Upload
lydieu
View
217
Download
0
Embed Size (px)
Citation preview
Neo-Empiricism and Intentionality
Steven Gamboa
Dr. Steven GamboaAssistant Professor of PhilosophyDepartment of Philosophy and Religious StudiesCalifornia State University, Bakersfield9001 Stockdale Hwy., Bakersfield, CA, 93311-1022
1
Abstract
A revival of empiricist theories in cognitive psychology, neuropsychology, and philosophy has been led by figures such as Antonio Damasio (1994), Lawrence Barsalou (1999), George Lakoff (1987), and Jesse Prinz (2002, 2004). Their work has served to connect familiar empiricist approaches to thought and reason with contemporary cognitive psychology and neuroscience. The work of Jesse Prinz is of special philosophical significance since it aims to bring together the work of neo-empirically minded theorists in the cognitive and neuro-sciences with main themes found in contemporary philosophical theories of intentionality and reference. In this paper, I examine Jesse Prinz’s efforts to synthesize a neo-empiricist theory of concepts with contemporary semantic theories of reference and intentionality. In part one, I analyze Prinz’s approach in some depth. In part two, I raise a question concerning the origins of intentionality. Specifically, I am interested in examining the minimum cognitive prerequisites for intentionality within Prinz’s theory of perception-based representation. In part three, I raise a problem case for Prinz’s account of the requirements for intentionality, and propose an adjustment in Prinz’s account to meet the challenge of the objection.
Keywords: neo-empiricism, concepts, intentionality, objectivity, Jesse Prinz, Tyler Burge
The last two decades have witnessed a revival of empiricist theories in cognitive
psychology, neuropsychology, and philosophy. Key figures in the resurgence of
empiricism include Antonio Damasio (1994), Lawrence Barsalou (1999), George Lakoff
(1987), and Jesse Prinz (2002, 2004). Their work has served to connect familiar
empiricist approaches to thought and reason with the most up-to-date work in cognitive
psychology and neuroscience. What unites these theorists (and what connects neo-
empiricism to its 17th and 18th century classical antecedents) is a commitment to the thesis
that all of our conceptual representations are perceptually based. In other words, neo-
empiricists claim that the same cognitive resources used in perception are utilized for
2
thinking as well. As a corollary, they deny that there are any conceptual representations
encoded in some distinct, non-perceptual system of representation.
Many philosophers will welcome the revival of empiricist theories of concepts,
especially those who viewed with suspicion the strongly rationalist orientation of much
cognitive psychology and philosophy in the years since Chomsky launched the cognitive
revolution. For philosophers sympathetic to neo-empiricism, the work of Jesse Prinz is
of special significance since it aims to achieve a synthesis, bringing together the work of
neo-empirically minded theorists in the cognitive and neuro-sciences with main themes
found in contemporary philosophical theories of intentionality and reference.
In this paper I examine Jesse Prinz’s efforts to marry a neo-empiricist theory of
concepts with contemporary semantic theories of reference and intentionality. In part
one, I analyze Prinz’s approach in some depth, and take note of its many virtues. In part
two, I raise a question concerning the origins of intentionality. Specifically, I am
interested in examining the minimum cognitive prerequisites for intentionality within
Prinz’s theory of perception-based representation. This question is of considerable
independent interest, but also provides a valuable means for exploring the options
available to a neo-empiricist account of the cognitive prerequisites for intentionality.1 In
part three, I raise a problem case for Prinz’s account of the requirements for
intentionality. I propose an adjustment in Prinz’s account to meet the challenge of the
objection, and consider whether the modification causes unwanted disturbances
elsewhere in Prinz’s account. I conclude that the modified account of the minimum
requirements for intentionality do not conflict with other major components in Prinz’s
philosophical project.
3
Part 1: Prinz’s synthesizing project
Furnishing the Mind is dedicated to developing and refining what Prinz calls
the “proxytype theory” of concepts. In describing proxytype theory, I will focus first on
what proxytypes per se are, and then discuss their semantic properties. Prinz defines
proxytypes as “perceptually derived representations that can be recruited by working
memory to represent a category.” (2002, p. 149) The definition requires unpacking. First,
that proxytypes are perceptually derived makes clear Prinz’s debts to the empiricist
tradition of imagism and the principle of “nihil in intellectu nisi prius in sensu”.
Imagism has long been out of philosophical favor, but that may in large part be due to its
association with an antiquated view of perception as something akin to the conscious
introspection of pictures in the head. More contemporary accounts of perception from
cognitive science offer a much wider range of highly structured representations for
empiricists to work with. For Prinz, perceptual representations are the products of
dedicated sensory input systems, but these can include dynamic, multimedia
representations that converge information from auditory, visual, tactile, and other sensory
modalities.
Secondly, Prinz’s definition establishes that proxytypes are recruited from long-
term memory networks. In our cognitive activities, we make use of an enormous amount
of diverse information. A crucial theoretical role for concepts is to encode and organize
this information. Prinz argues that all information organized conceptually can be
captured by links between perceptual representations.2 Prinz posits a number of links
(hierarchy, transformation, binding, situational, predicative) that serve to bind together
4
perceptually derived representations in a long-term memory network. The links are
established, in true empiricist fashion, on the basis of patterns of co-instantiation and co-
occurrence revealed in experience. For example, if one experienced a barking dog, a
visual representation of the dog will get grouped together via the binding link with the
auditory representation of the barking sound, such that henceforth these co-instantiated
features are linked in long-term memory.
The third component in Prinz’s definition of proxytypes concerns the limits
imposed by working memory. Our long-term memory networks contain myriad
perceptual representations linked in complex, overlapping, and connectionist ways. But
when one is engaged in thinking about some category, say dogs, it would be impossible
for the entire network of dog-related information to get activated since that would
seriously overload our working memory. One might be tempted at this point to introduce
amodal, non-perceptual symbols (indicators rather than detectors) to play the role of
concepts, but Prinz argues that such a move is unnecessary. There’s a simpler solution:
simply use a stored perceptual representation as a proxy for the entire category. On this
view, thinking is a process of re-enacting and manipulating perceptual states, and
proxytypes are used to stand in for (and thus represent) the objects involved in the
simulated scene.
Proxytype Theory is not exhausted by the claim that concepts are constituted by
proxytypes. Key to Prinz’s broader synthesizing project is his adoption of a ‘two factor’
account of conceptual content: proxytypes carry two distinct kinds of content, cognitive
and intentional. Prinz’s distinction between cognitive and intentional content
corresponds to the perhaps more familiar distinction drawn by many philosophers
5
between the ‘narrow’ and ‘broad’ content of mental states. Cognitive (narrow) content
refers to the sort of meaning that supervenes on the individual’s psychological states,
such that duplicates necessarily share all the same cognitive (narrow) content. Intentional
(broad) content refers to the sort of meaning that is not determined by what is in the head
because it includes those objects and kinds in the world that the concept picks out or
refers to. Cognitive psychologists have been more interested in the narrow meaning or
cognitive significance of concepts, whereas philosophers have been more concerned with
the intentional or broad relation between concepts and objects in the world. If successful,
Prinz’s two-factor account would bring together an impressive body of results on the
cognitive content of concepts in psychology with contemporary philosophical theories of
intentionality and reference.
The advantages of a two-factor theory of cognitive content are considerable.
First, it handles familiar Frege Hesperus/Phosphorus (1893) cases in a satisfying way:
while the concepts of <Hesperus> and <Phosphorus> have the same intentional content
(they both refer to the same heavenly body, viz., the planet Venus), the two concepts
have different cognitive contents and are thus distinguishable on that basis. Likewise,
Putnam’s ‘twin Earth’ case (1975) is more grist for the mill. In Putnam’s thought
experiment, the stuff that flows in rivers and fills oceans and lakes on twin Earth, while to
all appearances just like our familiar Earthly water, has a different molecular structure
from water on Earth (XYZ rather than H2O). In such a scenario, the Earthly and Twin-
Earthly <water> concepts have different intentional contents (since they refer to different
kinds), but they have the same cognitive content (which helps account for the intuition
6
that there is at least some sense in which duplicates must share the same beliefs and other
mental states).
Assuming that concepts bear two kinds of content, how do they relate?
According to the familiar Fregean line, sense determines reference. Using Prinz’s
terminology, we would say that a concept’s cognitive content determines its intentional
content. But Prinz doesn’t follow the Fregean model. Instead of cognitive content
determining intentional content or vice versa, Prinz offers independent accounts of the
factors that determine the semantics for each dimension. There are a number of
advantages to keeping these two accounts independent, but the most important for Prinz
is that by doing so we can avoid placing any epistemic conditions on successful
reference. Prinz captures the point this way:
Ordinary people believe that they can refer to a category even if they cannot distinguish it from another categories [sic] whose members are superficially similar. This is a fundamental antiverificationist conviction. Limits on one’s ability to distinguish things should not carry limits on one’s ability to refer to things. Recent philosophical efforts to find an adequate theory of intentionality can be regarded as efforts to make sense of antiverificationist convictions. A good theory makes sense of those convictions by making them come out true. (2002, p. 260-261)
Decoupling intentional content from cognitive content means that Prinz can avoid placing
any epistemic conditions on successful reference, such that one can successfully refer to
water without knowing anything about its chemical composition or to whales despite
believing that they’re fish.
Prinz straightforwardly identifies the cognitive content of concepts with the
proxytypes themselves. Two individuals have concepts with the same cognitive content
when they have type-identical proxytypes. Proxytypes are type-identical when they are
constituted by sensory detection mechanisms cued to the same sets of appearances. For
7
example, two individuals share the same dog concept (in the sense of cognitive content)
if it is activated by the same set of dog detecting appearances (e.g., furry, barks, fetches,
etc.). In such a case, not only do the two individuals track the same category (assuming
they do), they conceptualize it in the same way as well.3
For intentional content, Prinz endorses a modified covariational (aka,
informational semantic) theory of intentionality, influenced mainly by the work of
Dretske (1981).4 Covariationist accounts traditionally work on the assumption that
concepts are amodal, unstructured symbols (as in Fodor’s informational atomism or
Dretske’s indicator semantics), but Prinz argues that there are no compelling reasons why
structured proxytypes couldn’t play that role directly, i.e., enter into reference-endowing
co-variation with individuals and kinds, eliminating the need for amodal symbolic
middlemen.
Assuming that concept C is a proxytype, Prinz offers the following conditions for
C’s intentional content:
X is the intentional content of C iff
1. X’s nomologically covary with tokens of C, and 2. an X was the incipient cause of C.
An example to flesh out the definition: my <C> concept refers to carrots because it is
reliably tokened when in the presence of carrots. It doesn’t refer to vegetables (even
though all carrots are vegetables) because vegetables wouldn’t cause <C> to be tokened
in proximate counterfactual worlds that lacked carrots. Prinz’s definition requires
nomological covariance, which is a counterfactual supporting relation. Importantly, <C>
doesn’t refer to carrot appearances (or the retinal image caused by carrots) because <C>
would still be tokened in a proximate counterfactual world where carrots looked different
8
(and thus cause different retinal images). In this case, the concept <C> would have a
different cognitive content in the proximal counterfactual world since it would use
different appearances to track carrots, but it would have the same intentional content and
thus be ‘the same concept’ in that sense. And since carrots were the incipient cause of
my forming <C>, it refers to carrots even if I can’t tell the difference between carrots and
parsnips and so routinely mistake parsnips for carrots.5
While the identity conditions for cognitive and intentional contents are
independent in the sense laid out above, each sort of content can be said to refer to a type:
the cognitive content picks out appearances, while the intentional content picks out kinds
and individuals in the world. And these two types are obviously connected in the
psychological explanation of human cognition and behavior: we use appearances
(cognitive content) to track individuals and kinds in the world (intentional content).6
Part 2: minimum requirements for intentionality
One can generalize Prinz’s definition so as to leave open the question of what sort
of states can be bearers of intentional content: a state M has intentional content X iff M
nomologically co-varies with instances of X. The general formulation raises the
following question: what sort of M states are suitable candidates for entering into
intentionality-conferring nomological co-variation? I’ll review a number of prominent
candidates for necessary conditions for intentionality, including propositional attitudes,
concepts, and perceptions. As we’ll see, Prinz rejects each of these more demanding
candidates. Taking the long way through the alternatives that Prinz rejects is worth the
effort since it will help make sense of the positive position he eventually endorses.
9
One popular proposal for distinguishing internal states capable of bearing
intentional content from mere sensory reactions is to claim that genuine intentionality can
occur only within a psychology that has propositional attitudes, i.e., beliefs and desires.
Dretske (1988, 1995) is representative of this approach. Dretske distinguishes mental
representations capable of bearing intentional content from mere sensory registrations by
claiming that the former are found only in cases where the organism can learn. Learning
requires categorization, and categorization for Dretske requires propositional judgment.
Concerned with the over-application of intentionality to inappropriate subjects, Dretske
requires that the organism not merely be directed toward its goal by means of sensory
registration of environmental cues and innate behavioral response, but that it be directed
in order to get its goal via appropriate beliefs and desires, including instrumental beliefs
to the effect that the current activity is a way to achieve the goal (1988, p. 116).
A similar take on intentionality is found among contemporary philosophers
working on the evolution of cognition, such as Peter Godfrey-Smith (1996) and Kim
Sterelny (2001, 2003). They link the evolution of intentional agency to the emergence of
a distinction among the organism’s cognitive states between belief-like representations
and desire-like representations of preference. In the following passage, Sterelny
identifies what he takes to be the key difference between genuine intentional agency and
sub-intentional responses (whether innate or learned) to environmental cues:
[What makes genuine intentional agency different is that] “within our mental representations there is a distinction between reports and instructions. Intentional agents have decoupled representations. That is to say, we have internal states that track aspects of our world, but which do not have the function of controlling particular behaviors. Beliefs are representations that are relevant to many behaviors, but do not have the biological function of directing any specific behavior.” (Sterelny, 2003, p. 29)
10
Sterelny contends that the capacity to generate representations decoupled from any fixed
behavioral response is the cognitive prerequisite for the sort of flexible, non-stereotyped
mode of connectedness to the world that we identify as intelligent action. Only in cases
of intelligent action is attribution of intentional content properly motivated and
constrained. Deciding in any particular case whether an organism’s behavior qualifies as
intelligent action can be difficult, and disputes over competing explanations (cognitively
sophisticated vs. deflationary) are common and often empirically rather intractable. But
cognitive ethologists have also identified a plethora of examples of flexible, intelligent
behaviors among non-human animals (principally primates and other social mammals,
but also many bird species) for which we are clearly warranted in positing decoupled
representations and employing something analogous to belief/desire explanations.
The claim that perceptual representations embedded in explanations that invoke
propositional attitudes carry intentional content clearly succeeds in identifying a
sufficient condition for intentionality. But is it a necessary condition? In other words, is
belief required for intentionality? Given Prinz’s avowed anti-verificationist inclinations,
it is not surprising that he rejects the notion that supporting beliefs and other
propositional judgments are required for intentionality. If ties to beliefs and other
propositional attitudes were required for intentionality, then beliefs would be required for
concepts to have intentional content. Prinz rejects this dependency: “In fact, there is no
reason to think that concept possession requires having any beliefs at all. Concepts seem
to be ontogenetically and semantically prior to beliefs. Any account that tries to define
the content (intentional or narrow) of concepts in terms of beliefs will have to prove this
priority illusory.” (2002, p. 269)
11
The natural next step is to consider concepts as the most basic bearers of
intentional content. In Furnishing the Mind (p. 197, 2002) and, more extensively, Gut
Reactions (pp. 45-49, 2004), Prinz offers the resources for developing a proposal along
these lines. Prinz’s stated aim in the relevant passages is to distinguish concepts from
both percepts and emotions. If intentionality did necessarily involve concepts, a theory
that identified the conditions for concept possession could serve double duty as a theory
that identified constraints on which mental states could count as genuinely intentional.
As noted above, concepts are theoretical entities posited to explain higher order
cognitive abilities, or ‘cognition’ for short. As Prinz uses the term, ‘cognition’ refers to a
class of conscious mental processes or operations, but not all conscious mental processes
count as cognitions. To distinguish cognition from other mental faculties, Prinz invokes
the Kantian distinction between spontaneity and receptivity: “Cognitive states and
processes are those that exploit representations that are under the control of the organism
rather than under the control of the environment.” (2004, p. 45) While we spontaneously
engage in cognition (and it is thus under our endogenous control), perception and
emotion are largely a response to (and thus controlled by) exogenous stimuli from the
environment (including our bodies). Organismic control is dispositional; your <dog>
concept counts as a concept because you “can willfully form thoughts using your dog
concept”, even if its activation on any given occasion is prompted by an encounter with a
dog in the environment.
Prinzean “organismic control”, with its talk of conscious willing and deliberate
activation, requires naturalistic explanation to avoid homuncular regress (i.e., explaining
cognition in terms that presuppose cognition itself). To do so, Prinz contends that states
12
being under organismic control can be naturalistically explicated in terms of
representational states being under the “top-down control” of specific psychological
systems and neural structures. Prinz cites work in cognitive neuroscience that identifies
executive centers in the prefrontal cortex as playing the key role in manipulating and
maintaining representations in working memory. Prinz speculates that these executive
centers “may also play a special role in initiating the formation or activation of
representations when those representations are not elicited by environmental
stimulation.” (2004, p. 47) Prinz proposes that cognition be defined as a state that
“includes representations that are under the control of structures in executive systems,
which, in mammals, are found in the prefrontal cortex.” (2004, p. 47)
Given this understanding of concepts as “representations under organismic
control”, does Prinz see concepts as the most primitive bearers of intentional content?
Furnishing the Mind might leave that impression (since his discussion of intentional
content in that work is focused solely on concepts), but in his article “Beyond
Appearances” (2006), Prinz rejects the suggestion that concepts are strictly required for
intentionality. As with belief, conceptualization is sufficient for intentionality, but it is
not necessary. The sort of case Prinz asks us to consider involves perceptual recognition
without conceptualization:
Concepts are representations that can be actively tokened by an organism. In recognition, we often use representations that can only be tokened passively. I can recognize certain things that I cannot bring readily to mind in imagination or reflection. Recognition outstrips conceptualization. (2006, p. 437)
Face recognition offers a clear case where recognition outstrips conceptualization in that
we can recognize many more faces than we can deliberately recall in imagination and
13
reflection. The percepts stored in mental files may not be under endogenous control, but
they can be triggered by current perception. When such matching occurs, the current
percept inherits the semantic properties of the mental file, including its referential
content.
While Prinz argues that mere percepts can refer, even without the involvement of
concepts, some sort of recognition seems required for intentionality. Given Prinz’s views
on the conditions that determine intentional content, recognition of some sort looks
promising as a baseline minimum requirement for objective reference. In order for
Prinz’s nomological co-variation account of intentionality to apply in any given case,
there need to be mental files of percepts stored in long-term memory networks set up
(either by evolution or learning) to be set off by individuals and kinds in the world.
In the neurological hierarchy of perceptual processing, sensation precedes
recognition. So it would seem to follow from the recognition requirement that we cannot
literally sense tigers and dogs; we can at best sense their appearances, but sensations
themselves would lack objective reference. Nonetheless, Prinz contends that sensations
can refer to things beyond appearances. Sensations for Prinz are representations in
dedicated input systems that are consciously experienced. Current cognitive neuro-
scientific theorizing places conscious perception at an intermediary level between low-
level subsystems processing information from sensory receptors and high-level
subsystems that utilize relatively context invariant representations for object recognition.
As such, sensation does not require recognition. However, perceptual processing is not a
one-way street. Neural pathways allow for significant back propagation of
representations from high level perceptual processing subsystems into the intermediate
14
level where conscious sensation occurs, thereby enriching the representational potential
of sensations.
After back-projection, sensations are no longer purely bottom-up. They are blends of incoming signals and mental images produced by centers further down the processing stream. The representations used downstream, which have been matched with representations stored in memory, can inherit semantic properties from the stored representations. […] Sensations can take on new meaning once they intermingle with representations coming down from on high. (2006, p. 456)
Despite initial appearances, sensations can support objective reference and intentional
content once they are blended with higher order representations. The intentional content
of high-level perceptual representations can be exported to earlier processing subsystems
so that sensations can refer to objects in the world: “If representation amounts to
functional detection, then representation goes all the way down.” (2006, p. 456)
Part 3: Assessment
How should we assess the viability of Prinz’s position? It’s hard to escape the
suspicion that, in rejecting the more cognitively demanding proposals as identifying
sufficient but not necessary conditions for intentionality, Prinz has over-corrected by
offering necessary but not sufficient conditions. We can illustrate this point by
considering the widespread practice of taxis in the animal world. A taxis is an innate
directional response to a detected stimulus gradient. A mosquito uses its ability to detect
and differentially respond to carbon dioxide concentrations (chemotaxis) in its blood-host
search behavior. Protozoa of the genus Euglena use their eyespot and photosensitive
structures to detect light intensity and direction, and use their flagellum to move towards
the light source (phototaxis) to enhance photosynthesis. The point of considering such
15
behaviors is not to rehearse well-worn objections to naturalistic accounts of
intentionality. Indeed, I do not see these cases as counter-examples to nomological co-
variance definitions of intentional content per se. Important here is the distinction
Stalnaker (2003) draws between the conditions for determining the intentional content of
a suitably qualified internal state (a task for semantics), and the conditions for
determining whether a given internal state is qualified for bearing intentional content (a
task for cognitive theory). Thus, even if we maintain Prinz’s semantic theory unchanged,
we still need to enquire as to the requirements for internal states to be capable of carrying
intentional content at all. Taxis behavior makes vivid the need for co-variational theories
of intentionality to offer minimum criteria for internal states to qualify as capable of
bearing intentional content. The type of sensory registration of environmental stimulus
involved in taxis nomologically co-varies with environmentally important (to the
organism) kinds (e.g., blood-hosts for mosquitoes, light sources for euglena). Without
some basis by which to distinguish tactic sensory registrations from states capable of
genuine intentionality, we would have to grant that they do bear intentional content.
There is a strong presumption against attributing intentionality (and thus heavy
duty semantic properties such as reference) to the internal states of such unlikely
organisms as protozoa and mosquitoes; while the organisms can clearly detect and
differentially respond to environmental cues, these states remain sub-intentional. Nor is
this just a bias of common sense opinion against allowing such ‘primitives’ into the
intentionality club. Scientific explanation of tactic behavior need not make any appeal to
representations of distal objects, such that attributions of intentionality would be
superfluous. Further, a certain type of semantic indeterminacy is a fairly sure sign that
16
the attribution of intentionality and reference doesn’t fit well. In the case of the mosquito
using its ability to sense and track carbon dioxide concentrations in its blood-host search
behavior, we can characterize its state in the indicative mode as saying “blood host
ahead”, but an equally reasonable alternative is to interpret the state in the imperative
mode as delivering something like a command to its motor control system to “fly in that
direction”. The sort of indicative/imperative indeterminacy illustrated in the mosquito
case is reminiscent of Millikan’s (1996) discussion of ‘pushmi pullyu’ representations
that combine both detector (descriptive) and motor command (directive) functions in one
state. In my view, such states are representations in a sense, but their semantic
indeterminacy makes them incapable of supporting attributions of reference.
Consideration of taxis behaviors motivates the demand that Prinz’s theory of
intentionality provide a basis for distinguishing genuinely intentional states from mere
sensory detectors. The cost of not doing so would be a severely degraded notion of
intentionality. After rejecting a series of more cognitively demanding proposals, Prinz
has settled on the following baseline prerequisite for intentionality: objective reference
requires the involvement of representations stored in high-level perceptual sub-systems,
but these representations can exert their influence at early stages of perceptual
processing, such that perceptual and sensory states can bear intentional content.
Clearly, the states of protozoa during phototaxis, and of eurkaryotes and bacteria
during chemotaxis, will not qualify as carrying intentional content under Prinz’s criteria;
since these organisms lack a central nervous system, there are no mental files stored in
higher-level neural subsystems on hand to contribute their semantic properties. But other
cases don’t offer such simple solutions. It’s not a matter of merely separating organisms
17
with sophisticated neural architecture from those lacking such advantages. The same
organism can engage in some behaviors that warrant intentional characterization of their
internal states, and other behaviors that don’t. In his (2010), Tyler Burge gives much
attention to the olfactory homing behavior of salmon. After years in the open ocean,
mature salmon make an extraordinary return to their fresh water natal stream to spawn.
This amazing feat is accomplished roughly as follows: during the smolting period,
salmon "imprint" to the chemical properties of their natal tributary; on the return journey,
salmon use their incredible olfactory sensitivity and serial sampling of the environment to
find their way back to the natal stream. Extensive scientific study of salmon homing
behavior (see Hasler and Scholz, 1988) has established that homing depends on
imprinting of the cues that identify the home stream at the time the young salmon begin
their downstream migration. After three or four years at sea, the salmon recall what they
learned as smolts from their long-term memory.
Salmon homing behavior appears to instantiate all of Prinz’s requirements for
intentionality. A representation from a mental file stored in long-term memory is
matched with a sensory input. The stored olfactory representation nomologically co-
varies with the natal stream, and was functionally set up to be set off by that particular
stream. In a counter-factually proximate world where the stream had a different chemical
profile (a different olfactory appearance), the salmon would still use its olfactory
detectors to home towards the stream. Further, the functional co-variation depends on a
number of invariances: between the natal stream and its ‘odor’ or chemical signature, and
between the odor plume and the salmon’s imprinted memory of that cue.
18
Despite meeting all Prinz’s requirements, it is doubtful that the salmon
intentionally represents the location of its natal stream. As Dretske (1988) might put it,
salmon homing behavior may well be directed towards the stream without being directed
to the stream. Burge’s verdict on the salmon case focuses on objectification and is worth
quoting in full:
A capacity to localize a distal source of stimulation without serial sampling is a reliable sign of perceptual objectification. Localization is a capacity to determine direction and distance. The salmon has a breathtaking capacity to localize the home stream. These localizations are not representational. The salmon localizes its molting site only by combining sensory and motor capacities to sample serial intensities of proximal stimulation. As far as is known, the salmon’s olfactory system lacks any sensory state that determines both direction and distance of any object or property. The salmon stores a sequence of sensory registration types. Localization through the salmon’s action is just a reliable effect of the system’s serial sampling of intensities and serial responses to registrations of the types of proximal stimulation. (2010, p. 427)
It is not that salmon are too neurologically primitive to have states with intentional
content. Salmon visual systems can objectify and thus produce perceptual states that
refer to attributes in their environment. Burge denies, however, that salmon olfactory
sensing can refer to what they have the biological function of delivering. In the salmon
olfactory case, “[t]here is no systematic, structured capacity in the sensory system that
separates registration of distributions of proximal stimulation from states that specify, and
are as of, environmental attributes.” (2010, p. 426) On the plausible assumption that
Burge’s verdict in the salmon homing case is correct, Prinz’s account is guilty of over-
applying intentional content. Prinz needs to say more to avoid the over-application
problem.
I will close by offering a suggestion as to where Prinz’s account can be
supplemented so as to avoid the over-application problem. Granting Prinz’s claim that
19
the semantic properties of high-level invariant representations can extend all the way
down to sensory subsystems, the homing case makes clear that not all blending of high
and low succeeds in creating states capable of bearing intentional content and reference.
Thus, we must examine the links between higher-level perceptual subsystems and
sensations in order to distinguish those links that can establish referential relations to
environmental individuals and kinds from those that can’t. It may matter what kind of
high level invariant representation is blended with intermediate level sensation; not just
any kind of blending will do. In describing the sort of back propagation that allows for
sensations to refer to objects in the environment, Prinz holds that representations from
high-level subsystems are invariant across different contexts and perspectives. Such
invariance is key, since only relevantly invariant representations can acquire semantic
properties via co-variance with individuals and kinds in the environment. But it may be
that only certain invariances will do the job, and that the invariances involved in olfactory
homing are not the right sort to support objectifying reference.
What sorts of invariances are required for objective reference? In discussing the
invariances and constancies generated by high-level perceptual subsystems, Prinz relies
mainly on examples from the visual system. (2006, p. 454) This is not surprising since
the invariances encoded in visual perception are especially apt for purposes of
distinguishing objects from their environmental surround: color constancy, shape and
boundary constancy, size and depth constancy derived from binocular disparity. Burge
suggests that only vision, touch, and hearing provide the right sort of structured
representations required for objectification. (2010, p. 429) A behavior that fulfills its
biological function utilizing sensory states and stored representations that lack
20
objectifying invariances will not warrant intentional characterization. As many
philosophers (e.g., Quine, Strawson, Evans) have maintained, intentionality presupposes
objectification, and objectification requires specific sorts of perceptual constancies.
Again, not just any sort will do the job.
The flavor of these recommendations is distinctively Kantian. The Kantian
association should not be unwelcome for Prinz, given that he claims both Hume and Kant
as philosophical predecessors for his brand of neo-empiricism: “This story is not just
Humean; it is Kantian. Kant tells us that concepts without percepts are empty.” (2006, p.
450) But Kant made a point of distinguishing percepts from mere sensations. Percepts
are structured by the empirical intuitions of space and time. Without the objectification
offered via empirical intuition, there could be no objects of experience. Prinz’s account
of intentionality needs an analogous requirement in order to pull off his ambitious project
of reviving concept empiricism for a new age.
21
Works Cited
Barsalou, L.W. (1999). Perceptual Symbol Systems, Behavioral & Brain Sciences, 22, 577-660.
Burge, Tyler. (2010). Origins of Objectivity (Oxford: Oxford University Press).Damasio, A. (1994). Descartes’ Error: Emotions, Reason, and the Human Brain (New
York: Avon Books).Dretske, Fred. (1988). Explaining Behavior (Cambridge, MA: MIT Press)______. (1995). Naturalizing the Mind (Cambridge, MA: MIT Press)Godfrey-Smith, P. (1996). Complexity and the Function of Mind in Nature (Cambridge:
Cambridge University Press).Frege, Gottlob. (1893). On Sense and Meaning. M. Black, trans. In P. Geach and M.
Black, eds., Translations from the Philosophical Writings of Gottlob Frege. (Oxford: Blackwell, 1953).
Hasler, A.D. and A.T. Scholz. (1988). Olfactory Imprinting and Homing in Salmon. (New York: Springer-Verlag Berlin Heidelberg)
Lakoff, G. (1987). Women, Fire, and Dangerous Things (Chicago: University of Chicago Press).
Machery, Edouard. (2007). Concept Empiricism: A Methodological Critique, Cognition, 104, 19-46.
Millikan, Ruth G. (1996). Pushmi-Pullyu Representations, in Philosophical Perspectives, James Tomberlin, ed. (Ridgeview Publishing).
Prinz, J.J. (2002). Furnishing the Mind: Concepts and Their Perceptual Basis (Cambridge, MA: MIT Press).
______. (2004). Gut Reactions: a Perceptual Theory of Emotion (Oxford: Oxford University Press).
______. (2006). Beyond Appearances: The Content of Sensation and Perception, in Perceptual Experience, Tamar Gendler and John Hawthorne, eds. (Oxford University Press).
Putnam, Hilary. (1975). The Meaning of “Meaning.” In H. Putnam, Philosophical Papers, vol. 2: Mind, Language, and Reality. (Cambridge: Cambridge University Press).
Stalnaker, Robert C. (2003) Ways a World Might Be. (Oxford: Clarendon Press).Sterelny, K. (2001). The Evolution of Agency and Other Essays (Cambridge: Cambridge
University Press).______. (2003). Thought in a Hostile World (Oxford: Blackwell)
22
1 In the texts under review, Prinz does not directly address the question of the origins of intentionality in the manner raised here. However, he does address a number of related issues, such that the position developed in part 2 is suitably constrained by Prinz’s stated views.22 Prinz’s proxytype theory integrates elements from the three leading psychological theories of concepts: exemplar theory, prototype theory, and theory-theory. Exemplar theory holds that concepts are constituted by collections of representative instances of a category; prototype theory contends that concepts are constituted by a representation of the best (most typical) instance of a category; theory-theory holds that concepts are constituted by mini-theories about the essential features and causal-explanatory relations of the categories they represent. Each of the three leading psychological theories of concepts can point to impressive empirical support, and each is best suited to explaining some cognitive capacities, but has difficulty with others. They have typically been considered competitors, but Prinz views them as each offering insight into some aspect of the problem.3 The cognitive content of distinct concepts can be similar rather than strictly identical: to the extent that individuals have differing degrees of overlap in the sets of appearances that trigger their respective concepts, the cognitive content of their concepts will be more or less similar. 4 Traditionally, empiricists have offered similarity or resemblance-based accounts of intentionality, whereby concepts refer to whatever they resemble. The many intransigent problems with the resemblance-based account of intentionality have led many to reject empiricism tout court. But Prinz contends that empiricism about concepts is not intrinsically committed to a resemblance-based theory of intentionality, such that the latter can be jettisoned without having to abandon the former.5 The last point illustrates the theoretical virtues of Prinz’s incipient cause condition; by including this etiological component in his account of intentional content, Prinz is able to explain cases of representational error while at the same time adhering to the antiverificationist desideratum that prohibits epistemic constraints on successful reference.6 Prinz would prefer to substitute the Lockean terms “nominal kinds” and “real kinds” for “cognitive content” and “intentional content” respectively, since the Lockean usage avoids the implication that “cognitive content” is non-referential. For Prinz, concepts are better viewed as referring in two ways (or doubly referential): they refer in one way to appearances, and in another way to the kinds appearances are used to detect.