the main clause (as shown in 49a) or figure in
'accusative and infinitive constructions' (as illustrated in 49b) is a direct
result of their embedded status as well (Davidse 1991: 354). Because fac-
194 F active nominalization: Towards a descriptive position
tive clauses are rankshifted into nominal slots as constituents of the clause,
they cannot move beyond the boundaries of the nominal element of struc-
ture.
2.4. Conclusion
Kiparsky and Kiparsky (1971: 360-361) point out that there exist numerous
predicates that are ambiguous and have both a factive and a non-factive
use. A sentence like that in (50a), for instance, is said to have a factive and
a non-factive reading, evidence for which is provided by the minimal pairs
in (50b) and (50c) (Kiparsky and Kiparsky 1971: 360):
(50) a. They reported that the enemy had suffered a decisive defeat.
b. They reported the enemy to have suffered a decisive defeat, [non-
factive]
c. They reported the enemy's having suffered a decisive defeat,
[factive]
The sentence in (50c) implies "that the report was true in the speaker's
opinion" (Kiparsky and Kiparsky 1971: 360). It is factive and consists of a
systematic agnate of the factive type of ¿/¿«/-structure, viz. a gerundive
nominalization. The to-infinitive in (50b), on the other hand, "leaves open
196 F active nominalization: Towards a descriptive position
the possibility that the report was false" (Kiparsky and Kiparsky 1971:
360). As I argued before, the accusative and infinitive construction is only
possible with non-factive clause complexes, because it requires a direct
relationship between two clauses. The Kiparskys' claim that the projected
clause is not presupposed true and may be 'false' picks up on the hypotactic
nature of the clause complex: the clause the enemy to have suffered a deci-
sive defeat is hypotactically projected by the process of the main clause and
represents the projection of the main clause Subject, rather than that of the
speaker. The hypotactic clause in (50b) thus implies that the fact that the
enemy had suffered a decisive defeat is only what they reported; it does not
necessarily coincide with the speaker's opinion.
An example of a verb which can be used in different verb senses and
which, depending on the meaning in which it is used, is either factive or
non-factive is explain (Kiparsky and Kiparsky 1971: 361):
In (51), explain is used in the meaning of 'give reasons for', while in (52) it
means 'say that S to explain X' (Kiparsky and Kiparsky 1971: 361). In Hal-
liday's model, (51) is considered factive because the structure Adam's re-
fusing to come to the phone functions as embedded projection: rather than
being projected or directly asserted by the clause I explained, the I-person
expresses some assertion about the (pre-existent) proposition. The that-
clause in (52), on the other hand, is a non-factive, hypotactic projection of
the main predicate: it is directly projected by the Subject of the main proc-
ess and represents the locution that is implied in explain.
The Kiparskys rightly observe that the list of verbs with both a factive
and a non-factive reading is quite extensive: many constructions are am-
biguous, even those that contain what seem to be prototypically factive
verbs like regret·, the sentences in (53) and (54), for instance, differ in
meaning from the one in (55): regret is factive in (55) and expresses a kind
of affect, with the that-clause an embedded projection. In (53) and (54), on
the other hand, the ίΑαί-clause is hypotactically projected by regret, which
is used in the meaning of 'regret to say'. The /Aa/-clauses in it consequently
realize a locution rather than a fact.
(53) Irma regrets that she cannot answer your letters personally. (CB)
(54) The Times also regrets that enclosures accompanying letters cannot
be returned. (CB)
Delineation of the fact category 197
(55) He trained during the Gulf War and even regrets that he was never
called up to take part in any action. (CB)
Apart from ambiguous predicates, there exist also predicates which are
characterized by vagueness and which have features in common with both
the factive and the non-factive category. Among the processes that have in
the literature been argued to belong to this type are verbs describing "proc-
esses of knowing or coming to know" (Hooper 1975: 117), such as know,
discover, find out, learn, and predicates such as deny, doubt, it is possible,
it is true, it is false (see, among others, Kiparsky and Kiparsky 1971; Kart-
tunen 1971; Hooper 1975; Wierzbicka 1988).
(56) Anyone who has seen Internet knows that it uses amazing technology.
(CB)
(57) ... she also found out that her father has been in touch with Derek all
this time. (CB)
(58) ... he discovered that Sharon's parents, both known drug takers, had
been repeatedly attacking their daughter. (CB)
(59) The agent of a sacked Coronation Street star denies that she is on the
verge of a nervous breakdown. (CB)
(60) It is true that in the Seventies there was no shortage of 'stupid gym-
nasties', as the Bhagwan once called the act of lovemaking. (CB)
(61) ...it is possible that this year we may suffer again. (CB)
It has repeatedly been pointed out in the literature that predicates like the
ones in (56) to (61) are located somewhere in between 'true' factives and
non-factives. The process know has thus been described as "semantically
factive", but "syntactically non-factive" (Kiparsky and Kiparsky 1971:
348).58 It is, for instance, impossible to say *I know the fact that John is
there or *I know John's being here, which seems to suggest that know is
non-factive. Also, know takes the accusative and infinitive construction,
which is normally only possible with hypotactic or non-factive structures
(e.g., I know him to be there, based on Kiparsky and Kiparsky 1971: 348).
Similar observations have been made with regard to it is possible and it is
true·, like factive predicates, they cannot have their Subject raised to the
Subject-position of the main clause (e.g., It is possible that he will accom-
plish even more cannot be turned into *He is possible to have accomplished
even more). Like factives, moreover, they may or may not take extraposi-
tion (e.g., It is possible that we may suffer again or That we may suffer
again is possible). Unlike factives, however, clauses construed with predi-
cates such as it is possible and it is true do not alternate with the fact that-
198 F active nominalization: Towards a descriptive position
constructions and gerundives (e.g., *The fact that we may suffer again is
possible/true·, *John's being ill is possible/true) (Kiparsky and Kiparsky
1971).
For one thing, structures like these show that the distinction between
hypotactic projection and embedded projection is not always clear-cut and
that analyses of particular configurations are likely to confront us with the
existence of categories that lie in between hypotactic and embedded projec-
tion. Constructions like the ones in (56) to (61) thus reveal the need for fur-
ther clarification of the way in which factive /^-structures are related to
the clause with which they combine and ultimately, their semantic and
structural properties will have to be integrated in a description of factive
nominalization that is aimed at being exhaustive. My own analysis of fac-
tive nominalization, however, will focus on the internal functional proper-
ties of the factive category, rather than on the delineation of the external
contexts in which it occurs.
The problems related to the delineation of the factive category within the
group of gerundive nominalizations were already pointed out by Lees
(1960) in his groundbreaking study on nominalization in English. In his
analysis of what he calls 'gerundive nomináis', Lees (1968: 58) shows that
a distinction should be made between the following two subtypes:
(64) I think my only regret, as I say, was not being able to train for the job
I wanted to do. (CS)
(65) Threatening people will get you nowhere. (Halliday 1994: 248)
(66) Tearing up my new paper dolls was mean. (Thompson 1973: 378)
(67) Gathering quietly around the campfire each evening was nice.
(Thompson 1973: 377)
The situation becomes even more complicated when it turns out that, unlike
what Lees claims, the action type of gerundive nomináis (i.e. the type of
gerundive nominal that is not factive but 'action-referrent') can in some
cases take a possessive or periphrastic Subject, as in:
(68) Techniques for doing this include the caregiver's exposing the infant
to various contingent and discrepant experiences. (CB)
(69) My husband speaks very well, but his job involves my answering the
phone on his behalf quite a bit of the time. (CB)
(72) a. Zelda's reluctant signing of the contract surprised the entire crew.
b. Zelda's reluctantly signing the contract surprised the entire crew.
c. That Zelda reluctantly signed the contract surprised the entire crew.
Compare this with the meanings in which the simple noun party can be
used (Langacker 1991: 33):
(75) Harvey's cruelly taunting the bear was a severe blow to his campaign
for presidency.
(76) Harvey's cruelly taunting the bear is something that could simply
never happen.
Towards an internal, nominal analysis 205
instance, with the consequence that the derived noun is inherently definite".
Like pronouns and proper names, therefore, /Aai-clauses do not require
"separate grounding" to function as nomináis (Langacker 1991: 148). I will
refer to the nominal strategy which embedded /Aa/-structures employ as the
'proper name strategy'.
The unique combination of clausal and nominal properties which that-
factives realize can now be mapped out in more detail. With ¿Aaí-structures,
the nominalization process applies to a finite clause. By being downranked,
this finite clause is reified and construed as an abstract entity, with its proc-
ess scanned holistically rather than sequentially. Rather than effecting fun-
damental changes to the clausal internal properties of the finite clause, the
downranking mechanism reclassifies the clause into a nominal, capable of
discharging nominal functions such as grounding according to the proper
name strategy. Like proper names, factive ίΑαί-clauses establish an entity
which is definite in itself and can therefore function as a fully grounded
nominal.
Thus, by following Langacker's analysis of factive iAa/-constructions,
one can analyze them as nominal-like not only externally, but also inter-
nally. This counters both Huddleston's and McGregor's analysis of embed-
ded ίΑαί-structures. McGregor (1997: 127) argues that finite clauses cannot
be rankshifted because rankshift or reclassification always implies that the
downranked unit "takes on the grammatical and semantic properties inher-
ent to the lower ranking unit". Finite clauses "show none of the formal
properties ... which are characteristically associated with nominalizations"
(McGregor 1997: 131). It is especially the finite or fully grounded status of
/Aai-clauses which, according to McGregor, prevents them from being
downranked as nomináis: he argues that finite clauses "always represent
situations as situations, not as entities" (McGregor 1997: 131).
Against McGregor (1997), it can now be argued that that-
nominalizations do function as nomináis internally. First, as pointed out by
Langacker (1991: 448), it is essential to acknowledge that with embedded
/Aa/-structures "there are multiple levels of organization to consider, and
multiple paths through which the contents of a /Aa/-clause are accessible
from the ground". Reclassification need not be marked by means of a re-
duction of the grounding options that are available within the downranked
clausal unit. Nor does it always require that the reclassified unit adopts par-
ticular morphological and syntactic properties that are characteristic of the
accommodating class. Instead, reclassification is junctional in nature.
In the following section, I will zoom in on the fact ¿/¡^/-constructions
and show how they provide additional evidence for the 'proper name'
analysis of í/zaí-structures.
the fact ihat-constructions: A case of apposition 211
As pointed out in Chapter 8 (Section 4.1), Halliday analyzes the fact that-
constructions like those in (1) and (2) as that-factives functioning as post-
modifiers of the head noun:
(1) He was, I think, troubled by the fact that the Democratic Party
couldn't put together a majority on the national level.... (CB)
(2) And I don't like the fact that it's been carnivalized. (CB)
Quirk et al. (1972: 647, 1985: 1260) likewise treat the that-clauses in the
fact ¿Aírf-constructions as postmodifiers of the fact-noun. To distinguish
them from the restrictive type of relative postmodifying clauses (as in 3),
they describe /Aai-structures in the fact iAaf-constructions as 'appositive'.
(3) The news that appeared in the papers this morning was well received.
(Quirk et al. 1985: 1244)
(4) the thought that she might one day be a queen (Halliday 1994: 264)
(5) the news that the team had won (Quirk et al. 1985: 1244)
Quirk et al. (1985) group these //¡ai-structures with the appositive type of
postmodifying clauses. Halliday (1994: 263) calls them "embedded locu-
tions and ideas", functioning as postmodifiers within a nominal group. He
212 F active nominalizations as nominal constructions
argues that, like ί/ζαί-clauses postmodifying the noun fact, they represent
projections or metaphenomena, but, unlike in the fact iAaf-structures, "the
projecting element is the noun that is functioning as Thing" (Halliday 1994:
263). In the nominal in (4), for instance, it is the noun thought which pro-
jects the that- clause that follows it.
The analyses which Quirk et al. (1972, 1985) and Halliday (1994) sug-
gest for the relationships between embedded /Aa/-clauses and the noun
classes that precede them can now be summarized as follows:
restrictive relative
the fact that the thought/claim that
clauses
apposition relative apposition
head-postmodifier head-postmodifier head-postmodifier
/ expansion projection
the ίΑαί-clause is an em- the that-clause is an em- the /Aa/-clause is an em-
bedded projection or bedded direct represen- bedded projection or
metaphenomenon tation of experience metaphenomenon
Quirk et al. (1972: 647, 1985: 1260-1261) analyze the fact ίΑαΖ-clauses as
consisting of two units which are related to each other in a restrictive, ap-
positive way (see also Langacker 1991: 149). But what characterizes re-
strictive apposition? Apposition is said to be a relationship between "con-
stituents of the same level" which are "identical in reference" (Quirk et al.
1972: 620). In, for instance, Paul Jones, the distinguished art critic, both
Paul Jones and the distinguished art critic refer to the same person: the
relation between them can therefore be described as appositive. Appositive
units are restrictively linked when they form a single information unit, i.e.
when they are uttered on a single tone unit and are not separated by com-
mas in writing: the apposition between Paul Jones and the distinguished art
critic in Paul Jones, the distinguished art critic can thus be analyzed as
non-restrictive. Examples of restrictive apposition are the famous critic
Paul Jones, the singer Robeson and the River Thames.
The characterization of the fact ίΑαί-structures as appositive nominal
groups thus implies that they establish a relation between two units which
are of equal status. At the same time, however, Quirk et al. (1985: 1261)
classify the /¿¿¡/-clauses in the fact /Aû/-constructions as postmodifiers. I
argue that the appositive nature of the fact /Aaf-constructions is not recon-
cilable with their being described in terms of modification (see also Biber et
al. 1999: 645) and that, consequently, the use of the notion of postmodifica-
tion in the description of the fact /Aai-constructions is unjustified.
Let us have another look at Halliday's account of the structural relations
that may exist between members of a complex. Halliday (1994: 221) argues
that the notions of hypotaxis/parataxis and expansion/projection (described
in Chapter 8, Sections 2.1 and 2.2) "define complexes at any rank" and can
also be used to describe complex relations at the level of the nominal. In
general, parataxis is the "linking of elements of equal status", while hypo-
taxis is the "binding of elements of unequal status" (Halliday 1994: 221).
The initiating and continuing element in a paratactic relationship are 'free'
"in the sense that each could stand as functioning whole", while in a hypo-
tactic relationship only the dominant element is free (Halliday 1994: 221).
Parataxis thus forms an alternative to a relationship of modification, which
always involves one element modifying another (Halliday 1994: 218).
As far as I can see, the the fact ίΑαί-nominal and the ίΛαί-clause are
paratactically and not hypotactically linked to each other: they are not re-
lated through modification, but form elements on the same level, with each
of them having the potential to stand on its own. The construction the fact
that Caesar was dead can, for instance, be replaced by the fact or by that
214 Factive nominalizations as nominal constructions
Caesar was dead without affecting the acceptability of the clause in which
it is embedded:60
Each of these expansion types can be combined either with parataxis (as in
all the examples in 7) or with hypotaxis. In fact, in considering the various
possible mappings, Halliday defines 'apposition' as a combination of
elaboration and parataxis (Halliday 1994: 225): when two items are in ap-
position, the second item elaborates the first and specifies it in more detail.
The type of semantic relationship which the fact ίΑαί-constructions es-
tablish can now be identified. Halliday (1994) only gives a semantic char-
acterization of embedded /AaZ-clauses: they are projections without explicit
projecting clause. However, when a that-factive comes to function as the
the fact that-constructions: A case of apposition 215
(10) the assertion that/ They asserted that Caesar was ambitious.
216 F active nominalizations as nominal constructions
(11) the thought that/ They thought that Caesar was ambitious.
(12) the claim that/ They claimed that Caesar was ambitious.
In view of the fact that there is no reason to assume that a noun can serve a
projecting function, I propose to analyze constructions of this type also as
establishing a semantic relationship of elaboration. Like the fact that-
structures, the /Aa/-clauses in the claim/thought /Aai-constructions restate
and specify the nominal that precedes it. Like the fact /^/-constructions,
the claim/ thought /^/-constructions thus fit into the paradigm of apposi-
tion and realize a combination of parataxis and elaboration.
It follows that the appositive relationship established in a nominal such
as the thought that Caesar was ambitious does not differ from that in the
fact //^/-constructions: both are based on embedded projections, both para-
tactically relate the two units which they consist of, and in both the that-
clause restates and specifies the nominal that precedes it. In general, the
relationships that are established in the fact //¡^/-constructions, restrictive
relative clauses and constructions of the type the claim that can now be
summarized as follows, with the shaded areas indicating that the proposed
analysis deviates from the analysis suggested in Quirk et al. (1972, 1985)
and Halliday (1994) (see Table 4):
the fact Ubat-constructìons: A case of apposition 217
restrictive relative
the fact that the thought/claim that
clauses
head-postmodifier, i.e.
parataxis parataxis
hypotaxis
expansion: extension/
expansion: elaboration expansion: elaboration
elaboration
the ίΑαί-clause itself is the that-ciause is an the f/zaf-clause itself is an
an embedded projection embedded direct repre- embedded projection or
or metaphenomenon sentation of experience metaphenomenon
What constitutes a proper name, then, is derived from the way in which an
item functions in a construction and units which are not proprial lemmas
may also be used as proper names (Van Langendonck 1999: 96). The close
apposition test thus identifies not only proprial lemmas (e.g., Robeson in
the singer Robeson), but also structures of which the primary function is
not that of proper names. Van Langendonck (1999: 118) gives the follow-
ing examples:
(18) The ugly fact that he was holding a gun indicated his guilt. (Quirk et
al. 1985: 1262)
nominal. Like the second nominal in the appositive structure the poet
Burns, then, the /AaZ-clause in the fact //¡aZ-structures is more specific and
specifies which fact is being referred to. Because the finite //ia/-clause is
uniquely grounded, it functions as a full nominal (see Section 1). In terms
of Van Langendonck's analysis of restrictive appositive constructions, the
//¡¿¡/-clause can be said to serve the function of proper name in appositive
the fact fAa/-constructions.
The position of the fact that-structures in the overall system of factive
nominalization can now be clarified: they classify the (factive) /Aa/-clauses
which they contain explicitly in terms of the nominal class of fact; they
identify the class which the embedded /Aa/-constituent belongs to as being
that of facts. It is here that the main difference with embedded locutions
and ideas lies: in the thought/claim /Aa/-structures, the //¡«/-clause is identi-
fied as realizing the category of locutions and ideas, rather than that of
facts. In contexts that allow for a factive as well as a non-factive reading, a
/Aa/-clause which stands on its own as nominal may therefore be inter-
preted either as an embedded locution or as a fact:
The inclusion of the fact-noun, moreover, helps to "activate one single idea,
because nouns have a higher potential for reification than nominal that-
clauses" (Schmid 2000: 367). Because they include an explicit and nominal
classification of the //¡«/-clause, it can be argued that the fact //¡«/-structures
come closer to non-nominalized or ordinary nomináis, which might explain
why they can be used in clausal contexts that do not allow ordinary that-
factives (see, for instance, Davidse 2003), as in (20) and (21):
(20) a. We are pleased by the fact that you work so hard. (Davidse 1994:
270)
b. *We are pleased by that you work so hard.
(21) a. ... it's going to take some adjusting to the fact that Anchorage is a
one-paper town. (CB)
b. *... it's going to take some adjusting to that Anchorage is a one-
paper town.
positive relation to the fact. As such, they enter into a relation with the
nominal paradigm that normally functions in that particular position. Be-
cause in non-nominalized cases of appositives (like the poet Burns), the
type of nominal that figures as second element in a restrictive appositive
relation has the properties of a proper name (e.g., Burns), the /Aa/-clause in
the fact ί/ιαί-constructions can be said to connect with proper names. The
analysis of the nature of the relationship which the fact /¿«/-constructions
realize thus confirms what I suggested in my analysis of ίΑαί-factives, viz.
that /Aa/-factives are paradigmatically related to nomináis realizing the
proper name strategy.
(22) Since old Nanapush had saved her from death ..., her visiting the man
stood to reason. (CB)
(23) She appreciated the clerk's taking the time to be so friendly ... (CB)
222 Factive nominalizations as nominal constructions
It is fairly obvious that the main difference between iAai-factives and ge-
rundive factives is that the former are based on a finite clause, while the
latter are not. Which clausal level of assembly it is that gerundive nomi-
nalizations derive from, however, is unclear. In his analysis of nominaliza-
tions, Langacker (1991: 33) argues that gerundives start from an "interme-
diate level which profiles an instance that is left ungrounded". A gerundive
nominalization such as Zelda's signing the contract is thus claimed to de-
rive from the processual expression sign the contract. I will propose an al-
ternative analysis (see also Heyvaert 2000) which views gerundive nomi-
nalizations as being based on an 'atemporal clausal head' (as discussed in
Chapter 4, Section 2.1). A nominalization such as Zelda's signing the con-
tract will be analyzed as deriving from the atemporal clausal head signing
the contract. Rather than being an ungrounded instance, the processual
starting point of gerundive nominalizations is thus an atemporalized proc-
ess type specification. This is relevant for the analysis of gerundives as
nominal constructions: I will argue in Section 3.2. that it is the iype-status
of the clausal head which determines the nominal strategies that gerundive
nominalizations can adopt.
We will start from a number of well-known observations concerning the
internal status of gerundive nominalizations. As Lees (1968: 65-66) points
out, gerundive nominalizations can take auxiliaries (as in 24a), they can be
used with an adverbial (illustrated in 24b) and they are construed with par-
ticipants which are not periphrastically realized but as clausal constituents
(see 24c) (see also, among others, Fraser 1970; Chomsky 1970; Pullum
1991; Hye Suk Yoon 1996; Malouf2000):
(30) Those who ate the apples ran off wild and mad into the forest - the
apples having been poisoned by a woman who had loved Merlin ....
(CB)
Importantly, the modal auxiliaries have to in (31) and be able to in (32) and
the perfect auxiliary in having been declared in (33) do not themselves
ground the clausal head: they are atemporalized by the -ing suffix and
therefore essentially belong to an atemporalized process type specification
in which they further modify the simple verb stem.
In short, the lexicogrammatical properties of gerundive nominalizations
suggest that the processual level of assembly which they start from is that
of an atemporal clausal head. This analysis ties in with well-known obser-
vations about the clause-like status of gerundive nominalizations. As to the
precise effect of the -ing suffix on the verb it is added to, then, I would ar-
gue that, as in atemporal clausal heads, -ing imposes an atemporal profile
on the process and forms an alternative to grounding it. Rather than relating
the process to the speech event by means of modality/tense, the atemporal
profile of the -ing suffix confirms the ungrounded status of the process type
specification and serves to derive a higher-order type specification. -Ing
only changes the way in which the process is looked at: rather than scan-
ning the component states of the process in the step-by-step or sequential
way that is characteristic of finite clauses, -ing scans the component states
of the process type holistically. As also argued in Chapter 4 (Section 2.1.1),
it activates the component states of the verb cumulatively, through sum-
mary scanning.
Because they realize summary scanning rather than sequential scanning,
atemporal clausal heads are likely candidates for nominalization: as pointed
out before in the analysis of that-factives (Section 1.1), the nominalizing of
clause-like units (such as finite clauses) turns them into unitary entities of
Gerundive facts 225
(34) a. Last year, Mgr. Jacques Gaillot ... was ousted by the Vatican
for openly promoting the use of condoms to prevent the spread of
Aids. (CB)
b. ... was ousted for the fact that he openly promotes the use of
condoms to prevent the spread of Aids.
c. *... was ousted for that he openly promotes the use of condoms to
prevent the spread of Aids.
fore, Langacker (1991: 33) analyzes the gerundive nominal Zelda's signing
the contract as being derived from the 'ungrounded instance' sign the con-
tract. It is the inclusion of the Object which is held responsible for the in-
stantiated nature of the structure sign the contract·. "The fact that the nomi-
nalized structure incorporates a fully specified direct object supports the
claim that it represents an instance and not just a type" (Langacker 1991:
34). Langacker adds that even though "-ing attaches to the verb from which
this structure inherits its processual profile, it is the structure as a whole -
with its clause-like internal organization already established - that is nomi-
nalized" (Langacker 1991: 32).
I see two main differences between Langacker's analysis and the one
which I propose. First, my analysis considers the Object as part of the
clausal head or type specification and assigns an instantiating role to the
Subject only. Arguments in favour of these claims were given in Chapter 4,
Section 2.2 (see also Davidse 1997). Secondly, in my description of gerun-
dive nominalizations, the functional level of assembly at which the nomi-
nalization process takes place is that of signing the contract rather than sign
the contract. Of course, the atemporal clausal head signing the contract
itself is based on the simple process type specification sign the contract.
Crucially, however, sign the contract is not the structure which the nomi-
nalization applies to. Langacker's analysis remains vague about the precise
nature of the -ing of gerundive nominalizations and the difference with the
nominalizing -ing in action nomináis such as the signing of the contract. By
describing gerundive nominalizations as being based on an atemporal struc-
ture which includes the atemporal marker -ing, I explicitly identify the na-
ture of the -ing suffix in them as being verbal rather than nominalizing.
It can be pointed out here that the verbal status of the -ing suffix in ge-
rundive nominalizations such as Zelda's signing the contract is not recog-
nized by everyone. While in many of the descriptions of gerundive nomi-
nalizations it is posited that the suffix -ing is verbal (e.g., among others,
Horn 1975; Schachter 1976; Quirk et al. 1985; Pullum 1991; Hye Suk
Yoon 1996), some have claimed that it has nominalizing value (Baker
1985; Milsark 1988). As Milsark (1988: 622), for instance, puts it: "The
structures that have classically been called gerundives - that is, those that
occur in subject position and in object position of prepositions and of verbs
such as regret and enjoy - are ... those whose verbs have assumed nominal
categorization as a result of the affixation of -ing". At the same time, Mil-
sark considers the -ing form in gerundive nominalizations to be functioning
as head of a gerundive clause, and he argues that "the nominal features of
the derived verb form percolate up its projection path, ultimately relabeling
the entire gerundive clause as NP" (Milsark 1988: 613). In my opinion,
Gerundive facts 227
Milsark's analysis confuses the atemporal effect of the -ing suffix with
nominalization. As I pointed out before, the holistic perspective which -ing
offers on the component states of the process comes close to the holistic
viewpoint which nominalization in general has on clause-like structures
(including finite structures) (see Langacker 1991). The two should not,
however, be confused: the atemporal status of -ing does not itself effect
nominalization, witness the occurrence of atemporal -ing in subordinate (or
non-downranked) non-finite clauses. The atemporal status of -ing does,
however, seem to be conducive to nominalization and an atemporal struc-
ture can, when nominalized, function in contexts that are more typically
'nominal' in nature than can nominalized expressions which are finite.
Milsark's analysis may confuse atemporalization with nominalization,
but it does make clear that it remains to be explained how an atemporal
clausal head can be nominalized and fulfil nominal functions in the clause.
What is it that turns an atemporal clausal head into a gerundive nominaliza-
tion with nominal-like behaviour? In Hye Suk Yoon (1996), it is suggested
that the atemporal clausal component which gerundive nominalizations de-
rive from is nominalized by a process of 'zero derivation', which leaves
intact the internal processual status of the component but enables it to func-
tion as a nominal in larger configurations. I do not see any reason for ana-
lyzing the nominal behaviour of atemporal clausal heads in gerundive
nominalizations as resulting from what is basically a morphological proc-
ess, i.e. that of zero-derivation. Rather, I propose that, like the finite clause
in rAa/-factives and the fact ί/ιαί-constructions, the atemporal clausal head
in gerundives is downranked, with its internal processual status left intact.
More specifically, it is only when Halliday's construct of downranking
is interpreted as functional reclassification that the lexicogrammatical
properties of factive gerundives can be accounted for and the differences
and similarities with 'action' nominalizations can be explained. Both turn
out to involve the reclassification of a processual structure, but they realize
two different types of reclassification. The -ing of an action nominalization
such as the signing of the contract realizes the reclassification of a verb
stem (sign) into a noun (signing): it nominalizes the verb and changes the
class it belongs to, but it does not alter the rank which the verb functions
on: both the base verb and the derived noun belong to the rank of the word.
Action nominalization thus realizes a type of reclassification which does
not involve downranking. In the gerundive nominalization signing the con-
tract, in contrast, the nominalization process is not realized by -ing: the -ing
suffix is merely an atemporal verbal marker and does not shift the class of
the verb it attaches to from verb to noun. Rather, the entire atemporal
clausal head is downranked within the structure of the nominal.
228 Factive nominalizations as nominal constructions
(44) a. Only that could account for their having a suitcase with them.
(CB) [possessive pronoun, factive]
b. Techniques for doing this include the caregiver's exposing the
infant to various contingent and discrepant experiences. (CB)
[genitive, non-factive]
(45) a. ... showing one's feelings too clearly (or being highly sensiti-
ve) is regarded as a 'feminine' trait. (CB) [no determiner, non-
factive]
b. Having attended Harvard can be helpful. (Schachter 1976: 215)
[no determiner, factive or non-factive]
(46) a. Of course I regret getting pregnant. (CB) [no determiner, factive]
b. Going there was fun. (Postal 1970: 478) [no determiner, non-
factive]
(47) a. This burning the midnight oil of yours has got to stop. (Schachter
1976: 218) [definite determiner, non-factive]
b. Any talking loudly on your part will be punished. (Schachter
1976: 218) [indefinite determiner, non-factive]
232 Factive nominalizations as nominal constructions
It is this type that is most often referred to when the co-occurrence of nomi-
nal and clausal properties in gerundive nominalizations is pointed out. It is
also this type of gerundive nominalization which most clearly shows that
the atemporal clausal head has been reclassified, in this case as a nominal
head. As such, it adopts the common noun strategy and can function like
any other nominal head: it can be preceded by either a possessive pronoun
(as in 48) or a genitive (see 49), which serve to identify a definite and spe-
cific subset of it. The instance which gerundive nominalizations like these
refer to is part of the concrete, physical world of instances and is identified
in terms of the system of person, defined from the standpoint of the speaker
(Halliday 1994: 181). As regards the use of the genitive in factive gerun-
dive nominalizations, it should be pointed out that factive gerundive nomi-
nalizations do not seem to occur with the indefinite type of genitive illus-
trated in (50):
clerck 1991b: 497) - has always been overshadowed by the factive gerun-
dive prototype. Consider the following examples:
(51) Techniques for doing this include the caregiver's exposing the infant
to various contingent and discrepant experiences. (CB)
(52) My husband speaks very well, but his job involves my answering the
phone on his behalf quite a bit of the time. (CB)
(53) And then all of the sudden, it fractures again and a third time, and the
third time, there's an enormous ripping sound, and that is her having
the stroke. (CB)
(54) Some researchers have felt that having a mental illness would lead to
the individual's occupying a position in a lower social class .... (CB)
Similar to proper names, moreover, the use of the bare nominal is sufficient
to identify which kind is being referred to: definite grounding is thus part of
the internal structure of a generically used mass noun (Davidse 1999: 211).
Rather than referring to an indefinite instance in physical space (as does
milk in 59) or to a uniquely identified instance in physical space (as do or-
dinary proper names), however, generically used mass nouns designate
specific instances in type space. They are basically names for kinds that are
presented as instances. Notice that the definite status of generic action ge-
rundives is supported by the fact that they can occur as second nomináis in
close apposition:
Schachter (1976: 215) argues that this nominal "seems to name a type of
condition resulting from some completed activity". I believe Schachter's
claim to be essentially correct: some action gerundives contain the auxiliary
have. In my opinion, however, the clause in (62) can receive not only a ge-
neric reading, but also a reading in which a specific instance is being re-
ferred to. Rather than being necessarily generic, in other words, the nomi-
nal is ambiguous. On the generic interpretation, no particular instance is
envisaged. The atemporal clausal head having attended Harvard - which
designates a type specification - then comes to function as a definitely
grounded nominal with generic reference. Notice that, as I argued in Chap-
ter 4 (Section 2.2), the presence of the auxiliary of perfect aspect have does
not itself instantiate or ground the process type specification. Of course, the
secondary tense which have realizes is indirectly situated with respect to
the ground. Yet, it does not itself instantiate or ground the process type.
The nominalization having attended Harvard in (62) can, in my opinion,
also be analyzed as designating a specific instance, the identity of which is
retrievable from the context of speech. If, for instance, I am telling some-
one that I have attended Harvard, and that person answers that Having at-
tended Harvard can be helpful, the gerundive nominalization is factive in
meaning and designates a specific instance. As such, it agnates with the
nominal the fact of your having attended Harvard. Factive gerundives of
the type having attended Harvard will be discussed in more detail later on
in this section.
An example of a gerundive nominalization which resembles that in (62)
in that it is ambiguous between a generic, action reading and a factive in-
terpretation is that in (63):
Firstly the Subject of this nominal may be coreferential with the Subject of
the matrix clause; the speaker then asks the yow-person how he or she feels
about the fact that he/she has not been to art school. While this reading
might be the most obvious one here because there is a nominal in the ma-
trix clause, viz. you, which can readily be interpreted as identifying the
Subject of the gerundive nominalization (which is not the case in 62), a ge-
neric reading cannot be excluded. One can, for instance, imagine the ques-
tion being directed to the head of an art school, who has been to art school
himself: the nominalization in that case does not refer to a specific instance,
but it designates a kind of condition. The nominal not having been to art
school thus comes to be interpreted as generic, on a par with having at-
tended Harvard in its non-specific meaning. A generic reading, finally,
236 Factive nominalizations as nominal constructions
Schachter's analysis of these nomináis thus implies that they are indefinite
and take cardinal quantification. However, it is impossible to add quantifi-
ers such as some and much to the gerundives in (65) and (66). (As I will
show later on in this section, however, action-referrent gerundives can in
some very marked cases take indefinite quantifiers.) Conceiving them as
indefinite moreover seems to conflict with the fact that they designate spe-
cific instances of a process type. Against Schachter, I therefore argue that
nominalizations like the ones in (65) and (66) function as proper names,
much like the factive ίΑαί-structures that were discussed in Section 1 of this
Gerundive facts 237
(68) ... had decided to boycott me for the mere fact of having written a
book on Franco that was not a denunciation. (CB)
(69) But, as he was now recognizing, the fact of having registered as a
conscientious objector effectively closed the door .... (CB)
(70) Getting together quickly for coffee would be fun. (Thompson 1973:
377)
(71) Adjourning immediately at 4:00 was impossible. (Thompson 1973:
377)
As with the factive gerundives in (65) and (66), the specific instance of the
process type that is being referred to can be retrieved from the context. Ac-
tion gerundives of this type are also definite, and they can therefore func-
tion in appositive constructions:
238 F active nominalizations as nominal constructions
(72) Prue drew a momentary blank, still flushed with guilt over being
caught in the act of searching Luke's shack. (CB)
(73) Ayatollah Khomeini had said the act of severing ties with the United
States was a correct move. (CB)
(76) That burning the midnight oil of yours has got to stop.
Remarkably, the head can take postmodification (from you, on your part
and of yours), which specifies the instantiator of the downranked clausal
process type. The gerundive nominalizations in (74) to (76) therefore des-
ignate, as observed by Schachter, instances of a process type, rather than
being generic in focus. However, gerundives of this type which do not
specify the instantiator of the process periphrastically seem to viate more
towards generic reference, as in
It should also be noted that this - rather marginal - type of gerundive seems
possible only with action gerundives, not with factives.
Gerundive facts 239
reference to an
individual situation;
Adjourning at 4:00 was I resent being tagged a
impossible. problem boy.
proper name strategy
common
noun This burning the mid-
strategy definite or night oil of yours has got
indefinite, to stop. /
specific
Some reporting from you
would be welcome.
I have thus far argued that factive gerundive nominalizations are based on
atemporal clausal heads which are functionally reclassified and down-
ranked either as the common noun head of a nominal (with a possessive or
a genitive preceding it), or as proper names (with no determiner). Impor-
tantly, even when gerundives come to function as common nouns, they do
not behave as prototypical members of the common noun category. Pullum
(1991) rightly observes that gerundives fail to display certain properties
that characterize non-nominalized common noun heads (e.g., adjectival
modification and restrictive relative clauses are impossible; see Section
3.2). I claim that it is only by analyzing how the processual functional or-
ganization of the downranked clausal head is integrated with the nominal
structure of factive gerundives that we can understand such restrictions and
the phenomenon of factive gerundives as such.
Let us recapitulate the nominal forms of determination which factive ge-
rundives can establish. We have seen in the previous section that factive
gerundives always function as definitely grounded nomináis: they function
either as common nouns that are grounded by a possessive or a genitive in
front of the atemporal clausal head (e.g., Only that could account for [their
having a suitcase with them], CB); or they function as proper names and
the instantiator of the specific instance which they designate is retrievable
from the matrix clause or the context (e.g., [Being dropped by Scotland
before the Five Nations] was a disappointment...., CB). Strikingly, factive
gerundive nominalizations cannot take any of the other ordinary nominal
determiners: structures like [Caroline's/her opening the door] surprised me
danài regret [opening the door] do not relate to
(79) As matters stand, Tory MPs who resent John Major outmanoeuvring
them in last year's leadership coup will always find journalists more
than willing to indulge their conspirational ambitions. (CB)
(80) I don't approve of the crew's only remaining man going there. (Lees
1968:72)
(81) I think the country's making a statement is a lot bigger than partisan
politics. (CB)
(82) Its volume alone would be impressive, but its originality and sheer
brilliance vindicate completely her having been declared Doctor of
the Church. (CB)
NOMINAL CLAUSAL
instantiation, quantification, grounding instantiation
from both perfective and imperfective verbs and, if realized, their Subject
could be mapped onto both agentive and non-agentive participant roles.
Importantly, the analyses that were presented of the internal functional
organization of factive nominalizations point to the need to consider not
only prototypical or canonically coded nomináis but also more peripheral
realizations of the nominal paradigm. I mean two things by this. First, when
analyzing the nominal behaviour of nominalizations, it is essential that one
does not only look for canonical types of nominal coding. The discussion of
the internal nominal behaviour of nominalized constructions has often
foundered because nominalizations were expected to behave as it were
more 'typically nominally' than do ordinary, non-nominalized nomináis
themselves. Gerundive factives preceded by a possessive or a genitive, for
instance, have thus generally been considered to display some degree of
nominal internal behaviour, and this because they realize the canonical type
of nominal coding whereby the functions of type specification and ground-
ing are realized iconically, by separate components. Bare gerundive fac-
tives and í/zaí-factives, in contrast, do not realize grounding iconically and
have therefore hardly been recognized as being nominal-like internally as
well. It is only by taking into account the existence of non-canonical nomi-
nal codings such as non-separate symbolization of a function that one can
move beyond the existing descriptions of nominalizations and come to a
coherent analysis of their external and internal functional properties.
However, it is not only vital to consider non-canonical nominal codings
in the analysis of nominalizations; it should also be acknowledged that
nominalizations which realize the reclassification of a clausal structure will
always remain peripheral members of the nominal paradigm, i.e. members
which do not realize all of the properties that characterize prototypical
nomináis. Nominalized constructions cannot be expected to behave exactly
like non-nominalized nominal structures: their processual starting point
necessarily influences much of their internal organization. To conclude
from this that nominalizations like the ones which I discussed in this chap-
ter are not nominal-like internally, however, is too drastic. Nominalizations
can be said to realize the same schematic construction or configuration of
functions as do ordinary nomináis. The functions invoked throughout the
analysis of that-factives and factive gerundives are those of type specifica-
tion, instantiation, quantification and grounding.
What about the notion of factivity, then? What do the various types of
factive nominalizations have in common semantically, and how can this be
related to their lexicogrammatical properties? I believe that the obligatory
presence - explicit or implicit - of a Subject in gerundive factives sheds a
new light on the notion of factivity. Langacker (1991: 34) already noted
The internal organization of factive nomináis: Conclusion 247
In this study I have tried to show that it is in the dialectic between theory
and description that lies the key to an analysis of nominalization which
moves beyond existing descriptions and is able to shed a new light on spe-
cific nominalization types and on the process of nominalization in general.
My study has been based on the assumption that an accurate description of
specific nominalization patterns presupposes a coherent theoretical view on
the language system and an insightful description of nominal and clausal
categories. A significant part of it has therefore been devoted to the discus-
sion of a number of basic theoretical notions and to the introduction of
nominal and clausal functional categories. However, this study also makes
clear that it is descriptive analyses of nominalized constructions which
identify the theoretical constructs that are pertinent to the analysis of nomi-
nalization. It is also only through the description of nominalization patterns
that we can come to a better understanding of the constructional mecha-
nisms that lie behind the process of nominalization in general. The largest
part of this book has therefore been devoted not to theory, but to the de-
scription of two specific systems of nominalization.
The analysis of deverbal -er nominalization and of the various types of
factive nominalizations has provided new insight into the general process
of deverbal nominalization. Deverbal nominalization can now be defined as
the functional reclassification of a specific functional level of assembly in
the clause into a construction which does not only serve nominal functions
in larger configurations but internally makes use of nominal means of in-
stantiation, quantification and grounding. The analysis of nominalization
requires the following major steps: firstly, the level of assembly which the
nominalization derives from has to be identified; secondly, it has to be de-
termined whether the reclassification that applies to it involves rank shift or
not; thirdly, we have to identify the exact nominal strategy which the nomi-
nalized unit adopts; and, finally, the precise lexicogrammatical integration
of the clausal starting point with the nominal end-status of the nominalized
construction has to be considered: by bringing together their internal lexi-
cogrammatical features and their external functional behaviour, we can de-
scribe the unique semantic import of specific nominalization systems.
First, the various clausal levels of assembly which nominalization can
apply to turn out to coincide with the main functional levels in the organi-
250 Conclusion
zation of the clause (as identified in Chapter 4): nominalization can apply
to a simple process type specification or verb stem (e.g., bake -» baker,
sign -» the signing of the contract)·, it can start from a complex process
type specification or atemporal clausal head (e.g., playing the piano -»
[Her playing the piano] surprised us); nominalization can be based on a
non-finite clause (e.g., John playing the piano —» [John playing the piano]
surprised us), or it can apply to a finite clause (e.g., John plays the piano —>
[That John plays the piano] surprises us). The clausal categories which
deverbal nominalizations, unlike non-nominalized nomináis, imply (think,
for instance, of the categories of modal and temporal grounding in deverbal
-er nominalizations) are closely related to this clausal starting point of the
nominalization process.
Through nominalization, then, the various clausal levels of assembly are
turned into nominal constructions, i.e. they are reclassified into the nominal
class. While the nominal status of nominalizations derived from verb stems
is unproblematic, the internal nominal character of nominalizations based
on more elaborate units of clausal patterning is contentious. As my analysis
of gerundive nominalizations and /Aa/-factives reveals, the internally re-
classified nature of clause-like nominalizations shows up only when, firstly,
their downranked status is acknowledged and, secondly, when non-iconic
types of nominal coding are considered. The notion of downranking and,
behind it, the idea that language consists of levels of functional organiza-
tion which together make up a hierarchy or rank scale, is vital to the analy-
sis of nominalizations which contain units of clausal patterning: such
nominalizations establish reclassification by shifting a clausal configuration
of functions - with its internal clause-like structure intact - from clausal to
nominal rank. Through rank shift, a set of functions which normally figures
in the clause thus comes to function at nominal rank and is fitted into the
functional organization of the nominal. Generalizing, we can say that
deverbal nominalization always involves reclassification: in some cases
reclassification relates two units of the same rank (i.e. reclassification from
verb to noun). In other cases, reclassification is realized through rank shift.
To recognize the precise nominal paradigm in which rankshifted clausal
units come to function, a view of nominal organization is needed which is
radically functional and describes nominal composition in terms of compo-
nent functions which are not necessarily realized by discernible component
structures (see Chapter 2, Section 3.1). As the description of gerundive and
iAaZ-nominalizations makes clear, nominalizations, just like 'ordinary'
nomináis, do not necessarily realize the prototypical, iconic form of coding
that is characteristic of common nouns. An analysis of nominalizations
which tries to force them into what is felt to be the nominal prototype (i.e.
Conclusion 251
the common noun strategy) is therefore bound to fail and will not be able to
reveal the internal nominal status of the majority of nominalizations that
involve downranking.
The descriptive breakthrough needed to account for both iconically and
non-iconically coded nomináis and, consequently, to identify the internal
nominal properties of downranked clausal structures, was shown to be pro-
vided by Langacker's functional analysis of the nominal in terms of type
specification, instantiation, quantification and grounding. I have then
shown that, by systematically analyzing the internal organization of nomi-
nalizations from this functional perspective, their classification as nominal
can be based not only on their external functioning, but also on their inter-
nal functional behaviour.
The precise nominal paradigm which reclassified units fit into varies ac-
cording to the clausal level of assembly which is being nominalized. If a
verb stem or simple process type is nominalized, it is necessarily turned
into a common noun or nominal type specification (see Langacker 1991),
and this irrespective of the categories of instantiation and grounding that
may be implied in the relationship that is established in the resulting nomi-
nalization (think, for instance, of the analogy between the semantics of -er
nouns and the Subject-Finite unit). An atemporal clausal head is either
turned into a common noun (which is then instantiated and grounded by
means of determination) or, when instantiation is implied, it functions as a
full nominal or proper name. If the atemporalized process type specification
is construed as a unique instance, but then in type space, it follows the
proper name strategy of generic mass nouns. Likewise, a downranked finite
clause is reclassified and comes to function as a proper name, as also
pointed out by Langacker (1991). The same can, in my opinion, be argued
for nominalizations which apply to the fourth level of assembly which I
have identified, viz. that of ungrounded instances or non-finite clauses with
their Subject in the objective case (e.g., John playing the piano —» [John
playing the piano] surprised us).
Importantly, the functional characteristics of the clausal level of assem-
bly are thus integrated with those of the nominal construction in which the
reclassified unit comes to function. Downranked clausal units adopt in
some cases the common noun strategy (i.e. in gerundive nomináis of the
type his opening the door and that burning the midnight oil of yours), but
far more often, they function as full nomináis or proper names. When
nominalization involves rank shift, in other words, the default nominal
strategy seems to be that of proper names, which provide a unique name for
an instantiated process or for a unique process type (in generic gerundives).
In gerundive nomináis, it is only when the Subject-entity instantiating the
252 Conclusion
baker 'a food (meat, fruit or vegetable) that is suitable for baking'
bestseller 'a book or other item that sells well'
boiler 'a chicken suitable for boiling'
broiler 'a young chicken suitable for broiling'
cooker 'an apple for cooking*
diner 'aplace where you can dine'
dipper 'vegetable/fruit/other snack that has to be dipped before
being eaten'
dunker 'a doughnut that is preferably dunked before being eaten'
front-loader 'an appliance, such as a washing machine or a clothes drier
which allows for material to be inserted at the front'
fryer 'a young chicken or rabbit suitable for frying'
gusher 'a flowing oilwell'
jotter 'a small notebook'
keeper 'something that should be kept'
killers 'cattle suitable for killing'
kneeler 'a kind of chair, which you have to kneel on'
knocker 'a piece of metal on the front door of a house with which
you can knock on the door, to attract the attention of the
people inside*
loaner 'something that is loaned to someone'
low-loader 'a vehicle in which the carrying platform is kept low for
ease in loading'
poker 'a metal bar to move coal or wood in a fire'
reader 'a (course)book consisting of a compilation of literature '
roaster 'a kind of meat suitable for roasting'
scratcher 'a lottery ticket that one has to scratch to reveal the win-
ning patterns'
sipper 'a drink that one has to sip'
sleeper 'a train you can sleep in'
sleeper 'a bed or sofa in a train, in which you can sleep'
slipper(s) 'a kind of shoes that can be slipped into'
squeezer 'a kind of container that you have to squeeze in order to
extract something from it'
squirter 'a container that ejects liquid in a jet from a narrow open-
ing'
steamer 'an edible clam'
stepper 'a raised surface on which you have to put your foot, as a
way of working out'
256 Appendix
functions of Subject and Object, for instance, rather than being treated as se-
mantically 'empty', are given a semantic interpretation.
11. Like Halliday, moreover, McGregor (1997) relates the logical metafiinction to
the syntagmatic relation of dependency.
12. Importantly, in Cognitive Grammar, as in Systemic-Functional Grammar, it is
posited that there is "no fundamental distinction between morphological and
syntactic constructions, which are fully parallel in all immediately relevant
aspects" (Langacker 1987a: 82; see also Halliday 1961). Both morphological
and syntactic structures realize meaning in a natural or non-arbitrary way.
13. A similar distinction can be observed in the system of -ing nominalization: the
nominal opening in (1) is lexicalized, whereas opening in (2) forms an ad hoc
formation based on a schematic unit:
(1) He squeezed through a narrow opening in the fence. (CB)
(2) It entails a kind of patient regard, an opening of oneself toward
the subject.... (CB)
14. However elucidating the analysis of a construction's order of assembly may
be, it does not (always) tell the whole story: as pointed out by Langacker
(1987a, 1991, 1998, 1999), composite structures tend to exhibit partial rather
than full compositionality and, secondly, the contribution of their component
parts is not always recognized, i.e. they are not always fully analyzable. First,
as far as compositionality is concerned, the fact that components can be rec-
ognized within an assembly of symbolic units "does not entail that these com-
ponents exhaust its characterization", i.e. the composite whole "may have
properties above and beyond those of its components, which may in turn be
manifested in it only imperfectly" (Langacker 1987a: 87). The composite
structure is then either more specific than what can be computed from its
components, or it forms an extension of the expected compositional value
(Langacker 1999: 153). Nominalizations - particularly nominalizations at the
level of the word - are constant reminders of the partial nature of the compo-
sitionality of constructions. A structure like ruler, for instance, is more fre-
quently understood as a device for measuring things than as a device used for
ruling lines and its meaning thus significantly extends the expected composi-
tional value; similarly, a computer is much more than just 'something that
computes' (Langacker 1999: 128). In short, any assembly of symbolic struc-
tures must be treated as a separate entity in its own right and "with many pos-
sible sources of extracompositionality" (Langacker 1998: 4), such as dis-
course and context, particular domains of knowledge, mental spaces, meta-
phor and metonymy. The discrepancy between an expression's expected com-
positional value and the actual meaning that becomes its conventional value
facilitates a decline in the analyzability of the construction (Langacker 1999:
128): the language user becomes less cognizant of the components of the as-
sembly or no longer co-activates the composite structure and its components
(Langacker 1987a: 462). Analyzability is a matter of degree: while in a nomi-
nalization like teacher, the components may still be perceived, this is less true
for, for instance, nomináis like propeller and drawer. Many of the -er nomi-
260 Notes
nalizations that designate items of clothing likewise seem to have lost much of
their analyzability, to the extent even that nomináis like jumper, sneakers and
loafers no longer seem to be recognized as being derived from the processes
ofjumping, sneaking and loafing.
15. The systemic-functional view of nominalizations is closely tied up with the
concept of 'grammatical metaphor': nominalization is presented as a major re-
source for the creation of metaphorical rather than 'typical' or congruent lexi-
cogrammatical realizations of semantic categories. A nominalization such as
the cast's brilliant acting, for instance, is thus viewed as the metaphorical
counterpart of the clause the cast acted brilliantly (Halliday and Matthiessen
1999: 229). To describe the relationship between the congruent and the meta-
phorical realizations, Gleason's concept of agnation is used and it is claimed
that every metaphorical structure has "an agnate form corresponding to its
congruent realization" (Ravelli 1988: 141).
16. Agnates that help to distinguish gerundive nominalizations from action nomi-
nalizations are, for instance, non-finite modifiers (e.g., Any boy [cleaning the
kitchen] will be rewardedfor it) and non-finite clauses (e.g., With [Tom clean-
ing the kitchen], we can be sure that it will be very clean). Both structures ag-
nate systematically with gerundive nominalizations but not with action nomi-
náis: like gerundive nomináis, for instance, they allow for imperfective proc-
esses (conpare, for instance, any boy [being prepared to help] and with [Tom
being prepared to help] with the gerundive nominal [Tom's being prepared to
help] surprised me and with the action nominal * Tom's being of prepared;
*Tom's being of a good husband.
17. Plurals and non-plural mass nouns are grouped together by Langacker (1991)
because they behave alike in numerous respects. They can, for instance, both
occur as full nomináis without quantifier or overt grounding predication (e.g.,
They're looking for diamonds/ gold); they do not tolerate the indefinite article
(e.g., *a diamonds/ gold); and both can be used with a number of quantifiers
not permitted with count nouns (e.g., most diamonds/ gold/ *diamond; a lot of
diamonds/ gold/ *diamond) (see Langacker 1991: 77).
18. Note that for a plural, "the size of the instance will be given by the number of
its component entities", but collectively, these entities "constitute just one in-
stance of the plural-noun type" (Langacker 1991: 81).
19. As pointed out by Langacker (1991: 54), even in those three black cats, the
different structural levels are not realized entirely separate from each other:
number is indicated not only by the quantifier three, but also by the grounding
predication those and by the head noun.
20. Only when the clausal head is progressive in profile, is -ing impossible: the
auxiliary pair be playing can thus not be turned into an -ing form (*being
playing). Langacker (1991: 232) attributes this to the fact that "simpler alter-
natives are available that do essentially the same job, i.e. they provide effec-
tively equivalent information without creating potentially awkward ambigui-
ties". The -ing form playing is a simpler alternative for being playing and
therefore makes the latter redundant.
Notes 261
21. Important evidence in favour of the type status of atemporal clausal heads
comes from gerundive nominalizations and the way in which they function in
the structure of the nominal. I will not, however, go into their properties here
and refer to Chapter 9 for a discussion of the status of gerundive nominaliza-
tions as atemporalized process type specifications.
22. While both Halliday (1994) and Langacker (1991) agree on the grounding
function of the finite element in the clause, they differ as to where exactly
they draw the boundary between the grounding and the non-grounding part of
the verb group: Langacker considers only the actual markers of tense and mo-
dality as grounding; in Halliday's analysis, auxiliaries like did/does/was/
had/have are considered to be part of the finite or grounding element, and this
primarily because they are picked up in tags, such as didn't you?, has he?,
wasn't she?. In Langacker's model, the verb group has responded can thus be
divided into -s (which grounds the verb group) and have responded (which
constitutes the non-grounding, head-part of the verb group). In Halliday
(1994), that same verb group is analyzed as has (grounding) and responded
(non-grounding). When no auxiliaries are present, Halliday's analysis of the
verb group coincides with that in Langacker (1991): a verb group such as re-
sponded is divided into -ed and respond.
23. In Davidse ( 1997) and Heyvaert (2000), the view is defended that the clause is
instantiated (as well as grounded) by the Subject-Finite complex, rather than
by the Subject only. Especially on the basis of my analysis of gerundive
nominalizations (presented in Chapter 9), I have come to an analysis which is
slightly different in that it considers only the Subject to fulfill an instantiating
role. I view the function of the finite or grounding element of the verb group
as being that of grounding the instance in the speech event (which, of course,
presupposes instantiation, but crucially does not foreground it). As argued by
Davidse (1997), the Subject then co-operates in the grounding by establishing
objective or explicit person deixis (see Section 2.3.2). The Subject thus serves
two distinct, though related functions: it instantiates the process type specifi-
cation of the clause and, if the clausal head is finite, it co-operates in the
grounding.
24. The notions of 'subjective' and 'objective' as they are used here in the context
of modality should be distinguished from the use which Langacker (1985,
1990b) makes of them: in Langacker's work they do not refer to whether or
not a linguistic element is related to the speaker, but they relate to how explic-
itly the speaker is present in an utterance. The more explicit the speaker fig-
ures in the utterance, the more 'objective' its status is said to be. As Verstraete
(2001: 1513) points out, the notions of subjectivity and objectivity used by
Langacker are not entirely unrelated to the notions used in the classification of
modality: "they can be regarded as an additional dimension for the subjective
pole, depending on how explicitly the speaker-relatedness figures in the utter-
ance".
25. The subjective deontic modality of obligation/permission is expressed by the
prototypical uses of deontic auxiliaries such as must, should, shall, may.
262 Notes
26. The objective deontic modality of obligation and permission is typically real-
ized by non-Speaker-Hearer-oriented deontic modal auxiliaries, such as have
to, ought to, can.
27. Verstraete (2002: 65) observes that, when the objective modal auxiliary is a
periphrastic modal such as have to or be able to, the schematic, epistemic mo-
dality that accompanies objective modals can surface in the form of an epis-
temic modal auxiliary, as in He may have to leave his house; She might be
able to come over after all.
28. Note that the notions of 'subjective' and 'objective' are used here in a mean-
ing that is related to the way in which they are used by Langacker (1985,
1990b): the Subject is said to establish person deixis objectively because it
puts the person deixis 'on stage' as a maximally salient entity; the person
deixis that is realized in the finite element, on the other hand, is subjective, or
minimally salient: it remains implicit (based on Langacker 1991: 93).
29. In Langacker (1991: 247), a similar claim is made with respect to Subject-
verb agreement: "what is traditionally known as 'subject-verb agreement' is
analyzed as being part of the grounding predication".
30. An extensive list of English non-agentive -er nominalizations has been in-
cluded in the appendix.
31. The distinction between lexicalized and 'schematic' or 'ad hoc' nominaliza-
tions was introduced in Part I, Chapter 3, Section 1.2.
32. This view on English transitivity, in which a fundamental split is posited be-
tween transitive and ergative clause construals is proposed in Davidse (1992b,
1998b) and Davidse and Geyskens (1998). Davidse's view is rooted in the
British structural-functional tradition of Halliday (1967, 1968), Lyons (1969)
and Cruse (1972, 1973, as mentioned in Davidse and Geyskens 1998). It
should be distinguished from Perlmutter's (1978) distinction between 'unerga-
tive' and 'unaccusative' verbs: broadly speaking, one can say that Perlmut-
ter's 'unaccusative' category includes all the ergative verbs of the structural-
functional tradition, as well as intransitive verbs such as fall, die and stumble
(for a detailed comparison of the two traditions, see Davidse and Geyskens
1998).
33. Following Levin and Rappaport (1988), Lemmens claims that Goal-profiling
-er nomináis are related to the middle use of the transitive base verb: "as with
the middle construction the focus is on the properties of the Affected" (Lem-
mens 1998: 137). Lemmens's account of the relation between Goal-profiling
-er nomináis and middle formation will be discussed in Chapter 7, Section 2.
34. Interestingly, a category which is not included in Lemmens's account (it was
not accounted for in Ryder's approach either) is that of -er nomináis that pro-
file so-called 'oblique participants' such as jotter, stepper, kneeler, viewer,
stroller (I will explain the notion of 'oblique participant' (based on Laffut
2000) in Section 1.2). Together with the Goal- and Medium-profiling group of
-er nomináis, they make up the non-agentive type of -er nominalization. I will
go more deeply into this type of -er nominalization in Section 1.2.1.
Notes 263
35. In fact, it has been pointed out by Laffut (2000: 145) that the Locatum is the
Patient or primary affected of the process, but "because of the involvement in
the locative process, the Location can be analyzed as a second affected", both
when it is realized prepositionally and when it functions as a bare noun
phrase.
36. Deverbal nominalization types which have been analyzed in the literature as
offering an active and a passive voice option are those of -ion and -ment suf-
fixation, as in:
(1) a. the critics' hostile reception of the play (active, 'they received
the play')
b. the play's hostile reception by the critics (passive, 'the play was
received') (Quirk et al. 1985: 1289)
(2) a. the committee's appointment of John (active, 'they appointed
him')
b. John's appointment by the committee (passive, 'he was appointed')
(Lees 1968: 68)
37. Many aspects of the description of the middle proposed in this chapter are also
discussed in Heyvaert (1997), Heyvaert (1998) and Davidse and Heyvaert (in
press).
38. Structures like these have been variously referred to as activo-passives (Jes-
persen 1914-1929, 3: 347), medio-passives (e.g., Declerck 1991b: 203), proc-
ess-oriented passives (Halliday 1967), facility-oriented passives (Fawcett
1980: 148), patient-subject constructions (Van Oosten 1977: 459) and pseudo-
intransitives (Lyons 1969: 366; Smith 1978: 103). I will use the term 'middle
construction' (as used by, among others, Keyser and Roeper 1984; Fellbaum
1986; Hale and Keyser 1987; Fagan 1992; Kemmer 1993).
39. Hence labels like activo-passive, medio-passive and patient-subject construc-
tion.
40. Even if The door closed slowly is analyzed as a middle construction, it is
turned into an ergative one-participant construction as soon as the phrase all
by itself is added to it: The door closed slowly all by itself czanot but be inter-
preted as an ergative one-participant construal.
41. Note that the necessarily unexpressed nature of the Agent in middle construc-
tions has inspired some authors to call them pseudo-intransitives (Lyons
1969; Smith 1978).
42. For the notion of 'representational' or 'ideational' fonctions, see Chapter 2,
Section 4.
43. The intransitive type of middle is, in fact, well established in Dutch and has
received a fair amount of attention in the literature on Dutch middle
constructions (e.g., De Vries 1910; Holierhoek 1980; Van den Toorn 1982;
Hoekstra and Roberts 1993; Ackema and Schoorlemmer 1994; Peeters 1999).
44. The IDL corpus is the 38 million word corpus of the Instituut voor Neder-
landse Lexicografie 'Institute for Dutch Lexicography' in Leiden.
45. Intransitive middles in Dutch agnate with impersonal constructions in which
the oblique participant is realized as a prepositional phrase:
264 Notes
gacker 1987a, 1999) and are no longer recognized as -er derivations: are
nomináis such as jumper, loafers and sneakers still analyzable as deriving
from the processes ofjumping, loafing and sneaking? Or have they lost their
verbal meaning and evolved into simple nouns?
53. The idea that food is raised to be processed in a particular way is clearly be-
hind the following corpus example: ... can you supply me with any informa-
tion about this apple? Is it an eater or cooker and how long does it keep?
(CB).
54. The nominal singer can of course also have its lexicalized meaning here and
designate a professional singer.
55. I follow Noonan's (1985) definition of (clausal) complementation here, which
is in line with the literature on factive complementation in general (e.g.,
Rosenbaum 1967; Kiparsky and Kiparsky 1971; Stockwell, Schachter and
Partee 1973; Menzel 1975): "By complementation we mean the syntactic
situation that arises when a notional sentence or predication is an argument of
a predicate. For our purposes, a predication can be viewed as an argument of a
predicate if it functions as the subject or object of that predicate" (Noonan
1985: 42). Clausal complements differ in this respect from non-clausal or
nominal complements, which are prototypically conceived of as referring to
Objects, rather than to both Objects and Subjects (e.g., Radford 1997: 498,
519; Haegeman and Guéron 1999: 21, 124).
56. As Delacruz (1976: 180) puts it, "Possibly whenever the use of a predicate in
a sentence involves a presupposition one could say that the sentence as a
whole has or carries a presupposition. And if a sentence carries a presupposi-
tion, one might say that whoever asserts it presupposes whatever is presup-
posed by the sentence".
57. As pointed out in Davidse (1994: 280-281), there is a special sense associated
with fronted locutions. The sentence That he had been wrong was definitely
said by him, for instance, has a semantic value which differs from that of He
said that he had been wrong: "On the one hand, ... it specifies the content of
the Sayer's speech act. On the other hand, the embedded status of the locution
entails speaker commitment to the 'fact' that this locution was produced by
the Sayer" (Davidse 1994: 281). Also structurally speaking, fronted locutions
come close to facts: they can function as Subjects of the passive; they cannot
be substituted, but only be referred to (That was definitely said by him). They
can function as second nominal in appositive constructions (The words: 'Iwas
wrong' were definitely said by him, embedded locution: quote). Davidse
therefore proposes to consider the fronted type of locutions as embedded con-
stituents, rather than as tactically related clauses.
58. As correctly pointed out by Wierzbicka (1988: 49), this is "a contradiction in
terms", which "undermines the whole idea that semantic factors (such as fac-
tivity) 'account for the complement system of English, and other languages as
well'".
266 Notes
59. This has inspired Noonan (1985: 60) to coin the term 'nominalized comple-
ments' for them, to be distinguished from fAai-structures and infinitivals
which are termed 'complements'.
60. As I will point out in Section 2.2, not all clausal contexts allow for replace-
ment of a the fact ίΑαί-clause by a that-clause. This has primarily to do with
the higher degree of 'nominality' of the fact /Aai-clauses: they can be fitted
into more clausal contexts than can /AaZ-clauses. In Subject position, however,
the fact that- structures are typically replaceable by í/taí-structures.
61. Structures can be ambiguous between a restrictive relative reading and an ap-
positive, factive one, witness the nominal the fact that she forgot in The fact
that she forgot is unimportant (Langacker 1991: 432). On the relative clause
interpretation, the fact is interpreted as a participant in the relative clause: she
forgot the fact. In its factive reading, in contrast, that she forgot is what con-
stitutes the fact.
62. The possessive pronouns in gerundive nominalizations can also be turned into
objective case (see, among others, Sweet 1898; Poutsma 1929; Lees 1960;
Declerck 1991b): e.g., We have always been a close couple and shared every-
thing, so I never resented [him leaning on me when he was depressed]. (CB);
I do not like [him coming here so ofien], (Sweet 1898: 120).
References
Databases:
COBUILD corpus, Birmingham
INL corpus, Leiden
Croft, William
2000 Parts of speech as language universale and as language-particular
categories. In Approaches to the Typology of Word Classes, P.
M. Vogel, and B. Comrie (eds.), 65-102. Berlin/New York:
Mouton de Gruyter.
Dalton-Puffer, Christiane
1994 Are Shakespeare's agent nouns different from Chaucer's? - On
the dynamics of a derivational sub-system. In Studies in Early
Modern English, D. Kastovsky (ed.), 45-58. Berlin/New York:
Mouton de Gruyter.
Davidse, Kristin
1991 Categories of experiential grammar. Ph. D. diss., Department of
Linguistics, University of Leuven.
1992a On the real function of Subject. Paper delivered at the 19th Inter-
national Systemic-Functional Congress, Sydney, 13-18 July
1992.
1992b Transitivity/ergativity: the Janus-headed grammar of actions and
events. In Advances in Systemic Linguistics: Recent Theory and
Practice, M. Davies, and L. Ravelli (eds.), 105-135. London:
Pinter Publishers.
1994 Fact projection. In Perspectives on English. Studies in Honour of
Professor Emma Vorlat, K. P. Carlon, K. Davidse, and B.
Rudzka-Ostyn (eds.), 259-286. Leuven: Peeters.
1996 Turning Grammar on itself: Identifying Clauses in Linguistic
Discourse. In Meaning and Form, Systemic Functional Interpre-
tations, Meaning and Choice in Language: Studies for Michael
Halliday, M. Berry, C. Butler, R. Fawcett, and G. Huang (eds.),
367-393. Norwood, New Jersey: Ablex Publishing Corporation.
1997 The Subject-Object versus the Agent-Patient asymmetry. Leuven
Contributions in Linguistics and Philology. 86 (4): 413-431.
1998a Agnates, verb classes and the meaning of construals. The case of
ditransitivity in English. Leuven Contributions in Linguistics and
Philology. 87 ( 3 ^ ) : 281-313.
1998b On transitivity and ergativity in English, or on the need for dia-
logue between schools. In English as a Human Language: to
Honour Louis Goossens, J. Van der Auwera, F. Durieux, and L.
Lejeune (eds.), 95-108. Muenchen: Lincom.
1999 The semantics of cardinal versus enumerative existential con-
structions. Cognitive Linguistics 10 (3): 203-250.
2000a Semiotic and possessive models in relational clauses: thinking
with grammar about grammar. Revista Canaria de Estudios In-
gleses 40: 13-35.
2000b A constructional approach to clefts. Linguistics 38 (6): 1101—
1131.
References 269
Fagan, Sarah
1992 The Syntax and Semantics of Middle Constructions. Cambridge:
Cambridge University Press.
Fawcett, Robin P.
1980 Cognitive Linguistics and Social Interaction. Towards an Inte-
grated Model of a Systemic Functional Grammar and the Other
Components of a Communicating Mind. Exeter/ Heidelberg:
Julius Groos Verlag and Exeter University Press.
Fellbaum, Christiane
1985 Adverbs in agentless actives and passives. Papers from the
Parasession on Causatives and Agentivity at the Twenty-first Re-
gional Meeting of the Chicago Linguistic Society. 21-31.
1986 On the Middle Construction in English. Indiana: Indiana Univer-
sity Linguistics Club.
Fellbaum, Christiane, and Anne Zribi-Hertz
1989 The Middle Construction in French and English: A Comparative
Study of its Syntax and Semantics. Indiana: Indiana University
Linguistics Club Publications.
Francis, Gill, Susan Hunston, and Elizabeth Manning
1996 Collins COBUILD Grammar Patterns 1: Verbs. London:
HarperCollins.
Fraser, Bruce
1970 Some Remarks on the Action Nominalization in English. In
Readings in English Transformational Grammar. R. A. Jacobs,
and P. S. Rosenbaum (eds.), 83-98. Waltham, Massachusetts.
Gleason, Henry A.
1965 Linguistics and English Grammar. New York: Holt, Rinehart
and Winston.
Haas, William
1954 On defining linguistic units. Transactions of the Philological
Society. 54-84.
Haegeman, Liliane, and Jacqueline Guéron
1999 English Grammar. A Generative Perspective. Oxford: Blackwell
Publishers.
Haeseryn, Walter, K. Romijn, Guido Geerts, J. de Rooij, and M. C. van den Toorn
(eds. )
1997 Algemeen Nederlandse Spraakkunst [General Dutch Grammar].
Groningen: Martinus Nijhoff, Deurne: Wolters Plantyn.
Haiman, John, and Sandra A. Thompson
1984 'Subordination' in Universal Grammar. Proceedings of the Tenth
Annual Meeting of the Berkeley Linguistics Society. 510-524.
Hale, Kenneth, and Jay Keyser
1987 A View from the Middle. Lexicon Project Working Papers 10.
Cambridge: MIT.
References 271
Halliday, Michael A. K.
1961 Categories of the Theory of Grammar. Word 17 (3): 241-292.
1966a Some notes on 'deep' grammar. Journal of Linguistics 2: 57-67.
1966b The concept of rank: A reply. Journal of Linguistics 2: 110-118.
1967 Notes on transitivity and theme in English. Part I. Journal of
Linguistics 3 (1): 37-81.
1968 Notes on transitivity and theme in English. Part III. Journal of
Linguistics 4 (2): 179-215.
1970a Functional diversity in language as seen from a consideration of
modality and mood in English. Foundations of Language 6 (3):
322-361.
1970b Language structure and language function. In New Horizons in
Linguistics, J. Lyons (ed.), 140-165. Harmondsworth: Penguin
Books.
1976 System and Function in Language. G. R. Kress (ed.). London:
Oxford University Press.
1979 Modes of meaning and modes of expression: types of grammati-
cal structure, and their determination by different semantic func-
tions. In Function and Context in Linguistic Analysis: Essays Of-
fered to William Haas, D. J. Allerton, E. Carney, and D. Hold-
croft (eds.), 57-79. Cambridge: Cambridge University Press.
1985 An Introduction to Functional Grammar. London: Edward Ar-
nold.
1988 On the ineffability of grammatical categories. In Linguistics in a
Systemic Perspective. J. D. Benson, M. J. Cummings, and W. S.
Greaves (eds.), 27-51. Amsterdam/Philadelphia: John Benja-
mins.
1992 How do you mean? In Advances in Systemic Linguistics. Recent
Theory and Practice, M. Davies and L. Ravelli (eds.), 20-35.
London: Pinter.
1994 An Introduction to Functional Grammar (2nd ed.). London: Ed-
ward Arnold.
Halliday, Michael A. K., and Ruqaiya Hasan
1976 Cohesion in English. London: Longman.
Halliday, Michael A. K., and Christian M. I. M. Matthiessen
1999 Construing Experience Through Meaning. A Language-based
Approach to Cognition. London: Cassell.
Haugen, Einar
1953 On resolving the close apposition. American Speech 28: 165-
170.
Hengeveld, Kees
1989 Layers and operators in Functional Grammar. Journal of Linguis-
tics 25: 127-157.
272 References
Heyvaert, Liesbet
1997 Patientive -er nomináis. Leuven Contributions in Linguistics and
Philology 86 (4): 433-456.
1998 Nonagentive deverbal -er nominalization in English and Dutch: a
contrastive analysis. Languages in Contrast 1 (2): 211-243.
2000 Gerundive nominalization: from type specification to grounded
instance. In Constructions in Cognitive Linguistics. Selected Pa-
pers from the Fifth International Cognitive Linguistics Confer-
ence, Amsterdam, 1997, A. Foolen, and F. van der Leek (eds.),
103-121. Amsterdam/Philadelphia: John Benjamins.
2001 Nominalization as an interpersonally-driven system. Functions of
Language 8 (2): 287-329.
in press Nominalization as grammatical metaphor: on the need for a radi-
cally systemic and metafìinctional approach. In Metaphor in Sys-
temic-Functional Perspectives, L. Ravelli, A.-M. Simon-
Vandenbergen, and M. Taverniere (eds.). Amsterdam/ Philadel-
phia: John Benjamins.
Hjelmslev, Louis
1961 Prolegomena to a Theory of Language. Translated by F. J. Whit-
field. Madison: University of Wisconsin Press.
Hoekstra, Teun, and Ian Roberts
1993 Middle constructions in Dutch and English. In Knowledge and
Language. Volume 2. Lexical and Conceptual Structure, E. Reu-
land, and W. Abraham (eds.), 183-220. The Netherlands: Kluwer
Academic Publishers.
Holierhoek, C. M. G.
1980 Werkwoorden van Waarneming. Aspekten van hun Systematiek
en Ontwikkeling [Verbs of perception. Aspects of their function-
ing and development]. Den Haag: Universitaire Pers Leiden.
Hooper, Joan B.
1975 On assertive predicates. In Syntax and Semantics. Volume 4,
J.P. Kimball (ed.), 91-124. New York: Academic Press.
Hopper, Paul J., and Sandra A. Thompson
1980 Transitivity in grammar and discourse. Language 56: 251-299.
Hopper, Paul J., and Elizabeth C. Traugott
1993 Grammaticalization. Cambridge: Cambridge University Press.
Horn, George M.
1975 On the Nonsentential Nature of the POSS-ING Construction.
Linguistic Analysis 1 (4): 333-387.
Huddleston, Rodney
1969 Review article on Madeline Ehrman, 'The meanings of the mo-
dals in present-day American English'. Lingua 23: 165-176.
1988 Constituency, multifunctionality and grammaticalization in Hal-
liday's functional grammar. Journal of Linguistics 2: 137-174.
References 273
Lees, Robert Β.
1960/1968 The Grammar of English Nominalizations. Bloomington: Indiana
University Research Center in Anthropology, Folklore and Lin-
guistics, publication 12.
Lemmens, Maarten
1998 Lexical Perspectives on Transitivity and Ergativity. Causative
Constructions in English. Amsterdam/Philadelphia: John Benja-
mins.
Levin, Beth
1993 English Verb Classes and Alternations. Chicago: University of
Chicago Press.
Levin, Beth, and Malka Rappaport
1988 Nonevent -er nomináis: a probe into argument structure. Linguis-
tics 26: 1067-1083.
Lyons, John
1969 Introduction to Theoretical Linguistics. Cambridge: Cambridge
University Press.
1977 Semantics. Volume 2. Cambridge: Cambridge University Press.
Mackenzie, Lachlan J.
1990 First Argument Nominalization in a Functional Grammar of Eng-
lish. Linguistica Antverpiensia 24: 119-147.
Malouf, Robert
2000 Verbal gerunds as mixed categories in head-driven phrase struc-
ture grammar. In Syntax and Semantics. Volume 32. The Nature
and Function of Syntactic Categories, R. D. Borsley (ed.), 133—
166. San Diego: Academic Press.
Marchand, Hans
1969 The Categories and Types of Present-day English Word-
formation. A Synchronic-Diachronic Approach. 2nd edition.
München: C. H. Beck'she Verlagsbuchhandlung.
Massam, Diane
1992 Null objects and non-thematic subjects. Journal of Linguistics
28: 115-137.
Matthews, Peter H.
1966 The concept of rank in 'Neo-Firthian' grammar. Journal of Lin-
guistics 2: 101-109.
Matthiessen, Christian M. I. M.
1995 Lexicogrammatical Cartography: English Systems. Tokyo: In-
ternational Language Sciences Publishers.
Matthiessen, Christian M. I. M., and Sandra A. Thompson
1988 The structure of discourse and 'subordination'. In Clause Com-
bining in Grammar and Discourse, J. Haiman, and S. A. Thomp-
son (eds.), 275-329. Amsterdam/Philadelphia: John Benjamins.
276 References
McGregor, William B.
1991 The concept of rank in systemic linguistics. In Functional and
Systemic Linguistics: Approaches and Uses, E. Ventola (ed.),
121-138. Berlin/New York: Mouton de Gruyter.
1997 Semiotic Grammar. Oxford: Clarendon.
Menzel, Peter
1975 Semantics and Syntax in Complementation. The Hague: Mouton.
Milsark, Gary L.
1976 Existential Sentences in English. Bloomington: Indiana Univer-
sity Linguistics Club.
1977 Toward an explanation of certain peculiarities of the existential
construction in English. Linguistic Analysis 3: 1-29.
1988 Singl-ing. Linguistic Inquiry 19 (4): 611-634.
Munro, Pamela
1982 On the transitivity of 'Say' verbs. In Syntax and Semantics. Vol-
ume 15. Studies in Transitivity, P. J. Hopper, and S. A. Thomp-
son (eds.), 301-318. New York: Academic Press.
Noonan, Michael
1985 Complementation. In Language Typology and Syntactic Descrip-
tion. Volume 2. Complex Constructions, T. Shopen (ed.), 42-140.
Cambridge: Cambridge University Press.
Partee, Barbara H.
1973 The syntax and semantics of quotation. In A festschrift for Morris
Halle, P. Kiparsky, and S. R. Anderson, 410-418. New York:
Holt, Rinehart and Winston.
Peeters, Roger J.
1999 The Adjunct Middle Construction in Dutch. Leuven Contribu-
tions in Linguistics and Philology 88 (3—4): 355-401.
Perlmutter, David M.
1978 Inpersonal passives and the unaccusative hypothesis. Proceed-
ings of the Fourth Annual Meeting of the Berkeley Linguistics
Society. 157-189.
Postal, Paul M.
1970 On Coreferential Complement Subject Deletion. Linguistic In-
quiry 1 (4): 439-500.
Poutsma, H.
1929 A Grammar of Late Modern English. 2nd edition. Groningen:
Noordhoff.
Pullum, Geoffrey K.
1991 English nominal gerund phrases as noun phrases with verb-
phrase heads. Linguistics 29: 763-799.
Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik
1972 A Grammar of Contemporary English. London: Longman.
1985 A Comprehensive Grammar of the English Language. London:
Longman.
References 277
Radford, Andrew
1997 Syntactic Theory and the Structure of English. A Minimalist Ap-
proach. Cambridge: Cambridge University Press.
Rappaport Hovav, Malka, and Beth Levin
1992 -ER nomináis: implications for the theory of argument structure.
In Syntax and Semantics. Volume 26: Syntax and the Lexicon, T.
Stowell, and E. Wehrli (eds.), 127-153. New York: Academic
Press.
Ravelli, Louise J.
1988 Grammatical metaphor: an initial analysis. In Pragmatics, Dis-
course and Text. Some Systemically-inspired Approaches, E.
Steiner, and R. Veltman (eds.), 133-147. London: Pinter.
Roberts, Ian G.
1987 The Representation of Implicit and Dethematized Subjects.
Dordrecht: Foris.
Rosenbaum, Peter S.
1967 The Grammar of English Predicate Complement Constructions.
Cambridge, Massachusetts: MIT Press.
Rosta, Andrew
1995 'How does this sentence interpret?' The semantics of English
mediopassives. In The Verb in Contemporary English. Theory
and Description, B. Aarts, and C. F. Meyer (eds.), 123-144.
Cambridge: Cambridge University Press.
Ryder, Mary E.
1991 Mixers, mufflers and mousers: the extending of the -er suffix as
a case of prototype reanalysis. Proceedings of the Berkeley Lin-
guistics Society 17: 299-311.
1999a Complex -er nomináis. Where grammaticalization and lexicaliza-
tion meet? In Between Grammar and Lexicon, E. Contini-
Morava, and Y. Tobin (eds.), 291-331. Amsterdam/Philadelphia:
John Benjamins.
1999b Bankers and blue-chippers: an account of -er formations in pre-
sent-day English. English Language and Linguistics 3 (2): 269-
297.
Saussure de, Ferdinand
1972 Cours de linguistique générale. Edition critique préparée par
Tullio de Mauro. Paris: Payot.
Schachter, Paul
1976 A nontransformational account of gerundive nomináis in Eng-
lish. Linguistic Inquiry 7 (2): 205-241.
Schmid, Hans-Jörg
2000 English Abstract Nouns as Conceptual Shells. From Corpus to
Cognition. Berlin/New York: Mouton de Gruyter.
278 References
Thompson, Sandra A.
1973 On subjectless gerunds in English. Foundations of Language 9:
374-383.
Vachek, Josef
1966 The Linguistic School of Prague. Bloomington: Indiana Univer-
sity Press.
Vandenberghe, Wim
2001 Instigative Setting-constructions: Force Dynamic Research on
'New' Types of Agency. Leuven Contributions in Linguistics
and Philology 90: 365-390
Van den Toorn, M.C.
1982 Nederlandse Grammatica [Dutch grammar]. Groningen: Wolters
Noordhoff.
Van Langendonck, Willy
1988. -ER als akkusatief en -LING als ergatief suffix in een Zuidbra-
bants dialekt [-ER as an accusative and -LING as an ergative suf-
fix in a South-Brabant dialect], Spektator 17 (4): 317-327.
1999 Neurolinguistic and syntactic evidence for basic level meaning in
proper names. Functions of Language 6 (1): 95-138.
Van Oosten, Jeanne
1977 Subjects and agenthood in English. Papers from the 13th Annual
Regional Meeting of the Chicago Linguistics Society. 451-471.
1986 The Nature of Subjects, Topics and Agents: A Cognitive Expla-
nation. Bloomington: Indiana University Linguistics Club.
Vendler, Zeno
1967 Linguistics in Philosophy. Ithaca/New York: Cornell University
Press.
Verstraete, Jean-Christophe
2001 Subjective and objective modality: Interpersonal and ideational
functions in the English modal auxiliary system. Journal of
Pragmatics 33: 1505-1528.
2002 Interpersonal grammar and clause combining in English. Ph. D.
diss., Department of Linguistics, University of Leuven.
Wasow, Thomas, and Thomas Roeper
1972 On the subject of gerunds. Foundations of Language. Interna-
tional Journal of Language and Philosophy 8: 44—61.
Wierzbicka, Anna
1988 The Semantics of Grammar. Amsterdam/Philadelphia: John Ben-
jamins.
Williams, Edwin
1980 Predication. Linguistic Inquiry 11 (1): 203-238.
1981 Argument Structure and Morphology. The Linguistic Review 1:
81-114.
280 References
Yoshimura, Kimihiro
forthcoming The Middle Construction in English: A Cognitive Linguistic ana-
lysis. Berlin/New York: Mouton de Gruyter.
Index