What is a Phenomenal
Concept?
Janet Levin/USC
I.
Something about Mary
Consider
Mary, begins a well-known argument, a brilliant neurophysiologist who has
learned all the physical and functional facts about color vision and has
perfect inferential acumen, but who (though having the potential for normal
color vision) has been born and raised in a black-and-white room. Despite her theoretical knowledge, the argument
continues, Mary cannot know what itÕs like to see red unless she can somehow
escape her black-and-white prison and see such things as boiled lobsters and
strawberries. And this, the argument
concludes, provides a threat to physicalism, since there must therefore be
facts about color experience--namely, what itÕs like to see various
colors--that go beyond physical or functional facts. This is the "knowledge argument" against physicalism,
first advanced by Frank Jackson.[1]
Most
physicalists have attempted to respond to this argument by asking for a
clarification of its second premise.
There is something Mary lacks in her black-and-white room that
can be acquired only when she begins to have normal color experience, most
physicalists acknowledge, but what?
What exactly does it take to know what it's like to have a
certain sort of experience--and does the fact that Mary lacks that sort
of knowledge in her black-and white room really threaten physicalism?
One
of the first suggestions from physicalists (Nemirow, Lewis) was that all Mary
lacks in the B&W room is a certain set of abilities--abilities, for
example, to imagine having color experiences, or perhaps to remember them, or
predict what she will think or do once she has them. Knowing what it's like to have a certain experience, they
suggest, is just to have one ormore of these abilities wrt the experience in
question. After all, they maintain, the
only relevant facts there could be about what it's like to have certain
experiences are facts about their causal-functional roles; indeed, about the
causal-functional roles that common sense takes them to have. And Mary has access to all these
facts, even in her B&W room. But if
all Mary gains when she begins to have color experience is a kind of practical
knowledge, then physicalism is safe, since acquiring practical knowledge of
this sort is not to acquire access to any non-physical, irreducibly phenomenal,
facts.
This
view, however, did not make much headway in the debate with the dualists. First, they argued, it doesn't seem as if
there could be enough "common sense" functional descriptions to
exhaust all the relevant facts about the experiences we can distinguish in
introspection. What, for example, do we
commonly know about the causal roles of the experiences of red and green that
can distinguish us from our counterparts with an inverted spectrum; and is
there anything we know at all about that inconsequential pain I had yesterday
in my left elbow? I myself think that
the common sense functionalist can get farther than one may first think; that a
little creative tweaking of our notion of "common knowledge" to
include judgments about experiential similarities and dissimilarities, wrested
from us by an aggressive Socratic questioner, may help a lot. But there's a
second, more serious worry, namely, that even if future analyses provide subtle
and sophisticated descriptions that do manage to make these distinctions, it just
seems implausible that what Mary acquires when she leaves her B&W room
is merely a set of practical abilities, rather than something more cognitive,
something that can figure in judgments and inference--in short, something that
seems very much like knowledge of facts.
These
days, moreover, many physicalists concur.
Most recent physicalist responses to the knowledge argument agree that,
in coming to know what it's like to see colors, Mary has indeed acquired a
new set of facts that were not available to her in her B&W room. But,
they add--contrary to the claims of the dualists, her acquisition of these
facts does not threaten physicalism.
Knowing what it's like to see red, they argue, does not involve access
to a set of irreducibly phenomenal properties of color vision, but
merely to a new set of concepts that denote some of the very neural or
functional properties that Mary has learned about, under different
descriptions, in her black-and-white room.
And insofar as the use of these new concepts or representations in one's
knowledge claims counts as knowledge of new facts, Mary can be said to have
learned new facts. Terence Horgan was
among the first to propose a view of this sort, and Brian Loar has worked it
out in perhaps the greatest detail.
These
special phenomenal concepts, such theorists acknowledge, are radically
different from the concepts that standardly figure in intertheoretic reductions: they are irreducibly phenomenal--that
is, not equivalent to any physicalistic, topic-neutral, or other discursive
concepts--and we (normally) can acquire them only by having, and being aware
of, experiences with the properties they purport to denote. However, physicalists argue, since these
irreducibly phenomenal concepts denote old, familiar neural or functional properties,
then Mary's having acquired new concepts and thereby access to new facts will
not threaten physicalism.
Could
this physicalist strategy possibly work?
There is a line of argument that suggests it can't. The idea here is that if two distinct,
conceptually unconnected, concepts or representations denote the same property,
they must do so by describing distinct (higher-order) properties or modes of
presentations of it. Otherwise (as
Nagel might put it), there is no explanation of how the referential paths of
these unconnected concepts could converge.
But, by hypothesis, the concepts Mary acquires when she leaves the
B&W room are unconnected to any physical or functional
concepts--since otherwise she would have been able to acquire these concepts,
and thus figure out what it's like to see red, while still in her B&W
room--just by reflecting on the concepts she already possessed. Thus, these irreducibly phenomenal concepts
can pick out a physical state or property only by describing some non-physical
mode of presentation of it!
This
is sometimes called the Distinct Property Argument, which has most
recently been advanced by Stephen White. It derives from a Fregean theory of
reference, and has been regarded as plausible whether one thinks these
associated descriptions explicate the meaning or merely fix the
reference of the terms or concepts in question. This argument provides one of the prime motivations for property
dualism. But it's also (again, at least
arguably) the prime motivation for common sense functionalist views that
attempt to analyze, or explicate, our ordinary experiential concepts in
functional terms--and hold to the line that all Mary acquires, when
leaving her black-and-white room, are new abilities.
However,
the "new concept" physicalists argue, there is another model
of how irreducibly phenomenal concepts denote that undercuts the DPA. A better alternative, they suggest, is to
think of them as functioning somewhat like demonstrative terms or
concepts, which are also irreducible to anything discursive. That is, in
much the way that demonstratives like 'this' and 'that' pick out their
referents, irreducibly phenomenal concepts can refer "directly" to
the properties--presumably neural (or perhaps psychofunctional)--we
discriminate in introspection. This
means that they can denote without introducing either meaning explicating or
"reference-fixing" descriptions of higher-order properties, or modes
of presentation, whose nature (physical or non-physical) would in turn be open
to debate. In addition--and again, much
like demonstratives--in applying these concepts to ourselves in introspection,
we can acquire non-inferential, or epistemically "direct", knowledge
of the qualitative properties of our experiences.
Most
"new concept" physicalists (as I'll call them) suggest that these
demonstrative-like phenomenal concepts denote whatever physical properties are
causally responsible for our application of these concepts in various
introspective tasks (given of course, that there are such
properties; if not, we're stuck with either dualism or eliminativism). But which introspective tasks? This
depends, in part, on whether we're interested in type or token
phenomenal concepts. According to many
of these views, I can use a token phenomenal concept to pick out an instance
of the salient qualitative character of my current experience just by attending
to it in introspection: whatever
(neural) property-token causes me to make introspective note of some experience
I'm now having counts as the denotation of the token demonstrative
"that" (where I intend, of course, to pick out the experience I'm
having now).[2] And if our experience at a time exhibits
more than one instance of salient qualitative features (as would simultaneously
having a pain, hearing a noise, and seeing red, or seeing, simultaneously, two
different shades of red), then we can denote these property-tokens,
respectively, as 'this', 'that', etc., as long as we can discriminate among
them at thattime. Token demonstrative
concepts, like so many useful items, are to be used once and thrown away; they
don't have to be stored in memory, and their referential success implies nothing
about the user's ability to recognize other such experiences as
experiences of that kind. Because of
this, however, their usefulness in explaining our introspective knowledge of
phenomenal states--and thus, perhaps, our knowledge of what it's like to have them--may
be limited, since the function of token demonstrative concepts is to pick out instances
of phenomenal properties, and not those properties themselves.
Type
demonstratives, on the other hand, are different. These demonstratives
("that kind (of experience)") are supposed to pick out kinds
or properties from an introspective perspective, and not just instances of
them that occur at various times. What,
then, determines which kind or property a type-demonstrative used in
introspection denotes? The standard
answer to this question is that the denotation of a type-demonstrative concept
used to pick out a property in introspection is determined by the disposition
of the concept-user to recognize or re-identify further mental
states as experiences of that type.
(This is the way Loar talks about it, for example.) I f we manage,
fairly consistently, to pick out the same property when noting,
"Oh, that (twinge) again" or writing "S" in our diary, we
can be said to be using a phenomenal type-demonstrative.[3] Type-demonstrative concepts are the ones
that do the heavy lifting on this view, since we're interested in giving an
account of how we (including Mary) can have a special sort of first-personal access
to a property, a repeatable--so these have been the focus of this
promising new approach.
But
for all its apparent promise, it hasn't been clear that the demonstrative view
provides an adequate account of phenomenal concepts. There has been a rising tide of criticism of it--not surprisingly
by property dualists such as David Chalmers, but also, recently, by
physicalists such as Terence Horgan.
Criticism from both directions has focused on one particular worry,
namely, that these concepts are being invoked for what seem to be incompatible
tasks: To avoid the "distinct
property" objection, they must be purely recognitional, or
type-demonstrative, concepts--individuated solely by an individual's dispositions
to "go on" to identify more instances of that type. On the other hand, they must be robust
enough to account for what Mary seems to acquire when she leaves her B&W
room. If these concepts are
sufficiently "thin" to denote the way a demonstrative does--that is,
by serving merely as a pointer directed at a type of experience--the worry
goes, then maybe they are too thin to account for what seems to be the
rich and detailed knowledge of what it's like to have an experience which only
experience itself seems to provide.[4]
So,
according to the critics of this view, physicalists still haven't
captured what goes on when someone comes to know what it's like to have an
experience. The "ability
hypothesis" tried to avoid the distinct property objection by taking this
knowledge, many think implausibly, to be pure practical knowledge, while the
type-demonstrative theory tries to avoid this objection by taking this
knowledge, many think implausibly, to amount to nothing more than a pallid
pointing in. I've been worried about
this problem, and in a recent paper (PPR, May 2002), I suggest that phenomenal
concepts may best be viewed as hybrids with demonstrative and discursive
elements. But I've been thinking, more
recently, that a pure type-demonstrative theory may also be able to do better than it first may seem--if, that
is, it acknowledges a certain complexity in what goes on when we acquire
demonstrative concepts. These
speculations, interestingly, were prompted by a fairly recent article by Diana
Raffman, "On the Persistence of Phenomenology", which mounts a
serious and sustained criticism of the type-demonstrative view.
In
her article, Raffman cites some interesting facts about human psychology that
suggest that the type-demonstrative theory runs into a difficulty analogous to
what many take to be a major problem for the ability hypothesis, namely, that
there aren't enough type-demonstrative concepts to do justice to the
discriminations we can make among our experiences in introspection. Raffman takes these facts toraise insoluble
problems for the type-demonstrative account of knowing what it's like--and
therefore (on the assumption that this is indeed the best physicalist response
to the knowledge argument) for physicalism itself. What I'll argue, though, is that these observations don't make
trouble for the type-demonstrative view--or at least don't cause more problems
than the ones we've already identified.
In fact, as I'll argue, they may even suggest a way to combine the
virtues of the type-demonstrative and functional accounts of phenomenal
concepts to provide a more satisfying physicalistic explanation of the
knowledge Mary acquires when she comes to know what it's like to have an
experience of a certain type.
II.
Raffman's observation
Raffman's
worry is this: According to a number of
psychologists, normal subjects can discriminate (that is, discern just
noticeable differences (jnd's) among) far more shades of color than they can
re-identify over time; our "perceptual memory", that is, just isn't
that good.[5] For example (her example), subjects can
discriminate subtly different shades of red (say, red-31 from red-32 and
red-33) when presented with them simultaneously, but can't consistently pick
out red-31 (instead of red-32 and red-33) as "that (color I just
saw)" when the various shades are presented one-by-one. The same will be true, presumably, when
subjects attempt to reidentify the qualitative properties of their color experiences
as well.
Thus,
if having a (type-)phenomenal concept of a certain perceivable or experiential
property requires the ability to re-identify the property when encountering
it by itself (as well as being able to discriminate it from others), then
we can't have (type-)phenomenal concepts of such finely individuated shades as
red-31 and red-32. We may have discursive
or "theoretical" concepts of them, concepts such as "the
shade that normal subjects judge to be three jnd units from unique red";
we may have coarser-grained phenomenal concepts, which permit the
re-identification of shades in a band between red-31 and red-34; and, of
course, we may have token-demonstrative concepts of instances of these
determinate shades. But (just as many
claim to be the case on the functional view) the range of our phenomenal
type-demonstrative concepts of color experience falls short of the range of
color experiences we can discriminate in introspection (when they're
simultaneously displayed).
Now
Raffman argues that these facts about human perceptual memory cast doubt on the
claim that knowing what it's like to have experiences with various qualitative
characters is merely a matter of applying demonstrative-like concepts to neural
states that can also be identified in other ways. I'm not clear, however, why this should be so. Perhaps the worry is this: if knowledge of what it's like to see red-31
depends on our ability to identify that property consistently upon
presentation, then there can be no physicalistic account of how a person who is
currently discriminating red-31 from other shades in introspection knows what
it's like to see red-31. But clearly such a person would have this
knowledge: if Mary were to be presented
with a sample of red-31 upon leaving her B&W room, she'd describe herself
as now (finally!) knowing what it's like to see red-31. So physicalism can't account for what Mary
comes to know.
Physicalists,
however, have a response to these worries. (suggested by, though perhaps not
equivalent to, the line Michael Tye takes on the ability hypothesis in his new
book.[6]) They could respond by suggesting that our
notion of knowing what it's like is ambiguous. In one sense, anyone who can in fact discriminate the
experience of some particular shade--say, red-31--from others in introspection
counts as knowing what it's like to experience that shade. All that this sort of knowledge requires is
the ability to apply, in introspection, a token-demonstrative concept to
an instance of a particular neural property.
As long as there's a physical distinction between the property-tokens we
can discriminate in simultaneous presentation by applying token-demonstratives
to them, we've managed to denote them in a physicalistically acceptable way.
But
in another, more substantive sense, those who cannot consistently reidentify
the experience of a shade as fine-grained as red-31, or recognize instances of
it when it occurs by itself in introspection, do not know what it's like
to see red-31 (rather than, say, what it's like to see an instance of a broader
band of the red spectrum that includes red-31). This sort of knowledge--that is, knowledge of a phenomenal repeatable--requires
more than just an ability to point to an instance of it in introspection; it
requires, at very least, the ability to reidentify, or recognize, other
instances of it at different times, in different contexts. In this sense, of course, Mary may
not know what it's like to see red-31 even after she leaves her B&W room and
stares, attentively, right at a rose of that color. And, if Raffman's facts are correct, the same is true of most
of us. But though this inability
may reflect badly on human memory and categorization capacities--and thus on
our capacity to know what it's like to see colors in the more robust,
substantive sense--I can't see why it should reflect badly on physicalism.
Raffman
might argue that it just does seem that, in looking at that red-31,
red-31 rose, Mary can grasp a repeatable phenomenal property of her
experience; that is, she can know what it's like to see red-31 in the
second,.more substantive sense. And if
physicalists can't explain this, they must concede that what Mary has grasped
is a new. non-physical property of experience.
Thus, a (type-demonstrative) physicalist must deny that Mary has this
second sort of knowledge. But this, on
reflection, shouldn't be so hard:
physicalists, after all, can acknowledge that
Mary can know what it's like to see red-31
in the first sense, since she can pick out the experience with her
token-demonstrative, and also--in defense of their claim that Mary does not
have more substantive knowledge--point to how quickly, once the rose is removed
from sight, the memory fades.
Nor
should it bother physicalists that our ability to reidentify colors (and thus,
presumably, color experiences) seems to exhibit unexpected asymmetries. Raffman points out that a physicalist can't
just claim that our type-phenomenal concepts of color experiences (and
consequently, the color experiences we can know what it's like to have, in the
second, substantive sense) are uniformly coarse-grained; that though we
have phenomenal concepts of a fairly broad band of red experiences and green
experiences--or even maybe indigo and chartreuse experiences--we just don't
have concepts as fine-grained as red-31 or red-32. But this claim, she continues, would be equally untrue to the
facts, since people are in fact very good at re-identifying--not just
discriminating in simultaneous presentation--the "unique" shades of
red, green, yellow and blue, even though they occupy an equally fine-grained
place in our color quality space.
("Unique" red, by the way, is what we might call "pure"
red; a red with no yellow or blue in it at all; same for the other three
"unique" shades.)
This
might, though Raffman doesn't herself bring it up, suggest the following
worry: Here's Mary, just out of her
B&W room, being presented (congratulations!) with two roses, one that's
red-31 and one that's unique red. As
she's looking at them and reflecting on her new experiences, it may seem that
she has equal grasp of their phenomenal properties. But, on the type-demonstrative view, she has knowledge, in the
second, substantive, sense only of what it's like to see unique red. And this may seem counterintuitive, at
best. As in the last case, however, the
physicalist ought to be able to dispel this illusion by pointing out that these
intuitions can be upheld by the observation that Mary does have equal
knowledge of what it's like to have the experience tokens in question,
and predicting that as soon as the roses are taken away, and Mary is left with
the ability to conjure up the memory of red, but not red-31, our intuition that
she possesses equal knowledge of the types will fade.
It's
true that if we take the ability to recognize or reidentify some property as
necessary for having a (type) phenomenal concept of it, then these concepts
comprise what may seem to be an odd amalgam of the coarse- and fine-grained. But once again, this result, though perhaps
surprising, should not be damaging to physicalism, since the demonstrative
theorist has an account of how both token- and type-demonstratives--whatever
their range--can denote physical properties.
If we're to take psychology seriously in developing an account of
phenomenal concepts--and of knowing what it's like--then these asymmetries are
just the facts of life. And they should
not be all that surprising, since there are many other familiar cases (smell,
touch) in which the differences we can discriminate among various physical
stimuli imperfectly reflect the differences that--from a physical point of
view--there are.[7]
Still,
the asymmetries Raffman cites have interest independent of their consequences
for physicalism. It may be true, given
these observations, that on the type-demonstrative view (just as on the common
sense functional view) we can't know what it's like, in the substantive sense,
to have all the experiences we can discriminate in introspection, since
these outstrip the experiences for which we have type-demonstrative
concepts. But reflection on a related
phenomenon involving discrimination and recognition, also discussed by Raffman,
may prompt us to take a different view of what's involved in the acquisition
and use of type-demonstrative phenomenal concepts, and thus suggest that
physicalistic accounts of knowing what it's like may be more intuitively
satisfying than one may initially think.
The
phenomenon in question is that we seem to be able to increase the range
of our phenomenal concepts, at least to some degree, through explicit
instruction. For example, people
who have taken a course in music appreciation or wine-tasting will report that
they are now able to recognize, or identify consistently, sounds or tastes that
they were never able to recognize before.
Raffman herself acknowledges that our capacities for re-identification
can be enhanced, given the proper instruction, although she maintains that they
can't be expanded enough to permit us to re-identify all the shades we
can discriminate in simultaneous presentation.
Still, if they can be expanded all, this would mean that, though there
may be fixed psychological limitations to our discriminative abilities (and
thus, if you will, our token-phenomenal concepts), we can, through
training, enhance our knowledge, in the second, more robust sense, of what it's
like to have an experience. My aim, in the next section, is to ask what this
training or instruction might involve, and what consequences it may have for
our views about phenomenal concepts and "knowing what it's like".
III.
Increasing our stock of phenomenal concepts
Let's
suppose that the "quality spaces" associated with each type of
perceivable property are innate; that is, that anyone with our psychology (and
no particular training) can distinguish just the same colors, tastes, sounds,
etc. in the standard jnd experiments. And let's suppose--just a
supposition!--that I'm not, naturally, particularly good at remembering tastes
or sounds--or colors. Still, after a
music appreciation or wine-tasting (or interior decorating) course, I may be
able to re-identify more sounds or tastes or colors than I can right now (if
not all the ones I can discriminate).
Suppose,
then, that after a class in interior decorating, I become able to recognize
more fine-grained shades of red than I was able to recognize before.[8] What could I have been taught in this class
to have enhanced my perceptual memory--and thus my recognitional capacities--in
this way? The fact that we can make
fine-grained comparative discriminations between red-31 and red-32 in simultaneous
presentations seems important to this process, but clearly it can't be
sufficient, for if it were I'd have no recognitional deficiencies to remedy.
As
I see it, there are two (broadly characterizable) ways our recognitional
abilities could be enhanced. On one
view, a subject is given explicit theoretical information about just where in
our color quality space a certain target shade lies, and just what its
relations are to other shades. For
example, suppose that, when a number of colors are presented simultaneously, we
learn to describe or think of red-31 as 'the shade closer to unique red (along
some dimension) than red-32'--and then perhaps as 'the shade n jnd steps from
unique red'. On this view, an
individual attempting to expand her range of color concepts can be seen as
attempting to re-identify the experience of a certain target shade of red
(red-31, say) by consciously applying information she has learned about
its position in color quality space to her current experience. That is, she might come to recognize a new
instance of red-31, eventually, as the same shade she focused on before by
imagining or remembering an experience of unique red (by hypothesis, easy and
natural for everyone), and comparing it to her current experience. (This is the sort of thing, I take it,
that's sometimes taught in wine-tasting courses: you're first given a wine with some easily identifiable taste,
and then shown how to compare it to others that differ along various
dimensions.) To do this, of course,
she'd not only have to know that red-31 was, say, six jnd steps away from
unique red, but also be able to remember or imagine the experience of unique
red and the interval scale between unique red and red-31. The extent to which imaginative memory can
reconstruct the intervals along some qualitative dimension by which various
shades can fall away from unique red, and keep track of how far away a shade is
from some reference point, will determine the scope and limits of our ability
to reidentify new color properties, or, analogously, new phenomenal properties
in introspection. Presumably, some
people will be better at it than others.
But it also, on this view, requires the application of explicit
information about the interrelationships among the shades.
Now
suppose that someone gets good at this procedure, and becomes able to
reidentify the experience of red-31 consistently, without having to haul out
this explicit comparative information or those imagined paradigms. Such things do happen, after all.
Music appreciation and wine tasting and even, I would imagine, interior
decorating classes do, at least sometimes, serve to provide people with the
ability to home in, quickly and smoothly, on experiences they were incapable of
reidentifying before. Can we thus think
of them as deploying new purely recognitional concepts, rather than applying
theoretical concepts to the cases at hand?
I want to argue that the answer, plausibly, is "yes".
One
might raise the worry that the reidentifications one learns to make in this way
could not be sufficiently "direct" to count as the deployment of pure
recognitional concepts, since, first, it presumably takes some time to
develop these abilities, and second, they are, at least initially, informed
by the explicit comparisons I described above. The first worry, however, does not seem so serious, since there
are many cases in which one's abilities to recognize a certain type of thing
has been shaped by explicit instruction, even when the identification is made
solely on the basis of its "naturally" salient features--its look
or its feel. Think, for example,
of experienced doctors who can diagnose cancer before the biopsy results come
in.
It's
true, of course, as the example above suggests, that in many cases in which it
takes awhile to identify some object as belonging to a particular kind, the
kind in question is physically complex.
And one might think that this does not hold for my hypothetical
identification of red-31. Or, at the
very least, one might resist the suggestion that red-31 (or the experience of
it) is any more complex than unique red (or the experience of it), which
presumably can be easily remembered and recognized right away. These asymmetries, however, may have a
plausible explanation, since it shouldn't be surprising that what counts as recognitionally
simple and complex may not be determined by what's physically simple and
complex. Differences in recognitional
abilities for a uniform physical domain may be produced by idiosyncracies in
processing, evolutionary effects on detection mechanisms, or various other
phenomena--both for external phenomena and one's own experiential states.
But
in the end, these asymmetries, whatever their explanation, shouldn't matter to
the question of whether an ability to reidentify red-31, acquired in the way I
described, is as "direct" as the more (allegedly) natural ability to
reidentify unique red if they end up working in the same way. After all, the acquisition of many
kinds of skills involves a progression in which one analyzes the skill in
question, breaks it down into a sequence of components, then practices and
practices until it becomes second nature.
A skill that's second nature, of course, is not initially natural
and automatic; hence its name. But it becomes
natural and automatic in time; one comes, after practice, to be able to perform
the task, or exhibit the skill, quickly and unthinkingly. If the ability to reidentify shades like
red-31 becomes second nature in this sense, then there's reason to think that
we've acquired a new recognitional concept.
Or
better, there's reason to think so if the reference of these concepts is
determined in a certain way. Concepts
acquired as I've described might be taken to determine their referents in a way
that is part demonstrative and part discursive; that is, they may be taken to
refer to whatever property in fact bears the relevant relation to some
paradigm--whether or not the subject using the concept would identify that
property as the kind she had in mind.
In this case, the concept wouldn't be a pure type-demonstrative (though
it may include a demonstrative element).
However, if the reference of the concept is determined solely by the
subject's disposition to reidentify items as another one of those, then
it should count as a pure demonstrative concept, regardless of how much
explicit comparison, or other application of theory, was involved in its
acquisition. In such a case the
explicit information will be solely of heuristic value, merely functioning to
shape a new recognitional ability which alone determines the referential reach
of its associated (pure) type-demonstrative concept.[9]
The
former view--that is, that these concepts may be partially demonstrative and
partially discursive--is a view of phenomenal concepts I suggested in a recent
paper (PPR, May 2002), hoping to capture the intuitive force of the demonstrative
analogy, while reserving a role for what seems to be the robustness of the
knowledge that experience provides. But
it now seems to me that the latter view--that is, that these concepts are pure
demonstratives, underlain by dispositions shaped by explicit information--does
just as well on both these fronts, if not better.
That
is, if theoretical information of the sort I've discussed figures in the
acquisition and deployment of pure demonstrative concepts, the physicalist can
make the following claims. First, what
someone gains from experience, in coming to know what it's like to have it,
is--as the type-demonstrative theorists suggest--a new set of concepts that are
not equivalent to any concepts she already possesses, and which denote directly,
on the model of demonstratives, without needing to invoke metaphysically
suspicious modes of presentation. And
second, in at least some (and maybe most) cases, one has to know a lot
about the properties in question, in particular their interrelations with
others in the relevant quality space, to acquire the concepts--a
phenomenon which accounts for the intuition that what one gains in knowing
what it's like to see red or feel pain is interesting and substantial. So, if this view of how we acquire new
recognitional concepts is plausible, the physicalist may be able to have it
both ways: phenomenal concepts can have
the referential role of type-demonstratives, but their acquisition may still
require a person to know a lot about the interrelations among the properties
they denote.
But,
one might object, the view is not plausible in the least. What happens when we acquire new
recognitional concepts is something much less cognitively complex, something
that doesn't require anything like these explicit comparisons or other
applications of theory that I've described.
What goes on in wine-tasting or interior decorating classes--let alone
what goes on when our recognitional dispositions are "naturally"
shaped by our interactions with the world--is closer to a
"paradigm-foil" model of generalization learning than to anything
like what I've described.
That
is, even in getting explicit instruction in how to reidentify a color or
taste or sound, one is shown an example of the target item, and perhaps one
(the foil) that's quite different along a certain dimension, and told, when one
tries to generalize to further instances, whether or not one did so
correctly--without being told just why one's attempt to generalize
was correct or incorrect. In this case,
one could increase one's range of recognitional abilities, and thus phenomenal
concepts, without appeal to any explicit comparisons or other theoretical
information at all. Maybe this is a more reasonable view of how recognitional
abilities are enhanced by instruction.
And maybe it's also a reasonable view of how we 'naturally"--that
is, without explicit instruction--come to recognize items in the world as
belonging to one or another kind (and even, perhaps, of how innate proclivities
to generalize may have been shaped by selective pressures).
If
so, however, then the smooth deployment of recognitional concepts would still
require a significant amount of knowledge on the part of the subjects who
deploy them. This wouldn't be explicit
knowledge of the salient features of the items recognized as "one of that
kind", or even of the similarities and differences between that kind and
others, but rather the implicit knowledge of these similarities and differences
that shaped (by the paradigm-foil method) the recognitional dispositions in
question. Nonetheless, possession of these recognitional dispositions--and thus
the pure type-demonstrative phenomenal concepts they determine--brings a lot to
the table; enough, I'd venture, to explain how Mary, in acquiring such dispositions
after leaving her black and white room, manages to know so much. Thus to the question, "So this
is a phenomenal concept?", the answer may in the end be a tentative
"yes".[10]
[1]See Frank Jackson, ÒEpiphenomenal QualiaÓ, ÒWhat Mary DidnÕt KnowÓ.
[2]Focusing on, or taking introspective note of, some feeling (or property of it) can, I presume, be given a functional characterization; I will leave this issue aside for now. Also, I don't want to suggest that a creature who uses a token demonstrative to pick out an instance of some feeling must have anything but a demonstrative concept of an experience.
[3]Clearly, there seem to be affinities between the ability hypothesis and the view that what Mary acquires is a special set of type-demonstrative concepts, given that these concepts are individuated by a subject's dispositions to recognize or reidentify experiences of a certain type. Here, too, it seems that what Mary acquires upon leaving her B&W room is an ability--not necessarily to imagine seeing red, or predict what she'll think or do when she sees it, but to recognize or re-identify red--and the experience of it--when it presents itself again. Indeed, some versions of the ability hypothesis (like the one Michael Tye advances in a discussion of it in his new book) may just amount to notational variants of the type-demonstrative view sketched above.
To be sure, to call this a version of the ability hypothesis may seem to violate the principle from which its classical versions arose, namely, that what we gain from experience is merely the ability to apply to oneself--non-inferentially, in introspection--(implicitly known) functional descriptions which (at least more or less) explicate our concepts of experience. On these views, then, not only would Mary not gain access to new properties when she leaves her B&W room, but she wouldn't even gain a new set of concepts. This, I take it, is Lewis's view, and I endorsed it in my "Heatwave" paper, primarily because of worries about the "distinct property" argument against physicalism that I just described. As I've suggested in the text, however, the view that phenomenal concepts work like type-demonstratives--referring "directly" to experiential properties--may put these worries to rest. And I'll argue, in the final section, that there are important (and perhaps surprising) affinities between even the classic version of the ability hypothesis and the type-demonstrative account of phenomenal concepts.
[4]One might wonder if even sufficiently "thin" type-demonstrative concepts can avoid some version of the Distinct Property Objection. As John Hawthorne has suggested, a rational person, told that the qualia on the right side of her visual field may "dance" without her knowledge during a certain time period, may wonder whether even the statement "this (pointing in to a reddish quale on the left) = this (simultaneously pointing in to a reddish quale on the right)"--and so a Frege-problem may arise for this case. (Hawthorne himself uses his case to argue a different point, namely, the incompatibility of direct reference theories of phenomenal properties with the possibility of zombies.)
It seems to me, however, that not every case in which one can rationally doubt a true identity statement of the form "x=y" is a case in which 'x' and 'y' denote their referents via distinct modes of presentation. What makes the subject hesitant in the case described above is not (necessarily) some difference in how the reference of the terms is "mediated" or "presented", but only because one's capacities for perception or judgment have been challenged by someone it's rational to trust. (Compare Descartes's self-described situation with respect to God and the "eternal truths" in his Letter to Mersenne.) In such a case, doubt about the identity statement, and even the possible falsity of the statement-- is compatible with its a priority: if we know that x=y, we know it a priori.
[5]See her (1995), where she poses the problem I discuss in the text as a problem for materialism.
[6]He suggests that knowing what it's like to see red 31 is a matter of actually discriminating a token of it in introspection (when one is actually having thatexperience) or (in its absence) having the ability to recognize, remember or imagine that shade when presented on its own. See Tye (2000).
[7]Noting asymmetries if this sort may also be of interest for the question of whether we have cross-modal perceptual concepts of "primary qualities" such as shapes. It may be, that is, that--as a matter of psychological fact--a "man born blind and made to see" a la Molyneux can identify, upon first viewing them, certain shapes, but not others, that he's touched before: squares and circles, say, vs. lines and curves. If this is so, then any recognition-based test for the cross-modal transfer of concepts would have to be constructed quite carefully.,
[8]Raffman reports that psychologists agree that we can't even learn how to recognize all the shades we can discriminate in simultaneous presentation, but I'll nonetheless refer to the shades in question as 'red-31' and 're-32'; in fact, what we can learn to recognize may well correspond only to broader bands of the spectrum.
[9]Contrast this view with the "classic" ability hypothesis (as well as other views of the self-ascription of mental states such as Sellars's and Armstrong's), in which recognitional dispositions prompted by various properties of the states in question, serve merely as heuristics for the identification of phenomenal states which can be overruled if they conflict with the results of a functional description.
[10]An earlier version of this paper was read at UC Riverside in November, 2001. I'm very grateful to Eric Schwitzgebel for his helpful comments and criticisms.