A SEMIOTIC THEORY OF MATHEMATICAL TEXT

Paul Ernest

University of Exeter
p.ernest @ ex.ac.uk

In this paper I consider the content and function of mathematical text from a semiotic
perspective. My enquiry takes me beyond the written mathematical text for I also consider the
spoken text and texts presented multi-modally. I explore both the reading/listening and the
writing/speaking dimensions of mathematical text in this broader sense. In addition to making
the enquiry more extensive this necessitates the inclusion of a further vital dimension of
written mathematical text in use, namely, the social context. For texts do not exist in the
abstract, but are always and only present via their utterances and instantiations. Thus to open
up the mathematical text from a semiotic perspective is to explore its social uses and
functions, as well as its inner meanings and textures. Mathematical text is unlike fiction, for it
is not merely a doorway to a world of the imagination. It is not just a told tale rendered into
written language. This insight is readily available to schoolchildren. But it is one that
philosophers, mathematicians and educational researchers have to struggle to attain. Why is
this? How do mathematical texts differ from fiction? This brings me back to the main
questions of my enquiry. What do mathematics texts ‘say’? What light do the tools of
semiotics shed on the content and functions of mathematical text?

These are not simple questions, and they are rendered all the more obscure because
mathematical texts are not viewed or read neutrally. Irrespective of the reading subject who is
needed to make sense of the text, mathematics is thickly overlaid with ideological
presuppositions that prevent or obscure a neutral or free reading. So my first enquiry is into
some of these ideological presuppositions that distort a reading of mathematical texts, that is,
into what mathematics texts do not say.

WHAT MATHEMATICS TEXTS DO NOT SAY

Rorty (1979) has described the ideology gripping traditional philosophy that sees it, the text,
and the human mind as ‘mirroring nature’. In other words, he critiques the traditional
assumption that there is a given, fixed, objective reality and that mind, knowledge and text
capture and describe it, with greater or lesser exactitude. This traditional philosophy reached
its apogee in Wittgenstein’s (1922) Tractatus with its picture theory of meaning.
Wittgenstein’s doctrine asserts that every true sentence depicts, in some literal sense, the
material arrangements of reality. Language, when used correctly, floats above material reality
as a parallel universe and provides an accurate map or picture of it. However, a claim this
strong is hard to sustain, and even the logical positivists withdrew from this overly literal
position about the relationship between language and reality. They adopted instead the
verification principle which states that the meaning of a sentence is the means of its
verification (Ayer 1946).1 For without this revised view of meaning the predictive power and
1
In both the cases of Wittgenstein and the Logical Positivists mathematics is treated differently from empirical
or scientific claims or texts. In the Tractatus, Wittgenstein argues that mathematics is a by-product of the
grammar of world description. The logical positivists argue that mathematics is analytic a priori knowledge,
derived purely by logic and telling us nothing new. Both of these accounts trivialize mathematics, and focus on
the tasks of accounting for scientific knowledge of the empirical world and the rejection of metaphysics. From
the perspective of mathematical philosophy these accounts are fundamentally unenlightening.
generality of scientific theories is compromised. Ironically, Wittgenstein (1953), in his later
philosophy, also rejected this position himself having pushed the picture or mirror view to its
limits in the Tractatus.2

This ideology applied to mathematics remains very potent in Western culture. Mathematics is
seen to describe an objective and timeless superhuman realm of pure ideas, the necessity of
which is reflected in the ineluctable patterns and structures observed in our physical
environment. The doctrine that mathematics describes a timeless and unchanging realm of
pure ideas goes back to Plato, if not earlier, and is usually referred to as Platonism. Many of
the greatest philosophers and mathematicians have subscribed to this doctrine in the
subsequent two plus millennia since Plato’s time. In the modern era the view has been
endorsed by many thinkers including Frege (1884, 1892), Gödel (1964), and in some writings
by Russell (1912) and Quine (1953). According to Platonism, a correct mathematical text
describes the state of affairs that holds in the platonic realm of ideal mathematical objects.
Mathematical texts are nothing but descriptions or mirrors of what holds in this inaccessible
realm. 3

Although I shall reject it, quite a lot is gained by this view. First of all, mathematicians and
philosophers have a strong belief in the absolute certainty of mathematical truth and in the
objective existence of mathematical objects, and a belief in Platonism validates this. It posits a
quasi-mystical realm into which only the select few – initiates into the arcane practices of
mathematics – are permitted to gaze. Secondly it puts epistemology beyond the reach of
humanity’s sticky fingers and earth(l)y bodies, locating mathematical knowledge in a safe and
inaccessible zone. Because of this strategy of removal, any claims that mathematics is socially
constructed are disallowed tout court without having their specific merits or weaknesses
considered.

But it seems to me that the strategy of saying the text ‘2+2=4’ is true because 2+2=4, i.e.,
there are ideal abstract numbers 2 and 4 and when you combine 2 with 2 by addition the result
is equal to 4, is not very enlightening. The real question is what does addition mean, and why,
when you add 2 to 2 does the answer happen to be equal to another number and that number
is 4 and not something else? I can, of course, answer this particular question, but that’s not the
point. The point is that the translation from realm of signs and texts where ‘2+2=4’ is located,
to the realm of meaning (i.e., the interpretation of the sign in the platonic universe of number
where 2+2=4 holds), is not illuminating, per se, in providing the meaning of the expression.

Alfred Tarski (1935) in his famous paper on the semantic interpretation of truth argues the
sentence ‘Snow is white’ is true, if, and only if, Snow is white. Per se, this is trivial and
uninformative. However, Tarski’s project is fundamentally technical. He is concerned with the
definability of the concept of truth in formalized languages, and as a consequence of his
formal explication of ‘truth’ provided the foundations of model theory and arrived at one of
the important limitative results of modern mathematics. Namely, that truth is indefinable in
formal languages, on pain of inconsistency. For technical purposes he clarified the notion of
syntactic expressions mirroring states of affairs in the semantic realm, with valuable
mathematical and foundational results. But this does not lend support to the general
2
Wittgenstein (1953) went on to develop what might be termed one of the first postmodernist epistemologies,
with his doctrines that ‘meaning is use’, and that all text and knowledge emerges from language games
embedded in human forms of (social) life, with mathematics made up a multiplicity of language games.
3
Otte (1997: 50) contrasts Cartesianism, the view that the (mathematical) world is already known in principle,
and that in time scientific deduction will fill all the lacunae, with Pascal’s ‘subtle intelligence’ according to
which “the world can be appropriated cognitively only by the constructive incorporation of ever new types of
objects into thought”.
ideological ‘mirroring’ presupposition ‘as below, so above’ (to invert the fundamental
principle of astrology) that I am critiquing.

Philosophers have long been concerned about mathematical ontology: the study of the
existence and nature of mathematical objects and abstract entities in general. Balaguer (2004)
provides a state of the art survey of ‘Platonism in Metaphysics’, and concludes that
mathematical objects can be accounted for in three ways: physical (in the world), mental (in
the mind), and abstract (in Plato’s realm of ideal objects or some equivalent). His account is
of course much more subtle and nuanced than this one sentence summary. It is an impressive
display of intellectual prowess, although I get the feeling that professional philosophers like
Balaguer turn the virtue of subtle reasoning, through excessive application, into a vice. These
three possibilities seem limited in that mathematical objects cannot obviously be regarded as
purely physical or mental4, which leaves open only the possibility that mathematical objects
are abstract objects, confirming Platonism, or some variant of it. Balaguer himself
acknowledges some concerns about his categorization, perhaps as a gambit to stave off the
possible criticism that mathematical objects could also be seen as social constructions.

There are a couple of worries one might have about the exhaustiveness of the
physical-mental-abstract taxonomy. First, one might think there is another
category that this taxonomy overlooks, in particular, a category of social objects,
or perhaps social constructions. It seems, though, that social objects would
ultimately have to reduce to either physical, mental, or abstract objects. Balaguer
(2004: note 3)

Here he makes the unsubstantiated assumption that any other category, such as that of socially
constructed objects, can be reduced to one of his three categories. This introduces another
ideological presupposition, namely that of reductionism. By reductionism I mean the
positivist doctrine, much beloved of the Logical Empiricists, that that there is an hierarchy of
objects, theories or disciplines, and it is possible to translate and replace objects, theories or
disciplines further up the hierarchy into the objects, terms, concepts and theories lower down
the hierarchy without loss of generality or scientific significance. Thus according to this
scheme, sociology can be translated (reduced) to psychology, psychology to biology, biology
to chemistry, chemistry to physics, and finally physics to mathematics. (A further step, the
reduction of mathematics to logic, the goal of Logicism in the philosophy of mathematics,
was shown to be a failure. See, for example, Ernest 1991, 1998, Hersh 1997, Kline 1980).5

One of the most notable critics of reductionism is Feyerabend (1975), who argues that there is
semantic instability within and across theories or disciplines. In Feyerabend (1962) he claims
that meanings of the same terms and concepts in different theories are not only different but
4
I put to one side Intuitionism in the philosophy of mathematics, which argues that the concepts and results of
mathematics are purely mental constructions. Intuitionism is problematic in that it rejects much of classical
mathematics (the non-constructive and infinitary parts) and has never managed to explain how different minds
all arrive at identical mathematical results, especially as it holds that mathematics is a prelinguistic activity.
Lakoff and Núñez (2000) also claim that mathematical concepts are mental in origin, but unlike Intuitionism
their theory of embodied mind rejects the separation of mind and body.
5
There has been strong critical and negative reaction among many mathematicians and philosophers to social
philosophies of mathematics such as proposed in Hersh (1997) and Ernest (1998). Part of the common critique
strategy is to mis-represent such philosophies as saying that mathematics can be reduced to sociology (i.e., that
mathematical matters are decided by ‘mob rule’). If these critics subscribe to reductionism, which is not
uncommon, then the threat is that by ‘joining the ends’ mathematics is knocked off its pedestal as the foundation
of knowledge, and the whole chain of disciplines closes on itself in a vicious circle, knowledge eating itself like
the worm Ouroboros. No wonder social constructivist philosophies of mathematics are viewed as a nightmarish
threat!
are also incommensurable, and that ‘theoretical reduction’ is not reduction but the
replacement of one theory and its ontology by another. The same year Kuhn (1962) also
published his seminal work on the structure of science revolutions in which he argues that the
concepts of competing theories are incommensurable. Although the claims of strong
incommensurability did not stand up to criticism, there is a powerful argument from holism
that refutes the reductionist claims.

Consider for example, the reduction of psychology to biology. We know that human minds are
ultimately based in a material organ. Furthermore, specific areas and functions of the organ
(the brain) can be correlated with particular mental activities. A range of different chemicals
have profound and sometimes predictable impacts on thinking. Nevertheless, there is no
forseeable possibility that the full complexity of human thinking and behaviour as a whole
could be reduced to a biological model. The kinds of mental elements that can be found to
correspond to biological processes are so very simple and disconnected that there is no
prospect that current biology could explain human thoughts, feelings and behaviour as a
whole. The earlier attempt by Behaviorism to ignore the mind and to try to explain behaviour
scientifically is a well known failure.

I have suggested that both epistemological and ontological reductionism are fatally flawed.
We cannot simply define away complex objects or bodies of knowledge in terms of simpler
ones. In particular, I strongly believe that social constructions cannot be reduced to either
physical, mental, or abstract objects, for they combine elements of all three. This tripartite
ontology mirrors Popper’s (1979) definition of three distinct worlds, each with its own type of
knowledge.

We can call the physical world ‘world 1’, the world of our conscious experiences
‘world 2’, and the world of the logical contents of books, libraries, computer
memories, and suchlike ‘world 3’. (Popper 1979: 74)

Ironically, this also mirrors the tripartite division, much beloved of ‘New Agers’ into Body,
Mind and Spirit. I draw this parallel out of more than naughty playfulness. For it seems to me
that the Platonic realm of abstract entities, World 3, the world of the Spirit, and even Heaven
and the Kingdom of God, all require irrational belief or faith. Positing them does not simplify
the task of understanding the mathematical text. Rather it defers a key element of that
understanding, removing it to what I regard as an inaccessible realm. What is needed instead
is what Restivo (1993) has aptly termed the Promethean task of bringing mathematics to
earth.

One approach, which has some currency if not widespread acceptance, that I have applied
elsewhere (Ernest 1998), is to redefine ‘objective’ mathematical knowledge (using the term
‘objectivity’ without subscribing to absolutism in the epistemological realm or idealism in
ontology) as social and cultural knowledge that is publicly shared, both within the
mathematical community, and more widely as well. In mathematics, this includes all that
Popper counts as objective knowledge, including mathematical theories, axioms, problems,
conjectures, proofs, both formal and informal. However, I also want to include the shared but
possibly implicit conventions and rules of language usage, and a range of tacit understandings
that are acquired through participation in practices. These are shared in that they are deployed
and learned in public, for persons to witness. But they may remain implicit if only their
instances and uses are made public, and any underlying general rules or principles are rarely
or never uttered. According to such an account, explicit objective knowledge is made up of
texts that have been socially constructed, negotiated and accepted by social groups and
institutions. Naturally, such texts have meanings and uses both for individuals and for social
groups.

What I have briefly indicated is a social theory of objectivity that resembles, at least in part,
proposals by Bloor (1984), Harding (1986), Fuller (1988) and others. In some variant or
another, such a view also underpins much work in the sociology of knowledge and in post-
structuralist and postmodernist epistemology.6 By subscribing to an approach that demystifies
‘objectivity’ I am suggesting doing away with the ontological category of abstract objects,
which often amounts to Plato’s world of pure forms. That is, I have applied a principle of
ontological reduction. I am in good company here. Quine (1969) has argued for ‘ontological
parsimony’, the principle that we should apply Occam’s razor and not allow entities and their
types to multiply beyond what is necessary. Ryle (1949) has also argued convincingly that the
ontological separation of mind and body is also a mistake. Although the mental cannot be
simply reduced to the physical, this is an epistemological non-reducibility, not an ontological
one. So if we reject separate ontological categories for the mental (mind) and the physical
(body), we end up with a unified realm of being.

In my view, the universe is made up of not three types but one type of ‘stuff’, namely the
material basis of physical reality. Within this unified world there are among the myriads of
things and beings, humans with minds and groups of humans with cultures. Human minds, the
seat of the mental, are not a different kind of ‘stuff’, but are a complex set of functions of self
organising, self aware, feeling moral beings. Mathematical knowledge, like other semiotic and
textual matters, is made up of social objects. These are simultaneously materially represented,
given meaning by individuals and created and validated socially.

This discussion may seem like an excursus on the way to opening up the mathematical text,
but its function is to show that there is a very deeply entrenched ideology, all the way up to
the highest intellectual levels, that regards mathematical text as a vehicle that describes
superhuman and objective mathematical reality. According to this view, when the text is
correct, and it thus counts as expressing mathematical knowledge, it truly describes this
reality. So from this perspective, mathematical knowledge texts are mirrors that reflect a true
state of affairs in a timeless, objective, superhuman Platonic realm.

Richard Rorty (1979) argues that the assumption the knowledge or the mind mirror nature is a
major stumbling block in philosophy, and in rejecting this, identifies himself as postmodern,
“in the rather narrow sense defined by Lyotard as 'distrust of metanarratives'.” (Rorty 1991:
1). He goes on to argue that mathematical knowledge, for example, the Pythagorean Theorem,
is accepted as certain because humans are persuaded it is true, rather than because it mirrors
states of affairs in ‘mathematical reality’.

If, however, we think of "rational certainty" as a matter of victory in argument

rather than of relation to an object known, we shall look toward our interlocutors
rather than to our faculties for the explanation of the phenomenon. If we think of
our certainty about the Pythagorean Theorem as our confidence, based on
experience with arguments on such matters, that nobody will find an objection to
6
Some thinkers in postmodernity, e.g., Nel Noddings (1990) and Richard Rorty (Ramberg 2002) have styled
themselves as post-epistemological, because they want to reject the absolutist and foundationalist
presuppositions of traditional epistemology. However, just as I want to retain but redefine ‘objectivity’ in a non-
absolutist and non-idealistic way, so too do I wish to retain the term ‘epistemology’ for the theory of knowledge
without any similar presuppositions. I don’t believe we need to reject important terms because we don’t agree
with the ways they have been used by others. Terms are all the while growing, changing in meaning, and in mid-
renegotiation, as Lakatos (1976) demonstrates for mathematical concepts.
 the premises from which we infer it, then we shall not seek to explain it by the
relation of reason to triangularity. Our certainty will be a matter of conversation
between persons, rather than an interaction with nonhuman reality. Rorty (1979:
156-157)

Thus Rorty shares the view expressed above that it is social agreement, albeit in a complex
and non-whimsical way, that provides the foundation for mathematical knowledge. The truth
or otherwise of a mathematical text lies in its social role and acceptance, not its relation to
some mysterious realm.

WHAT IS MATHEMATICAL TEXT?

In keeping with modern semiotics I want to understand a text as a simple or compound sign
that can be represented as a selection or combination of spoken words, gestures, objects,
inscriptions using paper, chalkboards or computer displays, as well as recorded or moving
images. Mathematical texts can vary from, one at extreme, in research mathematics, printed
documents that utilize a very restricted and formalized symbolic code, to at the other extreme,
multimedia and multi modal texts, such are used in kindergarten arithmetic. These can include
a selection of verbal sounds and spoken words, repetitive bodily movements, arrays of sweets,
pebbles, counters, and other objects, including specially designed structural apparatus, sets of
marks, icons, pictures, written language numerals and other writing, symbolic numerals, and
so on.

The received view is that progression in the teaching and learning of mathematics involves a
shift in texts from the informal multi-modal to the restrictive, rigorous symbol-rich written
text. It is true that, for some, access to the heavily abstracted and coded texts of mathematics
grows through the years of education from kindergarten through primary school, secondary
school, high school, college, culminating in graduate studies and research mathematics. But it
is a myth that informal and multi-modal texts disappear in higher level mathematics. What
happens is that they disappear from the public face of mathematics, whether these be in the
form of answers and permitted displays of ‘workings’, or calculations in work handed in to
the school mathematics teacher, or the standard accepted answer styles for examinations, or
written mathematics papers for publication. As Hersh (1988) has pointed out, mathematics
(like the restaurant or theatre) has a front and a back. 7 What is displayed in the front for public
viewing is tidied up according to strict norms of acceptability, whereas the back where the
preparatory work is done is messy and chaotic.

The difference between displayed mathematical texts, at all levels, and private ‘workings’ is
the application of rhetorical norms in mathematics. These concern how mathematical texts
must be written, styled, structured and presented in order to serve a social function, namely to
persuade the intended audience that they represent the knowledge of the writer. Rhetorical
norms are social conventions that serve a gatekeeper function. They work as a filter imposed
by persons or institutions that have power over the acceptance of texts as mathematical
knowledge representations. Rhetorical norms and standards are applied locally, and they
usually include idiosyncratic local elements, such as how a particular teacher or an
examinations board likes answers laid out, and how a particular journal requires references to
other works to be incorporated. Thus one inescapable feature of the mathematical text is its
style, reflecting its purpose and most notably, its rhetorical function.

7
Hersh draws his analogy from Goffman’s (1971) work on how persons present themselves in everyday life.
Rhetoric is the science or study of persuasion, and its universal presence in mathematical text
serves to underscore the fact that mathematical signs or texts always have a human or social
context. I interpret signs and texts as utterances in human conversation, that is within
language games embedded in forms of life (Wittgenstein 1953) or within discursive practices
(Foucault 1972). Texts exist only through their material utterances or representations, and
hence via their specific social locations.8 The social context of the utterance of a text produces
further meanings, positionings and roles for the persons involved. Thus, in any given context,
a mathematical text or sign utterance, like any utterance, is indissolubly associated with a
penumbra of contextual meanings including its purpose, its intended response, the positioning
and power of its speaker/utterer and listener/reader. 9 Such meanings are both created and
elicited through the social context and are also a function of the meanings and positions made
available through the text itself. Different types of meanings and intentions are intended, but
perhaps the most central and critical function (and hence meaning) of mathematical texts in
the mathematics classroom is to present mathematics learning tasks to students. A
mathematical learning task:

1. Is an activity that is externally imposed or directed by a person or persons in

power representing and on behalf of a social institution,
2. Is subject to the judgement of the persons in power as to when and whether it is
successfully completed,
3. Is a purposeful and directional activity that requires human actions and work in the
striving to achieve its goal,
4. Requires learner acceptance of the imposed goal, explicitly or tacitly, in order for
the learner to consciously work towards achieving it,
5. Requires and consists of working with texts: both reading and writing texts in
attempting to achieve the task goal.10

A more general concept of mathematical task includes self-imposed tasks that are not
externally imposed and not driven by direct power relationships. However, in research
mathematicians’ work, although tasks may not be individually subject to power relations,
particular self-selected and self-imposed tasks may be undertaken within a culture of
performativity that requires measurable outputs. So power relations are at play at a level
above that of individual tasks. Even where there is no external pressure to perform, the
accomplishment of self-imposed task requires the internationalization of the concept of task,
including the roles of assessor and critic, based on the experience of social power-relations, to
provide the basis for an individual’s own judgement as to when a task is successfully
completed.

8
Especially in the age of mechanical reproduction we speak as if different utterances (tokens) represent the same
text (type). In such cases much can be shared between the different utterances, e.g., when the students in a class
each has a copy of a set school mathematics textbook, because they are located in a shared social context. Even
in this example there can be significant differences in reader reception of the text utterances. When the tokens
are located in different social contexts it makes less sense to refer to them as the ‘same’ text, even if this is
common parlance.
9
Clearly these are attenuated if not lost when a text is taken beyond its intended audience or appears accidentally
in some social context. Such appearances are a different utterances with a different social contexts, meanings and
intensions. For example, a school mathematical text taken from the shelf of a bookshop does not have the
specific associated directives and positionings for the customer as it would have for a reader in a particular
classroom (although it might elicit some comparable meanings through memories of schooling). A Babylonian
clay tablet would be unlikely to have any of the same meaning for a US Army soldier finding it in Iraq as it did
for the ancient scribes who created and used it.
10
Gerofsky (1996) adds that tasks, especially ‘word problems’, also bring with them a set of assumptions about
what to attend to and what to ignore among the available meanings.
Mathematical learning tasks are important because they make the bulk of school activity in
the teaching and learning of mathematics. During most of their mathematics learning careers,
which in Britain continues from 5 to 16 years and beyond, students mostly work on textually
presented tasks. I estimate that an average British child works on 10,000 to 200,000 tasks
during the course of their statutory mathematics education. This estimate is based on the not
unrealistic assumptions that children each attempt 5 to 50 tasks per day, and that they have a
mathematics class every day of their school career.

A typical school mathematics task concerns the rule-based transformation of text. Such tasks
consist of a textual starting point, the task statement. These texts can be presented
multimodally, with the inscribed starting point expressed in written language or symbolic
form, possibly with accompanying iconic representations or figures, and often accompanied
by spoken instructions from the teacher, typically a metatext. Learners carry out such set tasks
by writing a sequence of texts, including figures, literal and symbolic inscriptions, etc.,
ultimately arriving, if successful, at a terminal text the required ‘answer’. Sometimes this
sequence involves a sequence of distinct inscriptions, for example, the addition of two
fraction numerals with distinct denominators, or the solution of an equation in linear algebra.
Sometimes it involves the elaboration or superinscription of a single piece of text, such as the
carrying out of 3 digit column addition or the construction of a geometric figure. It can also
combine both types of inscriptions. In each of these cases there is a common structure. The
learner is set a task, central to which is an initial text, the specification or starting point of the
task. The learner is then required to apply a series of transformations to this text and its
derived products, thus generating a finite sequence of texts terminating, when successful, in a
final text, the ‘answer’. This answer text represents the goal state of the task, which the
transformation of signs is intended to attain.

Formally, a successfully completed mathematical task is a sequential transformation of, say, n

texts or signs ('Si') written or otherwise inscribed by the learner, with each text implicitly
derived by n-1 transformations ('ði').11 This can be shown as the sequence: (S0 ð0) S1 ð1
S2 ð2 S3 ð3 ... ðn-1 Sn. S1 is a representation of the task as initially inscribed or recorded
by the learner. This may be the text presented in the original task specification. However the
initial given text presenting the task (S0) may have been curtailed, or may be represented in
some other mode than that given, such as a figure, when first inscribed by the learner. In this
case an additional initial transformation (ð0) is applied to derive the first element (S1) in the
written sequence. Sn is a representation of the final text, intended to satisfy the goal
requirements as interpreted by the learner. The rhetorical requirements and other rules at play
within the social context determine which sign representations (Sk) and which steps (Sk ðk
Sk+1, k<n) are acceptable. Indeed, the rhetorical features of the transformed texts, together
with the other rules at play and the final goal representation (Sn), are the major focus for
negotiation (or correction) between learner and teacher, both during production and after the
completion of the transformational sequence. This focus will be determined according to
whether in the given classroom context the learner is required only to display the terminal text
(the answer) or a sequence of transformed texts representing its derivation.
11
Normally learners of school mathematics are not expected to specify the transformations used. Rather they are
implicitly evidenced in the difference between the antecedent and the subsequent text in any adjacent (i.e.,
transformed) pair of texts in the sequence. In some forms of proof, including some versions of Euclidean
geometry not generally included in modern school curricula, a proof requires a double sequence. The first is a
standard deductive proof and the second a parallel sequence providing justifications for each step, that is
specifications for each deductive rule application. Only in cases like this are the transformations specified
explicitly.
The final transformational sequence of texts displayed by the learner, and the actual
transformations derived during the work on the task may not be identical. The former may be
a ‘tidied up’ version of the latter, constructed to meet the rhetorical demands of the context,
rather than the working sequence actually used to derive the answer (see, e.g., Ernest 1993).
This distinction is most clearly apparent during the construction of a proof by a research
mathematician, as in the distinction between the ‘front’ and the ‘back’ of mathematical
activity (Hersh 1988). Here the proof as first sketched and the final version for publication,
both transformational sequences of texts, will almost invariably be very different. Lakatos
(1976) and others have criticized the pedagogical falsification perpetrated by the standard
practice of presenting advanced learners with the sanitized outcomes of mathematical enquiry.
Typically advanced mathematics text books conceal the processes of knowledge construction
by inverting or radically modifying the sequence of transformations used in mathematical
invention, for presentational purposes. The outcome may be elegant texts, but they also
generate learning obstacles.

Typically there are strict limitations on the modes of representation employed in a research
mathematicians’ published proofs, although these will vary according to the rhetorical
standards of the particular journal and mathematical subspecialism. Nevertheless, there may
well be written linguistic text, mathematical symbolism, arrays of signs, and diagrams, for
example, combined in the final text. An even greater number of different modes of
representation may be employed singly or together in a school mathematics text, including
combinations and selections of symbols, written language, labelled diagrams, tables, sketches,
models, and so on (even including arrayed objects and gestures, etc., where the text is
spoken). In school mathematics it is common for the transformational sequence of texts
produced by students during problem solution activities to use more modes of representation
than the starting text of the task specification, or the final text, ‘the answer’. Furthermore, the
transformational text sequence produced during task directed activities may be neither
monotonic nor single branched. Solution attempts may result in multi-branched sequences
with multiple dead ends and only one branch terminating in the final ‘answer’ text, and this
only appears when the activity is completed successfully.

My claim is that virtually all mathematical activity, throughout schooling, but also in graduate
and research mathematics, can be understood in terms of the production of sequences of texts
through the application of textual transformations. What distinguishes lower level from higher
level school mathematical activity in this description are the types of texts and
transformations involved. Routine mathematical activity typically involves relatively simple
initial texts and the deployment of restricted (algorithmic) transformation rules in the
production of sequences of texts. Non-routine mathematical activities, such as problem
solving, applications, or investigational work typically involve more complex task
formulations and require some novelty and insight in selecting which transformations to apply
and which elements to apply them to, in producing the sequence (Arcavi 2005).

Semiotic Systems
One of the central characteristics of mathematics both in school and in research, is the rule
based production and use of signs. However, except in degenerate cases, the use of signs and
rules is always underpinned by meaning.12 Even in degenerate cases, such as blind rule
following in school mathematics, there are meanings to the signs and processes, it is just that

12
Of course electronic computing systems need no recourse to meanings as all processes/processing is carried
out by unambiguous and fully specified rules and functions (in the mathematical sense).
the student involved is not accessing them. Thus signs, rules and meanings are the three
components of a semiotic system.

Semiotic system is the main theoretical tool developed in this paper. The theory of semiotic
systems provides a structure for describing the signs (texts) and the transformational rules
applied to them in both school and research mathematics. This model includes both the
publicly observable features of mathematical activity and the underlying meanings that
underpin the activities, especially the rules of textual transformation.

A semiotic system is defined in terms of three components:

1. A set of signs;
2. A set of rules for sign use and production;
3. An underlying meaning structure, incorporating a set of relationships between these signs
and rules. (See also Ernest 2005, 2006)

Signs
The set of signs comprises both elementary signs and compound signs made up of
concatenated sequences of signs. These signs constitute abstract types (after Peirce), but their
tokens (i.e., instances) can be spoken or uttered via various media: written, drawn, encoded
electronically or represented by any material means.

Semiotic systems can have multimodal sets of signs, such as the semiotic systems of nursery
or kindergarten arithmetic. These can include a selection of: verbal sounds and spoken words,
repetitive bodily movements, arrays of sweets, pebbles, counters, and other objects, including
specially designed structural apparatus, sets of marks, icons, pictures, written language
numerals and other text, symbolic numerals; and these inscriptions can be represented on the
chalkboard, in printed texts and charts, on computer and other ICT displays, and in children’s
own work on paper.13

In contrast, the semiotic system of school algebra at the lower secondary school level has for
its signs constants (numerals), variable letters (x, y, z, etc.), a 1-place function sign (-), 2-place
functions signs (+, -, x, /), a 2-place relation sign (=), and punctuation signs (parentheses,
comma, full stop).14 These are typically represented as textual inscriptions on the chalkboard,
in printed texts or worksheets or in student written work. In practice, the set of signs changes
over the course of schooling. Early on, in the introduction to algebraic notions during the
elementary school years, a blank space ‘ ’, empty line ‘_’ or empty box ‘ □’ may be used
instead of a variable letter. Later, after the introduction of school algebra in secondary school
including the signs listed above, further primitive function signs are introduced, including
sin, cos, tan, 1- and 2-place function signs (x c, xy), etc. At all levels in school algebra, the
formal signs may also be supplemented with written language (e.g., English, French, etc.).

13
Unlike some researchers, e.g., Duval (1995), Hayfa (2006), I do not see the need to create separate semiotic
systems for different types of sign/register/mathematical topic. I prefer a simple all-encompassing definition of a
semiotic system because it is more flexible and offers more generality than multiple system types. Although
there are complexities involved in coordinating different registers within one system, especially for learners, the
ultimate educational goal is for unity to triumph over difference.
14
Technically I should put, e.g., ‘=’ for =, in this account, but instead I am following common usage to allow = to
stand ambiguously for both a 2-place relation sign in the object language and for the metalinguistic sign (‘=’)
that names it in the metalanguage, where my discussion takes place.
Strictly speaking, if we add new primitive signs to a semiotic system we have changed it to a
new semiotic system. However, in practice we often act as if we are extending a single
semiotic system, or uncovering further parts of a single semiotic system.15

Rules
The rules for sign use, combination and production in a semiotic system can be analysed into
3 types, syntactic, semantic and pragmatic, after Morris (1945). The syntactic rules are based
only on the signs qua signs, such as rules for producing well formed formulas (WFFs), etc.
Thus ‘2x4=8=’ is not a WFF whereas ‘2x4=8’ is one, because ‘=’ is a 2 place relation sign that
for syntactic correctness must be combined with 2 well formed term (WFT) signs. The
complexity property of WFFs and WFTs described below is also syntactically defined.

Semantic rules concern the dimension of sign interpretation and meaning(s). Thus, for
example, deriving ‘2x=4’ from ‘2x+3=7’ in a semiotic system incorporating school algebra
can be justified in terms of the meanings of the signs ‘2’, ‘3’, ‘4’, ‘7’, ‘x’, ‘+’ and ‘=’. In
terms of significance the dominant sign in these expressions is the binary equality relation ‘=’
and this has an underpinning informal meaning of balance that must be respected to preserve
truth. The import of this is that whatever operation is applied to one of the binary relation
sign’s arguments (one of its ‘sides’) must also be applied to the other. Another feature at play
in this example is an implicit heuristic of simplification. This seeks to reduces the complexity
of terms in an equation en route to solution.16

Pragmatic rules include contingent and rhetorical rules and these are determined purely by
social convention. Examples include teacher requirements that students label answers with the
prefix ‘Ans. =’, and double underline the answer and this prefix (Ernest 1993). Likewise, in
university and research mathematics the end of a proof is commonly signified with the
Halmos bar ‘’ (analogous to the classical QED). Such pragmatic rules are socially imposed
or agreed conventions which are immaterial to syntactic and semantic correctness.

Semiotic systems in school and research mathematics extend from highly informal systems
with rules that are largely implicit or tacit, to at the other extreme, totally formalized systems
in which the rules are fully explicit. An example of an informal system is the semiotic system
and practice of arithmetical word problems. The symbols are alphanumeric; letters, numbers,
words, and elementary arithmetical operations (+, -, =); and most rules are implicit;
concerning the translation of written sentences into numerals and operations, and then
computing the answer. For example, “Mary has four sweets. Jim has three sweets. How many
15
Within limits this practice is justifiable, as we can embed a ‘substructure’ (a structure in which each of the
constituents sets of a semiotic system is a subset of the corresponding sets within a greater ‘superstructure’)
conservatively, within the said superstructure. However, the enlargement of a semiotic system, such as extending
the semiotic system of natural numbers to that of integers can lead to new metatheoretical results (e.g.,
multiplication can produce products less than either multiplicand) contradictory to the state of affairs in the
system of natural numbers, leading to epistemological obstacles (Ernest 2006). Arzarello (2006) extends the
notion of a semiotic system to that of a semiotic bundle. This comprises a collection of semiotic systems and a
set of relationships between the systems. This notion would allow a natural way of treating families of semiotic
systems related by the addition of further signs, rules or meanings, as discussed in the text.
16
The complexity C of an expression (term or formula) is defined inductively in terms of its syntactic structure.
The complexity of an atomic expression t, denoted C(t) = def 1. Given a set of k expressions t1, t2, …, tk, the
maximum complexity of which is n, and a k-place function or relation symbol F, the complexity of the
expression Ft1t2…tk, denoted C(Ft1t2…tk) = def n+1. Complexity is used as a measure of the impact of
transformational rules on a term constituting one of the arguments (‘sides’) of an equation. When a legitimate
(i.e., rule following) transformation of the equation achieves a reduction of the complexity of the equation or its
terms, normally, the task is closer to completion. The simplification heuristic motivates the use of rules to reduce
the complexity of terms or expressions, and it is normally implicated throughout the solution of algebraic and
arithmetical equations, for task goals are typically based on maximal simplification of texts.
do they have altogether?”. This can be translated into “Four and three; Add”, or ‘4+3=_’. But
this translation is based on heuristics, such as ‘How many altogether’ suggests ‘add’. This
cannot be turned into an explicit rule or algorithm, as the correctness of this interpretation
depends on the context; the sense of the whole set of statements. As has been noted in the
literature (Brown and Küchemann, 1976, 1977, 1981), children’s attempts to algorthmicize
this heuristic leads to incorrect surface rules such as ‘more’ or ‘total’ translate to ‘+’.

An example of a totally formalized system in which the rules are made as fully explicit as
possible is given by the formulation of the propositional calculus in Church (1956). Here the
rules of deductive inference are specified explicitly and completely. 17 Even in totally
formalized semiotic systems, the rules are based on the underlying meaning structure. It is just
that the process of distilling these meanings into explicit rules has reached a stage of
completeness, and so the rules can be applied in the generation of signs without reference to
the underlying meaning structure. The possibility of neglecting meanings in formalized
mathematics was noted three centuries ago by Berkeley (1710: 59): “in Algebra, in which,
though a particular quantity be marked by each letter, yet to proceed right it is not requisite
that in every step each letter suggest to your thoughts that particular quantity it was appointed
to stand for.” Thus meaning gives rise to rules with the result that the meaning structure can
be neglected (sometimes only temporarily) during the use of a semiotic system. The entire
edifice of mechanical calculation upon which digital computing is based depends on this
property. Computers apply rules to signs in completely formalized semiotic systems in which
there is no possibility of meaning once the computing structure is finally realized by humans.

Historically, the development of semiotic systems involves the distillation of the implicit
meanings of sign use, with their constraints and affordances, embodied in meaning structures,
into the more explicit system rules. But it should be noted that any translation of rules and
meanings from a implicit system (the meaning structure) into a more explicit form (the rules
of a semiotic system) is at best a homomorphism and can never be an isomorphism. In
translating relatively vague meanings into something more explicit, some elements of
meaning must always be lost and further elements must always be added. Refinement and
explication of meanings involves the social construction of new meaning, and considerable
human creativity and ingenuity may be involved.

But except for artificial examples of totally formalized systems, virtually none of the semiotic
systems at play in mathematics education or mathematical research have fully explicit sets of
rules. This because semiotic systems in use, equivalent to Saussure’s systems of parole, are
typically based on implicit or tacit rules inspired by the underlying meaning structure. In
semiotic systems, transformations of signs are typically permitted provided that they preserve
key meanings in the underlying meaning structure. In school algebra this is the truth (balance)
of equations, and also the relevance of transformations. In arithmetical word problems this is
the arithmetical meanings and values of the inherent arithmetic in the task statements. In
propositional calculus it is the truth value of the propositional signs. However, for each of
these three types, the rules are in many cases implicit, and are very often acquired by learners
or practitioners as ‘case law’ from the social use of semiotic systems.

17
But note that the rules specifying, for example, which axioms may be inserted into deductive sequences must
either be given as metalinguistic schemas, e.g., P(QP), where P and Q are metalinguistic variables ranging
over and replaceable consistently by any well formed formulas, or as an infinite set of all possible instances of
replacements in this expression within the object language (i.e., the signs of the semiotic system), or where there
is a single instance included as a privileged sign (axiom), e.g., p(qp), where p and q are elementary
propositional variables, coupled with a metalinguistic rule of replacement for Q by Q(p/q), whereby all instances
of p occurring in Q are replaced by instances of q in Q(p/q), such that if Q is true or assertable, so is Q(p/q).
In research mathematics, especially in the foundations of mathematics, the ideal semiotic
system aims to dispense with all uses of the underlying meaning structure, so that the rules for
sign use and production can be specified fully explicitly in syntactic terms. However, this can
never be fully achieved except in trivial cases because the specification of syntactical rules
requires metamathematical formulation. Thus even the specification of the propositional
calculus in purely syntactical terms (as discussed above) requires metamathematical
constants, variables and expressions, and their deployment depends on an underlying meaning
structure. Hilbert (1925), in his contributions to the philosophy of mathematics, freely
acknowledged that a meaningful but finitistic metamathematics is ineliminable when
mathematical theories are presented fully formally, i.e., with all their rules presented explicitly
in syntactical form (Neumann 1931).

Meaning Structure
The underlying meaning structure of a semiotic system is the most elusive and mysterious
part, like the hidden bulk of an iceberg. It is the repository of meanings and intuitions
concerning the semiotic system which support its creation, development, and utilitization
(Arcavi 2005). For individuals it can range from a collection of tenuous ideas and fleeting
images (Burton 1999), to something more well defined, akin to an informal mathematical
theory.

The meaning structure of a semiotic system can be described in three (equivalent) ways, as:
1. A set of mathematical contents;
2. An informal mathematical theory;
3. A previously constructed semiotic system.

A meaning structure is a loosely associated set of mathematical contents that can include:
signs, concepts, objects, properties, functions, relationships, rules, procedures, methods,
heuristics, classifications, problems, examples, ideas, images, metaphors, models, structures,
representations, propositions, theorems, arguments, proofs, theories, etc. It is a reservoir of
meanings that can be drawn upon in formulating, developing and operating a semiotic system,
such as the metaphor “equality is balance” for simple algebra (Sfard 1994).

An informal mathematical theory can serve as the meaning structure for a semiotic system in
which the sets of signs and rules constitute a formal mathematical theory. All formal
mathematical theories, I claim, have an underlying informal theory serving as its meaning
structure. Lakatos (1978) argues that such an informal theory provides the touchstone for
evaluating a formal mathematical theory. The theorems of the informal theory are potential
‘heuristic falsifiers’ through which the success of the formal theory can be judged, according
to whether it captures or contradicts them.

Two different formal theories are generally regarded as equivalent if their signs and rules are
inter-translatable, and they both share the same informal mathematical theory as their
meaning structure. Thus various formulations of Peano arithmetic, which may vary in their
sets of signs or rules, e.g., the choice of 0 or 1 as the first numeral sign, are regarded as
equivalent in this way, even though they are distinct semiotic systems.

Additionally, a previously constructed semiotic system can serve as a meaning structure for a
new semiotic system. Since the meaning structure of a semiotic system can include signs and
rules, this possibility is already inherent the first of the descriptions given above. It is also
possible for more than one existing semiotic system to be drawn upon or combined to make
up a meaning structure, and not all of these need to be mathematical systems. The
incorporation of entire or elements of non-mathematical semiotic systems into the meaning
structure provides the potential for links between mathematics and other human ways of
representing our experiences and environment.

This third case sounds circular, but it is not. Rather it is a case of what Peirce terms ‘unlimited
semiosis’, in which the interpretant of a sign is itself a new sign, and so on, ad infinitum. For a
semiotic system as a whole can be regarded as a sign, with its underlying meaning structure
constituting its interpretant (from the triadic, i.e., Peircean, perspective of signs) or its
signified (from the dyadic, i.e., neo-Saussurian perspective of signs). Like any other sign this
is linked via its origins, its uses in practice, its meanings and its associations with other signs,
or in this particular case, with other semiotic systems.

Often it will be the case that a newly developed or elaborated semiotic system (for an
individual or group of learners) will be more formally and explicitly specified than the
previously constructed or utilized semiotic system which serves as its meaning structure. In
the teaching and learning of mathematics it is commonly one of the goals of instruction to
increase the abstraction, complexity and formality of the semiotic systems to which learners
are introduced and inducted, over the course of time. Hence this gradient of increased
formality and explicitness.

Such processes are evident in institutionalized mathematics teaching at all levels. In school
mathematics the study of number properties and manipulations in numerical calculation
precedes and provides the meaning structure for elementary algebra. Operations in number
systems not only grow in complexity in the passage from the Natural Numbers, via Integers,
Rationals and Algebraic Numbers to Real Numbers, but also each of these transitions takes
the semiotic system as the meaning structure for the next semiotic system of number that is
sequentially developed. Typically in university mathematics the study of concrete structures
such as sets and number systems precedes the study of algebraic structures such as group, ring
and field theory, and the study of informal or ‘naïve’ set theory (Halmos 1974) precedes the
study of axiomatic set theory. In each these examples the study of an relatively informal
semiotic system precedes the study of its relatively more formal counterpart, and I claim,
provides a central part of the meaning structure of the subsequently developed, more formal
semiotic system.

The theory of semiotic systems provides a model for describing the teaching and learning of
mathematics in school. In learning any school mathematics topic in the form of a semiotic
system, learners are inducted into a discursive practice involving the signs and rules of that
system. Teachers present tasks in the form of signs, and present rules for working or
transforming the signs for accomplishing the tasks. Most commonly the rules will be
exhibited implicitly through worked examples, particular instances of rule applications, rather
than explicit rules stated in their full generality. Through observing the examples, working the
tasks, and receiving corrective feedback, learners internalize, build and enrich their personal
meaning structures corresponding to the semiotic system.

Trying to teach rules explicitly rather than through exemplification can lead to what I term the
‘General-Specific paradox’ (Ernest 2006). If a teacher presents a rule explicitly as a general
statement, often what is learned is precisely this specific statement, such as a definition or
descriptive sentence, rather than what it is meant to embody: the ability to apply the rule to a
range of signs.18 Thus teaching the general leads to learning the specific, and in this form it
18
This also applies to any general item of knowledge that is applicable in multiple and novel situations, such as a
mathematical concept, rule, generalised relation, skill or strategy.
does not lead to increased generality and functional power on the part of the learner. Whereas
if the rule is embodied in specific and exemplified terms, such as in a sequence of relatively
concrete examples, the learner can construct and observe the pattern and incorporate it as a
rule, possibly implicit, as part of their own appropriated meaning structure. This is how
children first acquire the grammatical rules of spoken and written language. Thus the paradox
is that general understanding is achieved through concrete particulars, whereas limited and
specific responses may be all that results from learning general statements. This resembles the
Topaze effect (Brousseau 1997), according to which the more explicitly the teacher states
what it is the learner is intended to learn, the less possible that learning becomes. For the
learner is not doing the cognitive work (meaning making) that constitutes learning, but
following surface social cues to provide the required sign – the desired response or answer.

The pattern whereby a learner first learns the use of signs through observation and
participation in public sign use in discursive practices embodies the well known dictum of
Vygotsky 1978: 128) “Every function in the child’s cultural development appears twice, on
two levels. First, on the social and later on the psychological level; first between people as an
interpsychological category, and then inside the child as an intrapsychological category.” This
Vygotskian scheme can be represented as a cyclic pattern for learners’ appropriation of signs
and the rules of sign-use through participation in a discursive practice. The pattern is
illustrated in Figure 1 (after Ernest 2005).

Figure 1. Model of Sign Appropriation and Use in Learning

SOCIAL LOCATION
Individual Collective

Learner’s public Social (teacher & others)

Public Conventionalisation
utilization of sign to negotiated and conventionalized
express personal meaning  (via critical acceptance) sign use
(Public & Individual) (Public & Collective)

MANIFESTATION Publication   Appropriation

Private Learner’s development of

personal meanings for
sign and its use
(Private & Individual)
 Learner’s own unreflective
response to and imitative use of
Transformation new sign utterance
(Private & Collective)

The cycle shown in Fig. 1 has four phases: Appropriation, Transformation, Publication and
Conventionalisation, any one of which can be taken as the beginning, for the cycle repeats
endlessly. In the figure Vygotsky’s two levels are shown, first, by the top right quadrant,
where the socio-cultural is represented as both public and collective, and second, by the
bottom left quadrant, where the (intra)psychological is represented as both individual and
private. This latter constitutes the notional space where a learner constructs his or her meaning
structure. The other two quadrants are crossing points on the boundary between these two
levels, and these are the locations where the learner semiotic agency is acted out.

In the development of a personal meaning structure a learner draws on further resources

beyond those indicated in Figure 1. These include existing meanings and the meaning
structures of other semiotic systems already partly mastered, as well as meta-discussions of
sign production and use. The latter are partly included in Fig. 1 in terms of social negotiation
and critical acceptance.
The model indicates schematically the route through which learners appropriate the rules of
sign-use, mostly through observing their exemplification in practice. The earliest uses of a
sign or rule can be based on simple imitation, corresponding approximately to Skemp (1976)
and Mellin-Olsen’s (1981) notion of instrumentalism, because of the performativity involved.
Later, after a sequence of related appropriations, performances and conventionalisations, the
use of the sign is transformed through the development personal meanings including a sense
of where and how the sign is to be used acceptably, and a whole nexus of other associations.
The successful appropriation and transformation of a sign, with its nexus of associated
meanings and meta-discourse, parallels Skemp’s (1976) notion of ‘relational understanding’.
This involves both being able to use the sign correctly, corresponding to conventionally
accepted usage within the micro-community of the classroom under the authority of the
teacher, and being able to offer a rationale or explanation for the usage.

The next phase is that of publication, in which individual learners engage in conversational
acts of sign utterance. These can vary from quick, spontaneous verbal, gestural or written
responses to a question or other stimulus, through to constructing extended texts elaborated
and revised over a period of time, prior to offering them to others. A group of learners can
elaborate such texts co-operatively, but such processes subsume several or even many sub-
cycles in which individuals utter signs to others in the group in an extended conversation
giving rise to jointly elaborated, negotiated and agreed texts.

The cycle is completed through the process of conventionalisation, in which signs are uttered
within the classroom conversation and are subjected to attention, critique, negotiation,
reformulation and acceptance or rejection, primarily by the teacher. Teacher approval is
normally the final arbiter of acceptance, because of the power and authority relations in the
classroom. Typically the conventionalized sign utterance that is accepted will satisfy three
criteria:

1. Relevance. The sign or text is perceived to be a relevant response or putative solution (or
an intermediary stage to one) to a recognised (i.e., sanctioned) starting sign which has
the role of a task, question or exercise. This might be teacher imposed or otherwise
shared and authorized.
2. Justification. The mode and steps in the derivation of the sign from the authorized starting
point will normally be exhibited as a semiotic transformation of signs, employing
acceptable rules or sign transformations within the semiotic system, or justified meta-
linguistically.19
3. Form. Both the signs and their transformations will normally exhibit teacher-acceptable
form, thus conforming to the rhetoric of the semiotic system involved as realized and
defined in that classroom. This system could be that of spoken verbal comments, drawn
and labelled diagrams, numerical calculations, algebraic derivations, or some
combination of these or other sign types (Ernest 1993).

An overall schematic model of a semiotic system within its social context of use is given in
Figure 2. This summarizes the three parts that make up a semiotic system; the signs, rules and
meaning structure. The signs can be elementary, or composite, and indeed sequences of signs
will often be all that is explicitly exhibited in the operation of the semiotic system. Many rules

19
Sign transformations do not always mean the replacement of one or more parts of a compound sign by
different parts, with the retention of the unreplaced parts. It may involve the construction of a wholly new sign in
the sequence. For example, axiom use in a logical proof, can involve the insertion of a new sign with no
components shared or overlapping with the previous step.
will often be implicit, exhibited only via specific applications as sign transformations, as well
as explicitly presented rules. For learners the meaning structure will be developed as the
semiotic system is utilized, although learners will also draw on other, partially mastered,
semiotic systems as well as their general repertoire of meanings, including some of the items
listed, from signs and concepts to the relevant informal mathematical theory.

Figure 2. Model of Semiotic System within its Social Context of Use

SOCIAL CONTEXT OF USE

SEMIOTIC SYSTEM

SIGNS RULES
Implicit Rules
Elementary Signs (Exhibited via applications in
Composite Signs)

Composite Signs Explicit Rules

MEANING STRUCTURE

Selection of: Signs, Concepts, Properties,

Functions, Relationships, Procedures, Methods,
Classifications, Problems, Examples, Images,
Metaphors, Models, Structures, Propositions,
Arguments, Proofs, Informal Mathematical
Theory, Other Semiotic Systems

Meta-Discussion of Semiotic System in Use

Aims, Goals, Purposes, Concept of Task
Roles, Positions, Power Relations, Relations with Social Institutions

The use of semiotic systems always takes place within a social context. Within social settings
there are persons and their roles, positions, power relations, and relations with social
institutions such as schools. An important dimension of social understanding as it relates to
semiotic systems is the concept of school learning task and the aims, goals, and purposes of
school work, that is presupposed by operating semiotic systems in school settings. The
transformation of signs in semiotic systems is directional, and the understanding of
directionality in general is socially acquired in a variety of social settings including home and
school.20 Ultimately, directionality in activities results from directions given by a person in
20
Note that the simplification heuristic described above plays a central role in operationalizing directionality in
mathematical tasks. That is, a significant part of the appropriation of directionality is associated with the implicit
some powerful role. Conversation is a major channel for the communication of such matters
(Ernest 1998). Where it concerns working matters pertaining to the semiotic system in use,
i.e., the signs produced, the rules employed, and the accomplishment of mathematical tasks, I
term this meta-discussion of the semiotic system. This plays an important role in correcting
and shaping sign production and rule deployment, as well as enabling the development of the
public meaning structure within the particular social context.

The semiotics of mathematical text

Mathematical text is a peculiar form of text on account of its subject matter. Mathematical
text does not refer to the embodied world of our experiences, with its frameworks of space
and time. Instead it is understood to refer to a different realm, one in which time stands still or
does not exist. I have earlier critiqued the myth that this other realm is a Platonic universe
beyond our social and material world (Ernest 2008). Instead mathematics refers to a semiotic
space, a socially constructed realm of signs and meanings. Human beings are sign using and
sign making creatures, and most humans can participate in the semiotic space of mathematics,
even if only to a limited extent. However, this participation involves the deployment of
semiotic resources and tools and the assumption of certain identities. Entry to this semiotic
space is via the texts of mathematics, but it is important to assert that it is not a timeless zone
(Mazur 2004). While there is no universal timepiece ticking away in semiotic space,
nevertheless individual and group engagement in mathematical activity is always over time
(Mason et al. 2007).21 What this means is that accessing mathematical texts always has a
sequential nature. Sequential development is the semiotic or logical analogue of time. While
reorderings of mathematical texts are always possible, and often effected, just as they are in
the construction and editing of films and fictional texts, as well as academic texts like this
one, at any stage the text is made up of sequentially ordered contiguous signs. This sequential
contiguity is the analogue of time, and is often enacted through the use of space. 22 Any array
of semi-permanent material signs, i.e., marks on a page rather than sounds in the air, is
ordered by spatial conventions of access into a beginning, middle and end. In any form of
representation, there is always an ordering present, and this structures the access and role of
readers.

To explore the semiotics of mathematical text including the roles of readers and writers it is
necessary to look at the language of mathematics. Here it is important to look beyond
mathematics to other disciplines which have analysed language and text without the
ideological presuppositions often present in discussions about mathematics (Ernest 2008). An
obvious area to turn to is linguistics, whose subject matter is language and text.

The linguistic theorist Halliday (1985) has developed the theory of Systemic functional
grammar which provides a illuminating tool for this analysis. In some of his publications he
and his colleagues have examined the linguistics of mathematical and scientific text (Halliday
and Martin 1993, UNESCO 1974). Halliday distinguishes three metafunctions of text in use,
and these can usefully be applied to mathematical text. These are the ideational, interpersonal
and textual functions.

understanding of the simplification heuristic as a technique for goal-directed activity.

21
Note that there is no universal time in empirical space either, as Einstein’s Relativity Theory tells us. However,
there are standardized, socially agreed regional time conventions in the local physical world in a way that there
are not in mathematical semiotic space.
22
Spatial or sequential contiguity also correspond with the syntagmatic and metonymic axes referred to in
linguistics; whereas equivalence and other substitution relations correspond with the paradigmatic and
metaphoric axes.
1. The ideational or experiential function concerns the contents of the universe of
discourse referred to, the subject matter of the text, the propositional content. This
includes the processes described and the objects or subject matters involved in the
process. Morgan (1998: 78) relates this to mathematical questions such as “What does
this mathematical text suggest mathematics is about? How is the mathematics brought
about? What role do human mathematicians play in this?”.
2. The interpersonal function concerns the position of the speaker, the interaction
between speaker and addressees, and their social and personal relations. The related
mathematical questions suggested by Morgan (1998: 78) include “Who are the author
and the reader of this mathematical text? What is their relationship to each other and to
the knowledge constructed in the text?”
3. The textual function is about how the text is created and structured, and how it uses
signs, and so on. Morgan (1998: 78) relates this to mathematical questions such as
“What is the mathematical text attempting to do? Tell a story? Describe a process?
Prove?”

As Morgan (1998) points out, these cannot be treated as wholly separate for mathematical
text, as, for example, the ideational function of representing the universe of discourse overlaps
with the textual function of binding linguistic elements together into broader texts. Both treat
the semiotics of mathematics, where subject matter and form are indissolubly bound up
together.

The ideational function

In exploring the ideational function of mathematical text the following questions arise. What
is the propositional content of mathematical text, and what is mathematics/mathematical text
about? What objects and processes are described? The answer from a semiotic perspective is
that mathematics is about mathematical signs and the operations applied to them (Radford
2002). The meaning of the signs of mathematics resides primarily in their uses and functions.
Some signs, such as numerals, appear to represent objects, numbers in the case of numerals.
Some signs represent operations on signs, and through them appear to represent processes
applied to objects. Thus the 2-place addition operation sign ‘+’ is applied to numerals, and
thus appears to act on numbers (objects). However, the classification of signs as representing
objects or processes is relative. It depends on the function that is foregrounded and the
perspective to which it gives rise. In fact all mathematical signs represent both objects and
processes.23 A sign represents an object when viewed or used as a unified entity in itself, or as
a single signifying entity. A sign represents a process when viewed in terms of its parts or in
its actions on, or relations with, other signs. For the coordination or structuring of parts into a
whole, even when apparently static, is a process or relation. For example, the number 3 is on
the one hand, a basic mathematical object. But on the other hand, in Peano arithmetic it is
formally defined as the successor of 2, and ultimately defined only in terms of the two
primitive signs S and 0 (3 =def S2 = SS1 = SSS0). This analysis, which incidentally mirrors
both historical and psychological developments of number (Ernest 2006), reveals the number
3 as constituted by a process.

Mathematics is constituted by semiotic systems, and by being made up of signs and sign-
based activities it appears to be at one remove from its subject matter, the objects and
processes of mathematics. For signs, by their nature, are always distinct from the objects they

23
Several authors have remarked on the dual nature of mathematical objects from philosophical, psychological
and sociological perspectives, and stressed the role of reification in the construction of objects from processes in
mathematics (Ernest 1991, 1998, Machover 1983, Radford 2002, Restivo 1992, Sfard 1987, 1994).
signify.24 Thus one might say that if the objects and processes of mathematics are to be found
anywhere, it is in the underpinning meaning structures of semiotic systems. For this is the
domain to which the signs within a semiotic system refer or ‘point’. At first blush, all appears
well and good, until we remember what makes up the meaning structures, notably the objects
and processes of mathematics and their constitutive signs, concepts, functions, and so on, up
to and including other semiotic systems. But this then raises again the question of the nature
of the objects, processes and other entities that make up the meaning structures. What is this
domain to which the signs of mathematics refer? It is nothing but the semiotic realm of
mathematics itself, for mathematical signs and texts do not represent or refer to some reality
beyond our material world, as I argued above. Nor do they represent our material world itself.
What mathematical texts and signs refer to is other mathematical signs, and these are cultural
objects, not material objects. The vast array of mathematical semiotic systems created,
communicated and sustained by human activity both make up the sign systems of
mathematics and the subject matter to which these signs refer. Far from being a vicious, self-
contradictory cycle, this is a virtuous cycle that creates and brings into being an ever growing
universe of mathematical signs, objects and semiotic systems.

Just as human imagination brings into being the countless characters, situations and worlds of
fiction, so too it brings into being the unending realms of mathematical objects, including all
of the constituents of the meaning structures of semiotic systems. Unlike the former, which
are usually the product of individual imaginations, the latter are social constructed. For it is
not enough for just one mathematician to assert that mathematical theories and objects exist.
They need to be shared and accepted before they can be said to be properly mathematical, as
indicated in Fig. 1 above.25

Thus the signs and texts in a mathematical semiotic system refer both to the signs of the
semiotic system itself, and to objects and processes in the meaning structure of the system,
which are themselves sign-based. The meaning structure draws on signs and meanings created
in other semiotic systems, but also grows in meaning and complexity as the semiotic system
develops and creates its own domain of signs during its development and use.26

Mathematics is a special subject matter in that its signs refer only to other signs and sign-
based processes and operations. It is also unique in the depth and complexity of its sign
formation operations whereby sign-based processes are condensed and reified into objects,
and so transformed and constituted as further sign objects. 27 In most of the semiotic systems
in mathematics these operations create a potentially endless supply of signs of increasing
structural complexity and abstraction. Mathematical theories sometimes encompass such
unending processes as a whole and in one bound turn them into new, yet more complex signs.
24
This follows from both de Saussurian and Peircean theories of semiotics, as well as theories of the sign (e.g.,
Morris 1952). Even in the extreme case of formal mathematical systems, in which it is possible to have a sign
signifying itself, such as in Henkin’s (1959) proof of the completeness of the first-order functional calculus, this
involves two wholly distinct roles for the sign in different domains, serving as signifier in the metalanguage (the
mention-function) and signified in the object language (the use-function).
25
Perhaps this is also true of the objects of fiction. Maybe they do not exist in any meaningful way until they
have been both described in fiction and interpreted by a reader.
26
Semiotic systems have a dual nature as is illustrated above, in Figure 1, both public/collective and
private/individual. Culturally semiotic systems grow and develop historically as their creators develop them as
human socio-cultural artefacts. For educational purposes these are reconfigured, recontextualized, and presented
to learners for them to meet on the interpsychological plane and internalize and appropriate on the
intrapsychological plane (Vygotsky 1978). Through such processes learners reconstruct private/individual
meaning structures, although their sign utterances and rule applications are primarily public.
27
Elsewhere I argue for the importance and prevalence of the reification in the semiotics and philosophy of
mathematics (Ernest 1991, 1998). See also Note 11.
By these processes the semiotic systems of mathematics can represent and incorporate infinite
objects and processes.

The Social Function of Semiotic Systems

This is primarily to represent and to solve mathematical problems and to work mathematical
tasks. These operations require the transformation of texts employing sign-based processes
following the rules of the semiotic system. Explicit and implicit rules are the key operative
mechanisms and principles through which new signs are formed and composite texts are
constructed and elaborated. The overt function of these rules is to provide a technology for the
transformation of mathematical signs in a goal directed way. 28 That is, a means for bringing
the signs closer to some desired (and sometimes locally defined) canonical state, and in so
doing preserving key invariants within the meaning structure. The two best known types of
transformations, corresponding to two dominant problem solving activity types, are numerical
calculation and proof.

Calculations are sequences of numerical terms that, through a sequence of numerical value-
preserving transformations, lead to the derivation of new terms that are relatively simpler 29
and increasingly approach one of the canonical forms of numerals. For example, the standard
canonical form for numerals is the decimal place value representation, which can be



represented as 0 10nxkn, for nN, where kn=0 except for a finite number of values of n,
where 0<kn<10. The key characteristic of this representation is that for any numerical value, it
is unique.30 Because of this property many calculational tasks, and the sequences constructed
in accomplishing the tasks, have unique endpoints (answers), which is a special and defining
characteristic of school mathematics. In a semiotic system encompassing arithmetical
calculations, there are an infinite number of terms that can be transformed onto any given
numerical value.31 Thus for any answer there is an unlimited corresponding set of tasks.

This discussion illustrates how the rules for the transformations of terms in numerical
calculations are based upon the principle of preservation of numerical value. There are also
rules used in inequality transformations that, for example, do not increase (they preserve
identically or decrease) the numerical values. Similarly, deductive proof sequences draw upon
rules for the transformations of propositional signs that do not decrease the truth values
attached them. As in these cases, the rules of the semiotic systems typically originate in or
refer to certain interpretations or valuations of the signs in the underlying meaning structure
and preserve these attributes as invariants.

In addition to providing a technology of sign transformations, semiotic systems have a further

function in providing representations of mathematical problems. Mathematical texts provide
users with a language for representing structural aspects of a range of situations. However,
situations are never given directly, they are always mediated by a second semiotic system,
acting as part of the meaning structure of the mathematical semiotic system in which the new
28
The rules embody in operative form the structural meanings of objects, relations and processes in the meaning
structure of the semiotic system.
29
By simpler I mean having reduced complexity, as discussed above (see Note 4).
30
My illustration only concerns the numerals representing natural numbers, although with further elaborations it
can be extended to other domains of number including Rational and Real numbers, bearing in mind the
complexities introduced by infinite decimal representations.
31
I leave aside the problem, which encompasses almost the whole history of mathematics, of numerical terms
signifying objects that are not part of the semiotic system in which the terms have been constructed, e.g., given a
semiotic system for N, if m>n N and, then n-mN. The generation of such non-elements is often a motivating
factor in the extension of semiotic or structural systems in the history of mathematics to new systems to include
them.
texts are constructed. Typically such subsidiary semiotic systems are less formal, perhaps
more intuitive systems than that in which the ‘modeling’ text is constructed for application
and use. The wide range of objects that can be drawn upon in constituting the meaning
structure of a mathematical semiotic system includes existing semiotic systems, some drawn
from outside of mathematics. Thus the structural modeling function enables the construction
of texts that on the one hand enable the application of mathematics to extra-mathematical
situations, and on the other hand, serve as starting points for transformations within the
mathematical semiotic system. It enables semiotic systems to both ‘look outwards’ via
modeling and application functions, and to ‘look inwards’ towards the sign transformation
functions.

The modeling function utilizing the relation between a primary mathematical semiotic system
and a subsidiary semiotic system incorporated informally within the meaning structure of the
former can serve a range of functions. It can be the location for plans or schematic
representations to be implemented in the primary semiotic system. As a subsidiary domain of
signification and meaning, it can be used to formulate informal representations serving as
plans or outline ideas, to be used to guide sign constructions and transformations in the
primary semiotic system. For example, a mathematical proof is often based on an informal
proof idea (Rotman 1988). Such an idea, represented in a subsidiary semiotic system, serves
as the plan or schematic outline of the proof transformation constructed in the primary
semiotic system as a sequence of signs. This can be generalized because quite often, rather
than being generated ‘from scratch’ in semiotic systems, transformational sequences are
implemented ‘translations’, in some loose sense, of schematic ideas or plans developed
outside the semiotic system. Typically such outline ideas are generated and represented in a
subsidiary semiotic system incorporated in the meaning structure.

Halliday’s (1985) theory of Systemic functional grammar provides a illuminating tool for this
analysis. In some of his publications he and his colleagues have examined the linguistics of
mathematical and scientific text (Halliday and Martin 1993, UNESCO 1974). Halliday
distinguishes three metafunctions of text in use, and these can usefully be applied to
mathematical text. These are the ideational, interpersonal and textual functions of his theory
of social semiotics (Halliday 1975, 1985). Having previouslyconsidered the implications of
the ideational function of in his social semiotics in this section I focus on the on the social
contexts of mathematical text. In particular, looking at what the interpersonal and textual
metafunctions of his theory might mean in elucidating the semiotics of mathematical text.

The Interpersonal Function

The second metafunction of mathematical text according to Halliday’s scheme is the
interpersonal function. This concerns the positioning of the ‘speaker’, that is the author of the
text, the positioning of the reader, and the relationships between the author and addressees as
embodied in the text (Morgan 1996).32

From the perspective of semiotic systems, there are two levels of language and text. First
there are the signs or texts of a semiotic system, which is where mathematics is constructed,
32
Although I refer to the speaker, author, utterer, writer of mathematical text, if not interchangeably, without
drawing firm distinctions between the spoken and written text, I fully appreciate that there are significant
differences. As Derrida (1976) has argued, writing is not reducible to the spoken word. Rotman (1994) further
argues that mathematical text is not reducible to alphanumeric text, let alone the spoken word (see the discussion
of the textual metafunction below). However, I previously argued that the texts I refer to are constituted by
multimodal sets of signs, all of which (in token form) are physically embodied, some on the scribbled or printed
page/computer screen, and some gestural, spoken or otherwise physically embodied. Viewed in this way, the
spoken/written distinction loses its force.
utilized or otherwise enacted. Second there is the metalanguage employed within the social
context in which the semiotic system is utilized. This corresponds to actors discussing
activities related to the semiotic system, or its social context, rather than enacting the
mathematics itself. Rotman (1993) makes a comparable distinction between the Code and the
MetaCode of mathematics.

In the school mathematics context, both of these two levels of text are involved in the setting
of tasks, their performance by the students, and commentary on and evaluation of the texts
produced. Classroom texts spoken by the teacher, often in conjunction with other supporting
modes of representation, position the actors in a number of reciprocal and pairwise defined
roles including task setter – task performer, work manager – work producer, assessor –
assessee, knowledge giver – knowledge applier, knowledge owner – knowledge requester. 33 In
each case the student is in the second of the two roles, the less powerful position. This reflects
the fact that the teacher has two overlapping roles, namely as director of the social
organisation and interactions in the classroom, i.e., social controller, and as director of the
mathematical tasks and work activity of the classroom, i.e., task controller. This corresponds
to the traditional distinction between being ‘in authority’ – social regulator – and ‘an
authority’ – knowledge expert (Amit and Fried 2005, Lloyd 1979).

In written classroom texts only some of these listed personal positions and roles are embodied
in the text, including task setter – task performer, knowledge giver – knowledge applier, with
the student/addressee adopting the second of the two roles, as before. Most of the roles
prescribed for students, whether in spoken or written texts, are to a large extent implicitly
embodied and encoded at the level of semiotic system texts, i.e., at the Code level.

Halliday has argued that positionings in the text become a surrogate for social regulation.
They stand in for and reproduce social structures and power differentials as experienced by
children. Thus there is a “chain of dependence such as: social order – transmission of the
social order to the child – role of language in transmission of the social process – functions of
language in relation to this role – meanings derived from these functions” (Halliday 1975: 5).
Thus social structures and power relations are embodied in language uses (in discursive
practices), and in particular, in the uses of texts.

Post-structuralists like Henriques et al. (1984) assert the potency of the constitutive triad of
power-knowledge-subject. They challenge the concept of unitary human subject and argue
that through the confluence of power and knowledge embodied in socially located texts not
only positions but subjectivities are formed. In place of the human subject as unitary agent,
they see the “subject as a position within a particular discourse” (Henriques et al. 1984:
203).34

The formative import of text and discourse in the construction of subjects and selves can be
traced back to the works of G. H. Mead and Vygotsky, via such processes as are described
above in Figure 1. Such views are stressed by discursive psychologists including Gergen
(1999), Harré (1979), Harré and Gillett (1994), and Shotter (1993), who see distinct identities
being constructed for an individual within differing discursive practices, according to the
linguistic and social positionings in play.
33
Morgan (1998) also has identified a number (8) of roles for teachers and students, indicated by the language
used in assessment tasks, which overlap with but are distinct from those given here.
34
Henriques et al. (1984) go on to problematize of the tensions between different subject positions, and the
production of human subjectivity. However, my concern takes for granted a temporarily unproblematized
concept of a person to ask: How are new (subsidiary) subjectivities or identities constructed through engagement
with mathematical texts/contexts?
More broadly, the strong impact of attitudes and beliefs on the formation mathematical
identity is well known through a variety of studies, although commonly drawing on more
traditional theorizations. The social construction of mathematical ability (and inability) is
multiply theorized by psychological, sociological, educational and feminist researchers (see,
e.g., Burton 1988, Buxton 1981, Diener and Dweck 1978, Ernest 1995, 2005a, Evans 2000,
Fennema and Leder 1990, Walkerdine 1998). There is currently a growth of interest in
research on mathematical identity within the mathematics education research community
(e.g., Boaler 2002, Boaler et al. 2000, Black et al. (2006), Grootenboer et al. 2006).35 But
little of this work treats to role of mathematical text in positioning its readers (and writers)
and the impact of this on identity construction.36

Semiotics offers some tools that further this project. In analyzing the role of texts and their
effects on readers Eco (1984) theorizes the Model Reader presupposed by and produced by
the text.

The author has to foresee a model of the possible reader (hereafter Model Reader)
supposedly able to deal interpretatively with the expressions in the same way as
the author deals generatively with them.
At the minimal level, every type of text explicitly selects a very general model
of possible reader through the choice (i) of a specific linguistic code, (ii) of a
certain literary style, and (iii) of specific specialization-indices. …
Many texts make evident their Model Readers by implicitly presupposing a
specific … competence. … But at the same time text … creates the competence of
its Model Reader (Eco 1984: 17)

This raises the questions. How do mathematical texts impact on their readers, and what is
specific about these texts (and their readers) in the context of mathematics? What are the
characteristics of the model reader(s) of mathematical texts? What assumptions are made of
the model reader in order to engage with the text and what are the constitutive impacts on the
model reader of these engagements?

According to Radford (n. d.) the model reader’s interpretation of a text falls within a system
of knowledge, which when applied to mathematics, comprises:
1. systematic knowledge: knowledge of language as a vocabulary and set of grammatical
rules, and in particular, the signs and rules of the semiotic system(s) of mathematics in
play;
2. encyclopedia: encyclopedia of cultural knowledge and conventions, including general
knowledge of mathematics, its meanings and domains of interpretation and
application, both within and outside mathematics; and

35
It is worth noting that these projects draw on differing theoretical bases. Grootenboer et al. 2006) contrast
three different approaches: (1) the [cognitive] psychological/developmental; (2) the socio-cultural; and (3) the
poststructural. Black et al. (2006) draw upon three distinct theoretical bases in their group project investigating
identity, participation and mathematical relationships, namely (1) socio-cultural theory, (2) discourse theory and
(3) psychoanalysis, but view them as complementary. The present paper stems from a social constructivist/socio-
cultural perspective, but also draws on discursive/ poststructural and cognitive psychological concepts as tools
for semiotic analysis.
36
Walkerdine (1998) explicitly discusses the role of text in constructing subjectivities without detailed
elaboration, but most references in the literature refer to the discursive production of selves without explicitly
highlighting the role of written text in such processes (e.g., Evans et al. 2006, Davies and Harré n.d.).
 3. experience: history of previous interpretations and engagements with other texts,
primarily mathematics texts of various forms, but also including some popular texts or
others unrelated to mathematics.

Thus the reader of a mathematical text both draws on and develops their systematic
knowledge (of semiotic systems), their broader personal encyclopedia of knowledge, and their
experience of earlier engagements with mathematical tasks and other texts. 37 (And of course
the development of the reader’s knowledge in this way constitutes the major goal of
education, namely learning.)

Eco (1984) distinguishes between the model reader (a theorized and generalized addressee)
and the empirical reader (an actual reader of a text). Clearly Radford’s three types of
knowledge (systematic, encyclopedic, experiential) will vary between model and across the
range of actualized empirical readers. This is especially important in mathematics, where
levels of competent engagement with mathematical text vary greatly.

In addition to drawing on and impacting on readers’ knowledge, engagement with

mathematical text results in the development of different selves or identities within the
reader.38 I assume these to be different from the selves or identities developed in extra-
mathematical contexts and practices.

The deepest analysis of the different identities developed within mathematics is provided by
the semiotic theory of mathematics due to Brian Rotman. As part of his project Rotman
(1988, 1993) analysed the language of published research mathematics texts. He identified
sentences to be the main linguistic units, and these to be made up of symbols, terms (nouns)
and verbs. Following the traditions of literary and grammatical analysis, he takes the type of
verb case in mathematical sentences to be the main indicators of the roles of author and
addressee. He finds these verb cases to be of two main sorts.

First, there are verbs in the indicative mood, concerning the communication or indication of
information. In this case, “the speaker of a clause which has selected the indicative plus
declarative has selected for himself the role of informant and for his hearer the role of
informed” (Berry 1975: 166). Thus the speaker/author asserts to the hearer/addressee some
state of affairs that obtains, or more commonly in mathematics where texts describe
mathematical actions and processes, the outcomes of these processes. Such sentences not only
describe the outcome of past, contingent sequences of actions and procedures, a particular
transformation of signs, but also claim that when operating within the rules of the language
game (the semiotic system), the outcome described is what always must happen. The
descriptions of these outcomes resemble logical predictions, taking place in a timeless realm
but describing the logical outcomes of the processes involved. Thus indicative propositions
37
This can be compared to models of mathematical thinking and problem solving, such as that of Charles and
Lester (1982), that specifies three factors: experience, affect, and cognitive factors. This model does not
distinguish systematic from encyclopedic knowledge, and introduces the novel factor of affect (interest,
motivation, pressure, anxiety, etc.). Schoenfeld (1992) offers a more elaborated model with five components:
knowledge base, strategies, metacognition, beliefs and affects, and practices (knowledge of real-world contexts,
experience). This again introduces affect (including belief). It also distinguishes two types of systematic
knowledge: knowledge base and strategies. However, the version of encyclopedia is limited and is put with
experience into one category. A further novel component is metacognition, the self-knowledge and self-
regulation of the mathematical reader/enactor. These are typical of such models in mathematics education
research (e.g., Charles 1985), in adding further psychological details but underestimating the significance of the
encyclopedia.
38
It is in the development of these different selves or identities that the affective factors indicated in Note 37
importantly come into play.
might be said to describe thought experiments which persuade us to accept the validity of
their assertions (Peirce 1931-58).

Second, there are verbs in the imperative mood, requesting that an instruction or action be
carried out. There are two forms:
(i) the inclusive imperative (e.g. “Let us define ... .”, “Consider a language L...”), in
which the addressee is required to cooperate or collaborate in following the
speaker or carrying out the instruction in some imposed shared realm of discourse
jointly, and
(ii) the exclusive imperative (e.g. “Add…”, “Count the cases...”, “Integrate the
function...”) which demands that an action be carried out by the hearer alone in a
presupposed shared frame.
In both cases “the speaker of a clause which has selected the imperative has selected for
himself the role of controller and for his hearer the role of controlled. The speaker expects
more than a purely verbal response. He expects some form of action.” (Berry 1975: 166)

These findings are echoed in Shuard and Rothery’s (1984) analysis of school mathematics
texts. They found several types of text, each with its own purpose: Exposition, Instructions,
Examples and exercises, Peripheral writing, and Signals. These go beyond Rotman’s analysis,
presumably because the function of school texts goes beyond that of describing mathematical
results. They found expository writing utilizing the indicative mood in school mathematics
texts, although these typically provided exposition of concepts and methods, including
explanations of vocabulary, notation, and rules, rather than fully fledged proofs.

Another type of language in school mathematics texts utilizes the imperative mood. This
included instructions to the reader to write, draw or to perform some action, typically utilizing
direct imperatives. It also included examples and exercises for the reader to work on. Often
these are routine problems involving the application of specific procedures to sign, but they
can also include word problems, non-routine problems and investigative work requiring the
use of general heuristics to guide solutions. These tasks may be expressed with direct
imperatives, but can also utilize implied imperatives if they are in question form or contain
the substance of imperatives without the appropriate verb.

Beyond these two types of text, Shuard and Rothery also found peripheral writing, including
introductory remarks and meta-exposition encouraging the reader, giving clues, and so on,
that utilized the indicative mood but at the metalinguistic level. Lastly they found signals,
including headings, letters, numbers, boxes, logos. These are not assertions but meta-signals
to the reader to give the text structure and emphasis.

There is thus a good correspondence between the types of language employed in school
mathematics and research mathematics, as indicated by the verb forms. However, the roles of
reader and writer and the social power relations they embody are more sharply distinguished
in school texts than in research texts. This reflects the clear cut role and power relation
differentiation in the social context of schooling, where the teacher and pupil roles are not
interchangeable. The teacher is almost invariably the writer of the task text (or their surrogate,
its presenter), while the pupil is the reader/enactor of the text. In research mathematics texts
although the author role controls that of the reader, it is expected that mathematicians will
assume the roles of both reader and writer according to whose text is being read. 39

39
Indeed mathematicians, certainly in Anglophone countries such as USA and UK, are loathe to bestow the title
‘mathematician’ on anybody who is not a writer of research mathematics texts, irrespective of whether they
routinely read research mathematics texts, or work professionally with mathematics in other ways.
Furthermore such readers can also adopt the meta-role of critic of research mathematics texts,
in a way that is not normally encouraged ed in the context of the mathematics classroom.
However, Walkerdine argues that classroom indicators of mathematical success includes both
rule-following and rule-challenging.

To be successful, children must follow the procedural rules. However, teachers

perceive breaking set as the challenging of the propositional rules. They read it as
‘natural flair’. … To challenge the rules of mathematical discourse is to challenge
the authority of the teacher in a sanctioned way. Both rule-following and rule-
breaking are received — albeit antithetical — forms of behaviour. (Walkerdine
1998: 90)40

Thus the positioning of the reader of school mathematics text is a complex and contradictory
one if the reader is to both develop the powers that are available through engagement with the
text (including acting on its imperatives) and be perceived by the teacher as so doing.

Rotman (1993, 2003) continues his account of the semiotics of mathematics with a theory of
the mathematician as both an author/writer and addressee/reader of mathematical text, based
on the linguistic features of such text. He posits three aspects of the identity of the
mathematician, three ‘actors’, each with a corresponding and increasing level of
abstractedness, but with diminishing agency and subjectivity, in the sequence: Person, Subject
and Agent.

The Person is embedded in the material world and has full access to voluntary human
activities including the uses of language, metalanguage, and meaning-making in general. 41 In
the classroom, the student as Person hears the teacher’s instructions to engage in a specific
mathematical task. The Person is the only one of the three agentic roles that has full access to
metalanguage as well as the language of semiotic systems, to indexical signifying resources,
which are typically excluded from the signs of mathematical semiotic systems, and to the
meaning structure associated with the semiotic system.

A subsidiary, abstracted and restricted identity is that of the mathematical Subject. The
Subject has a restricted agency corresponding to the mathematical languages that Rotman
(1993) terms the Code (as opposed to the MetaCode) of mathematics, corresponding to the
signs and texts of the semiotic systems of mathematics. The Subject is circumscribed by sign
systems that lack indexical markers for time, physical place and other personal modes of
expression. Thus the Subject lacks the Person’s capacity for self-reference or self-expression
(not to mention feeling), and is only able to read and write within the semiotic system itself. A
student evokes/assumes the identity of mathematical Subject in carrying out a task within the
semiotic system, although she may punctuate this performance by assuming the full identity
of Person in making meta-remarks about the task, or in attending to events or other personal
matters and breaking out of the restricted functional identity of Subject (or of Agent).

The Agent is a minimal representative of the reader or writer which has no voice, no
subjectivity, only the power to carry out imagined instructions as defined in the text. The
Agent is a ‘skeleton diagram’ of self (Peirce, 1931-58, 2: 227), like the moving fingertip on a
map tracing out an imagined or past journey, but in the realm of signs. The Agent is

40
Walkerdine goes on to argue there are gendered presuppositions about the appropriateness of rule-breaking,
and that it was valorised by teachers for the boys that she studied in the 1980s and pathologised for the girls.
41
It is the Person (and only the Person) who owns and experiences beliefs, affects and feelings, and has
metacognition, subjectivity and voluntary agency.
 the actor associated with the domain of procedure who functions as the delegate
for the Person through the mediation of the Subject; the Agent executes a
mathematically idealized version of the actions imagined by the Person and it
does so formally since it lacks the Subject’s access to meaning and significance.”
(Rotman 2003: 3).

The Agent represents an unconscious agency that can only follow fully specified procedures
without any decision making. Because of these features it is possible to fully mechanize the
agency involved, which is what underpins digital computing. The Agent in mathematics could
only be imagined until electronic computing was developed, with its electronic mechanization
of the mathematical agent. A semiotic realm has been created, namely digital information
space, where a surrogate agent can act out its instructions mechanically.

In the context of schooling, when a learner has developed automaticity in a certain range of
rule-based transformations of text, the Agent is the sub-identity of the Person that enacts
them. The Agentic self carries out mechanical procedures such as those involved in ‘27+91’,
‘Simplify 3a+5b-2a+b’, or ‘Solve 39x-42=17’, once they have become automatic for the
student. In carrying out such tasks there are three stages. First the Person acknowledges or
recognises a task and its goal structure. Second the Subject identifies the procedures to apply.
Third the Agent applies the rules of the semiotic system involved to the text and generates the
sign sequence to which they give rise. The person monitors the activities of the Agent (and
Subject) to ensure the chosen actions fulfil the task goals. Thus movement between and
alternation of the three roles during the working of tasks is not only a possibility, but in many
cases is a necessity.

To be successful a student of mathematics needs to be able to make transitions between these

roles/positions. Ultimately a range of powers needs to be developed for each of the three
roles/positions, including the following.

The Agent must be able to:

 perform routine text transformations;
 obey basic imperatives in mathematical text.

The Subject must be able to:

 read mathematical texts and making sense of them as tasks, computations, derivations;
 access a repertoire of text transformations and apply them in completing tasks;
 write mathematical problems and tasks;
 judge whether mathematical texts follow the appropriate rules, i.e., read them critically.42

The Person must be able to:

 regulate the activities of the Subject (and Agent) through planning, monitoring, and
control;
 regulate the identity assumed and the subjectivity accessed, according to the social
context.

Although these are the powers of the different roles/positions making up the model reader in
mathematics, they do not form a fixed developmental sequence. Clearly what comes first in
the emergence of the empirical reader is the Person, although the development of its powers is

42
In some activities this capability might also or instead be part of the role of the Person.
an accomplishment achieved over several or many years, and for students its successful
completion is by no means guaranteed. In Ernest (2003) I discuss how the Agent may develop
before or at the same time as the Subject, in response to the types of mathematical texts/tasks
with which the learner engages, subject to and constituted by the specific contexts of
schooling and pedagogies in which it takes place. I also point out the existence of pathologies
in the empirical reader, with some students stuck with only an Agent and not a properly
formed Subject in mathematics, as evidenced by their ability to perform limited routine tasks
but not able to decide what processes and knowledge to apply to less routine or not recently
practiced tasks. This is a limited version of Skemp’s (1976) instrumental understanding.

Schoenfeld (1992) studied the patterns of work of novice and expert non-routine problem
solvers in mathematics and categorized it into seven stages: Read, Analyze, Explore, Plan,
Implement, Verify, Self-question. He found that many novices’ attempts comprised a minute
or two of reading the question, followed by an unbroken sequence of unreflective exploration
of the problem, seeking to solve without plan, typically through the generation of specific
example texts. This is commonly observed, and when persisted in usually leads to failure, due
to the lack of regulation and monitoring of outputs and their poor match with task goals.

In contrast, experienced mathematician problem solvers typically work through all of the first
6 levels or stages, cyclically, often more than once. Furthermore, they typically ask
themselves questions privately or out loud throughout the process, suggesting metacognitive
self-monitoring and self-regulation of their overall problem solving processes. Success in
mathematical problem solving is typically associated with such self-questioning behaviours.
This suggests that the regulatory powers of the Person as listed above are an essential part of
creative and non-routine problem solving in mathematics, if not all mathematical tasks.

The powers of these mathematicians resemble those intended for the model reader of
mathematical texts, as well as those at the powerful end of the spectrum of the actualized
capacities of empirical readers. In contrast, the capacities of the novices constitute those near
the other weaker end of the spectrum of powers of empirical readers, as in the pathological
cases mentioned above.

Although Eco’s concept of the model reader is a useful theorization to apply to the
interpersonal metafunction of mathematical text, there are two notable ways in which it falls
short. Mathematical identity, even just in theory, includes quite a lot more than the model
reader. Model readers of any genre of text engage in interpreting and actively constructing
meanings for texts (Corner 1983).43 However, in mathematics the model reader usually also

43
Hall (1980) has theorized three modes of reader response/sense making from engagement with text, based on
the assumption that texts encode a dominant ideology or reading. In brief, these modes are 1. dominant (or
'hegemonic') reading: the reader fully shares the text's code/ideology, which may seem 'natural' and 'transparent';
2. negotiated reading: the reader broadly accepts the preferred reading, but sometimes modifies it in a way
which reflects their own position, experiences and interests; 3. oppositional ('counter-hegemonic') reading: the
reader, who adopts an oppositional relation to the dominant code, understands the preferred reading but does not
share the text's code and rejects this reading, bringing to bear an alternative frame of reference. The extent to
which mathematical texts encode dominant ideologies and to which model readers can construct oppositional
readings needs further development. However note that in Part 1 (Ernest 2008a) I critique the dominant Platonist
ontology that constitutes the received ideology of mathematics, and in Ernest (1998) I offer a model of
mathematical development (the Generalised Logic of Mathematical Discovery) in which critical reactions to
published mathematical texts can lead to a global restructuring of the overall research practice, including
changed, methods, informal theories, proof paradigms, criteria and meta-mathematical views – many of the
elements that make up an ideological position. In Ernest (2005b) I distinguish between two text-related roles 1.,
the role of proponent (or friendly listener) presenting (or following sympathetically) a text, argument or thought
experiment (attempting to ‘share’ the constructor’s meaning, rather than looking for grounds on which to dismiss
writes and draws, constructing novel texts as well as interpretations of given texts (Rotman
1994). Thus in mathematics we need to include writing powers in theorizing the model reader/
(writer). Secondly, as discussed above, mathematical identity has an important affective
component. The mathematical model reader/writer does not just have
cognitive/epistemological capacities, but also emotions, feelings, attitudes, beliefs, and values
that play a central role both in the self-perceptions of subjective identity, and in facilitating or
blocking the functioning of the mathematical model reader/writer. Enlarged in this way, the
concepts of model reader/writer and empirical reader/writers have real potential for theorizing
differing mathematical identities and linking them to different level of mathematical
performance.

The Textual Function

Halliday’s third category, the textual function, is about what sort of text a mathematical
inscription/utterance is, how the text is created and structured, and how it uses signs and
textual components to fulfil its purposes. This function concerns issues such as what the
mathematical text is attempting to do, whether it is describing a process, communicating
relationships between signs, setting out to prove a claim, specifying a task, and so on. In
mathematical text, as in all subject domains, this function overlaps with the ideational and
interpersonal functions described above. In particular, the ideational, interpersonal and textual
components of mathematical texts form a mutually constitutive triad in the closed and self-
referential semiotic space of mathematics.

One of the special characteristics of mathematics is the range of signs and sign types utilized
in constructing mathematical text. Typically in printed or written texts these are of three types:
1. Alphanumeric signs used to make terms, words and sentences, encompassing
written language (e.g., French or English) and numeric and algebraic terms and
propositions;
2. Special mathematical symbols for relations, functions, objects, either in the form
of pictograms (e.g., , , ∞, ∑) or taken from a range of languages and scripts
(e.g., Gothic and Hebrew letters), and often employing added subscripts and
superscripts;
3. Diagrams and pictures, typically two dimensional drawings and line diagrams with
or without labelling. (Rotman 1994)

The simplest mathematical texts utilize only the elementary signs of type 1, which, in addition
to alphabetic letters (both lower and upper case) and punctuation signs also includes a small
selection of simple and familiar pictograms (=, +, -, ., /) as well as the numerals 0, 1, 2, 3, 4, 5,
6 ,7 ,8, 9. A very large range of compound signs can be made with this set of basic signs,
including all possible word problems and arithmetical tasks, and some elementary algebraic
tasks.44 Mostly, the texts written by students working word problems or routine mathematical
tasks will just employ a limited range of type 1 signs.

it). The role of proponent/friendly listener can be at two levels 1a, reflective or higher order, or 1b, at a lower,
passively attentive level (corresponding approximately to Hall’s first two roles – reversed. 2., the role of critic, in
which a text is examined for weaknesses and flaws. This can be at two levels 2a, local critique, and 2b, global
critique, as indicated in Generalised Logic of Mathematical Discovery (Ernest 1998). The latter partially
corresponds with Hall’s oppositional reading role, in the way that it offers an analogue of Kuhn’s (1970)
scientific revolution within mathematics.
44
For very young children some texts include pictures of, e.g., flowers or cars, to be counted, and multimodal
texts for them can also employ arrays of material objects, etc.
Texts that incorporate signs of type 2 are typically more complex and advanced mathematical
texts, and it would be unusual to find advanced mathematical texts that do not use a
significant number of such signs. Signs of type 3 (diagrams and pictures) typically occur
throughout school texts, but will only be part of some specialized mathematical task
presentations, e.g., in geometry.45 In advanced published (or equivalent) mathematical texts
signs of type 3 are usually sparingly used, except in, say, topology or geometry texts, and will
often be very abstracted and heavily labelled, such as in the arrow diagrams of category
theory.

In contrast, the texts written by students working non-routine mathematical tasks will often
employ signs of types 1 and 3, and possibly also signs of type 2 in advanced mathematical
studies. As indicated above, such texts may well utilize a broader range of sign types than
those employed in the problem or task specification or in the final text, the answer.

The two most important functions of teacher presented mathematical texts in the social
context of schooling are first, to present tasks for the reader to work, and second, to supply
expository text providing explanations (including definitions and terminology), or
demonstrations of procedures, calculations or deductive proofs. Putting to one side meta-
expository text, intended to help orientate, organise or instruct the reader, mathematical texts
will typically fall within well defined semiotic systems, so that the reader has a good idea of
the set of signs that is in play, and which it is appropriate to use, and the sets of rules and
methods to apply in completing mathematical tasks. The expository texts will be designed to
add to the reader’s meaning structure (i.e., their understanding) and their knowledge of the
rules and methods in play in the system (their competence in operating the semiotic system).
However, as we have seen in the previous section, such texts and tasks both elicit and
generate a complex array of reader/writer roles and positions, in both utilizing and
constructing mathematical identities.

Only advanced mathematical students or mathematicians will be expected to read extensive

mathematical texts with elaborate demonstrations of procedures or deductive proofs, as these
take specialized knowledge and reading skills (Mousley and Marks 1991). Although such
texts have epistemological significance, i.e., they are knowledge claims and validations, they
are fundamentally technical and non-narrative texts which are read to see if and how rules and
transformations are correctly applied to derive their final signs. 46 Great ingenuity and
imagination may have been deployed in their construction, and indeed considerable
knowledge and imagination may be needed in their reading, but nonetheless such
mathematical texts are technical inscriptions. Because of their dense and technical nature
successful readers of advanced mathematical texts typically imagine a universe of
mathematical objects that are described and acted on in the text. Although these correspond to
objects in the meaning structure of the associated semiotic system, they are typically
mythologised by mathematicians as entities in an ideal superhuman realm. Such realms are,
needless to say, imaginary, but nevertheless fulfil an important function in facilitating human
45
An exception is when diagrams, photographs or other pictures are attached to/included in mathematical tasks
to signify the task context or ‘real world’ situation which the task is supposed to model. Although a widespread
view is that such type 3 signs support the student’s meaning-making for the task, researchers have indicated that
that this role is far from proven and indeed can impact differentially on learners according to social class,
rendering some tasks less accessible (Boaler 1993, Cooper 1992, Dowling 1998).
46
Hanna (2000) has convincingly argued that mathematical proofs can have explanatory roles as well as their
demonstrative (epistemological) functions. Although mathematical proofs are fundamentally technical, in the
sense described, they nevertheless employ (and extend) a wide range of argumentative devices, including those
of deductive logic and rhetoric, used in all types of texts intended to convince or persuade. Such devices include
all of the traditional forms of argument, such as modus tollens and reductio ad absurdam, and add in further
modes such as mathematical induction.
understanding and meaning making (Burton 1999).47 Thus ironically, given my earlier
critique, Platonistic conceptions of mathematical objects may be an effective means of
comprehending mathematical signs via the meaning structures of semiotic systems, even if
they do not have the ontological significance they are often given (Rodd 1998). Thus in one
sense, the textual function of advanced mathematical texts is to tell tales about these fictional
mathematical objects.48

CONCLUSION

In this paper I have presented a tentative and evolving semiotic theory of mathematical text.
The aim has been to give an account of how mathematical text ‘works’, both in the contexts
of schooling (including university taught mathematics) and mathematical research. The theory
of semiotic systems provides a simple structural device for understanding the mechanics of
mathematical text, and its symbolic functions, although there is a pool of ineliminable
complexity in the underlying meaning structure of any semiotic system. A further area of
complexity concerns the relations between semiotic systems. Repeated minor changes to
semiotic systems results in a sequence of related, overlapping semiotic systems that are
treated as the same in some contexts but must be treated as different in others. Yet all learners,
to be successful in mathematics, must master such chains of semiotic systems, adding new
properties to say, semiotic systems concerning number, and ‘forgetting’ other properties, e.g.,
‘you cannot divide larger into smaller numbers’ (Ernest 2006).

This account also sketches elements of a theory of the mathematical subject, and its relations
with mathematical activity and mathematical identity. The simplicity of the theory of semiotic
systems, which looks at whole series of mathematical signs and derivations/transformations in
the performance of a mathematical task, is not supposed to diminish the difficulties and real
accomplishments of learners in being able to perform just a single transformation of text. Or
rather, in producing a single transformation of text in the context of a particular task, for to
‘be able’, along with capacities and competences, are unobservable theorizations inferred
(induced) from observed sequences of performances.

REFERENCES
Amit, M. & Fried, M. (2005) ‘Authority and Authority Relations in Mathematics Education: A View from an 8th
Grade Classroom’, Educational Studies in Mathematics, Vol. 58, No. 2 (February 2005): 145-168.
Arcavi, A. (2005) ‘Developing and Using Symbol Sense in Mathematics’, For the Learning of Mathematics, Vol.
25, No. 2 (July 2005): 42-47.
Arzarello, F. (2006) ‘Semiosis as a Multimodal Process’ in L. Radford and B. D’Amore, Eds., Revista
Latinoamericana de Investigación en Matemática Educativa, Número Especial Semiotics, Culture and
Mathematical Thinking, 2006: pp. 267-300.
Berkeley, G. (1710) The Principles of Human Knowledge. Reprinted in Fontana Library, Glasgow: W. Collins:
1962.
Berry, M. (1975) Introduction to Systemic Linguistics I, London: Batsford Books.
Black, L., Brown, M., Mendick, H., Rodd, M. & Solomon, Y. (2006) Mathematical Relationships: Identities and
Participation Project, consulted 1138 BST on 19 June 2008 at
<http://www.lancs.ac.uk/fass/events/mathematicalrelationships/index.htm>.
Boaler, J. (1993) ‘The Role of Contexts in the Mathematics Classroom: Do they Make Mathematics More
Real?’, For the Learning of Mathematics, Vol. 13, No. 2 (June 1993): 12-17.

47
Although imaginary, mathematical objects are no less real than any other cultural objects, and indeed might be
said to be more so, in view of their distinct and clear properties specified in the texts that generate them.
48
I find myself torn – on the one hand I strongly espouse nominalism in mathematical ontology, mathematical
objects are nothing but signs (Ernest 2004) – on the other hand I increasingly find myself a mathematical realist
– mathematical objects are real, cultural entities (Ernest 1998). Perhaps these traditionally opposed positions are
but two sides of the same coin.
Boaler, J. (2002) ‘The Development of Disciplinary Relationships: Knowledge, Practice and Identity in
Mathematics Classrooms’, For the Learning of Mathematics, Vol. 22, No. 1 (March 2002): 42-47.
Boaler, J., Wiliam, D. and Zevenbergen, R. (2000) The Construction of Identity in Secondary Mathematics
Education, consulted 1210 BST on 24 July 2006 at
<http://www.kcl.ac.uk/education/papers/construction.pdf>.
Brousseau, G. (1997) Theory of Didactical Situations in Mathematics: Didactique Des Mathematiques, 1970-
1990 (Translated and edited by, N. Balacheff, M. Cooper, R. Sutherland, V. Warfield), Dordrecht:
Kluwer Academic Publishers.
Brown, M. and Küchemann, D. E. (1976, 1977, 1981) Is It An Add Miss? Parts 1, 2 and 3, Mathematics In
School, Vol. 5, No. 5, 1976: 15-17; Vol. 8, No. 1, 1977: 9-10; Vol. 10, No. 1, 1981: 13-15.
Burton, L., (1999) ‘Why is Intuition so Important to Mathematicians but Missing from Mathematics
Education?’, For the Learning of Mathematics, Vol. 19, No. 3 (November 1999): 27-32.
Burton, L., Ed., (1988) Girls into Maths Can Go, London: Holt, Rinehart and Winston.
Buxton, L. (1981). Do you Panic about Maths? Coping with Maths Anxiety, London: Heinemann Educational
Books.
Charles, R. I. (1985) The role of problem solving, Arithmetic Teacher, February 1985: 48-50.
Charles, R. I. and Lester, F. (1982) Teaching Problem Solving, New York: Wiley.
Church, A. (1956) Introduction to Mathematical Logic, Vol. 1, Princeton, New Jersey: Princeton University
Press.
Cooper, B. (1992) 'Testing National Curriculum mathematics: some critical comments on the treatment of 'real'
contexts for mathematics', The Curriculum Journal, Vol. 3, No. 3: 231-243.
Corner, J. (1983): 'Textuality, Communication and Power', in H. Davis. & P. Walton (Eds.) Language, Image,
Media, Oxford: Basil Blackwell, 266-81.
Davies, B. And Harré, R. (n.d.) Positioning: The Discursive Production of Selves, consulted at 0825 BST on 14
June 2008 at <http://www.massey.ac.nz/~alock/position/position.htm>.
Derrida, J. (1976) Of Grammatology, (Trans. By Gayatri C. Spivak), Baltimore: The Johns Hopkins Press.
Diener, C. I. and Dweck, C. S. (1978) An analysis of learned helplessness: Continuous changes in performance,
strategy and achievement cognitions following failure. Journal of Personality and Social Psychology,
Vol. 36, 451-462.
Dowling P. (1998), The Sociology of Mathematics Education. London: Falmer.
Duval, R. (1995) Sémiosis et Pensée Humaine, Bern: Peter Lang.
Eco, U. (1984) The Role of the Reader, Bloomington: Indiana University Press.
Ernest, P. (1991) The Philosophy of Mathematics Education, London: Falmer Press
Ernest, P. (1993) Mathematical Activity and Rhetoric: Towards a Social Constructivist Account, in I.
Hirabayashi, N. Nohda, K. Shigematsu, F. L. Lin, Eds, Proceedings of PME-17 Conference, Tsukuba,
Japan: University of Tsukuba, Vol. 2, 1993: 238-245.
Ernest, P. (1995) ‘Values, gender and images of mathematics: a philosophical perspective’. International Journal
of Mathematics Education, Science and Technology, Vol. 26, No. 3 (1995): 449-462.
Ernest, P. (1998) Social Constructivism as a Philosophy of Mathematics, Albany, New York: SUNY Press.
Ernest, P. (2003) 'The Epistemic Subject in Mathematical Activity’, in M. Anderson, A. Saenz-Ludlow, S.
Zellweger, & and V. V. Cifarelli, Eds., Educational Perspectives on Mathematics as Semiosis: From
Thinking to Interpreting to Knowing, New York, Ottawa and Toronto: Legas Publishing, 81-106.
Ernest, P. (2004) ‘Nominalism and Conventionalism in Social Constructivism’, Philosophica No. 74: 3-29.
(Reprinted in The Philosophy of Mathematics Education Journal 19,
<http://www.people.ex.ac.uk/PErnest/pome19/index.htm>).
Ernest, P. (2005) 'Agency and Creativity in the Semiotics of Learning Mathematics’, Hoffmann, M., Lenhard, J.
and Seeger, F., Eds, Activity and Sign – Grounding Mathematics Education, New York, Springer, 2005:
23-34.
Ernest, P. (2005a) Gender and Mathematics: The Nature of Mathematics and Equal Opportunities (Doctoral
module coursebook), Exeter: School of Education and Lifelong Learning, University of Exeter.
Ernest, P. (2005b) 'Agency and Creativity in the Semiotics of Learning Mathematics’, M. Hoffmann, J. Lenhard,
& F. Seeger, Eds, Activity and Sign – Grounding Mathematics Education, New York, Springer, 2005:
23-34.
Ernest, P. (2006) 'A Semiotic Perspective of Mathematical Activity: The Case of Number ', Educational Studies
in Mathematics, Vol. 61, 2006: 67-101.
Ernest, P. (2008a) ‘Towards a Semiotics of Mathematical Text (Part 1)’, For the Learning of Mathematics, Vol.
27, No. 2, 2008: 2-8.
Ernest, P. (2008b) 'Towards a Semiotics of Mathematical Text (Part 2), For the Learning of Mathematics, Vol.
28, No. 2 (June 2008): 39-47.
Ernest, P. (2008c) ‘Towards a Semiotics of Mathematical Text, Part 3’, For the Learning of Mathematics, Vol.
28, No. 3, 2008: 42-49.
Evans, J. (2000) Mathematical Thinking and Emotions in Context: Adults, Practices and Numeracy, London:
Falmer Press.
Evans, J., Morgan, C., & Tsatsaroni, A. (2006) Discursive Positioning and Emotion in School Mathematics
Practices. Educational Studies in Mathematics: Affect in Mathematics Education: Exploring
Theoretical Frameworks, A PME Special Issue, 63, 2.
Fennema, E. and Leder, G., Eds, (1990) Mathematics and Gender, New York: Teachers College Press.
Gergen, K. (1999) An Invitation to Social Construction, London: Sage.
Grootenboer, P., Smith, T. & Lowrie, T. (2006) ‘Researching Identity in Mathematics Education: The Lay of the
Land’, consulted at 1445 BST 18 June 2008 at <http://www.merga.net.au/documents/symp12006.pdf>.
Hall, S. (1980) 'Encoding/decoding'. In Centre for Contemporary Cultural Studies (Ed.): Culture, Media,
Language: Working Papers in Cultural Studies, 1972-79, London: Hutchinson, 1980: 128-38.
Halliday, M. A. K. (1975) Learning How to Mean, London: Edward Arnold.
Halliday, M. A. K. (1985) An Introduction to Functional Grammar, London: Edward Arnold.
Halliday, M. A. K. and Martin, J. R. (1993) Writing Science: Literacy and Discursive Power, London: Falmer
Press.
Halmos, P. R. (1974) Naïve Set Theory, New York: Springer Verlag.
Hanna, G. (2000) ‘Proof, explanation and exploration: An overview’, Educational Studies in Mathematics, Vol.
44, Nos. 1-2, 2001: 5-23.
Harré, R. (1979) Social Being, Oxford: Basil Blackwell.
Harré, R. and Gillett, G. (1994) The Discursive Mind, London: Sage Publications.
Hayfa, N. (2006) ‘Impact of Language on Conceptualization of the Vector’, For the Learning of Mathematics,
Vol. 26, No. 2 (July 2006): 36-40.
Henkin, L. (1949) The Completeness of the First-Order Functional Calculus, Journal of Symbolic Logic, Vol. 14,
159-166. Reprinted in J. Hintikka, Ed. (1969) The Philosophy of Mathematics, Oxford: Oxford
University Press, 42-50..
Henriques, J. Holloway, W. Urwin, C. Venn, C. & Walkerdine, V. (1984). Changing the Subject: Psychology,
Social Regulation and Subjectivity. London: Methuen.
Hilbert, D. (1925) On the Infinite, Mathematische Annalen, Volume 95, 1925: 161-90. Translation in Benacerraf,
P. and Putnam, H. Eds, Philosophy of Mathematics: Selected readings, Englewood Cliffs, New Jersey:
Prentice-Hall, 1964: 134-51.
Kuhn, T. S. (1970) The Structure of Scientific Revolutions, (2nd edition), Chicago: Chicago University Press.
Lakatos, I. (1978) ‘A Renaissance of Empiricism in the Recent Philosophy of Mathematics?’ in I. Lakatos, The
Methodology of Scientific Research Programmes (Philosophical Papers Volume 2), Cambridge:
Cambridge University Press, 1978: 24-42.
Lloyd, D. I., Ed. (1979) Philosophy and the Teacher, London: Routledge and Kegan Paul.
Machover, M. (1983) ‘Towards a New Philosophy of Mathematics’, British Journal for the Philosophy of
Science, Vol. 34, 1983: 1-11.
Mason, J., Drury, H. & Bills, E. (2007). Explorations in the Zone of Proximal Awareness’, in J. Watson & K.
Beswick, Eds., Mathematics: Essential Research, Essential Practice: Proceedings of the 30th annual
conference of the Mathematics Education Research Group of Australasia. Adelaide: MERGA Vol. 1
(2007): 42-58.
Mazur, B. (2004) ‘On the Absence of Time in Mathematics’, For the Learning of Mathematics, Vol. 24, No. 3
(November 2004): 18-20.
Mellin-Olsen, S. (1981) ‘Instrumentalism as an Educational Concept’, Educational Studies in Mathematics, Vol.
12, 1981: 351-367.
Morgan, C. (1996) ‘“The Language of Mathematics”: Towards a Critical Analysis of Mathematical Texts’, For
the Learning of Mathematics, Vol. 16, No. 3 (November 1996): 2-10.
Morgan, C. (1998) Writing Mathematically: The Discourse of Investigation, London: The Falmer Press.
Morris, C. (1945) Foundations of the Theory of Signs, Chicago: University of Chicago Press.
Mousley, J. and Marks, G. (1991) Discourses in Mathematics, Geelong, Australia: Deakin University Press.
Neumann, J. von (1931) The Formalist Foundations of Mathematics, Erkenntnis, 1931, 91-121. Translation in
Benacerraf and Putnam (1964): 50-54.
Peirce, C. S. (1931-58) Collected Papers (8 volumes), Cambridge, Massachusetts: Harvard University Press.
CD-ROM version published in Past Masters series, Charlottesville Virginia: InterLex Corporation.
(Citations refer to paragraph number).
Radford, G. P. (n. d.) Beware of the Fallout: Umberto Eco and the Making of the Model Reader, Consulted at
1243 BST on 15 June 2008 at <http://www.themodernword.com/eco/eco_papers_radford.html>.
Radford, L. (2002) ‘The Seen, the Spoken and the Written: a Semiotic Approach to the Problem of
Objectification of Mathematical Knowledge’, For the Learning of Mathematics, Vol. 22, No. 2 (July
2002): 14-23.
Restivo, S. (1992) Mathematics in Society and History, Dordrecht: Kluwer.
Rotman, B. (1988) ‘Towards a Semiotics of Mathematics’, Semiotica, Vol. 72, No. 1/2, 1-35.
Rotman, B. (1993) Ad Infinitum The Ghost in Turing's Machine: Taking God Out of Mathematics and Putting
the Body Back in, Stanford California: Stanford University Press.
Rotman, B. (1994) ‘Mathematical Writing, Thinking and Virtual Reality’, in P. Ernest, Ed., Mathematics,
Education and Philosophy: An International Perspective, London, ‘ Falmer Press, 1994: 76-86.
Rotman, B. (2003) ‘Will the Digital Computer Transform Classical Mathematics?’, Philosophical Transactions
of the Royal Society, London, Series A, Vol. 361 (2003): 1675–1690.
Schoenfeld, A. H. (1992) ‘Learning to Think Mathematically: Problem solving, Metacognition, and Sense
Making in Mathematics’, in D. A. Grouws, Ed., Handbook of Research on Mathematics Teaching and
Learning, New York, Macmillan, (1992): 334-370.
Sfard, A. (1987) ‘Two Conceptions of Mathematical Notions: Operational and Structural’, in J. C. Bergeron, N.
Herscovics and C. Kieran, Eds, Proceedings of PME 11 Conference, Montreal: University of Montreal,
Vol. 3, 1987: 162-169.
Sfard, A. (1994) ‘Reification as the Birth of Metaphor’, For the Learning of Mathematics Vol. 14, No. 1
(February 1994), 44-55.
Shotter, J. (1993). Conversational Realities: Constructing Life through Language. London: Sage.
Shuard, H. and Rothery, A., Eds, (1984) Children Reading Mathematics, London: John Murray.
Skemp, R. R. (1976) ‘Relational Understanding and Instrumental Understanding’, Mathematics Teaching, No.
77, 1976: 20-26.
UNESCO (1974) Interactions Between Linguistics and Mathematics Education, UNESCO: Paris.
Vygotsky, L. S. (1978) Mind in Society: The development of the higher psychological processes, (Edited by M.
Cole et al.), Cambridge, Massachusetts: Harvard University Press.
Walkerdine, V. (1998) Counting Girls Out (second edition), London: Falmer Press.

Acknowledgement
A version of this paper was published in three parts in For the Learning of Mathematics as Ernest (2008a, 2008b
& 2008c)