From linguistic contextualism to situated cognition: the

case of ad hoc concepts

Jérôme Dokic

To cite this version:

Jérôme Dokic. From linguistic contextualism to situated cognition: the case of ad hoc con-
cepts. Philosophical Psychology, Taylor & Francis (Routledge), 2006, 10 (3), pp.309-328.
<ijn 00260970>

HAL Id: ijn 00260970

http://jeannicod.ccsd.cnrs.fr/ijn 00260970
Submitted on 5 Mar 2008

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
Philosophical Psychology
Vol. 19, No. 3, June 2006, pp. 309–328

From Linguistic Contextualism to

Situated Cognition: the Case of
Ad Hoc Concepts
Jérôme Dokic

Our utterances are typically if not always ‘‘situated,’’ in the sense that they are true or false
relative to unarticulated parameters of the extra-linguistic context. The problem is to
explain how these parameters are determined, given that nothing in the uttered sentences
indicates them. It is tempting to claim that they must be determined at the level of thought
or intention. However, as many philosophers have observed, thoughts themselves are no less
situated than utterances. Unarticulated parameters need not be mentally represented.
In this paper, I try to make precise the notion of representation at stake here. In one sense
of ‘representation’, something is represented if it is inferentially relevant. In another, less
demanding sense, something is represented if it is relevant to the construction of a context-
sensitive, ad hoc concept. Ad hoc concepts act as ‘‘proxies’’ for cognitively more demanding
representations. They ‘‘imitate’’ the latter’s epistemic and pragmatic roles while being
inferentially less sophisticated. Thus, there are two senses in which a thought can be said
to be situated: (1) its truth-value is relative to a non-represented contextual parameter,
(2) its truth-value is not itself relative, but it involves a context-sensitive, ad hoc concept.
Keywords: Contextualism; Situated Cognition; Relativism; Ad Hoc Concepts;
Unarticulated Constituents

1. Introduction
In a posthumous essay composed in 1897, Gottlob Frege observed that ‘‘if someone
says ‘It’s raining’ the time and place of utterance has to be supplied’’ (1979, p. 135).
He pointed out that nothing in the sentence indicates who made the utterance, and
where and when. Surely the phenomenon is quite general. Most if not all of our
utterances are context-dependent or ‘‘situated,’’ in the sense that they are true or false

Correspondence to: Jérôme Dokic, EHESS, Institut Jean-Nicod (CNRS-EHESS-ENS), 1bis, avenue de Lowendal,
F-75007 Paris, France. Email: Jerome.Dokic@ehess.fr

ISSN 0951-5089 (print)/ISSN 1465-394X (online)/06/030309-20 ß 2006 Taylor & Francis

DOI: 10.1080/09515080600690102
310 J. Dokic
relative to contextual parameters that are not indicated in the uttered sentences.
Frege himself claimed that these parameters must somehow be incorporated into the
thoughts expressed by the utterances. More recently, John Perry and others have
suggested, pace Frege, that thoughts themselves are as context-dependent as
utterances. My goal in this paper is to use insights from philosophy of language,
especially pragmatics, to evaluate Perry’s and other forms of contextualism and
eventually give a different account of the sense in which our thoughts are situated.
I shall contrast two varieties of contextualism. ‘‘Linguistic contextualism’’ is
specifically about public-language representations, and the way their semantic
evaluation draws on the extra-linguistic context of utterance. ‘‘Cognitive
contextualism’’ is the situated cognition claim that conceptual thoughts and other
mental representations (whether or not they are expressed in symbols) depend
heavily and systematically on the subject’s environment. As we shall see, no easy path
leads from the former to the latter, since there are quite different versions of both
linguistic and cognitive contextualism.
Eventually, a moderate form of cognitive contextualism will be put forward which
involves the notion of an ‘‘ad hoc concept,’’ borrowed from cognitive psychology and
relevance theory. Ad hoc concepts are mental representations whose instantiation is
sensitive to the context of the relevant cognitive task. I shall claim that these concepts
‘‘imitate’’ the epistemic and pragmatic behaviour that more sophisticated concepts
would exhibit if they were used in the same context. In other words, they act as
‘‘proxies’’ for cognitively more demanding conceptual representations.
I shall then argue that ad hoc concepts (or at least some of them) are best analysed
using the notion of a ‘‘cognitively unarticulated constituent.’’ In John Perry’s seminal
work, the proposition expressed by an utterance can have constituents that are
unarticulated in the sense that there are no morphemes that designate them.
Independently of whether there are such constituents at the level of language, I shall
introduce an analogous notion at the level of thought. Roughly, cognitively
unarticulated constituents, in contrast to cognitively articulated constituents, are not
associated with inferential roles; they are so to speak inferentially invisible. Finally,
I shall suggest that the distinction between cognitively unarticulated and cognitively
articulated constituents corresponds to two ways (a minimal and a more substantial
one) of representing the world in thought.

2. Moderate Linguistic Contextualism

Even in the most literal cases of communication, it seems that we can use a sentence
to convey a thought that intuitively goes beyond what the sentence alone expresses as
a matter of linguistic convention. Here are some well-known examples from the
literature, where square brackets indicate the part of the thought that is not
linguistically expressed:
1. ‘It’s raining [here/in Paris]’.
2. ‘Thank goodness it’s over [now]’ (Prior, 1976).
 Philosophical Psychology 311
3. ‘The book is to the left [from my perspective]’.
4. ‘The local bar [relative to where we are] is open’.
5. ‘Everybody [in the classroom] should be attentive’.
6. ‘You’re not going to die [from that cut]’.
7. ‘Here is John’s book [the book he wrote/read/owns, etc.]’.
8. ‘It’s 3 o’clock [relative to a particular time-zone]’.
9. ‘These events are simultaneous [relative to a particular inertial frame]’.
The position that I shall call ‘‘moderate linguistic contextualism’’ (for reasons that
will emerge shortly) is meant to be an explanation of this apparently widespread
phenomenon. According to moderate linguistic contextualism, the proposition
literally expressed by an utterance (in the Russellian sense of ‘proposition’) can (and
perhaps must) have constituents that are not represented formally, either
syntactically or sub-syntactically. These constituents are said to be unarticulated
(Corazza, 2004; Perry, 1993; Recanati, 2002, 2004), in contrast to those that are
formally represented, and said to be articulated. In the foregoing examples, the square
brackets indicate unarticulated constituents of the propositional contents of the
relevant utterances.
Critics of moderate linguistic contextualism do not believe in unarticulated
constituents. For instance, Jason Stanley claims that ‘‘there are no convincing
examples of what John Perry called ‘unarticulated constituents’’’ (2000, p. 391; see
also his 2002). On Stanley’s view, all propositional constituents turn out to be
articulated at the level of logical form, whether they are actually pronounced or not.
More generally, critics of moderate linguistic contextualism subscribe to the
‘‘principle of full articulation,’’ according to which the semantic structure of an
utterance is isomorphic with the syntactic structure of the sentence uttered.
There are different ways of implementing the principle of full articulation. Stanley
himself defends a position that has come to be known as ‘‘indexicalism,’’ according
to which what seem to be unarticulated constituents are in fact designated by
variables at the level of logical form which take their values in the context of the
utterance. For instance, an utterance of ‘It’s raining’ really involves a two-place
predicate ‘RAIN(t, l)’, where t and l are variables respectively for a time and
a location. Alternatively, Cappelen and Lepore (2005) have recently developed what
they call ‘‘semantic minimalism.’’ Instead of postulating hidden variables at the level
of logical form, these authors suggest that we take the truth-conditions of the
problematic utterances at face value. For instance, an utterance of ‘It’s raining’ is
true if and only if it is raining. The metaphysical fact that rain is a phenomenon
that concerns a (vaguely) bounded region in space should not affect our semantic
analysis.1
I cannot go into the details and subtleties of the recent debate between friends
and foes of moderate linguistic contextualism.2 The question I would like to raise
concerns how relevant the debate is to the issue of cognitive contextualism. And the
prima facie disappointing answer is: ‘‘not much.’’ Moderate linguistic contextualism
is compatible with the most extreme scepticism about there being interesting cases
312 J. Dokic
of context-dependence at the level of thought. Thus, one might argue that even if
there are unarticulated constituents at the level of language, they must be mentally
represented by the speaker (as well as by the hearer if she wants to understand the
thought expressed by the utterance). For instance, I can use the sentence ‘It’s raining’
to mean that it is raining at a particular place (typically where I am), only because
this is the place I have in mind when I talk to you. On this view, all propositional
constituents are ‘‘articulated’’ or made explicit at the level of thought.
Jerry Fodor seems to defend such a view (see, e.g., Fodor, 2001). Producing and
understanding an utterance are a matter of translating formulas of a language of
thought (mentalese) into public-language expressions, and conversely. Now the
thought expressed by an utterance can be at least partly constituted by mentalese
expressions that are not the translations of any public language expressions actually
used in the relevant context. These would correspond to ‘‘unarticulated
constituents,’’ even though the thought itself is of course completely explicit. The
phenomenon of ‘‘unarticulated constituents’’ would thus be strictly linguistic, and
irrelevant to the issue of cognitive contextualism.
I hope it is now clear why I call the form of linguistic contextualism under
discussion in this section ‘‘moderate.’’ This is because it does not have clear
implications for the assessment of cognitive contextualism. Moderate linguistic
contextualism is relatively neutral as to whether there are interesting cases of context-
dependence at the level of non-linguistic mental representations. Unarticulated
constituents are said to be provided by the ‘‘extra-linguistic’’ context, but the notion
of extra-linguistic context is underspecified.

3. Semantic Relativism
There is in recent philosophy of language a much more radical version of linguistic
contextualism. Up to this point, the issue has been whether the truth-value of
a complete, non-ambiguous, non-elliptical and non-indexical sentence is relative to
anything other than a possible world – such as a particular place in the case of ‘It’s
raining’. It was at least implicitly agreed on both sides that the truth-value of an
utterance is absolute (in the sense that what is uttered is not true or false relative to
anything but a possible world). Of course, this is consonant with a traditional
Fregean view, according to which the truth of an utterance corresponds to the truth
of a thought expressed by the utterance. Since, according to Frege, the thought has an
absolute truth-value, so has the utterance.3
Now some philosophers of language are prepared to abandon the orthodox view
that utterances have absolute truth-values. The idea at the core of ‘‘semantic
relativism’’ is that the content of an utterance can survive variations in truth-value,
even when the world of evaluation is fixed. Within the same possible world,
utterance-content can be true relative to one partial situation or ‘‘point of
evaluation’’ (Predelli, 2005) but false relative to another. It has been argued that
relativizing utterance-truth is independently needed to handle assertions about future
 Philosophical Psychology 313
states of affairs (MacFarlane, 2003), epistemic modals (MacFarlane, 2005b), scalar
predicates like ‘rich’ and ‘tall’ (Richard, 2004) and matters of personal taste such as
‘‘Roller coasters are fun’’ (Lasersohn, 2005).
What are the cognitive implications of semantic relativism? This depends on
whether the connection between utterance-truth and thought-truth is maintained or
not. Some relativists indeed sever this connection, by contending that the parameters
relative to which the utterance is to be evaluated are determined by the speaker’s
intentions, the topic of conversation or the expectations of the conversants (see, e.g.,
Predelli, 2005, p. 365). This seems to suggest that the thought expressed by the
utterance fixes these parameters and thus is not itself relative to them. On this
interpretation, the cognitive implications of semantic relativism are dim. As in
the case of moderate linguistic contextualism, what is not represented at the level of
language gets represented at the level of thought.

4. Cognitive Relativism
Maintaining the Fregean connection while denying the absoluteness of thoughts
yields a more ambitious version of semantic relativism, which implies an analogous
form of relativism at the level of thoughts. According to what can be called ‘‘cognitive
relativism,’’ the very same thought can be true in one situation and false in another
even though the same possible world is in question. This idea is implicit in John
Perry’s defense of the view that there can be ‘‘thought without representation.’’
On this view, a thought can ‘‘concern’’ (his terminology) a contextual feature that is
not mentally represented. For instance, the thought It’s raining is true or false only
relative to a particular place that need not be mentally represented (or representable)
by the thinker.4
Perry’s view is not restricted to conceptual thoughts. He holds a similar view with
respect to perceptual (visual) contents:
[S]imply to understand the fact that I duck when I see the ball, or the way hunger
and perception of a milk shake leads me to move my arm, we need not postulate
a self-representation. (Perry, 1993, p. 220)5
[O]ur most primitive knowledge about ourselves lacks any [component standing
for us]: basic self-knowledge is intrinsically selfless. (Perry, 1993, p. 205)

In general, perceptual spatial contents are correct or incorrect only relative to a non-
represented frame of reference centred on a particular viewer (at a given time).
What are the prospects of cognitive relativism? Note that a true cognitive relativist
should deny that there is a privileged or unique context relative to which a thought is
to be evaluated. John MacFarlane is among the few relativists who bite the bullet in
this respect, since he argues that utterance-content is true or false relative to a context
of assessment (see MacFarlane 2005a), and of course there is no a priori limit to such
contexts. However, rejecting the uniqueness of contexts in the cognitive case is
sometimes counterintuitive. Perry himself, commenting on the ducking example,
cites facts about the subject’s control systems as grounding the perceptual thought
314 J. Dokic
that a ball is approaching.6 These facts seem to provide the salient context relative to
which this thought is to be evaluated as true or false; it is true if and only if the ball is
approaching the subject. Other contexts simply seem irrelevant from the point of view
of understanding the subject’s rationality.
Now a cognitive relativist who accepts the uniqueness assumption faces a problem.
The problem is to explain why the facts that anchor a thought to its privileged
context of evaluation are not represented within the thought after all. Perry insists
that anchoring facts supplying the needed coordination between perception and
action are ‘‘external to the belief.’’ However, Perry’s insistence might be just the
remnant of an internalist prejudice. The fact that there is a privileged context
pertaining to the evaluation of a thought may show that the thought already takes
into account the relevant contextual parameters. This would mean that the thought
cannot be said to be true or false relative to these parameters. For instance, if the
subject is somehow represented within her thought that the ball is approaching, the
thought is not true relative to her; it is true tout court. I shall come back to this point
in the last sections of this paper.
Cognitive relativism is a bold and interesting approach, which may well be
appropriate to some cases of thoughts. The point raised in this section is that it
presupposes a substantial notion of representation, which has yet to be defined or at
least clarified further. In what sense does the perceptual thought or content that a ball
is approaching not represent the subject given that it is grounded on her visual
experience? Or should we say that the subject is incorporated into the thought
without being represented? In the absence of clear answers to these questions, the
prospects of cognitive relativism cannot be evaluated, which I think motivates
exploring other alternatives.

5. Relevance Theory
Relevance theory provides one such alternative. Relevance theorists (Carston, 2002;
Sperber & Wilson, 1998; Wilson & Sperber, 2002) contend that linguistic
understanding depends on grasping ad hoc concepts that are not lexically encoded
but constructed in an occasion-specific way constrained by expectations of relevance.
For instance, one may suggest that the thoughts typically communicated by the
utterances (1)–(9) involve the following ad hoc concepts, indicated here by hyphens:
10. It’s raining-here
11. It’s over-now
12. The book is to-my-left
13. The bar which is local-relative-to-here is open
14. Everybody-in-the-classroom should be attentive
15. You’re not going to die-from-that-cut
16. Here is the book written-by John
17. It’s 3 o’clock-relative-to-this-time-zone
18. These events are simultaneous-relative-to-this-inertial-frame
 Philosophical Psychology 315
Relevance theorists appeal to the notion of ad hoc concept in order to reanalyze the
relationship between language and thought. They claim that linguistically-encoded
meaning is far more schematic and fragmentary than is usually thought, and that
natural-language words are best seen as encoding ‘‘pro-concepts,’’ namely pointers to
an open set of ad hoc concepts. Interestingly, they also point out that their account of
linguistic understanding downgrades moderate linguistic contextualism’s notion of
unarticulated constituents. Consider an utterance of ‘Mary opened the door’, in a
context in which it is intuitively true if and only if Mary opened the door in a specific
way, namely with a key. The thought literally expressed by this utterance involves
the ad hoc concept open-with-a-key. There is no need to postulate the key as a
‘‘hidden’’ or ‘‘unarticulated’’ propositional constituent in order to complete a
‘‘neutral’’ linguistic meaning which would leave open Mary’s way of opening the
door (Wilson & Sperber, 2002).

6. Ad hoc Concepts
The notion of ad hoc concept used in relevance theory comes from the psychology of
concepts. One of the most intriguing claims of recent cognitive science is that mental
representations of objects and properties are very often constructed ‘‘on the fly,’’
depending on the current cognitive task. In several empirical studies, Larry Barsalou
(1983, 1987, 1999) has investigated so-called ‘‘ad hoc categories,’’ such as ‘‘things to
pack in a suitcase’’ and ‘‘things to take from one’s home during a fire,’’ which are
instrumental to achieving particular goals. Unlike common categories like ‘‘fish’’ and
‘‘apple,’’ ad hoc categories are not well established in long-term memory.
In Barsalou’s theory, they are temporary constructions activated in working memory:
People have the ability to construct a wide range of concepts in working memory
for the same category. Depending on the context, people incorporate different
information from long-term memory into the current concept that they construct
for a category. (Barsalou, 1987, p. 118)
For instance, I do not have to activate all of my encyclopaedic knowledge of
fish when I think about them in a particular context. I can use different clusters of
information when I am in a restaurant and when I am scuba diving. I may construct
ad hoc concepts of categories such as ‘‘fish that is ready to eat’’ and ‘‘fish to observe’’
out of the common category ‘‘fish.’’ These concepts are associated with different
‘‘packages of situation-specific inferences’’ (Barsalou, 2005). For instance, a
fish-that-is-ready-to-eat is typically cooked and placed on a plate, whereas a
fish-to-observe is typically swimming in the water. Barsalou and his collaborators
have convincingly shown that ad hoc categories behave in interesting ways like
common categories. For instance, they have a ‘‘graded structure,’’ in the sense
that some members of the category are judged to be more typical than others
(Barsalou, 1983).
Two caveats should be mentioned before the notion of ad hoc concept is put to
work. First, this notion is independent of Barsalou’s neo-empiricist view according
316 J. Dokic
to which ad hoc concepts are perceptually derived representations. This is a very
controversial view (also defended by Prinz, 2002), and it is not entailed by the very
idea of ad hoc concepts.
Second, the notion of ad hoc concept should not be thought to entail an epistemic
account of concepts of the kind Fodor (1998) strongly opposed. Indeed, what have
been called ‘‘ad hoc concepts’’ are in fact conceptions of various objects and
properties, and as such embody knowledge of the categories they represent. For
convenience, I shall stick to the term ‘ad hoc concepts’, but eventually we should
perhaps acknowledge that the same concept of dog can be associated with different
ad hoc conceptions depending on the relevant cognitive task.7

7. A Puzzle
Because it acknowledges that the semantic interpretation of an utterance heavily and
systematically depends on the extra-linguistic context, relevance theory is a form of
linguistic contextualism. Should we also think of it as embodying a form of cognitive
contextualism? Well, it depends on how ad hoc concepts, and thoughts composed of
them, are analysed. Consider an utterance of ‘Everybody should be attentive’ by
a teacher who wants to convey the thought that everybody in her classroom should
be attentive (i.e., the domain of the quantifier ‘everybody’ is implicitly restricted).
The teacher’s thought will typically involve the ad hoc concept everybody-in-the-
classroom – a useful category given her context.
Does it follow that the teacher has a mental representation of a particular
classroom? If the answer is ‘‘yes,’’ it seems that all semantically relevant aspects of the
utterance are articulated or made explicit at the level of thought, and there is no
room for cognitive contextualism. There is no interesting sense in which the ad hoc
thought Everybody-in-the-classroom should be attentive, once it has been formed,
is context-dependent. At least, it does not seem to be more context-dependent than
the thought Everybody in the classroom should be attentive, which is composed in the
usual (logical) way of the concepts everybody, classroom, being in, and so on.
If the answer is ‘‘no,’’ namely if the teacher need not form a mental representation
of a particular classroom, the ad hoc concept everybody-in-the-classroom is different
from the logically complex concept everybody in the classroom, which (by definition)
involves such a representation. Indeed, Barsalou’s findings show that ad hoc concepts
often behave as if they were logically non-composed, simple concepts. Now this
raises a puzzle that I do not think has been adequately dealt with. Surely, the ad hoc
concept everybody-in-the-classroom owes something to the universal quantifier
concept everybody. It does not seem acceptable to acknowledge that they are
different concepts, and leave the matter at that. One has the feeling that the former
concept is somehow derived from the latter. (Moreover, this feeling seems to be
independent of the issue of whether the English verb ‘everybody’ encodes a non-
restricted universal quantifier concept.) The puzzle, then, is to understand the nature
of the derivation, assuming that the ad hoc concept is not literally composed of the
 Philosophical Psychology 317
other concept. In order to solve this puzzle, one should enquire further into the way
ad hoc concepts work in the mind of a cognizer.

8. The Dual Role of Thoughts

According to a familiar picture (which goes back at least to Sellars, 1974), there are
two dimensions to the use of a particular thought. First, the thought has an inferential
role, which determines its ability to interact inferentially with other thoughts. The
relevant inferences can be purely formal (like the inference from Claire is asleep
to Someone is asleep) or material (like the inference from It’s raining to The streets
will be wet). Some notion of thought-constituent emerges at the former level. More
precisely, the constituents of a thought (its structural elements) are those on which
formal inferences involving it can hinge. For instance, it is part of the inferential role
of the thought Claire is asleep that it can interact with other thoughts in which the
constituents Claire and is asleep figure separately, as in the following inference: Claire
is asleep. Claire is my sister. So my sister is asleep. This formal inference hinges both on
Claire and on the property of being asleep.
Formal complexity corresponds to cognitive complexity. Drawing inferences that
hinge on some entity in the world is a substantial cognitive achievement. For
instance, in order to make sense of the inference about Claire, the subject must be
able to identify her across inferentially connected thoughts (Claire is asleep and Claire
is my sister), which depends on possessing some criterion of identity appropriate to
Claire.
There is an interesting analogy with the use of language here. Recanati (1997) has
pointed out that an entity that is explicitly referred to in using a word is selected from
a ‘‘paradigm’’ of other relevant entities in the discourse situation:
When I say ‘‘In Paris, it is raining’’, this makes sense only insofar as the location
Paris is virtually contrasted with some other location, such as London or the
country. This is a point which European ‘‘structuralism’’ has much insisted on:
whatever is singled out in speech is extracted from a ‘‘paradigm’’ or contrastive set.
(pp. 54–55)

An analogous point holds at the level of thought. In general, the entity on which an
inference hinges is extracted from a cognitive ‘‘paradigm’’ of several (postulated)
entities. Using a criterion of identity for Claire will also allow the subject to contrast
Claire with other persons relevant to the current cognitive task, as in the inference
Claire is my sister. The owner of this car is not my sister. So Claire is not the owner
of this car. In this inference, Claire is extracted from a paradigm that includes at least
one other person, described as the owner of a demonstrated car.
The second dimension of thought is its epistemic-pragmatic role. A thought is not
only inferentially related to other thoughts; it is also anchored to perception and
action. Of course, a thought’s epistemic-pragmatic role depends on its inferential
role, since it can be related to perception and action more or less indirectly, through
(formal and material) inferences involving other thoughts. It is epistemically
318 J. Dokic
connected to perception, in the sense that, together with other thoughts, experiences
can act as evidence for its truth. It is pragmatically connected to action, in the sense
that, together with other thoughts, it can have behavioural consequences.
On the whole, the evidential and the consequential features of a thought should be
coherent. More precisely, they should exhibit ‘‘harmony,’’ in Dummett’s generalized
sense (see Dummett, 1991, ch. 9). The notion of harmony was first used in logic, as
a requirement on the introduction and elimination rules associated with a given
connective. For instance, Prior’s (1960) famous ‘‘tonk’’ rule is not harmonious,
because, roughly speaking, it allows one to infer more than what one is initially given.
From p, one can infer p tonk q (introduction rule for ‘tonk’); and from p tonk q one
can infer q (elimination rule for ‘tonk’). Dummett has usefully generalized the logical
notion of harmony to deal with the meaning of non-logical words. For instance, an
introduction rule for the demonstrative thought That figure is round includes the
conscious perception of an object as being round, and an elimination rule includes
the motor instruction to reach toward this object (see Campbell, 2002). These rules
are harmonious only if they concern the same object in the world.
In the following sections, I shall develop a suggestion about the formation of
at least some ad hoc concepts. The suggestion is that these concepts imitate the
impact that more sophisticated concepts would have on perception and action were
they used in the relevant cognitive task. I shall first deal with thoughts typically
expressed by indexical utterances, which hopefully will provide a model or exemplar
for understanding other, non-indexical ad hoc concepts.

9. Proto-Indexical Thoughts
Consider the following thoughts, which are typically expressed by utterances
respectively of ‘It’s raining here’ and ‘It’s raining’:
19. It’s raining here
20. It’s raining-here
What is the difference between them? To begin with, (19) can participate in
inferences hinging on a particular place. For instance, together with the thought Here
is Paris, it implies the thought It’s raining in Paris. In order to make sense of this
inference, the subject needs to master a substantial conception of the relevant place,
which is cross-identified in the premises of the inference. Just as a particular place is
referred to, something is predicated of it. (19) involves our ‘‘official’’ concept of rain
as a spatially located event.
In contrast, the ad hoc thought (20) cannot participate in inferences hinging on
a particular place. It is inferentially encapsulated with respect to thoughts that
explicitly refer to raining places. It does not involve any predication of a place. At the
inferential level, a one-place concept of rain (rain at time t) is used instead of the
official two-place concept of rain (rain at time t and location l). The former is derived
from the latter in the following way: the subject’s criterion of identity for rain events
 Philosophical Psychology 319
is silenced, and she is left with a mere ‘‘criterion of application’’ (Dummett, 1973,
ch. 4), which is sensitive to the local presence of rain. We use the simpler concept
when the cognitive task does not involve any comparison with other places in
which it might rain or not. As Perry (1993) has emphasised, there are familiar tasks
of this sort:
Talking on the phone and reading the national weather reports are one thing,
talking to someone in the same room about the weather is a bit different. Our
reaction to the local statement ‘‘It’s raining’’ is to grab an umbrella, or go back to
bed. No articulation of the fact that the reporter’s place and our place are the same
is really necessary. (p. 216)

The ad hoc thought (20) has the inferential role of a feature-placing thought, in
something like Strawson’s (1959) sense. Still, it is not a mere feature-placing thought.
Its epistemic-pragmatic role is modelled on or mimics the behaviour the more
complex thought (19) would exhibit in the same situation. It is sensitive to what can
be locally perceived, and it is geared to action at the very same place (thus ensuring
harmony).8 I shall say that it is ‘‘proto-indexical’’.
In general, a proto-indexical thought mimics contextually relevant aspects of
an indexical thought’s epistemic-pragmatic role without being able to participate in
inferences hinging on what is indexed. The thoughts that can be expressed by
‘It’s raining’, ‘It’s over’ and ‘The ball is to the left’ are all proto-indexical relative to
the cognitively more sophisticated thoughts It’s raining here, It’s over now, The ball
is to my left. They involve the ad hoc concepts raining-here, being-over-now and
being-to-my-left.
Some indexical thoughts are based on mere dispositions to gather information
about what is indexed in the world. This is the case with a thought like I will live
forever, which can be formed in the absence of any self-related information actually
received by the subject. Other indexical thoughts are based on more actively keeping
track of what is indexed. This is the case with a thought like It’s hot in this place,
formed on the basis of visually attending a particular place (see Evans, 1982, ch. 6).
Similarly, proto-indexical thoughts can be more or less active. In some cases, the fact
that a proto-indexical thought is geared to action relative to the very contextual
feature to which it is perceptually sensitive is guaranteed as it were a priori, by virtue
of the embodied subject’s cognitive architecture. In other cases, this fact is guaranteed
by an active capacity to keep track of the relevant feature (I can think It’s hot-here
about a place I visually keep track of while moving around it). The thought
can remain proto-indexical because the mere practical capacity to keep track of
something is not enough to establish a cognitive contrast between that thing and
other potentially relevant things.

10. Sub-Inferential Monitoring

The claim so far is that ad hoc versions of indexical thoughts can be formed through
what I shall call ‘‘sub-inferential monitoring,’’ namely harmony-preserving control
320 J. Dokic
at the epistemic-pragmatic level that works independently of inferential mechanisms.
Can this claim be generalized beyond the sphere of indexicality? In this section, I shall
tentatively consider other potential cases with a similar structure.
Consider first the case of quantifier domain restriction. It can be effected either
explicitly, at the level of inferential role, or implicitly, directly at the level of
epistemic-pragmatic role. Consider the following thoughts:
21. Everybody in the classroom should be attentive
22. Everybody-in-the-classroom should be attentive
23. Everybody should be attentive
The thought (21) involves an explicit restriction of the domain of everybody. Its
inferential role licenses various inferences, including the inference to the more
complex thought Everybody in some classroom should be attentive. The subject can
make sense of this inference because her thought involves some conception of what
counts as a particular classroom as opposed to others. In other words, it involves
some criterion of identity for classrooms. Inferences involving the thought can then
hinge on a particular classroom. In general, drawing inferences that hinge on
something (an object or a property) is cognitively demanding. For instance, grasping
thought (21) requires possessing an identifying conception of a particular classroom
as opposed to other classrooms. This is useful if the thought is to interact inferentially
with thoughts about other students in other classrooms.
However, the teacher is not currently interested in comparisons with other
classrooms; she just wants her students to be attentive. She does not need an
identifying conception of the classroom she is presently occupying. This is why she
uses the ad hoc thought (22). I would like to suggest that it has the same inferential
role as the simpler thought (23), which involves an unrestricted universal quantifier.
In particular, neither (22) nor (23) can participate in inferences hinging on
a particular classroom (which is not even relevant in the case of the simpler thought).
Both are, so to speak, blind to classrooms.
If the thoughts (22) and (23) have the same inferential role, what is the difference
between them? On the present suggestion, the ad hoc thought involves an implicit
quantifier domain restriction. Some low-level cognitive mechanisms restrict the
epistemic-pragmatic role of the thought to particular students in a particular
classroom. Of course, holistic constraints imply that a whole chain of thoughts will be
thus restricted. As a consequence of this contextual restriction, ways of establishing
the thought involving students outside the room are simply never considered.
Similarly, no practical conclusion is reached which concerns students in irrelevant
classrooms. There might be relatively cheap inhibitory processes that select the inputs
and outputs of the thought’s inferential role (while preserving harmony between
them). With respect to her ad hoc thought (22), restriction of the domain of
quantification does not operate explicitly, at the level of inferential role. Rather,
it operates implicitly, at the level of epistemic-pragmatic role, in such a way that (22)
is inferentially encapsulated with respect to more complex thoughts such as
Everybody in some classroom should be attentive.
 Philosophical Psychology 321
As another illustration of the claim that sub-inferential monitoring underlies the
formation of (at least some) ad hoc concepts, consider an utterance of ‘Mary opened
the door’. Depending on the context, it can evoke one of the following thoughts:
24. Mary opened the door with a key
25. Mary opened-with-the-key the door
26. Mary opened the door
The thought (24) can participate in inferences hinging on ways of opening the
door. It can be formed in a context in which there is an issue as to how Mary opened
the door. She opened it with a key rather than, say, by just pushing an unlocked
door. The utterance can also convey the more general thought (26) in (rarer)
contexts in which no specific way of opening the door has been made salient. The
inferential roles of (24) and (26) are clearly different. For instance, (24) implies
that Mary did not open the door by just pushing it, while (26) does not have this
implication.
Now suppose that the context is such that a specific way of opening the door,
namely with a key, has been made salient, but no other way of opening the door is
relevant. In this context, the subject has no need for a thought like (24), which
involves an identifying conception of a way of opening the door. She has no need for
such a thought because her cognitive task won’t involve any comparison between
ways of opening doors.
One possibility, then, is that the subject uses a thought that satisfies the following
two conditions. First, it is inferentially on a par with (26): it cannot participate in
inferences hinging on a specific way of opening the door. It does not wear a reference
to a specific way of opening on its inferential sleeves, so to speak. In that respect, it is
conceptually less sophisticated than (24). Second, it nevertheless mimics relevant
aspects of the epistemic-pragmatic role of (26). The subject is disposed to restrict the
use of her thought to selected inputs and outputs. More precisely, she is disposed to
consider as evidential data relevant to her thought only cases of opening the door
with a key, and she is disposed to act on her thought only in ways compatible with
this way of opening the door. When these two conditions are satisfied, the subject
forms the thought (25), which involves the ad hoc concept open-with-a-key.
This may seem abstract, but the idea is in fact relatively simple. Prima facie at least,
there is no incoherence in supposing that the reference to a specific way of opening
shows up in the way the thought is used with respect to perception and action, rather
than in inferences in which more sophisticated thoughts like (24) can participate. For
instance, the subject who is in the relevant context would react to a request like
‘Please open the door’ only by opening the door with a key. Harmony implies that the
use of the thought is equally restricted on the evidential side. For instance, the subject
would answer the question ‘Has the door been opened?’ only by enquiring whether it
has been opened in the right way. Of course, there is room for the ad hoc concept
open-with-a-key only if sub-inferential monitoring of this sort is a less substantial
cognitive achievement than deploying an identifying conception of a specific way of
opening the door (see Dokic, in press, for further discussion of this point).
322 J. Dokic
Of course, these remarks are merely suggestive, but they point to a possible
solution to the puzzle about ad hoc concepts that was raised above. Ad hoc concepts
seem to owe something to other, more stable concepts. For instance, the ad hoc
concept everybody-in-the-classroom seems to be derived from the unrestricted
universal concept everybody. We now see that even though the former is not literally
composed of the latter, both concepts are inferentially similar. The unrestricted
concept is exploited but its use is modified in such a way that a different, implicitly
restricted concept is eventually grasped.
On the present suggestion, cognitively simple concepts can be used in some
contexts as proxies for more complex ones. However, this cognitive strategy has also
a drawback. Ad hoc concepts will often be less determinate than the concepts whose
epistemic-pragmatic role they mimic. For instance, the ad hoc concept open-
with-a-key mimics relevant aspects of the epistemic-pragmatic role of the more
complex concept open with a key, but the resemblance may be quite partial if this is
enough for the purposes of the current task. The indeterminacy will perhaps be less
severe when the subject can exploit an actual relationship with a contextual feature,
as in the case of proto-indexical thoughts.

11. Moderate Cognitive Contextualism

The foregoing account of ad hoc concepts is more moderate than cognitive
relativism. Unlike the latter, it is compatible with the standard Fregean claim that
thoughts are true or false absolutely. For instance, the thought typically expressed
by ‘It’s raining’ is not true or false relative to a particular place, since it involves the
ad hoc concept raining-here, which already takes into account the relevant place at the
sub-inferential level.
This account still belongs to the contextualist camp, and I shall accordingly call
it ‘‘moderate cognitive contextualism.’’ It exploits a distinction between cognitive
relevance and inferential relevance, and claims that some cases of cognitive relevance
involve context-dependent mechanisms. Let me elaborate on these points.
Let’s say that a worldly parameter P is inferentially relevant to a thought T when
T can participate in inferences hinging on P. For instance, the place the thought It’s
raining in Paris is about is inferentially relevant. Together with the thought Paris
is the capital of France, it formally implies It’s raining in the capital of France. If a
parameter P is inferentially relevant to a thought T, P is a propositional constituent
of T. The thought It’s raining in Paris is about Paris, which is a constituent of the
(Russellian) proposition associated with the thought.
The notion of inferential relevance also applies to ad hoc thoughts. Consider the
thought typically expressed by ‘It’s raining’, namely It’s raining-here. The complex
property of raining at a particular place (typically the thinker’s) is inferentially
relevant to it. This is why the inference from It’s raining-here to It’s not sunny-here is
licensed, where the complex property of raining at a given place is conceptually
contrasted with the complex property of being sunny at this very place. It does not
 Philosophical Psychology 323
follow that the place itself is inferentially relevant to these thoughts. Since they do not
involve an identifying conception of it, they cannot participate in inferences hinging
on it. This shows that a complex property can be inferentially relevant even though
its constituents are not.
When a parameter is inferentially relevant to a thought, I shall say that it is
cognitively articulated, more precisely that it is a cognitively articulated constituent of
the proposition associated with the thought. This terminology is justified by the fact
that, as we have seen above, the cognitive requirements for inferential relevance are
analogous to the cognitive requirements for a parameter to be linguistically
articulated.
Obviously, since inference is a cognitive achievement, everything that is
inferentially relevant is cognitively relevant. The foregoing account of ad hoc
concepts shows that the converse is not true. A particular place, typically the
thinker’s, is cognitively relevant to the ad hoc thought It’s raining-here. The thought is
anchored to this place thanks to a (more or less active) control loop that
independently narrows down its epistemic-pragmatic role. As a result, the thought
is sensitive to what is going on around the thinker, and is geared to action at
(roughly) the same place. This cognitive mechanism contributes to the thought’s
having a determinate content. Were it absent or malfunctioning, the thought would
not be about the complex property of raining at a particular place (at best it would be
about a spatially neutral feature registering the presence of rain). However, this place
is not inferentially relevant, for the ad hoc thought is not sophisticated enough to
participate in inferences hinging on it.
When a parameter is cognitively but not inferentially relevant to a thought, I shall
say that it is ‘‘cognitively unarticulated,’’ more precisely that it is a ‘‘cognitively
unarticulated constituent’’ of the proposition associated with the thought.
It is important to note that the notion of cognitively unarticulated constituent is
somewhat independent of the analogous notion at the heart of moderate linguistic
contextualism. Typically, a linguistically articulated constituent is also cognitively
articulated. For instance, if I utter ‘It’s raining in Paris’, I am probably deploying
some concept of Paris at the level of thought.9 However, an utterance’s unarticulated
constituent may not be cognitively unarticulated. I can utter ‘It’s raining’ while
thinking It’s raining here, i.e., having some identifying conception of the place where
I say that it is raining.

12. Two Modes of Representation

I have deliberately not used the vexed notion of representation in introducing
the distinction between cognitively articulated and cognitively unarticulated
constituents. Now the question can be raised of what constituents can be said to
be represented by the thought (or rather by its thinker). It is reasonable to say that a
thought represents its cognitively articulated constituents, but what about mere
cognitively unarticulated constituents? Does the thought represent them? On an
324 J. Dokic
austere view, what is represented coincides with what is inferentially relevant. This
might be the view preferred by friends of the language of thought hypothesis, since
part of the motivation for the latter is to explain the systematic role thought plays
in inferences.
On a different view, mere cognitively unarticulated constituents are represented,
albeit in a different way than cognitively articulated constituents. This might be the
view favored by cognitive scientists and philosophers who want to relax the notion of
representation, or distinguish between various forms of mental representation.
As Andy Clark (1997) suggests:
[W]e should not be narrow-minded about the nature of the inner events that help
explain behavioral success . . . . Stories invoking internal representations . . . may
come to coopt such highly complex, non-local, dynamical processes as the vehicles
of specific kinds of information and knowledge. (p. 174)

Coordinating mechanisms tampering with the roles of thoughts at the sub-

inferential level might be considered as ‘‘minimally’’ representing the world, at least
in cases where the indeterminacy is not too severe. For instance, the ad hoc thought
It’s raining-here involves a minimal representation of a particular place.10
Rob Wilson’s (2004) notion of ‘‘exploitative representation’’ also belongs here.
Here is how he defines this notion:
Exploitative representation is an efficient form of representation when there is a
constant, reliable, causal or informational relationship between what a device
does and how the world is. Thus, rather than encode the structure of the world and
then manipulate those encodings, ‘‘smart mechanisms’’ can exploit that constancy.
(p. 164)

The construction of the ad hoc thought It’s raining-here involves a smart

mechanism of this kind, which exploits the thinker’s spatial situation in order to
align the place of perception with the place of action.
A consequence of the relaxed view of mental representation is that the scope of
‘‘thought without representation’’ is limited. Consider again Perry’s example of
‘‘selfless self-knowledge.’’ His insight can be reframed as follows. I am not
represented in perception in the way the approaching ball is represented.
Both things are mentally represented, but in quite different ways. The ball is a
cognitively articulated constituent of perceptual content, while I am just a cognitively
unarticulated constituent. Perceptual content has the form of the ad hoc thought
A ball is approaching-me as opposed to the more complex thought A ball is
approaching me (without hyphens). Temporal and causal facts about my control
systems will ensure that the person ducking is the same as the person whom
the ball is rushing toward, namely me. There is no obstacle to considering these
facts as realizing a primitive form of self-representation. It follows that I am
represented in perception after all, although minimally, as a cognitively unarticulated
constituent.
If we accept this view, all thought-constituents are represented in some way
or other. A thought-constituent can be represented at the level of inferential role,
 Philosophical Psychology 325
through an identifying conception, which can be ad hoc or not. In this case, it is said
to be cognitively articulated. It can also be represented at a more implicit, sub-
inferential level, as a worldly parameter that plays a contextual role in channeling
the thought’s epistemic-pragmatic role. In contrast to an identifying conception,
such minimal representation does not allow for a cognitive contrast between the
constituent and other, potentially relevant ones. In this case, the constituent is said
to be cognitively unarticulated.

13. Conclusion
In this paper, I have tried to exploit insights from linguistic contextualism in an
account of thoughts that are tailored to a specific cognitive task—thoughts composed
of so-called ad hoc concepts. Different versions of linguistic contextualism have
been mentioned. Moderate linguistic contextualism introduces the important notion
of unarticulated constituents, but as it is, this notion is not very helpful in
explaining how an utterance’s unarticulated constituents are mentally represented.
Semantic relativism suggests that they are not propositional constituents of the
utterance after all. Cognitive relativism claims that they are not even mentally
represented, but this claim presupposes a more substantial notion of representation
than is usually given.
Reflection on the notion of ad hoc concept, introduced by cognitive psychologists
and at the core of relevance theory, suggests that a modest form of cognitive
contextualism may be enough to account for at least some of the cognitive
phenomena Perry and other contextualists have drawn our attention to. The key
distinction is between cognitively articulated and cognitively unarticulated
constituents. This distinction is independent of the corresponding distinction in
the linguistic case, and can be defined in terms that are neutral with respect to the
language of thought hypothesis. Roughly, cognitively articulated constituents are
inferentially relevant in a way cognitively unarticulated constituents are not, but
it does not follow (at least not without further argument) that they are referred to
by some kind of mental symbols.
While cognitive relativism claims that a given thought is true or false relative to
a contextual feature (other than a possible world) that is not represented, moderate
cognitive contextualism makes the point that this feature may be a cognitively
unarticulated constituent. Thinkers can exploit the context in which they perceive
and act in order to form ad hoc thoughts and dispense with formally identifying some
of their cognitively relevant constituents. Since cognitive exploitation can be seen as
a form of (minimal) representation, cognitively unarticulated constituents are
still mentally represented.
It does not follow that cognitive relativism is wrongheaded. On the contrary, it is
quite possible that there are cases of relative thought-truth. In fact, we can now give
a more precise formulation of cognitive relativism. It is the claim that thoughts can
be true or false relative to contextual features that are neither cognitively articulated
326 J. Dokic
nor cognitively unarticulated. In this paper, I leave open the question of what
thoughts have relative truth-values in this sense.

Acknowledgements
Various ancestors of this paper have been presented at the conference ‘‘Memory and
Embodied Cognition’’ (Macquarie University, Sydney, November 2004), the 5th
Prague Interpretation Colloquium (Czech Academy of Sciences, April 2005) and the
conference ‘‘La ciencia como proceso cultural’’ (UNAM, Mexico City, June 2005).
I thank Angeles Eraña, Axel Barcelo, Eros Corazza, John Sutton and a referee for
comments and/or encouragement.

Notes
[1] See also Borg (2004), which is also in the spirit of semantic minimalism.
[2] Recanati (2004) contains a very useful presentation of the relevant issues.
[3] See Evans (1985, ch. 12) for an influential defence of the Fregean view based on
apparently non-negotiable features of assertion. Full-blown relativists, such as John
MacFarlane, have opposed Evans’s argument and put forward a non-standard account of
assertion.
[4] I shall use italics when thoughts rather than linguistic representations are in question.
[5] Arguably, Perry’s examples are not on a par. Perhaps the ducking example does not
involve any representation; a fortiori, it does not involve any self-representation. The
case is different with the milk shake example, which involves some form of practical
reasoning.
[6] ‘‘The eyes that see and the torso or legs that move are parts of the same more or less
integrated body’’ (Perry, 1992, p. 219). See also Corazza (2004, especially ch. 2) who
contends, in a Wittgensteinian spirit, that our thoughts are anchored to non-represented
contexts determined by ‘‘language-games’’ and ‘‘forms of life.’’
[7] Interestingly, Barsalou (2005) calls ad hoc representations ‘‘conceptualizations,’’ and
defines a concept as a productive ability to generate many different situated
conceptualizations.
[8] Harmony is relative to thought-content. Consider a subject who utters ‘It’s raining’ on
the basis of her visual experience of the weather in Paris. She takes the train to Marseille
and, a few hours later, acts on the basis of her initial thought by opening her
umbrella there. The subject exhibits disharmony with respect to the ad hoc thought
It’s raining-here (since seeing the weather in Paris does not tell us anything about the
meteorological condition in Marseille), but she exhibits harmony with respect to a
mere feature-placing thought (which is indeed the only coherent thought-content she is
grasping).
[9] Here is a more controversial case. Suppose Wittgenstein is right and there are purely
expressive uses of ‘I’. For instance, an utterance of ‘I’m tired’ expresses the subject’s
tiredness, and is not associated with an identifying self-conception. The subject forms
a neutral or impersonal thought like There is tiredness. In other words, the subject is a
cognitively unarticulated constituent of the thought. Arguably, though, the subject is still
a linguistically articulated constituent of the utterance.
[10] The term ‘minimal representation’ is itself used by Clark in a consonant sense.
 Philosophical Psychology 327
References
Barsalou, L. (1983). Ad hoc categories. Memory & Cognition, 11, 211–227.
Barsalou, L. (1987). The instability of graded structure in concepts. In U. Neisser (Ed.), Concepts
and conceptual development: Ecological and intellectual factors in categorization (pp. 101–140).
New York: Cambridge University Press.
Barsalou, L. (1999). Perceptual symbol systems. Behavioral and Brain Sciences, 22, 577–660.
Barsalou, L. (2005). Situated conceptualization. In H. Cohen & C. Lefebvre (Eds.), Handbook of
categorization in cognitive science (pp. 619–650). St. Louis, MO: Elsevier.
Borg, E. (2004). Minimal semantics. Oxford, England: Clarendon Press.
Campbell, J. (2002). Reference and consciousness. Oxford, England: Clarendon Press.
Cappelen, H., & Lepore, E. (2005). Insensitive semantics: A defense of semantic minimalism and
speech act pluralism. Oxford, England: Blackwell.
Carston, R. (2002). Thoughts and utterances: The pragmatics of explicit communication. Oxford,
England: Blackwell.
Clark, A. (1997). Being there: Putting brain, body, and world together again. Cambridge, MA:
MIT Press.
Corazza, E. (2004). Reflecting the mind: Indexicality and quasi-indexicality. Oxford, England:
Clarendon Press.
Dokic, J. (in press). Situated representations and ad hoc concepts. In M. J. Frápolli (Ed.),
Saying, meaning and referring: Essays on François Recanati’s philosophy of language. London:
Palgrave.
Dummett, M. (1973). Frege: Philosophy of language. London: Duckworth.
Dummett, M. (1991). The logical basis of metaphysics. London: Duckworth.
Evans, G. (1982). The varieties of reference. Oxford, England: Clarendon Press.
Fodor, J. (1998). Concepts: Where cognitive science went wrong. Oxford, England: Oxford
University Press.
Fodor, J. (2001). Language, thought and compositionality. Mind & Language, 16, 1–15.
Frege, G. (1979). Logic. Posthumous writings. Oxford, England: Blackwell.
Lasersohn, P. (2005). Context dependence, disagreement, and predicates of personal taste.
Linguistics and Philosophy, 28, 643–686.
MacFarlane, J. (2003). Future contingents and relative truth. Philosophical Quarterly, 53,
321–336.
MacFarlane, J. (2005a). Making sense of relative truth. Proceedings of the Aristotelian Society, 105,
321–339.
MacFarlane, J. (2005b). The assessment sensitivity of knowledge attributions. In T. Szabo Gendler &
J. Hawthorne (Eds.), Oxford studies in epistemology (pp. 197–233). Oxford, England: Oxford
University Press.
Perry, J. (1993). Thought without representation. In J. Perry (Ed.), The problem of the essential
indexical and other essays (pp. 205–225). Oxford, England: Oxford University Press.
Predelli, S. (2005). Painted leaves, context, and semantic analysis. Linguistics and Philosophy, 28,
351–374.
Prinz, J. J. (2002). Furnishing the mind: Concepts and their perceptual basis. Cambridge, MA:
MIT Press.
Prior, A. N. (1960). The runabout inference ticket. Analysis, 21, 38–39.
Prior, A. N. (1976). Thank goodness that’s over. In P. T. Geach & A. Kenny (Eds.), Papers on logic
and ethics (pp. 78–84). London: Duckworth.
Recanati, F. (1997). The dynamics of situations. European Review of Philosophy, 2, 41–75.
Recanati, F. (2002). Unarticulated constituents. Linguistics and Philosophy, 25, 299–345.
Recanati, F. (2004). Literal meaning. Cambridge, England: Cambridge University Press.
Richard, M. (2004). Contextualism and relativism. Philosophical Studies, 119, 215–242.
Sellars, W. (1974). Meaning as functional classification. Synthèse, 27, 417–437.
328 J. Dokic
Sperber, D., & Wilson, D. (1998). The mapping between the mental and the public lexicon.
In P. Carruthers & J. Boucher (Eds.), Language and thought (pp. 184–200). Cambridge,
England: Cambridge University Press.
Stanley, J. (2000). Context and logical form. Linguistics and Philosophy, 23, 391–434.
Stanley, J. (2002). Making it articulated. Mind & Language, 17, 149–168.
Wilson, D., & Sperber, D. (2002). Truthfulness and relevance. Mind, 111, 583–632.
Wilson, R. (2004). Boundaries of the mind: The individual in the fragile sciences. Cambridge,
England: Cambridge University Press.