SWIF
Philosophy of Mind,
01 September 2001
http://www.swif.uniba.it/lei/mind/forums/levine.htm
Forums
Forum
2 Carruthers's reply
Commentary on
Carruthers, Peter. Phenomenal Consciousness. Cambridge:
Cambridge University Press, 2000.
Joseph Levine In Phenomenal Consciousness,
Peter Carruthers defends a dispositionalist higher-order theory of phenomenal
consciousness. On this view, to enjoy a phenomenally conscious experience
is to occupy an analog representational state that is accessible to a faculty
of higher-order thought. In the course of defending his view, Carruthers
takes on an impressive array of arguments and opposing positions. He addresses
"mysterians", who believe that no materialist theory is adequate to an
explanation of phenomenal consciousness, as well as first-order representationalists
and actualist higher-order theorists. Since his theory is itself a representational
theory, he also engages the debate over narrow content, arguing that narrow
content, in the form of a consumer semantics, is the appropriate theory
of content for psychological explanation.
As a mysterian myself, let me begin
my critical comments by addressing his response to that position. As he
notes, there are really two positions here, one metaphysical and the other
epistemological, though they both share a reliance on conceivability considerations.
The metaphysical argument goes like this. Since zombies (or zombie worlds)
are conceivable, they are possible. Though it is not a straightforward
matter to infer what is possible from what is conceivable, in this case
there is reason to think the inference goes through; so goes the argument,
at any rate. (Both Kripke 1980 and Chalmers 1996 defend the inference in
this case, using different, but related, arguments.) But if zombies are
possible, then the physical facts do not metaphysically determine the mental,
or phenomenal facts, and thus some form of dualism is true.
I am generally sympathetic to Carruthers's
response to the metaphysical argument (see Levine 1998 and 2001, chapter
2). He insists on distinguishing concepts from properties, at least if
by "property" one means "natural property" - the sort of feature by virtue
of which objects change and events happen. Distinct concepts can pick out
the same property, and so what seems conceivable may not be possible. For
instance, one might be thinking that some object could be F without being
G, but just not realize that the property of being F just is the property
of being G. The situation envisaged, though conceivable in the sense that
there is no incoherence or contradiction in the description of the situation,
is still impossible. Though there are complications here, I believe this
general line is basically right.
Still, I want to quibble with some
of what Carruthers has to say on this point. With respect to Kripke's argument
against the type-identity of pain and cfiber-firing, Carruthers claims
that we have two options by way of reply: First, maybe "pain" isn't a rigid
designator, so "pain = cfiber-firing" isn't necessary after all. Second,
maybe physicalism about phenomenal consciousness doesn't require an identity
claim, so "pain is constituted by/realized in cfiber-firing" is good enough.
I think neither of these replies works.
With regard to the first option,
it's supposed to go like this. "Pain" is akin to "manifest-water", which
picks out whatever substance in a world instantiates the manifest properties
of water. But "manifest-water" also expresses a property that is shared
across worlds, the property by virtue of which something in a world falls
under that concept. What is the corresponding property for "pain", if not
a phenomenal property? This is of course Kripke's (and Chalmers's) point.
Carruthers seems to think this point
is already addressed by his adoption of what he calls his "thick" theory
of properties. This comes up prominently in his reply to Chalmers as well,
when he denies significance to the fact that the primary intension of "pain
= cfiber-firing" is contingent. He argues that primary intensions are mere
functions from possible worlds to extensions, and therefore irrelevant
to the identity of honest-to-God properties; that is "natural" properties,
the mind-independent features of objects that figure in what happens to
them. He says that "grue" and "bleen" also determine such functions, yet
no one thinks of them as genuine properties (unless of course they are
just collapsing the concept/property distinction altogether).
But I don't think this quite gets
him off the hook. First of all, I think there is an intermediate level
of property in between the full-fledged natural property (the sort that
figures in causal generalizations, etc.) and the one that just collapses
into a concept. Being grue, after all, is an objective condition that objects
either satisfy or not, and this is so whether or not human beings had ever
thought of the property. I agree that it is not a causally significant
property, but it seems to me pretty clear that there is a determinate fact
of the matter whether or not something is grue. It's also the case that
we can form distinct concepts of grue: for instance, the concept expressed
by the canonical description "green before 2001 and blue after" and the
concept expressed by "has reflectance G [put in the technical details here]
before 2001 and reflectance B after".
The point is that it would be a real
problem for physicalism if there weren't a supervenience claim of the form
"Grue metaphysically supervenes on P", where "P" designated a physicalistically
respectable property. Of course in the case of grue there is such a property
(or group of properties). If you specify all the reflectance properties
of all the objects at all times, you have thereby metaphysically determined
the distribution of grue-facts. So even if "pain-feeling" is on a par with
"grue", it seems to me we still have a problem if Chalmers is right that
the pain-feeling facts do not logically (or metaphysically) supervene on
the physical facts.
So why do I think, nevertheless,
that by distinguishing properties from concepts one can avoid the anti-physicalistic
consequence of the conceivability argument? My long-winded argument is
cited above, but the short answer is this. I think one must reject the
assumption that we have a priori access to primary intensions. That is,
I endorse what Chalmers calls "strong metaphysical necessity" - the idea
that a primary intension can be both necessary and a posteriori. Though
Chalmers argues that this involves a bizarre metaphysics of modality, I
argue that it's quite benign in the end. For the details, see the works
cited above.
Let's turn now to the epistemological
side of the mysterian position. As I argue (see Levine 1983, 1993, and
2001, chapter 3), there is an explanatory gap between the physical (broadly
construed, so as to include functional or computational states as well)
and the phenomenal. Support for the existence of the explanatory gap comes
partly from the fact that zombies - physical and functional (including
representational) duplicates of me but without phenomenal consciousness
- are conceivable. Again, I don't infer that they are therefore possible
as well. But merely from the fact that they are conceivable, I argue, we
see that we can't explain phenomenal consciousness by reference to those
properties we and the zombie are stipulated to share. The point is, we
have a conception of phenomenal consciousness such that we can't really
understand how it could be constituted by (broadly) physical processes.
Carruthers employs two strategies
to overcome the explanatory gap. On the one hand, he lists what he takes
to be the principal desiderata on an adequate explanation of phenomenal
consciousness, and then argues that his theory meets them all. On the other
hand, whatever qualms remain over the conceivability of zombies he deals
with by appeal to the notion of a recognitional concept. In other words,
whatever legitimate explanatory demands there are can be met by his theory,
and what can't be met is explained away.
The legitimate explanatory demands
are these: What endows phenomenal states with subjectivity, there being
something it's like to occupy them? Why do they seem ineffable, private,
to involve intrinsic properties of experience, and to be accessible in
a privileged manner? I have my doubts that Carruthers's theory does successfully
meet these explanatory demands (my doubts chiefly concern the first one),
but let me put this aside for now and turn to the question of what seems
to remain.
One way to put the matter is this.
When I wonder how a (broadly) physical process could be like this (mentally
ostending my visual experience of red, say), I'm not wondering how such
a process could be such that I'm tempted to judge it to have this feature
or that. I can certainly imagine a computational architecture that would
lead the subject of that architecture to judge that certain of its states
were ineffable, say, or couldn't be doubted. But the explanatory gap isn't
between a description of the computational architecture and a description
of certain propensities to make judgments. (Notice that zombies would definitely
have the same propensities for judgments that we do, so the conceivability
argument doesn't even apply here.) Rather, it's between the description
of the computational architecture and a description of the experience itself.
It's the character of my visual experience of red that I want explained,
not my tendency to judge that it's ineffable, etc..
Fair enough, Carruthers might say,
but this is where the appeal to the notion of a recognitional concept does
its work. If what you want out of an explanation, goes the argument, is
that zombies should no longer be conceivable - in other words, that one
should be able to derive a description in first-person terms of someone's
having a visual experience of red from a description in third-person terms
of their physical or computational structure - then you won't get it. But
the reason is perfectly benign, from a naturalistic perspective. The problem
is that one's first-person concepts of phenomenal properties are recognitional
concepts, and much as one can't derive indexical or demonstrative descriptions
from purely non-indexical or non-demonstrative descriptions (see Perry
1979), one can't derive descriptions involving recognitional concepts from
descriptions not containing such concepts. So in this sense the explanatory
gap is not exactly bridged, but still explained away, and therefore rendered
harmless.
Perhaps the best way to make the
case, and also, to my mind, to see what's wrong with it, is by way of Carruthers's
own example of "bare color". He asks us to imagine that there exists a
creature who makes "bare-color" judgments, purely recognitional judgments
unconnected to any concepts about normal lighting conditions and normal
observers, or to any phenomenal properties. Like chicken-sexers, they just
judge "this is red" without having any idea how they do it, without there
being any basis they are aware of for their judgment. They just respond.
Such creatures would experience an explanatory gap, since they could conceive
of the physical story for red holding without something's being red, i.e.
without their applying their recognitional concept. Yet it is clear that
these creatures pose no problem for the explanatory reach of a physical
(or computational) theory of color recognition.
But what strikes me as the real import
of the example is how different bare-color responses are from genuine phenomenal
experiences, and this bears directly on the adequacy of the appeal to recognitional
concepts as a way of responding to the explanatory gap. After all, we don't
feel tempted in the chicken-sexing case, nor in the hypothetical bare-color
example, to see a problematic explanatory gap present. What, then, explains
the difference between these cases and genuine phenomenal experience?
When I wonder how this experience
(again, mentally ostending my current visual experience of something red)
could be constituted by such-and-such computational/physical processes,
I have a substantive and determinate conception of - I am, as it were,
directly acquainted with - a feature of my experience. The reddishness
doesn't have the character for me of a something-I-know-not-what that happens
to be prompting this response, as it does for the bare-color responder.
The point is it's the reddishness I am wondering about, not about my tendency
to regard it as intrinsic, nor about my tendency to respond to it with
a certain judgment, or anything else of that sort. In the bare-color case,
as in the chicken-sexer case, there isn't any substantive and determinate
content to my judgment over and above the response itself. Put another
way, "that again" carries a much richer content in the phenomenal case
than it does in the bare-color case. What's more, it is precisely what
corresponds to that richer content that is the focus of my puzzlement when
judging that there is an explanatory gap.
This last point is crucial, for it
is natural at this point to respond that our phenomenal concepts are not
in fact purely recognitional, and therefore of course they possess a richer
content than is possessed by the corresponding concepts of the bare-color
responder. In addition to the recognitional component, on Carruthers's
view, is a descriptive component that derives from our stock of folk-psychological
concepts. So we have concepts of perception, experience, and the like,
with which we can fill out our concept of "that again", to make it more
like "that perceptual experience of something red, again". But this doesn't
address the sense of richer content at issue here, for there is no explanatory
gap when it comes to folk psychological concepts, so long as they are understood
functionally, as Carruthers understands them. (Of course, if our concepts
of experience and perception are not understood in this third-person way,
but rather in the first-person manner of phenomenal concepts generally,
then the argument begins all over again with respect to them.)
Above I used the term "direct acquaintance",
which of course carries a lot of philosophical baggage. Let me hasten to
say that I have no theory of our epistemic access to our own phenomenally
conscious experiences. What does seem clear to me is that not only is the
explanatory gap a problem about phenomenally conscious experiences themselves,
but also about how we can have the sort of cognitive access to their content
that we seem patently to have. The two aspects of the problem of phenomenal
consciousness are just two sides of the same coin. How can we be subjects
of experience, where this involves our occupying states that are somehow
themselves cognitive apprehensions of themselves? In fact, this question
leads naturally to a discussion of Carruthers's positive view, since his
view can be seen as precisely an attempt to answer it.
However, there is one more point
about conceivability arguments I want to address first, and this has to
do with Carruthers's discussion of the possibility of inverted qualia.
On his view, two factors make a state phenomenally conscious: its being
an analog representational state and its being available to a faculty of
higher-order thought. The zombie, or absent qualia problem is a challenge
to that aspect of the theory. But a theory of phenomenal consciousness
has to tell us not only what makes a state phenomenally conscious to begin
with, but also what determines its precise phenomenal content; what distinguishes
reddish visual experiences from greenish ones. With respect to this question,
Carruthers maintains that phenomenal content is determined by representational
content, and it is at this point that the problem of inverted qualia becomes
relevant.
If inverted qualia are possible,
then two individuals might be occupying the same functional/intentional
state, and yet be having distinct experiences - one that is reddish, say,
and the other greenish. But then phenomenal content couldn't be determined
by intentional content. So Carruthers must deny the possibility. He employs
two strategies. First, he allows that inverted qualia are indeed conceptually
possible, but argues, as with absent qualia, that their conceptual possibility
does not entail their metaphysical, or "natural" possibility, and it is
only the latter that really matters to a naturalistic theory of phenomenal
consciousness. Second, he appeals to his version of narrow content theory
- an inferential role consumer semantics - to counter arguments to the
effect that inverted qualia are indeed naturally possible.
With respect to the first strategy,
I basically agree that if inverted qualia are only conceptually possible
that this doesn't constitute an objection to the intentional theory of
phenomenal content. But I'm puzzled by Carruthers's insistence on the distinction
between metaphysical and natural possibility, and his claim that it is
only the latter that matters here. I would have thought that it is metaphysical
possibility that mattered. Now, given the arguments Carruthers in fact
uses, the distinction may not be crucial. However, I can think of a situation
in which the distinction might matter, and it's unclear to me what Carruthers
would say about it.
So, suppose that as a matter of natural
law it's impossible for a perfectly symmetrical information processing
system to be physically realized. Let's also suppose that from the point
of view of the theory of computation, however, there is no obstacle to
such systems (and suppose they could include analog representations as
well). It's just that for some reason (which, by the way, it seems hard
to imagine what it could be) you can't actually build (or evolve) such
a device. So we know that our perceptual system couldn't be symmetrical,
and therefore invertible, and we know no naturally possible (I assume that
means physically possible) perceptual system could be, but there's no reason
to suppose that no metaphysically possible system could be.
Well, consider such a system in world
W1, call it S1. By Carruthers's theory, according to which phenomenal consciousness
supervenes on functional/intentional properties, S1 (assuming it's attached
to an HOT faculty) contains phenomenally conscious states whose phenomenal
contents are determined by their intentional contents. Let system S2, in
W2, have intentional contents inverted around the axis of symmetry. We
can of course insist that the phenomenal contents go with the intentional
contents, but we have the Shoemaker and Block scenarios, discussed by Carruthers,
to contend with. My only point is that if one found those scenarios convincing
in the first place, they ought to bother one in this case, and, by stipulation
one can't appeal to constraints on the realization mechanisms to get around
the problem. So it looks as if cases of inverted qualia can be metaphysically
possible even if not naturally possible.
As I say, it's not clear to me how
much the distinction between natural and metaphysical possibility matters
here because I take Carruthers in the end to be arguing that inverted qualia
are not even metaphysically possible, though they are conceptually possible.
But his argument for this is obscure, at least to me. At one point he says:
"if [the notion of narrow content] is legitimate, then the teeth can drawn
from all forms of experience inversion...". My problem is that I can see
how the appeal to narrow content insulates him from the sort of inversion
argument that troubles externalist representationalists - in particular
Block's Inverted Earth argument (see Block 1990) - but how does it help
with the sort of inversion argument, like Shoemaker's, which is aimed at
traditional functionalist theories? After all, these theories have usually
been understood to be internalist. (That's certainly how Shoemaker conceived
of functionalism when worried about inverted qualia in Shoemaker 1984.)
In fact there's a curious transition
in Carruthers's discussion. He begins by discussing Shoemaker's argument,
which involves memory switching, etc., and concludes that it's inconclusive
on the question of genuine possibility (I'll use this term to cover both
metaphysical and natural possibility), as opposed to conceptual possibility.
He then brings up Block's Inverted Earth argument as something that pushes
us in the pro-qualia direction, but then replies that if we have the requisite
notion of narrow content we don't have to worry about Block's argument
(that's where the quote above comes in). But what happened to the original
Shoemaker argument? Why does the legitimacy of the requisite notion of
narrow content "draw the teeth from all forms of experience inversion",
rather than just from those that address externalism, as Block's case does?
Finally, let me say why in the end
I think it's very plausible to think inverted qualia are genuinely possible.
First, I assume that symmetrical internal systems of functional roles are
genuinely possible. Again, it's hard to see how one could rule them out,
especially their metaphysical possibility. Second, I assume that symmetrical
external quality spaces - of the sort that would be the external contents
of some perceptual system - are also genuinely possible. Once we have these
two symmetrical systems - internal roles and external quality spaces -
how could one determine the identity of a single phenomenal content by
its location in either space?
Suppose one wanted to do it purely
internally. True, qualitative difference can be accounted for internally,
so red and green can be distinguished (or the analogues for red and green
in the symmetrical system under consideration). But appeal to location
in the relevant system of relations won't account for what makes red red
and green green, and not the other way around. To do that, it looks like
one would need to appeal to the identities of the external properties to
which each is responsive. But if the relevant quality space is itself symmetric,
then this won't work either, for the sorts of reasons anti-externalists
like Block point out. This problem - that you can get a functional account
of qualitative difference but not a functional theory of particular qualitative
characters - was precisely Shoemaker's point in the papers cited above,
and why he opted in the end for a physiological type-reduction theory for
individual qualia (not that I think that move works either; see Levine
1989 for a detailed discussion of Shoemaker's position).
Let's turn now to consider Carruthers's
positive theory. As mentioned above, Carruthers defends the view that phenomenally
conscious states are analog representational states that are available
to a faculty of higher-order thought. He distinguishes his view from a
number of alternatives, but two in particular take up the bulk of his discussion:
first-order representationalism (FOR) and actualist higher-order representationalism
(HOR). FOR is the view that phenomenally conscious states are analog (or
non-conceptual, the difference doesn't matter for our purposes) representational
states that are available ("poised", as Tye 1995 puts it) for cognitive
processes such as control of behavior, rational planning, etc.. Actualist
HOR theory is the view that to be phenomenally conscious is to occurrently
entertain a higher-order representation that one is in the state in question.
Carruthers's position differs from the latter in not requiring that there
actually be an HOR directed on the phenomenally conscious state, just that
it be accessible to the faculty so that one could direct an HOR on it.
His position differs from FOR in that mere availability to planning, behavior
control, and the like is not sufficient; it must be a capacity for forming
HOR's to which the state is available.
The main problem with FOR, according
to Carruthers, is that it can't provide a principled distinction between
"experiences" or perceptual states that are phenomenally conscious and
those that aren't. There is abundant evidence both from ordinary experience
and from scientific psychology - from absent-minded drivers to cases of
blindsight - to the effect that there are perceptual states that are not
phenomenally conscious. While advocates of FOR might want to appeal to
the "poised" condition to account for the non-conscious status of these
perceptual states, Carruthers argues that there is solid evidence that
they do contribute to behavior control, which includes interacting with
unconscious beliefs and desires as well. So why don't they count as phenomenally
conscious?
There is one point Carruthers makes
repeatedly in response to various attempts on the part of FOR advocates
to provide a principled distinction between conscious and non-conscious
states: it is not sufficient to merely find some difference that is extensionally
accurate, but the difference in question must actually explain why the
phenomenally conscious states have this feature of subjective feel and
the others don't. So, for instance, in response to the suggestion that
the crucial property is being accessible to the highest-level decision-making
faculty, Carruthers complains: "How can the mere fact that an analog content
is now in a position to have an impact upon the highest-level decision-making
processes confer on it the subjective properties of feel and "what-it-is-likeness"
distinctive of phenomenal consciousness?" (page 170) I am quite sympathetic
to this complaint, and it is crucial to the development of his own view.
Now some have pressed a similar complaint
against (actualist) HOR theory, but advocates reply, with some justice
I think, that this objection betrays a misunderstanding of their position.
The entire point of (actualist) HOR theory is that being phenomenally conscious
is not a monadic property of the first-order state, but a relation holding
between the higher-order state and its target state. This idea has problems
of its own (see Neander 1998, Byrne 1997, and Rosenthal 2000), but still
it has a principled answer to what makes some states conscious and others
not: namely, states are conscious when we are conscious of them, which
means that they are the intentional objects of higher-order thoughts.
The point I'm getting to is this.
Carruthers levels a serious objection against FOR, one to which, at least
on the surface, actualist HOR is not vulnerable. However, once Carruthers
abandons the actualist version of HOR for the dispositionalist version,
the objection comes back to rear its ugly head; why, after all, should
mere availability to a higher-order faculty render a first-order analog
perceptual state phenomenally conscious? The actualist response that for
a state to be phenomenally conscious is for one to be phenomenally conscious
of it only works if there is an occurrent higher-order state. A mere disposition
for there to be such an would seem to endow the first-order state with
at most a disposition to be phenomenally conscious.
Carruthers is of course aware of
this objection, and it is precisely to meet it that he employs his consumer
semantics. The idea is this. First-order analog perceptual states, when
not functionally attached to a higher-order faculty, have intentional contents
that are about the features of distal objects (though the contents are
narrowly individuated). But when serving within a mental system that contains
a higher-order faculty, one that, in particular, is capable of noting an
appearance-reality distinction, the analog perceptual states, by virtue
of their connection to this faculty "down-stream" (hence a consumer semantics),
acquire a new content, one that not only represents the features of distal
objects, but also themselves as representing these distal features. Instead
of "red at 4 o'clock", the content is "seems to be red at 4 o'clock", or,
maybe, even more explicitly, "I'm detecting red at 4 o'clock". We now have
back the analysis of "is conscious" in terms of "is conscious of", but
without the need to appeal to an occurrent accompanying higher-order state.
I have three problems with this view.
The first is actually not aimed at the dispositionalism of the view, but
applies to any version of HOR. It's crucial to Carruthers's case, especially
against FOR, that it be possible for there to be unconscious experiences.
Again, this is a feature of any version of HOR. Now Carruthers makes a
strong case that there are unconscious analog perceptual states - and of
course if you want to stipulate that these be called "experiences", there's
no stopping you. But I think it's clear that in conscious experience we
encounter a property that is essentially phenomenal, in the sense that
its instantiation is "for" a subject. Put another way, pains, tickles,
and sensations of red all seem to be essentially modes, or bits of conscious
experience. Of course you can have the representation of bodily damage
or distal color properties without any phenomenal feel, no one doubts that.
But whether such registrations of properties would be the same sort of
event as a painful feeling, or a reddish experience, seems highly doubtful.
I present this objection to register
my disagreement with Carruthers on this crucial point. But I mention it
only to put it aside, since Carruthers, with some justice, can accuse me
of begging the question against him. After all, he denies there are qualia
in the sense of intrinsic properties of experience, and my objection appeals
to such qualia. Unfortunately I can think of no clearly non-question-begging
way to put the objection. If I do, I'll let you know. (For a discussion
of how the burden of argument plays out here, see Levine 2001, chapter
5).
To state my second problem I refer
back to my remarks above about the conceivability argument. Appeal to a
disposition to form HOR's doesn't really explain phenomenal consciousness,
and the conceivability of having the former without the latter is evidence
of this. Of course Carruthers has his general reply to the conceivability
argument, and I have already commented on that. The reason I bring this
up again here is that it might seem as if his argument above concerning
the role of recognitional concepts in explaining the appearance of an explanatory
gap is strengthened by his account of the dual content expressed by analog
perceptual states when attached to an HOR faculty. So it is only this aspect
of his response to conceivability considerations that I want now to address.
Carruthers argues that just as one
can form a recognitional concept of distal features like red on the basis
of one's first-order analog states, so too one can form a recognitional
concept of experience-of-red on the basis of one's higher-order representations
of one's first-order analog states. These, then, are the recognitional
concepts appeal to which allegedly accounts for the appearance of an explanatory
gap.
Now part of my objection to this
move above had to do with what I claimed was a clear difference between
the cognitive relation that obtains between the recognitional concept of
"bare-color" and our standard cognitive relation to experiences of red
and green. I claimed that the former lacked the substantive and determinate
character of the latter, which pointed to a kind of cognitive intimacy
not provided for in the relation between recognitional concept and what
it is a concept of.
However, it might seem now that Carruthers
has a way of addressing this concern. It's not that the recognitional concept
which is applied when I am aware of my own experience of red is a distinct
representation from its first-order target, but rather it is one and the
same representation, though its secondary content. When my first-order
analog state simultaneously says "red at 4 o'clock" and "seems red at 4
o'clock", it is simultaneously applying the concept "red" to the distal
object (or spatial region) and the concept "is a red appearance" (or a
"red-seeming") to itself. This reflexivity in the secondary content might
plausibly account for the sort of cognitive intimacy that seems apparent
in our apprehension of our own sensory experiences.
It isn't clear to me if Carruthers
in fact takes this line, but, whether or not he does, I don't think it
holds up. The problem is that the thoughts we have when entertaining conceivability
thought experiments, or puzzling about the explanatory gap, are genuine
thoughts, with full-fledged concepts as constituents. That we might form
recognitional concepts of first-order perceptual states within the representational
system of the HOR faculty makes perfect sense. But then these representations
are distinct from those in the first-order perceptual faculty. When I think,
"Oh, another one of those sorts of experiences again", it isn't the experience
itself but a thought about it that's in play.
This leads naturally to my third
problem, which relates directly to the idea that because of their effects
down-stream, first-order analog perceptual states take on a secondary higher-order
content. The proposal just seems ad hoc. I don't see why, merely by virtue
of its availability to a higher-order faculty, a state's content
should thereby take on some of that higher-order character itself. What
basis is there for assigning this secondary higher-order content when interpreting
the system as a whole, instead of merely leaving the analog perceptual
states with their first-order contents and assigning the higher-order cognitive
faculty the content that there's been a representation of a certain distal
feature reported by the perceptual faculty? I see nothing in the consumer
semantics that forces the former interpretive scheme, and therefore it
seems merely ad hoc to insist on it.
The idea, as I understand it, is
that once the down-stream cognitive faculties are capable of discerning
the difference between something's seeming red (i.e. causing the red-registers
to go off) and really being red, then the consumer, these cognitive faculties,
come to interpret (as it were) the red register going off as meaning "seems
red". Consumer semantics, after all, is a matter of the significance of
the representation for the consumer - in this case, higher-level cognitive
faculties. But given the lack of any syntactic markers of reflexivity in
the analog perceptual states, it seems a much better overall interpretation
of the system to assign the usual distal contents to the first-order analog
states and locate the appreciation of their constituting a "seeming" in
the representational states of the higher-order faculty itself, where the
appearance-reality distinction is explicitly appreciated. What you get
then is a down-stream state saying "seems red at 4 o'clock", which, in
effect, can be glossed as, "a report of red at 4 o'clock coming from the
visual faculty". Pushing the "seeming" all the way down into the perceptual
contents themselves seems gratuitous.
Consider the following analogy. I
wear my watch all the time and rely on it constantly to give me the time
of day. Of course, I know that watches are notoriously prone to run fast
or slow, so I take its reading with a grain of salt. When asked what time
it is, I can imagine being in a careful enough mood to say, "well, by my
watch it's 2:15", or "it seems to be 2:15". Does that mean that my watch
is reporting on its own state when the big hand is on the three and the
little hand on the two? I think my watch is indicating the time, and then
I qualify the content of my own representation when I don't completely
trust it. Of course there is a somewhat metaphorical sense in which the
watch is also telling me that its hands are in the requisite positions.
But that isn't a matter of its content, but rather of its being so - its
having its hands where they are - and my ability to detect that fact. My
detecting that fact is where to locate the representational content that
says the hands are where they are. Similarly, that it's seeming to me as
if there's red at 4 o'clock is a content that is appropriately located
in the representation that detects the first-order perceptual state; not
in the first-order state itself.
Though I'm no fan of consumer semantics,
let me emphasize here that my objection is not an objection to this sort
of semantics per se. I take the point of positing narrow contents, including
consumer-based ones, to be to overcome the Frege problem. By appeal to
inter-representational relations one can more finely individuate contents
and thereby account for differences in cognitive significance that don't
track differences in reference. Let that be so. Indeed, let it be the case
that we can use these down-stream relations to distinguish the perceptual
contents of two creatures both of whose perceptions are tracking the same
distal feature. In particular, let it be that "red at 4 o'clock" in a creature
with an HOR faculty has a different narrow content from its corresponding
state in a creature without such a faculty, because narrow content is a
function of, among other things, intra-mental relations. Still, it's a
big leap from acknowledging this difference to then saying that the one
attached to the HOR faculty is itself partly higher-order. Why can't the
mere difference in its causal relations be the difference in content? What
makes it necessary to put a reflexive gloss on that difference? I don't
see the motivation, except that it helps with the theory of phenomenal
consciousness. But that's precisely why it's ad hoc.
In this commentary I've criticized
a number of aspects of Carruthers's view. As one with mysterian leanings,
it's not unexpected that I would find material with which to disagree.
But this shouldn't obscure my appreciation for the impressive achievement
this book represents. By pressing the anti-mysterian case as forcefully
and honestly as he does, with such a wide array of philosophical and psychological
argumentation to back him up, Carruthers has done the "consciousness business"
a great service.
Department of Philosophy
Ohio
State University
Homepage
References
Block, N. (1990). "Inverted Earth", Philosophical Perspectives, 4: Action Theory and Philosophy of Mind, J. Tomberlin, ed., Atascadero, California: Ridgeview Publishing Co., 53-80.
Byrne, A. (1997). "Some Like It HOT", Philosophical Studies, 86, 103-129.
Chalmers, D. (1996). The Conscious Mind. Oxford: Oxford University Press.
Kripke, S. (1980). Naming and Necessity. Cambridge, MA: Harvard University Press.
Levine, J. (1983). "Materialism and Qualia: The Explanatory Gap," Pacific Philosophical Quarterly 64, 354-361.
(1989). "Absent and Inverted Qualia Revisited", Mind & Language, vol. 3, no. 4, 271-287.
(1993). "On Leaving Out What It's Like", in Davies, M. and Humphreys, G., eds., Consciousness: Psychological and Philosophical Essays, Oxford: Blackwell, 121 - 136.
(1998). "Conceivability and the Metaphysics of Mind", Nous 32, 449-480.
(2001). Purple Haze: The Puzzle of Consciousness, Oxford Univ. Press.
Neander, K. (1998). "The Division of Phenomenal Labor: A Problem for Representational Theories of Consciousness", in J. Tomberlin, ed., Philosophical Perspectives 12: Language, Mind, and Ontology, 411-434.
Perry, J. (1979). "The Problem of the Essential Indexical". Nous 13, 3 - 21.
Rosenthal, D. (2000). "Metacognition and Higher-Order Thoughts", Consciousness and Cognition, 9, 231-242.
Shoemaker, S. (1984). Identity, Cause, and Mind. Cambridge: Cambridge University Press.
Tye, M. (1995). Ten Problems of Consciousness: A Representational Theory of the Phenomenal Mind. Cambridge, MA: Bradford Books/M.I.T. Press.
© Joseph Levine