François RECANATI: Philosophie du langage (et de l'esprit), Paris: Gallimard, 2008, 268 pp.

,
ISBN 978-2-07-042162-6.

Most introductions to the philosophy of language (and this is what François Recanati intends his
book to be, p.9) are presentations of philosophical doctrines of authors that are regarded as the
classics of the discipline: thus a typical textbook will contain chapters or paragraphs on Frege on
sense and denotation, Russell on definite descriptions, Carnap on intension and extension, Tarskian
semantics, Quine on the indeterminacy of translation, Dummett and Davidson on the theory of
meaning, Kripke on proper names, Kaplan on demonstratives, and more. Such presentations may be
more or less critical, but the prevailing tone will be one of neutrality, the leading assumption being
that the student is supposed to familiarize with such notions and doctrines as they were originally
introduced by their authors. With respect to this kind of literature, the present book stands out in
being "subjective" through and through: it is essentially an exposition of Recanati's own philosophy
of language (and, to some extent, mind). Classical doctrines are presented and discussed to the
extent that they are relevant -negatively or, most of time, positively- to what Recanati regards as the
correct view.
Such a radical choice has both advantages and liabilities. One kind of liability is obvious:
whatever does not fit with the author's main theoretical line goes unmentioned, whatever its
historical importance or theoretical merit. Readers of this book will not learn about Dummett's and
Davidson's semantic programs, formal semantics of natural language and Tarski's contribution to it,
the Chomskyan approach to language (or any other cognitive approach), Grice on implicatures,
Brandom's views on language and communication; they will get the scantiest information
concerning Austin on speech acts (p.111), and they will know of Wittgenstein only as having
suggested that meaning is use (p.86). One feels that the beginning student may not be learning about
contributions that many regard as constitutive of the discipline.
But, on the other hand, the choice has the advantages of coherence and fluency. These are
enhanced by Recanati's great clarity and precision, which make the book most readable in spite of
the inherent difficulty of some the topics it deals with - no trivial achievement. Moreover, the
book's very subjectivity makes it interesting and useful reading even for seasoned professionals,
that are provided with a synoptic (and short) presentation of the views of one of today's most
authoritative philosophers of language.
Those familiar with Recanati's views will know that he has been receptive to philosophical
influence of two kinds: first, pragmatics and communication theory, particularly in Grice's version;
secondly, direct reference theory, especially in Kaplan's systematization. Hence, his thought on
language and mind is a prima facie surprising, but in fact extremely well organized and well argued
combination of a radically non-cognitive referential semantics with a contextualist theory of
interpretation and communication, involving all sorts of cognitive factors.

The book starts by taking stand on the priority issue. Though language and thought share many
important features (both have a syntax and a semantics), thought comes first: representational
content is originally mental. We cannot study a sentence without making appeal to the thought with
which it is associated; understanding a sentence is understanding the thought it expresses. The plan
of the book is laid out accordingly. There are three approaches to meaning: a cognitive approach
that reduces meaning to association with mental representations, a (purely) referential approach, and
a pragmatic approach, on which meaning is the role an expression plays in a certain social activity.
The cognitive approach is flatly rejected (more on this later); the pragmatic approach "subsumes the
referential approach without reducing to it". The book's first and second part will deal with
reference and pragmatics, respectively. The third part will deal with intentionality, i.e. with the
property of mental states "of relating to other things than themselves though containing such
otherness in themselves, so to speak" (p.24). Recanati believes that such a property can be
naturalized, i.e. based on genuine relations between coexisting entities (as in causal theories of

1
reference). This will be explained in the fourth part, which Recanati regards as containing "the
book's message".
Part I, on reference, begins with an exposition of Frege's theory of denotation and sense
(here called 'mode of presentation'). A distinction is introduced between denotation, designation and
extension. In 'Snow is white', 'snow' both denotes and designates the snow, while 'white' denotes
whiteness but does not designate it. Extension (of a predicate) is, standardly, the set of objects that
possess the property denoted by the predicate. After a discussion of Frege's theory of concepts (here
called 'properties'), of Carnap's simplification of it, and a short excursus on Mill's theory of
reference, Recanati introduces the theory of reference he endorses, i.e. David Kaplan's. Here he
deals with direct reference and rigidity, scope ambiguities involving modal operators and
descriptions, character and content, the token-reflexivity of indexicals, and the distinction of
circumstance of utterance (= context) and circumstance of evaluation.
Recanati takes sentences' denotations to be states of affairs, not truth values as in Frege.
States of affairs are the sentences' truth conditions. Frege's argument to the effect that thoughts
cannot be sentences' denotations (for they are not compositional) is rejected, as it does not apply to
states of affairs but to Fregean thoughts, which are not truth conditions but truth conditions under a
certain mode of presentation. Here, Recanati appears to be mixing up two different arguments of
Frege's. On the one hand, there is an argument to the effect that sentences' denotations cannot be
thoughts (the Morning Star/Evening Star argument): this is a deductive argument (a reductio). On
the other hand, there is an argument to the effect that denotations are truth values (the Odysseus
argument): this is not a deductive argument but an inference to the best explanation. Frege has
neither a single argument, nor a conclusive argument to the effect that denotations of sentences are
truth values, not thoughts.
Recanati also rejects Frege's solution to the problem of the unity of the proposition, based on
the "unsaturated" nature of concepts. Instead, he proposes to base it on the distinction between
designative function (of singular terms) and predicative function (of predicates). "To the syntactic
cohesion of a complex expression involving a designative element and a predicative element there
corresponds, on the semantic level, the cohesion of the state of affairs denoted by the complex
expression" (p.76). I don't see this as an improvement on Frege's solution. The issue was (and is):
why does a sequence of expressions, each with its own denotation, constitute a complex expression
that designates a semantic whole, rather than a list of denotations? The answer that expressions in
the sequence have different functions does not explain the combination's effect, given what we
know about the constituents' semantics, unless we define the effect into the notion of predicative (or
designative) function. Recanati's solution looks as verbal as Frege's.
Part II, on semantics and use, begins with the traditional view according to which
pragmatics is secondary to semantics because force depends on meaning: it is by virtue of its
meaning that an expression comes to have a certain force when used in certain circumstances. The
opposite view -priority of pragmatics- relies on the objection that, while every expression belonging
to some language has a function, some expressions (e.g. "Thank you") have a function without
having representational content. The traditionalist, in turn, objects that subsentential expressions
have representational content but no (pragmatic) function, as only complete sentences can be used
to produce speech acts. But, the pragmaticist conclusively retorts, an expression may have force
without exercising it in every case (as with declarative sentences embedded in other sentences);
moreover, subsentential expressions can also be said to have a function (predicative expressions
have predicative function, etc.). Recanati believes that, all things considered, generality favours the
pragmaticist's viewpoint. As one would expect, he adds that, because of contextual dependence,
sentence types only possess schematic representational content but no genuine truth conditions; on
the other hand, he grants that illocutionary act potential is associated with sentence type, hence it is
also true that pragmatics presupposes semantics.
This is cashed out in the theory of interpretation. Understanding is, in part, computation of a
sentence (type)'s semantic value based on semantic competence, i.e. knowledge of both lexical

2
semantic rules and compositional rules. However, what we actually understand are utterances of
sentences. Interpretation of utterances is seen as a special case of interpretation of actions, i.e.
explanation of an agent's behaviour by ascribing him certain mental states (intentions, beliefs and
desires). Contrary to semantic computation, the interpretation of utterances is defeasible, hence
holistic. Successful communication is analysed along Gricean lines: it consists in the hearer's
recognition of the speaker's communicative intention. In Grice's analysis, this is the intention to
achieve a certain effect E by way of the hearer's recognition of the speaker's intention - not the
intention to achieve E, but the intention to achieve it by way etc. Communicative intention is thus
reflexive. Though Recanati presents this conception as alternative to attributing the speaker
infinitely many intentions of ascending order, I don't see the advantage: the hearer need not just
recognize that I, the speaker, intend to achieve E, but that I intend to achieve E by his recognizing
my intention i, i.e. the intention - not to achieve E, but to achieve E by his recognizing... and so on
ad infinitum. Packing the infinity into the reflexive i will not make it go away.
The first two parts of Recanati's book are an admirably clear defense of controversial, but
extremely well argued views on the semantics and pragmatics of language. The last two parts are
more intricate and, to my mind, less convincing, though almost equally clear. Part III is on the
semantics of mental representations. Mental representations, like sentences, have content: when one
thinks, there is something one thinks. Hence, thought, like language, has its own semantics. It also
has syntax, i.e. constituent structure, respecting Evans's constraints. Given the strong analogy
between language and thought, we can either regard thought as primary and see language as the
externalization of thought, or regard thought itself as essentially linguistic, as in Fodor's "language
of thought" hypothesis. Recanati doesn't consider a third alternative: regarding thought - the thought
that folk psychology and (some) philosophy of mind is about- as mere internalization of language,
bearing no obvious relation to cognitive processes implemented in the brain. But more on this later.
After a useful section on compositionality of thought vs. partial compositionality of
language, Recanati tackles the issue of intentionality of mental states. It appears that mental
intentionality cannot be a genuine relation, for we may think of nonexistent objects such as unicorns
or Santa Claus. However, if we take it to be a quasi-relation one term of which (the intentional
object) is internal to the other (the representing mental entity), we risk attributing to the mental the
mysterious power of representing objects whose reality is conferred upon them by the mind in the
act of representing them (p.132), which Recanati sees as a step in the direction of Cartesian
dualism. As an alternative, it is suggested that we start with Grice's natural meaning and the causal,
fully extensional relations on which it is based. To go over to non-natural meaning -the kind of
meaning that characterizes representations- we rely on Dretske's idea that A represents B if and only
if it is A's function to indicate B (A indicates B if it covaries with B). Whatever has a function can
malfunction, hence representations can be erroneous - misrepresent what they are supposed to
represent- without ceasing to be representations: "what is wrong, non-existent, or impossible can be
represented, whereas only reality can be indicated" (p.142). This should take care of unicorns. But
the analogy is weak: a faulty thermometer wrongly represents the temperature as being 20°C,
whereas it is 10°C; what does a thought of a unicorn wrongly represent? Being? The world?
Be that as it may, the functional account does not explain the representational properties of
mental states. Representing artifacts, such as thermometers, are attributed a function by their users,
who represent to themselves the artifact as having a certain function. What about the latter
representing? If we say that representing power is intrinsic to the mind, we once more fall into
dualism. Instead, we want to say that, in the case of mental representations, their function is
assigned to them by the system to which they belong: a mental representation represents, say, a
predator in that its overall function, as defined by its connections with perception and behaviour, is
to indicate that kind of predators. "A mental representation's content is accounted for...by the
relations that the vehiculating mental state entertains with its environment as well as with the
organism's other mental states" (p.149). One feels it's a long way from here to thoughts about Santa
Claus.

3
 The concluding Part IV is an almost self-contained essay on descriptivism vs. direct
reference theory. Descriptivism has obvious merits: it explains why 'Pegasus doesn't exist' is
meaningful and why true identities involving proper names ('Hesperus is Phosphorus') can be
informative. But it also has to face powerful objections. As Kripke noticed, descriptions associated
to proper names are not always individuating. Moreover, could descriptivism be extended to general
terms? Isn't it more plausible to think that words like 'extraterrestrial' directly refer to properties? (A
venerable tradition has it that such words refer by way of their definitions. To oppose such a view,
the example of 'extraterrestrial' is particularly unhappy. 'Yellow' would have been better). Finally,
as Evans pointed out, descriptivism does not account for our experience of singularity: "The
objects' singularity is a basic datum of our experience and representation of the world. When we
represent the world to ourselves, we refer, either linguistically or mentally, to individual objects"
(p.163).
The claim that reference is direct is initially introduced as meaning that "objects we
represent are directly given to us in experience". But then, it is pointed out that we think of objects
of which we have no direct experience (nor any individuating description); this is accounted for by
the Kripkean notion of a causal chain. It is concluded that "We can speak of direct reference even
when relations to the object are "indirect"" (p.173). This is confusing. The claim that reference is
direct is a semantic thesis, carrying no epistemological connotations: it means that reference is not
mediated by associated descriptions (indeed, it is not mediated by anything epistemological, least of
all by experience). This does raise problems, as it entails that (more often than desirable or
plausible) we don't know what we are talking about. However, mixing up the semantic and the
epistemological won't help.
After ably discussing attributive and referential uses of descriptions, Recanati wonders
whether descriptivism could be salvaged by integrating reference-determining relational factors into
descriptive content, i.e. whether the Kaplanian notion of character could be generalized beyond
indexicals and demonstratives. This cannot be done, in his opinion: in the general case, conditions
of reference cannot be seen as part of the character, as they are not accessible to the speaker's
competence. Here, Recanati presents the standard arguments for externalism, such as Putnam's
Twin Earth.
The book's concluding sections deal with neo-Fregeanism (the view that severs modes of
presentation from descriptivism by countenancing non-descriptive modes of presentation) and the
semantics of mental representations, particularly mental indexicals. In the latter case, we are invited
to distinguish between concept types (such as the concept of "here"), concept instances (this
particular "here", whose reference is determined by a subject's relation to some particular place),
and concept occurrences (the mental event of a concept's activation). Like linguistic indexicals,
mental indexicals have a character of sorts: this is a file (dossier), a collection of information
pointing to something in the world. The file's content -the information gathered about the referent-
does not determine reference, but the pointing does. Files are said to be "paradigms of non-
descriptive modes of presentation" (p.207). However, they are descriptive (they contain information
about the referent which fixes "conditions of conformity"), though their descriptive content does not
determine reference. This is how Recanati (like Evans) tries to reconcile externalism with some
analogue of Fregean sense.

I have two main objections to Recanati's views as presented in this book. First, I believe his
dismissal of cognitive semantics is unfair. He seems to think that in cognitive semantics linguistic
expressions are assigned uninterpreted mental representations, "pictures in the mind": he mentions
in this connection Lewis's criticism (in 1970) of Katz's and Postal's (1964) notion of semantic
interpretation as being mere translation into "markerese", a language that is itself in need of
interpretation. He also mentions Evans's and McDowell's general criticism of "translational
semantics" (in [1976], pp.ix-x), that claims to give the meaning of a sentence S by describing its
relation to S', a sentence of some other language. Whatever the merits of such arguments, they do

4
not apply to most versions of present day cognitive semantics. In cognitive views, the meanings of
linguistic expressions are mental entities whose connection with the world depends on perceptual
and motor processes, not on objective relations between a speaker and her environment as in
externalist semantics. The dividing issue is this: for externalism, the reference determining relations
may be, and often are cognitively inert: they may not affect the speaker's cognitive system in any
way. For cognitivism, cognitively inert relations cannot play any semantic role, as meaning must be
accessible to the mental processing of language: if not, then meaning is not what we understand.
Actually, it is Recanati who conceives of mental representations as formulas of a
syntactically structured lingua mentis that is in need of semantic interpretation (pp.119-122). Here
comes my second objection. Like Fodor, Recanati seems to think that such an interpretation must
consist in some sort of causal connection with entities in the world, patterned upon referentialist
semantics of language. But much neuroscientific research appears to be telling a different story.
Real life understanding -grasping the meaning of linguistic expressions- is a complex process which
involves, among other things, the activation of neural processes that are also active in perceptual
and motor performances. To the extent that understanding is something we do, such a partial
replication of the processes by which we are connected to the world (in perception and movement)
seems to be all the connection we have - and need. If we want to be semantic naturalists (we needn't
be), this is what we should try better to describe and understand, rather than looking for some
naturalistically acceptable emulation of the objective connection of word to object that is assumed
(though never analyzed) in referential semantics.

Diego Marconi

References

G.Evans, J.McDowell, "Introduction" to Truth and Meaning (Oxford: Oxford Univ.Press, 1976)
D. Lewis, "General Semantics", Synthese 1970(22), 18-67.