Peregrin - An Inferentialist Approach To Semantics

Philosophy Compass 3/6 (2008): 1208–1223, 10.1111/j.1747-9991.2008.00179.
An Inferentialist Approach to Semantics: Time

for a New Kind of Structuralism?
Jaroslav Peregrin*
Academy of Sciences of the Czech Republic
Abstract
The perennial question – What is meaning? – receives many answers. In this paper
I present and discuss inferentialism – a recent approach to semantics based on the
thesis that to have (such and such) a meaning is to be governed by (such and such) a
cluster of inferential rules. I point out that this thesis presupposes that looking
for meaning requires seeing language as a social institution (rather than, say, a
psychological reality). I also indicate that this approach may be seen as a new
embodiment of the old ideas of structuralism.
What is Meaning?
One of the most basic questions anybody who deals with language must
– sooner or later – face is the question: What is meaning? i.e., what is it
that makes some types of sounds or inscriptions meaningful? (It is therefore
somewhat surprising that a name for the ‘science of meaning’, semantics,
did not materialize until the late nineteenth century.)1 Moreover, even in
the twentieth century, investigation into the nature of meaning was held
as a more pressing task by philosophers than by linguists (particularly, of
course, following the linguistic turn which took place in the first half of
the twentieth century,2 and which stimulated the idea that to understand
meaning might be the task of philosophy).
Hence, what is meaning? It is helpful first to distinguish between various
senses this question may have. Let us start with the most obvious: taking
it to ask about the substance of which a meaning is made. What kind of
stuff makes up the chunk that must be glued to a type of sound or
inscription to make it meaningful?
The most exposed cases of such an alleged gluing are evidently the
events of baptism, through which a (proper) name becomes associated
with a tangible object (typically a human infant), and thus we may be
tempted to think that meanings are generally tangible things, elements of our
physical world. This answer, however, soon falls into disrepute; for the
physical world can not provide enough suitable entities to furnish all our
© 2008 The Author
Journal Compilation © 2008 Blackwell Publishing Ltd
An Inferentialist Approach to Semantics 1209
linguistic expressions with meanings. (Problems arise already with common

nouns, since the meaning of a general word like table cannot be any one
tangible thing, any one concrete table; and the situation is exacerbated
when we move to verbs, not to speak of prepositions and the like.)
Nevertheless, though meanings cannot be generally identified with tangible
things, it may still seem that naming these things, or referring to them, is a
central function of language in the first place; and indeed many philosophers
have used this to argue that at least some expressions of our language
(besides proper names also natural kind terms) cannot but mean tangible
things;3 or alternatively it has been used to argue that we should build
semantics around the relation of reference or designation, by-passing the
concept of meaning altogether.4
However, admitting that naming/referring/designating is a semantically
important enterprise does not alter the fact that if what we want to
capture is the concept of meaning (an entity that makes the difference
between a mere shriek and a word), then the physical world does not offer
us sufficient types of entities. This may prompt us to turn to another
world for assistance, one that would be more generous in this respect. And
here the obvious candidate seems to be the world of the mental. It contains
inexhaustible riches of entities, and, moreover, of entities apparently
qualified to be seen as meanings; for surely it is the human mind that
effects furnishing expressions with any semantics they may have. Many
theoreticians of language do embrace this answer;5 but there are others
who argue vigorously that meaning cannot be mental.6 The trouble with
mental entities is that they lack an important property that meaning
requires, namely intersubjectivity. The very point of meaning seems to lie
in its ability to be shared by many;7 whereas mental contents are inevitably
subjective, locked in one’s own mind.8
If we want to say that the sentence London is huge expresses the thought
that London is huge, it can be neither my thought (an episode within my
brain), nor your thought, nor indeed the thought of any other one individual,
it must be a type of thought which can be instantiated within many minds.
Hence the thought must exist somehow outside of the minds, possibly
with the capacity of being included into them, but so that such an
inclusion would not compromise its intersubjective existence and its
readiness to be included into other minds.
More generally, if the physical world is suitable in respect to its
intersubjectivity but insufficient as to its richness, and if the mental world
is, vice versa, suitable in respect to its richness but insufficient as to its
subjectivity, we would need a world that combines the positive qualities
of these other two worlds. And many philosophers have concluded
that such a world is the world of ideal entities, so colorfully described
at the dawn of philosophy by Plato and, during the last century, largely
re-housed under the legislation of set theory. Hence, many semanticists
inferred that meaning must be an ideal entity of the kind of, say, a
© 2008 The Author Philosophy Compass 3/6 (2008): 1208–1223, 10.1111/j.1747-9991.2008.00179.x
1210 An Inferentialist Approach to Semantics
number. Principal promulgators of this tradition were Frege, Carnap and

Montague.
Semantic Structure and Its Explication

But notably in the last quarter of the twentieth century, a different
answer to the question what kind of stuff meanings are made of started
to flourish: no stuff at all, for the talk of meaning is metaphorical, it is a
mere façon de parler. What we mean when we say that a word means
something is not that it has an entity glued to it, but rather that it has a
property, for example that it is employed by members of a society in a
certain way.
This answer has been endorsed by various kinds of theoreticians: by the
later Wittgenstein and his followers; by ordinary language philosophers
following Austin and Grice; and by those who, like Quine, revived American
pragmatism. I think that an apt name for the onslaught of this heteroge-
neous movement is the pragmatic turn.9
However, although these philosophers and linguists agreed that the
question what is meaning? in its substantial sense was unanswerable, this did
not mean they thought nothing could be asked about meaning – or at
least about semantics. An alternative to the substantial construal of the
question is one that I will call structural – it interprets the question as
asking not what meanings are made of, but rather how meanings (whatever
these may be) of grammatically related expressions are related to one another.
An example of a crucial statement belonging to the semantic enterprise
corresponding to this interpretation is the celebrated principle of composi-
tionality:10 ‘The meaning of a complex expression is uniquely determined
by the meanings of its components and the mode of their combination’.
But wait; is this supposed to be a construal of the original question? Is
it not, at most, only its part? Surely, to inquire what is meaning, is not to
call for a response citing how meanings relate one to another? It seems
that we must first find out what meanings are, in the substantial sense,
before being able to inquire about their properties and relations, e.g.,
whether they do compose.
However, there already exists a strong tradition which claims, in effect,
that this is not the case, and, indeed, that the structural reading of the
question is the only viable reading. Classical Saussurean structuralism was
the prototype for this stance; but as I have argued elsewhere (see my
Meaning and Structure), certain semanticists quite unrelated to the
structuralist tradition, such as W. V. O. Quine or W. Sellars, can also be
read as endorsing views which are structuralist in the broad sense entertained
here. And the idea I will pursue for the remainder of this article is that
insofar as there is something as meaning, it is a purely structural matter. But
I will zoom in on a particular place on the structuralist landscape; a place
where semantically relevant structure amounts to inferential structure.
Before I do so, let us illuminate the structuralist outlook by an analogy

with numbers. Over the last one and a half centuries, many philosophers
have wrestled with the question what is a number? (usually because they
take this to be a reduction of the question what is mathematics about?).
Many answers have been proposed. Thus, Edmund Husserl ventured that
numbers are general ideas, devoid of any qualitative properties and
reduced to pure quantity. Gottlob Frege and Bertrand Russell proposed,
in effect, seeing numbers as classes of equinumerous classes of things. John
von Neumann and Ernest Zermelo both identified the number zero with
the empty set, but von Neumann went on to identify any other number
with the set of its predecessors whereas Zermelo identified it with the
one-element set constituted by merely its predecessor. But as time passed,
it became increasingly clear that none of the answers really tell us what
natural numbers are, but only how to explicate them, in the sense of
explication introduced by Carnap (and elaborated by Quine).
This is worthy of attention. In his Foundations of Arithmetic, Frege,
before dealing with the question what is a number?, considers a simpler
question, namely what is a direction of a line? His conclusion was that
though there is no substantial answer to this question, what we may do is
identify the direction of the line a with the set of all lines parallel with a.
Saying that the direction of a is a set brings about some problematic
consequences (for example the direction then ‘contains’ lines), but, Frege
insists, if we are aware of this fact, in other contexts this identification
would be not only unproblematic, but helpful. And analogously, Frege
claims that, with this proviso, we can identify the number of objects falling
under a concept C with the set of all concepts equinumerous with this C – i.e.,
such that the objects falling under them can be mapped, in the one-to-one
manner, on those falling under C.
Carnap then generalized this method of attacking concepts and
introduced the term explication.
The task of explication consists in transforming a given more or less inexact

concept into an exact one or, rather, in replacing the first by the second. We
call the given concept (or the term used for it) the explicandum, and the exact
concept proposed to take the place of the first (or the term proposed for it)
the explicatum. The explicandum may belong to everyday language or to a
previous stage in the development of scientific language. The explicatum
must be given by explicit rules for its use, for example, by a definition which
incorporates it into a well-constructed system of scientific either logicomath-
ematical or empirical concepts . . . A problem of explication is characteristically
different from ordinary scientific (logical or empirical) problems, where both
the datum and the solution are, under favorable conditions, formulated in exact
terms (for example ‘What is the product of 3 and 5?’, ‘What happens when
an electric current goes through water?’). In a problem of explication the
datum, viz., the explicandum, is not given in exact terms; if it were, no
explication would be necessary. Since the datum is inexact, the problem itself
is not stated in exact terms; and yet we are asked to give an exact solution.
(Logical Foundations of Probability 3–4)
Applied to meaning, it gives us the approach articulated by David Lewis:

‘In order to say what a meaning is, we may first ask what a meaning does
and then find something which does that’ (173).
Above we have considered four kinds of answers to the question What is
meaning? construed in the ‘substantial’ way: (1) meanings are physical
entities; (2) meanings are entities of the mental world; (3) meanings are
ideal entities of a Platonic realm; and (4) there are no such entities as
meanings, talk about them is a metaphor. Now it is important to realize
that structuralism, broadly construed, can be seen as embracing answer (4),
as its message can be interpreted as there are no meanings, there is only
semantic structure. But the concept of explication alleviates the tension
between this answer and answer (3); the fact that there are, strictly speaking,
no such objects as meanings, does not prevent us from explicating meanings
as objects. And in many contexts it is not unreasonable to simply identify
a meaning with its explication.
However, it is important to keep in mind the basic difference between
answer (3) and the combination of answer (4) with explication, both of
which may foster a picture of language as a system of expressions mapped
onto a system of objects. If we embrace answer (3), then we will see the
interconnection between an expression and the corresponding object as a
result of activities of language-users and consequently we will want to ask
how the interconnection was established and how it is sustained (and here
we are likely to invoke concepts like naming, representing, etc.).11 On the
other hand, if we accept that the object is merely the result of explication,
then we will see the interconnection as fall-out from the theoretician
doing the explication, and will recognize that the pursuit of its inauguration
by language users is misguided.
explicatum vs. explicandum

Consider what may happen if we forget that the object we are referring to
as meaning is in fact its mere explication. Returning to the Fregean
example, imagine, for instance, how somebody who might have forgotten
that the set of all parallels of a is not the direction itself, but its explication,
might then wrestle with the question of how the direction of a line could
possibly have come to contain other lines. We may analogously imagine
a semanticist who forgets that an object is merely an explication of meaning
and subsequently feels the need to research how the speakers manage to
interconnect that object with the corresponding expression.
To flesh out this idea, let us consider the thesis, taken by many
semanticists for granted, that the English connective and denotes the
truth-function characterized in the well-known truth table
A B A∧B
T T T
T F F
F T F
F F F
(Disregard for now that the functioning of the English and has additional
complexities, that it, for example, often expresses temporal succession.
The point would remain the same even if and were taken to denote some
more complicated function, such as that proposed within various versions
of dynamic semantics.12)
The trouble is that this assumption could delude us into thinking that
and is the name of such a function in an analogous sense to which the
name Julius Caesar is the name of the historical person. And from this it
takes only a small step to wondering whether there is an act of baptizing
of the above truth-function by and, analogous to the act by which Julius
Caesar was baptized. (Not, of course, that anyone would be suggesting an
act quite analogous to Julius’s christening; but the usual wisdom is that
the interconnection of the word and the truth table is a matter of convention,
which does indicate some act of deliberate decision.)
The same may be the case if we move from merely logical words, like
and, to other parts of the vocabulary. Frege proposed explicating the
meaning of a predicative expression like (to be a) dog as a function mapping
individuals onto truth values, dogs on the truth and every other individual
on the falsity; a function that may obviously be identified with a set of
individuals, in the case of (to be a) dog with the set of all dogs. Carnap
urged that this would not yield us a feasible explication of meanings and
added possible worlds: the explication of the predicative expression became
a function mapping possible worlds onto their respective sets of dogs. And
then came others who have tried to further improve on this proposal.
In any case, Frege’s proposal clearly reflects the fact that a predicate like
(to be a) dog forms true sentences with some names (namely, names of
dogs), and false with others. Carnap’s improvement is then exposed as
reflecting the fact that what is and what is not a dog depends on the state
of the world; and further improvements may then reflect further semantic
aspects of predicates. Whatever the resulting function may be, it is not to
be seen as something that came to be named by the predicate, but rather
as something that tries to capture the functioning of the predicate.
Hence, asking how an expression has come to name the entity like the
truth function or the Carnapian intension is precisely the kind of misguided
question that might be engendered by not observing the distinction
between the explicatum, the explicating object, and the explicandum, the
explicated phenomenon. A function is a thing, and what comes naturally
when we consider the establishment of a relationship between an expression
and a thing is some relation of the kind of naming. This may lead to the
idea that the whole language is simply a huge system of interconnected
names. And indeed this view has informed the notion of language held by
many theoreticians of language (originally more philosophers and logicians
than linguists, but recently probably mostly linguists influenced by logic).
Meanings, Rules, and Inference

What alternative did the pragmatic turn offer? Ludwig Wittgenstein,
whose early Tractatus significantly contributed to the notion of language
as a great system of names, stimulated the turn when he later concluded
that we should see the meaning of an expression as the way in which the
expression is used by speakers, i.e., as its role within our language games:
‘For a large class of cases – though not for all – in which we employ the
word ‘meaning’ it can be defined thus: the meaning of a word is its use
in the language’ (Philosophische Untersuchungen §43). He also pointed out
that what is crucial to the constitution of our language games are (various
kinds of) rules. In this light, we can compare language with chess: the
‘meaning’ of the wooden pieces we use to play the game, their being pawns,
rooks, bishops, etc., is also determined by the rules of chess. This indicates
that meaning can be conceived of as a role conferred on an expression by
the rules of our language games. But what kind of rules are these?
Let us return to the connective and. What is it that we must grasp to
understand its meaning? The most straightforward answer may seem to be
that what we must grasp is that a complex sentence, arising from connecting
two sentences by its means, is true only when both the subsentences are
(which is what appears to be reflected by the truth table). But what has
truth to do with our language games? If what we are after is a role with
respect to the rules of these games, then this characterization is helpful
only in so far as it can be read as referring to the rules.
But it seems that truth may be seen as a sort of a correct assertability. This
should not be controversial: it seems that there is a sense of correctness in
which an assertion of a sentence is correct iff it is true (needless to say,
there are also other senses of the correctness of assertions – an assertion
of a false statement may, for example, be correct in the sense that it saves
somebody’s life).13 If we admit this, then we can say that and is characterized
by the rule that
A and B is correctly assertible if both A and B are,
i.e., by the pair of rules
if A and B is correctly assertible, then both A and B are
if both A and B are correctly assertible, then A and B is
If we now write, as usual,
A1, . . . , An ⵫ A
in the sense of A is correctly assertible whenever A1, . . . , An are, we may
further rewrite this as
A and B ⵫ A
A and B ⵫ B
A, B ⵫ A and B
Now the truth table above can be seen as summarizing these rules: the
first row says that if A and B are true, i.e., correctly assertible, then also
A and B is; whereas the other three say that if either A or B is false (not
correctly assertible), then also A and B is (not correctly assertible); in other
words that if A and B is correctly assertible, then both A and B are. Hence
the claim may be that seeing the connective as a name of the truth
function is misguided (though in many contexts it does not raise any
problems), for in fact the truth function is merely the explication of the
expression of the inferential role.
Considerations of this kind are well known from the philosophy of
logic, where we have been witnessing, for several decades, discussion
between those who are convinced that the semantics of logical constants
is essentially inferential (and is to be studied by proof theory) and those for
whom the constants must be seen as standing for something (and hence
must be accounted for by model theory).14 This, however, is not what
interests us now; our interest is whether the inferential paradigm can be
extended outside the boundaries of logical constants.
Perhaps And; But What About Dog?

It may seem that the proposal to construe the meaning of even empirical
expressions in inferential terms is preposterous. What may work for and
would hardly work for dog – it would seem imperative that empirical
vocabulary, to become meaningful, must represent something. Whereas
with logical words there may be an issue over the relative merits of
grasping their semantics in inferentialist or in representationalist terms, for
empirical words there seems to be only the latter option.
However, is this truly so? Consider a person looking into the sky and
saying ‘The sky is blue’ and a parrot repeating the same sounds. What
makes the difference between the former act, which is an assertion, and
the latter one, which is a mere emitting of sounds? One answer might be
that what makes something an assertion, rather than just the sounds, is the
fact that it is a move in a certain game, namely a language game. Just like
what makes kicking a ball through the goal posts scoring a goal is the
fact that it was an act within the space constituted by the rules of a
football game.
However, the obvious objection is that what makes the difference

between the asserter and the parrot is that the former is thinking and hence
can associate his sounds with a certain thought – and thus gives them their
meanings. But even if we waive the doubts of the possibility of the
mentalist construal of meanings voiced at the beginning of this paper –
what does it mean to think? As Alan Turing observed, there is really no
way to find out whether somebody is thinking other than to check
whether she behaves in a certain way (‘reasonably’), i.e., whether she talks
and behaves so that it ‘makes sense’. So, though the claim that what
differentiates between a human speaker and a parrot are the former’s
thoughts, is surely true, it is problematic to use it as an answer to our
question (what makes one’s emitting of sounds into an assertion?), for we may
need to proceed with the explanation the other way around.
So this is why we may prefer the answer that to talk meaningfully is to
take part within certain language games. What kind of language games?
It is clear that not any would do. Shouting ‘Go, go!’ at an ice hockey
match, though perhaps a kind of a language game, clearly would not be
acceptable as a hallmark of thinking. Also, reciting poems would be easily
imitable by a non-thinking device, such as a tape-recorder. It seems that
if we want to know whether we face a thinking being, we should check
whether it is capable of reasoning. Hence we would probably ask questions,
and along with receiving the answers, we would check whether the adept
of thinking is able to give reasons for what she says. And we would try
to challenge some of her claims to find out whether she is capable of
defending them. In the course of this, we would probably expect her to
challenge our claims (our challenges to her claims) and to require us to
give reasons.
In short, it seems that a particularly suitable language game for the role
of a touchstone of a thinker would be the game Brandom calls giving and
asking for reasons. It is this game that seems to ‘bring thinking into the
open’. For this reason, Brandom considers this game the very basis of our
language; and thus he moves inferential rules, i.e., rules that lead us from a
reason to what it is a reason for, to the centerstage of our on-going
language game jamboree.
This indicates that even contentfulness of empirical words must be
underpinned by certain inferences. No empirical word is meaningful in
the distinctively human way (i.e., expresses a concept) unless it is a potential
token in the game of giving and asking for reasons. A word does not
express the concept of dog unless it can be used as part of sentences which
can in turn be used for reasoning, i.e., from which other sentences can
be inferred and which can be itself inferred from other sentences. The
English word dog would not express our concept of dog if it could not be
used to reason from This is a dog to This is not a cat etc.
Hence I have argued that there is no meaning without inference. But
it may still seem that there is, at least for empirical words, also no meaning
without a representation. For how could a word like dog come to express
the concept of dog without, at least inter alia, representing dogs? And
though there is undoubtedly some truth in this, the inferentialist answer
is that the concept of representing leads to a very odd way of capturing
what is going on between our empirical vocabulary and the world.
As it is only sentences that may be used to make a move in a language
game, any contact between a word and (a part of ) the world must be
mediated by sentences. Beside this, what matters is not what the speakers
really do with the sentences, but what they take to be correct to do – the
relation is normative. Thus, the link between the word dog and the world
is a matter of such facts as that it is correct to use the sentence This is a
dog in certain situations, and incorrect in others.
True, the usage of This is a dog may be ‘non-inferential’ in the sense
that its correctness is a matter of directly the extralinguistic circumstances,
and hence what is in question is not an inference in the standard sense
(from language to language), but an ‘inference’, as it were, from the world
to language. (Similarly, at the other ‘end’ of language, there are ‘inferences’
from language to action.) This means that if we want to extend the
inferentialist treatment of meaning from expressions like and to expressions
like dog, we have to generalize the concept of inference.
But talking about ‘generalized inferences’ may not be the best way of
seeing the situation. Imagine chess. The move I make responds exclusively
to the moves made by my opponents. It cannot respond to anything else,
for there is, in fact, nothing else to respond to. The pieces, board and
other equipment, strictly speaking, are not necessary – it is clear that we
can play chess completely without them. Thus, the rules of chess spell
out a pure, disembodied structure. However, as Lance pointed out,
language is more similar to a sport like football than to a game like chess.
Notice that football is less ‘disembodied’ than chess, in that its rules must
take into account the physical properties of the ball or the goalposts.
Similarly, the rules of language must reflect the fact that our language
games are not games in the sense of being self-contained; they are an
important way for us to interact with the world. Thus most of our language
games involve the world, and hence also the rules reflect the involvement.
As a result, even if you construe semantics in the inferentialist way, we
must keep in mind that the inferential rules governing it and conferring
meanings on expressions will involve the world. (Brandom stresses that
our linguistic practices cannot be seen as ‘hollow, waiting to be filled up
by things’, but rather as ‘as concrete as the practice of driving nails with
a hammer’ (332).) Hence, to understand dog, we must know not only how
the sentences containing dog (This is a dog, Every dog is a mammal and
others) can be correctly played within the game of giving of asking for
reasons in response to utterances of other players (that This bird is a dog
counts as a challenge to Every dog is a mammal, which than can be
defended by But this bird is not a dog), but, more broadly, how they are
correctly used also vis-à-vis non-linguistic circumstances (that This bird is not
a dog is correctly played only when what one is pointing at is a bird etc.).
Where Does Semantic Structure Come from?

Talking about the structuralist attitude to semantics I stated that, according
to this view, there is no meaning over and above an explication of semantic
structure. But, we may now ask, where does semantic structure come from?
Especially with Chomsky’s publishing of his path-breaking Syntactic Structures
and his subsequent increasing conviction that the structures are imple-
mented by nature within the language faculty of our mind/brain, many
theoreticians of language simply took for granted that a semantic structure,
or a ‘logical form’, must be another kind of such implementation.
However, though it is undoubtedly important and fruitful to investigate
the structures of the human mind/brain that underlie human linguistic
competence, we have come to the conclusion that it may be useful to see
language as a social institution. We concluded that from the viewpoint of
semantics it is essential to disclose the nature of our language games, and
especially the game of giving and asking for reasons, as a basically inter-
subjective enterprise. It is clear that though the rules of football exist only
insofar as there are thinking humans who engage in this game (and in so far
as their brains are wired up in such a way that they are able to do so), studying
the rules is far removed from studying the minds/brains of the players;
and studying the rules of language and the semantic structures they institute
may be equally far removed from studying our language faculty.
De Saussure claimed that every important linguistic structure is based
on binary oppositions; but where do they come from? An answer may be
that they are simply the similarity standards that natural selection has
implemented into our brains and that make us find both some pairs of
sounds (or inscriptions) and some pairs of worldly things, in contrast to
others, alike. Then all that is needed to create a language is to connect
the types of sounds/inscriptions with the types of things. But obviously
this would lead to a mere caricature of language, lacking the sophisticated
structure that any real language undoubtedly has.
Many theoreticians of language would surely say that it is not the
oppositions, but rather the whole sophisticated structure that is inborn.
But this answer is less satisfactory than it may seem at first sight; unless it
can be supplemented by a story of how the structure would wander from
the brains into the open, where language is at home. On the one hand,
it is quite clear that the fact that we use the complex kind of language
we do must be largely supported by the wiring of our brains, achieved by
natural selection. But on the other hand, this is not to say that the
structure of the public language is a mere copy of a structure present
within our brains. It is clear that, for example, the structure of a football
game (the fact that it consists of two halves, that it is played by two teams
of eleven players etc.) has clearly something to do with human brains, but
nobody would pursue the idea that it is a mere imprint of a structure of
the brain we are born with. The question is why the case of language
should be any different.
Moreover, we must distinguish between syntax (which has to do with
the individuation of the potential vehicles of meaning) and semantics
(which is a matter of different vehicles carrying the same or different
meanings). And even if the former might be seen as closely connected to
the very structures of our brain (it is true that the individuation directly
rests on the wired up similarity standards), the situation is very different for
the latter. It may be inborn that we perceive two sounds as instances of
the same word, but surely not that we perceive two different types of
sounds as meaning the same.15 So what is the basis of the oppositions, or
of the structure, which make up semantics?
The answer the (neo)structuralists – who were also (neo)pragmatists,
like Quine or Sellars – were bound to give is that it is the sameness and
difference of usability, of the role that a word is able to play within our
language games, that any semantic structure must rest on.16 Two sentences
are semantically alike or synonymous if they can be employed to make
similar or the same moves within our language games. And then two words
can be said to be synonymous if their replacement within a sentence never
changes the way the sentence may be employed within the games.
Inferentialism, which gives the use-theory of meaning a normative
twist, is based on the assumption that the role of a word within our
language game is not a matter of the actual moves their players make, but
rather of the rules governing the permitted usage of these words. (Of
course, the rules are also a matter of what the players do; however, not
of the moves they choose to make, but of their ‘taking’ certain moves as
correct, whereas others as incorrect. In chess, the rules are also not a
matter of the moves the players tend to make, but rather of the fact that
they take some moves for legal and others for illegal.)
The notion of synonymy, which results from these considerations, is
vague, and hence also the resulting semantic structure, and consequently
its materialization into meanings, is vague. Therefore we must be aware
of the fact that making the step from the structure to an explication of
meaning is a non-trivial one – it amounts to drawing sharp boundaries
there where there are really none. However, this should not be read as
saying that explication is a dubious enterprise. Replacing fuzzy phenomena
by their non-fuzzy explications, using idealized models, is a standard part
of the methods we use to account and make sense of our world.
Thus the resulting picture is that our linguistic games, especially the central
game of giving and asking for reasons, are governed by certain rules which are
implicit in our practices and hence are not distinct in the way explicit rules
can be, but which are nevertheless essential. What we call meanings, then, are
the roles individual expressions of our language acquire vis-à-vis the rules.
Conclusion: Language as a Social Institution

This brings us to an important moral: inferentialism is not mere preference
for one fundamental semantic concept (inference) over another (reference,
representation, or something else). It involves the conviction that to understand
language in its semantic aspect we must turn our attention to the social
background of language. Hence the idea is that though it is undoubtedly
interesting and important to study the psychology of the participants of
communication, to understand what semantics is about we must turn our
attention to language as a social institution.
Let us return to Wittgenstein, whom I listed among the initiators of
the pragmatic turn and who also stressed the key role of rules within our
language games. Why did he abandon the earlier elegant system of his
Tractatus and settled for his later haphazard theory of language games, as
presented in his Philosophical Investigations? One answer might be that
while in Tractatus he saw his whole language as a gigantic system of names,
later he was to realize that the concept of naming is too complex to be
used as an unexplained explainer.
What does it take to be a name? It is often assumed that a name of a
thing is something as a label stuck to the thing. But imagine a society
with a habit of doing literally this: sticking labels with inscriptions onto
things. By doing so, are they actually giving names to the things? Surely
not by the labeling alone: the labeling means various other things: the
labels may serve as mere decorations, they may bear advice for people
encountering them etc. So what decides whether they be considered as
names, or as something different? Undoubtedly the ways in which they
are treated by the members of the society, the larger practices into which
the practice of sticking labels is embedded. And hence I think that
Wittgenstein realized, in Coffa’s words, that
the ultimate explanatory level in semantics is not given by references to
unsaturation or to the form of objects or meanings, but by reference to the
meaning-giving activity of human beings, of activity embodied in their
endorsement of rules. (267)
And it is just in this spirit that Brandom, the initiator of contemporary
inferentialism, sees semantics as underlain by inference: inference is what
is needed to oil the wheels of the social practices that make us rational,
content-mongering creatures. Inferential structure is crucial (and not only
for the logical, but for any vocabulary) because it reflects the social back-
ground of language. Brandom is convinced that the inferential structure
of language is the result of the interplay of commitments we undertake and
entitlements we acquire when engaging within our language games. It is
from this viewpoint that it is important to look at language as a social
institution – for it follows that what we usually call meaning is a matter of
this very aspect.
Many initiators of the pragmatic turn, notably Quine, were quite hostile
to the very concept of meaning – they praised the turn for ridding us of
the concept as an excessive baggage and for letting us concentrate directly
on our linguistic practices. In the same spirit, Sellars criticized Carnap for
his inclination towards ‘formal semantics’.17 However, in this paper I have
indicated that the two enterprises, inferentialism as one of the outcomes
of the pragmatic turn, and formal semantics as a project of a logico-
mathematical explication of meaning, need not be seen as incompatible.
Indeed I am convinced that the interconnection of the two projects may
help us make sense of many traditional ideas on the boundary between
linguistics and philosophy: it may throw new light on some of the ideas
of the classical and newer structuralism, it may provide for a new and
illuminating way of representing semantics yielded by normative use theories,
and it may lead us to a reinterpretation of the Frego-Tarskian formal
semantics such that it survives the pragmatic turn.
Acknowledgment
Work on this paper was supported by grant No. 401/06/0387 of the
Czech Science Foundation.
Short Biography
Jaroslav Peregrin’s research is located at the intersection of logic, analytic
philosophy, and semantics; he has authored papers in these areas for
Australasian Journal of Philosophy, Erkenntnis, Journal of Philosophical Logic,
Pragmatic and Cognition, Semiotica, Studia Logica, and the Elsevier’s Handbook
of the Philosophy of Science (see his bibliography at <http://jarda.peregrin.cz/
mybibl/mybibl.php>). His book Meaning and Structure (Ashgate, 2001)
argues that recent and contemporary (post)analytic philosophy, as developed
by Quine, Davidson, Sellars, and Brandom, is largely structuralistic in the
very sense in which structuralism was originally tabled by de Saussure; it
also indicates that this view of language is not incompatible with formal
approaches to semantics. His current research focuses on both logical and
philosophical aspects of inferentialism, namely, the view that meaning is
essentially a matter of inference. He is a researcher at the Institute of
Philosophy of the Academy of Sciences of the Czech Republic and a
professor of logic at the Faculty of Arts and Philosophy of the Charles
University in Prague. As a visiting scholar, he worked at the University
of Konstanz in Germany and the University of Pittsburgh in the USA.
Notes
* Correspondence address: Institute of Philosophy, Academy of Sciences of the Czech
Rebublic, Jilská 1, 110 00 Prague, Czech Republic. Email: peregrin@ff.cuni.cz.

1
The first appearance of the term is usually taken to be Bréal.
2
See Rorty.
3
See Kripke; Putnam.
4
See, e.g., Devitt.
5
An important current semantic school favoring this line is the ‘cognitive’ school originating
with Schiffer and with Fodor, ‘Psychosemantics’.
6
The arguments go back to Frege (‘Der Gedanke’). Quine (Ontological Relativity) criticizes the
idea that semantics consists in the mind’s linking of signs to objects as the ‘museum myth’.
7
As Davidson puts it, ‘that meanings are decipherable is not a matter of luck; public availability
is a constitutive aspect of language’ (314).
8
Recently, there have been attempts to develop ‘externalist’ theories of mind according to
which this is not the case. But this attitude usually blocks the order of explanation from mind
to language, for it comes to rest mind on language.
9
Egginton and Sandbothe have employed this term is a slightly narrower sense.
10
See Peregrin, ‘Is Compositionality an Empirical Matter?’
11
Note that unlike in the case of the theories mentioned in the introduction of this paper,
which invoke such concepts to characterize relations between words and their referents, in this
case they would have to amount to the relation between words and their meanings, which is
much less feasible.
12
See, e.g. Dekker, ‘The Semantics of Dynamic Conjunction’.
13
As we seem unable to specify the kind of correctness in play here without recourse to the
concept of truth, this does not amount to a reduction of the concept of truth to other concepts,
hence to a theory of truth. Cf. Peregrin, ‘Brandom and Davidson’.
14
Its point of departure was the discussion between Prior and Belnap, where the former pointed
that an unrestricted trust into the ability of inferential patterns to generate meanings may lead
us astray, the latter opposed that indeed we do need a restriction, however, this restriction
can derived from the inferentialist framework itself. As for the opposite ways of formulating
the view of what logic is about, see Hacking (the proof-theoretic perspective); Tarski (the
model-theoretic one).
15
For discussion of the concept of semantic structure from the viewpoint of inferentialism, see
Peregrin, Meaning and Structure ch. 9; ‘Nature of Meaning’.
16
Quine’s views of language are most clearly expressed in his seminal Word and Object; for
Sellars’s see especially ‘Language, Rules and Behavior’; ‘Some Reflections on Language Games’;
‘Language as Thought and as Communication’.
17
See Peregrin, ‘Semantics without Meaning?’, for a detailed discussion of this movement
towards ‘semantics without meanings’.
Works Cited
Belnap, Nuel. ‘Tonk. Plonk and Plink’. Analysis 22 (1962): 130 – 4.
Brandom, Robert. Making It Explicit. Cambridge, MA: Harvard UP, 1994.
Bréal, Michel. Essai de sémantique. Paris: Hachette, 1897.
Carnap, Rudolf. Introduction to Semantics. Cambridge, MA: Harvard UP, 1942.
——. Logical Foundations of Probability. Chicago, IL: U of Chicago P, 1950.
——. Meaning and Necessity. Chicago, IL: U of Chicago P, 1947.
Chomsky, Noam. Syntactic Structures. The Hague: Mouton, 1957.
Coffa, Alberto. The Semantic Tradition from Kant to Carnap. Cambridge: Cambridge UP, 1991.
Davidson, Donald. The Structure and Content of Truth. Journal of Philosophy 87 (1990): 279 – 328.
Dekker, Paul. ‘The Semantics of Dynamic Conjunction’. Logic, Language and Computation III.
Eds. J. Seligman and P. Blackburn. Stanford, CA: CSLI, 2000. 105 –25.
Devitt, Michael. Designation. New York, NY: Columbia UP, 1981.
Egginton, William and Mike Sandbothe, eds. The Pragmatic Turn in Philosophy. New York, NY:
SUNY Press, 2004.
Fodor, Jerry. The Language of Thought. Scranton: Cromwell, 1975.
——. ‘Psychosemantics’. The Problem of Meaning in the Philosophy of Mind. Cambridge, MA:
MIT Press, 1987.

Frege, Gottlob. ‘Der Gedanke’. Beiträge zur Philosophie des deutschen Idealismus 2 (1918/19):
58 –77.
——. Grundlagen der Arithmetik. Breslau: Koebner, 1884. English translation: Foundations of
Arithmetic. Oxford: Blackwell, 1953.
——. ‘Über Sinn und Bedeutung’. Zeitschrift für Philosophie und philosophische Kritik 100 (1892):
25 – 50.
Hacking, Ian. ‘What is Logic?’ Journal of Philosophy 76 (1979): 285 –319.
Kripke, Saul. ‘Naming and Necessity’. Semantics of Natural Language. Eds. D. Davidson and G.
Harman. Dordrecht: Reidel, 1972. 253 – 355. Subsequently published as Naming and Necessity.
Cambridge, MA: Harvard UP, 1980.
Lance, Mark Norris. ‘Some Reflections on the Sport of Language’. Philosophical Perspectives 12
(1998): 219 – 40.
Lewis, David. ‘General Semantics’. Semantics of Natural Language. Eds. D. Davidson and G.
Harman. Dordrecht: Reidel, 1972. 169 –218.
Montague, Richard. Formal Philosophy: Selected Papers of Richard Montague. New Haven, CT:
Yale UP, 1974.
Peregrin, Jaroslav. ‘Brandom and Davidson. What Do We Need to Account for Thinking and
Agency?’. Philosophica 76 (2006): 43–59.
——. ‘Is Compositionality an Empirical Matter?’. The Compositionality of Meaning and Content.
Eds. M. Werning, E. Machery, and G. Schurz. Frankfurt: Ontos, 2005. 135 – 50.
——. Meaning and Structure. Ashgate: Aldershot, 2001.
——. ‘The Nature of Meaning. Brandom vs. Chomsky’. Pragmatics and Cognition 13 (2005):
39 – 57.
——. ‘Semantics without Meaning?’. Prospects of Meaning. Ed. R. Schantz. Berlin: de Gruyter,
forthcoming 2008.
Prior, Arthur. N. ‘Conjunction and Contonktion Revisited’. Analysis 24 (1964): 191– 5.
——. ‘Runabout Inference Ticket’. Analysis 21 (1960/61): 38 – 9.
Putnam, Hilary. ‘The Meaning of “Meaning” ’. Language, Mind and Knowledge. Ed. K. Gunderson.
Studies in the Philosophy of Science 7. Minneapolis, MN: U of Minnesota P, 1975. 131– 93.
Quine, Willard Van Orman. Ontological Relativity and Other Essays. New York, NY: Columbia
UP, 1969.
——. Word and Object. Cambridge, MA: MIT Press, 1960.
Rorty, Richard, ed. The Linguistic Turn. Chicago, IL: The U of Chicago P, 1967.
de Saussure, Ferdinand. Cours de linguistique générale. Paris: Payot, 1931. English translation:
Course in General Linguistics. New York, NY: Philosophical Library, 1959.
Schiffer, Stephen. Meaning. Oxford: Clarendon Press, 1972.
Sellars, Wilfrid. ‘Language as Thought and as Communication’. Philosophy and Phenomenological
Research 29 (1969): 506 –27.
——. ‘Language, Rules and Behavior’. John Dewey: Philosopher of Science and Freedom. Ed. S.
Hook. New York, NY: Dial Press, 1949. 289 – 315.
——. ‘Some Reflections on Language Games’. Philosophy of Science 21 (1954): 204 –28.
Tarski, Alfred. ‘What are Logical Notions?’ History and Philosophy of Logic 7 (1986): 143 – 54.
Turing, Alan. ‘Computing Machinery and Intelligence’. Mind 59 (1950): 433 – 60.
Wittgenstein, Ludwig. Philosophische Untersuchungen. Oxford: Blackwell, 1953. English translation:
Philosophical Investigations. Oxford: Blackwell, 1953.
——. Tractatus Logico-Philosophicus. London: Routledge, 1922. English translation: London:
Routledge, 1961.


Peregrin - An Inferentialist Approach To Semantics

Uploaded by

Copyright:

Available Formats

Peregrin - An Inferentialist Approach To Semantics

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Peregrin - An Inferentialist Approach To Semantics

Uploaded by

Copyright:

Available Formats

Philosophy Compass 3/6 (2008): 1208–1223, 10.1111/j.1747-9991.2008.00179.

An Inferentialist Approach to Semantics: Time

linguistic expressions with meanings. (Problems arise already with common

number. Principal promulgators of this tradition were Frege, Carnap and

Semantic Structure and Its Explication

Before I do so, let us illuminate the structuralist outlook by an analogy

The task of explication consists in transforming a given more or less inexact

Applied to meaning, it gives us the approach articulated by David Lewis:

explicatum vs. explicandum

Meanings, Rules, and Inference

Perhaps And; But What About Dog?

However, the obvious objection is that what makes the difference

Where Does Semantic Structure Come from?

Conclusion: Language as a Social Institution

© 2008 The Author Philosophy Compass 3/6 (2008): 1208–1223, 10.1111/j.1747-9991.2008.00179.x

© 2008 The Author Philosophy Compass 3/6 (2008): 1208–1223, 10.1111/j.1747-9991.2008.00179.x

© 2008 The Author Philosophy Compass 3/6 (2008): 1208–1223, 10.1111/j.1747-9991.2008.00179.x

You might also like