. 24
( 62 .)


wide. And my belief that Paris is the capital of France remains my belief
even though it has a wide realization. I think that this is also true even
in cases of radically wide realization, paradigmatically in those involving
social actions: Jane™s signing a check is her action because the core re-
alization of that action is realized in large part (even if not wholly) by
something that she does, such as moving her pen-grasping hand over
a piece of paper. To my mind this is a desirable feature of my view of
Individualism and Externalism in Cognitive Sciences

mental properties in particular because it preserves what is right about
an individualistic view of subjectivity and in so doing readily allows for
both third- and ¬rst-person perspectives on the mind.
One might well challenge this aspect of my view as unnecessarily and
unjusti¬ably conservative. Unnecessarily, for once realization goes wide
surely we are on our way to undermining subjectivity and the misplaced
position of privilege that the individual subject has in our thinking about
the mind. And unjusti¬ably, since in at least some cases of wide realiza-
tion, particularly those of radically wide realization, there is no non-
arbitrary way to single out individuals as the subjects or “owners” of the
corresponding mental properties. If we have wide realizations of mental
states, and thus wide mental states, so too we should have “wide subjects”
of those states. Andy Clark and David Chalmers suggest something like
this view of the self as a consequence of their endorsement of what they
call “the extended mind,” a view that Clark has more recently applied in
his thinking about rationality and human intelligence.16
Reply: There may, of course, be interesting science ¬ction or other fan-
ciful examples that pull our intuitions toward such radical conclusions,
but it is important not to lose sight of the fact that, at least in the world
that we actually inhabit, and being the creatures that we actually are,
there is a basis for marking out individuals as the subjects of properties,
even those properties with wide realizations. Individuals “ and here, as
always, our paradigms are individual people and individual organisms “
are spatio-temporally bounded, relatively cohesive, uni¬ed entities that
are continuous across space and time. Recall that the possibility of wide
systems was modeled on the actuality of systems that formed part of such
individuals as exempli¬ed by the variety of physiological systems theo-
rized about in biology and medicine. While these narrow systems (for
example, the circulatory system) share some of the features that make
individuals metaphysically distinctive and certainly have their own prop-
erties, they are not themselves individuals, and it seems strained or at best
derivative to view them as the subjects of the sorts of properties that we
would intuitively ascribe to the individuals they constitute. For example,
the visual system and its parts can be lesioned, can have imbalances in lev-
els of neurotransmitters, and have certain of its pathways blocked (either
experimentally or “in the wild”). But it is the individual who perceives,
who suffers from a visual agnosia, who experiences a hallucination. The
same is true of wide systems, and this provides a principled basis for as-
cribing mental properties in particular to individual subjects rather than
the wide systems of which those subjects are a part. In the actual world, it
Context-Sensitive Realizations 143

is individuals who form and maintain beliefs, experience emotions, and
wonder about what will happen next, even if those individuals form part
of what I have called folk psychological systems.

10 putting our metaphysics to work
The chief goal in this chapter and the last has been to develop a meta-
physics suitable for an externalist view of psychology. Importantly, the
argument here is not driven by a consideration of the special nature of
mental properties but, rather, by re¬‚ection on a range of properties from
the cognitive, biological, and social sciences.
Although the considerations here have been metaphysical, the exter-
nalist metaphysics developed is one with explicit ties to a strategy of ex-
planation, integrative synthesis. The explanatory strategy of constitutive
decomposition is well entrenched in a range of physical sciences, and
philosophers of the cognitive sciences have been quick to point to ways
in which it can be and has been used with respect to cognitive capaci-
ties. Integrative synthesis, while not quite novel, has a patchier history,
particularly in the physical sciences. But again, if I am correct about the
signi¬cance of the notion of wide realization, then we should expect in-
tegrative synthesis to ¬nd a central place in the sciences, including the
fragile sciences.
In the following chapters we will see where all of this philosophical
footwork leads us. In the next chapter I consider externalist views of the
mind in action in the cognitive sciences, extending this to noncomputa-
tional approaches to cognition in Part Three.

Representation, Computation, and Cognitive Science

1 the cognitive science gesture
We saw in Chapter 4 that as individualism was coming under attack in the
late 1970s, it was also being defended by Jerry Fodor and Stephen Stich
as a view of the mind particularly apt for a genuinely scienti¬c approach
to understanding cognition. In contrast to the original externalist papers
of Putnam and Burge, those in which methodological solipsism and the
principle of autonomy were introduced focused on the relevance of indi-
vidualism for explanatory practice in cognitive science. They appealed to
the computational nature of cognition in arguing that cognitive science
should be individualistic, and for substantive conclusions about its scope
and methodology.
These early individualist arguments of Stich and Fodor invoked what
I shall call the cognitive science gesture. They primarily pointed to general
features of cognition and theory in cognitive science that, they claimed,
revealed their individualistic nature. Cognition was computational (and
computation was individualistic), or cognitive processing was mechanistic
(and such mechanisms were individualistic). Neither used a sustained,
detailed examination of particular theories and explanations in cognitive
science to argue that they were or must be individualistic. Instead, Fodor
and Stich were, at least initially, content with a more abstract, general
gesture toward features they took to be central to cognitive science.
The gesture toward the developing cognitive sciences in defending
individualism served both to motivate and to buttress more purely philo-
sophical considerations in favor of individualism, such as appeals to func-
tionalism, physicalism, and causal powers, and the idea that it was narrow

Representation, Computation, and Cognitive Science 145

content that was truly explanatory of cognition. We have also seen how
the standard view of realization reinforced the idea that the metaphysics
of the mental pushed one to an individualistic view of the mind. Thus, a
happy con¬‚uence: A range of broadly accepted views of the place of the
mind in the world and empirical practice in our nascent sciences of the
mind both imply individualism about the mind. The rejection of indi-
vidualism was to come both at the expense of an acceptable metaphysics
and at the cost of forgoing the prospects for a thoroughly naturalistic
treatment of the mind.
My chief aim in this chapter will be to undermine this view of the rela-
tionship between individualism and cognitive science. After outlining the
accepted view of individualism and the cognitive sciences in sections 2“3,
in sections 4“7 I turn to reshape the received wisdom here, using David
Marr™s celebrated theory of vision as a focus. At the core of my discus-
sion are the notions of exploitative representation and wide computationalism.
With an alternative vision of representation and computation in hand, I
shall conclude by considering some recent examples within the cognitive
sciences that exemplify externalism in practice.

2 individualism in cognitive science
Although the cognitive science gesture is a gesture “ rather than a solid
argument that appeals to empirical practice “ it is not an empty gesture.
Even if the original arguments of Fodor and Stich did not win widespread
acceptance amongst philosophers, they struck a chord with psychologists,
linguists, and computer scientists. Indeed, the dominant research tra-
ditions in cognitive science have been at least implicitly individualistic.
Consider three expressions of individualism that bring out its attractions
for many cognitive scientists.
One attraction of individualism is its perceived connection to the rep-
resentational theory of mind, which holds that we interact with the world
perceptually and behaviorally through internal mental representations of
how the world is (as the effects of perceiving) or how the world should be
(as instructions to act). Ray Jackendoff expresses the connection between
such a view and individualism when he says:
Whatever the nature of real reality, the way reality can look to us is determined
and constrained by the nature of our internal mental representations. . . . Physical
stimuli (photons, sound waves, pressure on the skin, chemicals in the air, etc.) act
mechanically on sensory neurons. The sensory neurons, acting as transducers in
Pylyshyn™s (1984) sense, set up peripheral levels of representation such as retinal
Individualism and Externalism in Cognitive Sciences

arrays and whatever acoustic analysis the ear derives. In turn, the peripheral rep-
resentations stimulate the construction of more central levels of representation,
leading eventually to the construction of representations in central formats such
as the 3D level model. . . .

Jackendoff calls this view the “psychological” (versus philosophical) vi-
sion of cognition and its relation to the world. His central claim in this
paper is that only the psychological vision directs us to viable research pro-
grams in cognitive science. Provided that the appropriate, internal, rep-
resentational states of the organism remain ¬xed, the organism™s more
peripheral causal involvement with its environment is irrelevant to cog-
nition, since the only way in which such causal involvement can matter
to cognition is by altering the internal mental states that represent that
Jackendoff™s skepticism about the “philosophical” vision parallels the
disdain for “philosophical” approaches to language that Noam Chomsky
has expressed in drawing the distinction between the “I-language” and the
“E-language,” and Chomsky™s insistence that only the former is suitable
as an object of scienti¬c study. Chomsky takes a conception of language
as “internal,” “intentional,” or “individual,” as opposed to “external,”
“extensional,” or “social,” to be a condition of a serious, empirical in-
vestigation of language, to be what makes the question of language use
and acquisition a problem rather than a mystery. The I-language is the
generative procedure inside the individual that is causally responsible for
that individual™s linguistic output, and it is what linguists attempt to re-
construct when they postulate features of a universal grammar. Given
Chomsky™s nativism about language in particular and cognition more
generally, and in light of the relationship between strong nativism and
individualism that I argued for in Chapter 1, we could expect Chomsky
to be critical of externalist views of the mind. Indeed, in “Language and
Nature” Chomsky turns directly to the philosophical tradition of exter-
nalism, beginning with the Putnam-Burge thought experiments, to argue
for the poverty of the resulting conception of language and how it should
be studied, a critique extended in his New Horizons in the Study of Language
and Mind.2
To take a third example of an individualistic perspective on cogni-
tion, consider this extract from the Foreword to Simon Baron-Cohen™s
Mindblindness, written by Leda Cosmides and John Tooby:

Although it is a modern truism to say that we live in culturally constructed worlds,
the thin surface of cultural construction is dwarfed by (and made possible by)
Representation, Computation, and Cognitive Science 147

the deep underlying strata of evolved species-typical cognitive construction. We
inhabit mental worlds populated by the computational outputs of battalions of
evolved, specialized neural automata. They segment words out of a continual
auditory ¬‚ow, they construct a world of local objects from edges and gradients
in our two-dimensional retinal arrays, they infer the purpose of a hook from its
shape, they recognize and make us feel the negative response of a conversational
partner from the roll of her eyes, they identify cooperative intentions among indi-
viduals from their joint attention and common emotional responses, and so on.3

While Cosmides and Tooby clearly do assign the environment of the
organism a role in the evolutionary history of species-typical capacities,
the cognitive capacities themselves are individualistic.
These individualistic conceptions of the mind ultimately shape the
direction of research within the cognitive sciences through their effect
on how central notions in the ¬eld are construed. The most central of
these are representation and computation.

3 mental representation as encoding
Underlying the visions of cognitive science exempli¬ed by Jackendoff,
Chomsky, and Cosmides and Tooby are what I will call encoding views of
mental representation. Simply put, encoding views hold that to have a
mental representation, M, is for M to encode information about some
object, property, event, or state of affairs m. A well-known, protean ver-
sion of the encoding view is the picture or copy theory of mind, where to
have a mental representation of m is to have a mental picture or image
of m in your head, where the picture is “of m” just because it looks like
m. A version of the encoding view more prevalent in cognitive science
and that builds more elaborately on the idea that cognitive processing
takes place in a sort of code is the language of thought hypothesis: To
have a mental representation of m is to have a token in your language of
thought, M, that stands for or refers to m. Unlike the copy theory of men-
tal representation, on this view there need be no resemblance between
representation and represented, with the role that resemblance plays in
the picture theory being played by some putatively natural relation, such
as causation or teleology. At the time that the debate between individ-
ualists and externalists was being articulated, the language of thought
hypothesis laid some claim to be “the only game in town” for understand-
ing mental representation and computation.
On either view, discrete mental representations encode informa-
tion about particular aspects of the world. Existing in a code, mental
Individualism and Externalism in Cognitive Sciences

representations are governed by a kind of syntax that determines how ba-
sic units in the code can be manipulated, combined, and decomposed.
Thus, cognitive scientists can and should explore the properties of repre-
sentations and the rules that govern their internal dynamics, rather than
the relationships that exist between organisms and environments.
There are two prima facie problems with encoding views of mental
representation. These problems constitute not so much knockdown ar-
guments against such views as open the way to reconceptualize represen-
tation in ways amenable to externalist visions of cognition.
The ¬rst is that although there may have been a time at which encoding
views were the only game in town, over the last twenty years a number of
views of mental representation have been developed that break from the
encoding tradition. One of these is the notion of distributed representa-
tion. On this view of representation, what is in the head are not discrete
symbols, each encoding their own piece of information, but less content-
laden nodes which, in combination with the connection strengths linking
them, collectively represent information about the world. A related alter-
native conception of representation is that of subsymbolic computation,
whereby the units over which the computations are de¬ned, the repre-
sentations, are not themselves symbols (that is, codes). Both of these
conceptions of representation were developed within a connectionist
framework, but they have a basis in dynamic approaches to cognition
more generally. What these views share is the idea of thinking about rep-
resentation as ¬‚eeting, situated, dynamic, and interactive, and as such
they mark a departure from encoding views of representation.4
These alternatives to encoding views constitute an important break
in the grip that those views have had on how cognitive scientists have


. 24
( 62 .)