« Internalism and empty thoughts
Theories of Perception and the Existence of Qualia »

Roving Minds: Pereboom on Consciousness and Introspective Inaccuracy

Last semester I attended two conferences: the Physicalism and Beyond workshop at the Berlin GAP.6 conference, in September, and the Alabama Philosophical Society’s 2006 meeting, here at The University of Alabama, in October. A social highlight of the Berlin conference was finally meeting Sven Walter, with whom I co-edited Phenomenal Concepts and Phenomenal Knowledge: New Essays on Consciousness and Physicalism (OUP, 2007). The APS included presentations by philosophers from eight different states, including brainpains.com’s own Robert Howell. Not too shabby for a state philosophy conference in the Bible Belt!

The talks at both conferences made my mind rove all over the place. I’ve been thinking about one talk at the APS: Derk Pereboom’s keynote address, “Consciousness and Introspective Inaccuracy” (which is on his website). He claims that the knowledge argument depends on the accuracy claim: the claim that an introspective mode of presentation accurately represents the qualitative nature of the corresponding phenomenal property. In other words, the claim is that an experience’s phenomenal properties really are as they phenomenally seem (or as the experience presents them as being). He thinks that the physicalist can reasonably reject the accuracy claim.

According to Pereboom, it is an open possibility that the accuracy claim is false—that phenomenal properties are not as they are introspectively represented. (He qualifies that view, but not in ways that affect what I will say.) And he suggests that this open possibility provides the physicalist with a reasonable response to the knowledge argument:

While Mary is in the room, she does not represent a phenomenal-red sensation in the characteristic introspective way. But it is a serious open possibility that by virtue of her physical knowledge she nevertheless accurately represents the complete real nature of this phenomenal state and its properties. For the qualitative nature of the property of phenomenal redness, in particular, might not be as it is introspectively represented. Instead, it might be accurately represented by virtue of Mary’s physical knowledge. Accordingly, from her physical knowledge she might then be able to derive every truth about the real nature of the phenomenal state, despite the fact that this physical knowledge gives her no access to the phenomenal state as it is introspectively represented.

I suspect Pereboom is right that the knowledge argument depends on the accuracy claim (or on something very much like it). If it does, then in identifying that assumption he advances the discussion.

However, I do not think his response to the knowledge argument works, at least not as it stands. The accuracy claim is plausible. Phenomenal redness may fail to resemble any feature of ripe tomatoes; redness may fail to be the way it appears phenomenally. But how could phenomenal redness fail to be the way it appears phenomenally?

Pereboom thinks such a failure is an open possibility, in a sense of “open possibility” that he explains as follows:

Of the many notions of epistemic possibility, the sense I here have in mind is the usual possible for all we know; that is, possible given what we human beings now know. (The relevant “we” in this case are perhaps those who have thought carefully about these philosophical issues) For this sense of epistemic possibility, I will use the term ‘open possibility’.

I think we do now know that the accuracy claim is true. So, I do not think that it is an open possibility (in Pereboom’s sense) that the accuracy claim is false. Why does Pereboom think otherwise?

He describes various theories on which the accuracy claim comes out as possibly false. For example, he describes a causal account of phenomenal representation that “gives rise to the open possibility that phenomenal concepts are qualitatively inaccurate.” Indeed, he describes several accounts that do not support the accuracy claim. He also describes accounts that do support the accuracy claim, such as the view that phenomenal properties partly constitute phenomenal concepts. But, he thinks, we should not be confident that an account that supports the accuracy claim is correct. He observes that no such view has been developed thoroughly and convincingly enough to rule out the alternatives, such as a causal theory. He takes this to indicate that the falsity of the accuracy claim is consistent with what we now know.

However, the accuracy claim is independently plausible. It does not derive its plausibility from any particular theory of phenomenal representation, such as the view that phenomenal properties partly constitute phenomenal concepts. On the contrary, the latter view is plausible because it explains the accuracy claim. Further, causal accounts of phenomenal representation of the sort Pereboom describes are not very plausible, for familiar reasons. Indeed, the Mary case helps to reveal their implausibility. She knows all the relevant causal facts before leaving the room but, intuitively, she learns more truths about phenomenal representation when she leaves. Pereboom defends the viability of causal accounts by invoking analogies to natural kind concepts. But these analogies too have familiar shortcomings, which the Mary case also helps to bring out.

Thus, I think Pereboom must do more than he does in his paper to establish his premise that it is an open possibility that the accuracy claim is false.* Of course, we are not 100% certain that the accuracy claim is true. But that observation obviously cannot ground a physicalist response to the knowledge argument. If Pereboom could show that a causal theory of phenomenal representation is more plausible than the claim that Mary’s pre-release knowledge is incomplete (or more plausible than the claim that the complete truth about what it’s like to see in color is not a priori deducible from the complete physical truth), then that would be another story. I doubt that this can be shown. But I think he would have to do something along those lines in order to make his response to the knowledge argument convincing. Still, the paper is rich, innovative, and well worth careful study.

[*Footnote: Pereboom adduces a few other considerations in support of his premise that it is an open possibility that the accuracy claim is false. For example, he suggests that the intuitive force of the accuracy claim may derive from the fact that the discrepancies between the nature of a phenomenal property and how experience presents it are rare. I do not find that suggestion plausible, but I will not go into this here. He doesn’t place much weight on these other considerations and this initial blog entry is already long enough.]

Posted by Torin Alter on Wednesday, January 31st, 2007 at 11:52 am.

14 Responses to “Roving Minds: Pereboom on Consciousness and Introspective Inaccuracy”

Torin,

Thanks for the post – I think yours is a good way to argue against what I say in the paper. Here it is, by the way:

http://www.uvm.edu/~phildept/pereboom/INTROSP8.pdf

First, I consider it a mark in favor of my view that you don’t consider my qualitative inaccuracy response to be in some sense mistaken in principle, and that we’re instead arguing just about the relative plausibility of various hypotheses. I see the main point of the paper as putting a response to the knowledge argument on the table which isn’t subject to an objection that shows that it can’t work in principle, but is instead an open possibility for which the question left open is precisely its plausibility.

To address this issue, in the most recent version I add the following thoughts:

One might think that it remains implausible to claim that the qualitative inaccuracy for our introspective representations of phenomenal properties is as universal and extensive as would yield a promising materialist response to the knowledge argument, and this should provide resistance to the belief that this claim is true. But it may be that all of the developed positions on the metaphysics of consciousness have implausible features that should make for at least some resistance to belief. Karen Bennett, in “Why I am not a Dualist” (manuscript, here: http://www.princeton.edu/~kbennett/papers.html) argues that the traditional dualist position has such an implausibility: it accepts that there exist a fairly large number of psychophysical laws that are brute in the sense that there is no explanation as to why they hold, or for which the explanation we can envision is arbitrary divine preference (Locke suggests and Robert Adams endorses the divine preference explanation). Supposing this is right, should one be less resistant to the traditional dualist view than to the qualitative inaccuracy suggestion? Other non-physicalist hypotheses, such as neutral monism, Berkeleyan idealism, and panpsychism, whatever their merits, also have features that seem implausible to many. Should one be less resistant to such views than to qualitative inaccuracy? A thorough reflective equilibrium procedure requires us to take into account any implausibilities that accepting the accuracy claim would lead to endorsing. By analogy, in the free will debate, one might initially think that the consequence and manipulation argument against compatibilism are strong enough to dislodge it, but rejecting compatibilism on their basis forces one to accept either libertarianism or to deny free will, each of which is in some respect also seriously implausible. Here too one needs to count the overall cost. And when one counts the overall cost in the metaphysics of consciousness, it may be more plausible to deny qualitative accuracy than to endorse one of the positions that its acceptance leaves open. This is as much as I hope to establish in the paper.

Left by Derk Pereboom on February 1st, 2007

Derk and Torin, a question about the accuracy claim (but first, thanks for the interesting post, Torin; and thanks to both for the exchange!). Here are a couple of putative versions of the claim, ordered in terms of apparent strength:

(AC1) introspection does not misrepresent the qualitative nature of phenomenal property P.

(AC2) introspection (correctly) represents the essential features making up the qualitative nature of phenomenal property P.

(AC3) introspection (correctly) represents all the features making up the qualitative nature of phenomenal property P.

First, which of these is supposed to be at play in the knowledge argument?

Second, here’s an argument that clearly seems to undermine (AC3). But I suspect it undermines (AC2) as well, and it may be turned into an argument against (AC1). If this is right, and if the 3 claims above provide the best ways to cash out the accuracy claim, then, contra Torin, we have reason to think the accuracy claim is false after all. The argument seems fine, quite plausible in fact, and it doesn’t seem to rely on any assumption that’s at stake in the dispute between Physicalists and anti-Physicalists.

The argument exploits the familiar sorites series of coloured patches ranging from red to yellow, where each adjacent patch, although of a slightly distinct chromatic shade from its neighbours, is nevertheless visually indiscriminable from them. (This is meant to be analogous to Williamson’s pain-from-early-morning-to-noon example, which is relevant here too.)

Suppose you’re walking in front of the series from one side to the other (or if you worry about the temporal aspects of the case, suppose you’re seeing the whole series at once from a certain distance but still close enough). Presumably, as you walk along the series, you have distinct experiences, representing distinct colour shades. And presumably, each distinct experience (or pairs thereof) has a distinct phenomenal character.

The important point is that, when you’re walking along the series, just as the colours of adjacent patches are visually indiscriminable, it seems that the phenomenal character (or properties) of successive experiences is introspectively indiscriminable. And if this is possible, then two experiences could have distinct phenomenal properties, with distinct qualitative natures, even though you cannot introspectively discriminate between their qualitative natures.

But this suggests that (AC3) is false: since you cannot introspectively discriminate the distinct qualitative natures of two distinct phenomenal properties, then introspection must not be representing all the features making up such qualitative natures.

Provided that the features responsible for the difference in qualitative nature between two distinct phenomenal properties are essential for the individuation of the qualitative nature of these phenomenal properties, then this example seems to undermine (AC2) too.

But it could also be construed as undermining (AC1). If introspection doesn’t represent the difference in qualitative nature between two different phenomenal properties, it must be representing two such phenomenal properties as being the same. By hypothesis, though, they must be distinct. Hence, introspection must be misrepresenting at least one the qualitative natures of such phenomenal properties.

Hence, we’ve got a reason to reject the Accuracy claim. Now, there may be (familiar) ways to resist the argument. And it may be that this way to reject the Accuracy claim still doesn’t provide the means to resist the knowledge argument, as one of you mentioned, I think. Nevertheless, we have some pretty good reason to think the Accuracy claim is in fact false, contra Torin’s suggestion that we know it’s true.

Left by Philippe Chuard on February 1st, 2007

Derk,

Thanks for your reply. I agree that I haven’t tried to show that your qualitative inaccuracy response is incoherent or anything like that. And I agree that, in the end, we must assess theories by weighing their costs and benefits—and that if a premise in an argument leaves open only implausible theories, then the premise inherits their implausibility. But I stand by what I said. Here are some further reactions:

Granted, other responses are worse than implausible. For example, the acquaintance response may rest on an equivocation fallacy (I argue that it is, in my 1998 Phil Studs paper). However, the significance of the implausibility problem should not be underestimated. Arguably, the only real problem with the ability hypothesis response is that it’s implausible. Its defenders could always bite the bullet, but this may amount to denying the existence of that which needs to be explained. I have the same concern about the qualitative inaccuracy response.

I agree that, plausibility notwithstanding, the qualitative inaccuracy hypothesis is not mistaken in principle. But I do think that your *arguments* for the hypothesis (if I understand them) are mistaken. In particular, you seem to base the conclusion that an experience’s phenomenal properties might not be as they phenomenally seem partly on the premise that a causal theory of phenomenal representation might be true (where “might” expresses open epistemic possibility—consistency with what we know). I think that reasoning gets things the wrong way around. If causal theories have the consequence that (given what we know about the likely causes of experience) they undermine the accuracy claim, then that’s a reason to reject those theories. This isn’t just a point about plausibility. The point is that we shouldn’t let the phenomenal data be too beholden to our theorizing. And I take that data to include the idea that phenomenal redness corresponds, at least roughly, to how experiences of seeing red phenomenally appear.

It seems unsatisfactory to counter my criticism by bringing in the implausibility of non-materialist theories. This is not only because I did not here defend any particular theory, let alone non-materialism. It is also because the range of theories left open by the accuracy claim is extremely wide. In particular, this range includes many versions of materialism, including both a priori (type-A) and a posteriori (type-B) varieties. Further, I’m not sure that “the traditional dualist position” is relevant. Contemporary dualist views don’t tend to invoke a large number of basic psychophysical laws. In his 1996 book, Chalmers speculates that there may a small number of basic psychophysical law—maybe only one—from which others are derivable. This is not to deny that non-materialist theories have serious shortcomings; many, perhaps all, do. But why compare contemporary materialism with traditional dualism rather than to contemporary non-materialist theories?

To my mind, the qualitative inaccuracy response can’t be adequately defended by appealing to the virtues of materialism, the vices of non-materialist theories, the history of counter-intuitive theories prevailing in science, or anything of this general nature. To adequately defend the qualitative inaccuracy response, I think you need defend the qualitative inaccuracy view more directly—and not just as a coherent alternative, but as a credible one. Until you do that, this option will—for me, at least—remain nothing more than a mere theoretical possibility: a coherent but implausible option.

Of course, if Philippe argues that my claims about the qualitative inaccuracy view are wrong anyway. I disagree, but I’ll respond to that separately.

Best,
Torin

Left by Torin Alter on February 8th, 2007

Philippe,

Thanks for the thoughtful reply. Actually, the accuracy claim that I had in mind is weaker than any of AC1-3. AC1 says that introspection does not misrepresent the qualitative nature of phenomenal property P. I take this to mean that introspection does not *in any respect* misrepresent the qualitative nature of P. The accuracy claim says something weaker: that an experience’s phenomenal properties really are as they phenomenally seem. In other words, introspection doesn’t get it wholly wrong—phenomenal properties (non-trivially) resemble their introspective representations. This claim is consistent with there being some misrepresentation, including the sort that your Williamson-esque argument brings out. What the accuracy claim rules out is the possibility that phenomenal properties are utterly different (bear no non-trivial resemblance to) how they seem phenomenally, from the point of view of the experiencer. And that possibility, I think, we can safely rule out.

I probably should have clarified this in my first post. Derk clearly distinguishes the accuracy claim from AC3 in his paper (in a footnote about revelation). But I suspect he will agree that the relevant assumption is not AC1 or AC2. If I’m wrong, then I’ve misunderstood him.

I think your Williamson-esque arguments AC1-3 are pretty powerful. But I don’t think they create problems for the accuracy claim.

Left by Torin Alter on February 9th, 2007

Hi Torin, I agree entirely. So your claim is not:

(a) we know that introspection accurately represents the qualitative nature of phenomenal property P.

Rather, it’s something like:

(b) we know that introspection doesn’t completely misrepresent the qualitative nature of phenomenal property P (and represent it accurately to some important extent).

Left by Philippe Chuard on February 9th, 2007

Really interesting idea. The Knowledge Argument does seem to rely on the Accuracy Claim (AC), but I don’t see how there’s any way AC could turn out to be false. Phenomenal properties are, by definition, what they appear to be, no? Also, if AC’s false, it creates larger problems than it solves. If AC’s false, we’re further away from understanding consciousness than if the Knowledge Argument’s true, if not in principle prevented from ever understanding it.

Left by Dylan Murray on February 19th, 2007

Hi Dylan,

I agree with most of what you write. But a nitpick: I don’t think AC is true by definition. If it were, then Derk could simply use a different definition.

Left by Torin Alter on February 20th, 2007

Hey Torin,

Fair enough, but is any other definition feasible? The phenomenality of a property just is the seemingness. Say we float a different definition, ’schmenomenality’, that we stipulate isn’t necessarily about the seemingness of properties. It seems this case of the schm- strategy fails; haven’t we stipulated away precisely what was at issue?

Left by Dylan Murray on February 21st, 2007

Hi Dylan,

I think I agree, with two qualifications. First, the seemingness here must be understood as phenomenal appearance. I’m sure you intended this. But “seems” is notoriously ambiguous between phenomenal and epistemic senses, so I thought I’d mention it. Second, I don’t think the schmenomenality case shows that the accuracy claim is true by definition, unless “by definition” is understood in an unusually broad way (e.g. as including any a priori truth).

Left by Torin Alter on February 21st, 2007

Hi Torin,

Yes, “seemingness” in the sense of phenomenal appearance. Apologies for the ambiguity. I still don’t see how the accuracy claim isn’t true by definition, though. The schmenomenality case is intended to show this by eliciting just what a stipulation of the definition of “phenomenal appearance” would have to be in order to entail the accuracy claim’s possibly being false. To open up that possibility, wouldn’t the stipulation have to be something like:

1. phenomenal appearances aren’t necessarily how they appear

or:

2. what it feels like isn’t necessarily what it feels like

Left by Dylan Murray on February 21st, 2007

Hi Dylan,

Thanks for the thoughts. I’m not sure about your 1&2. How about:

(~D) Phenomenal properties aren’t by definition how they phenomenally appear.

Here are a couple of thoughts that might make what I’m saying clearer:

The accuracy claim concerns the relation between what Derk calls, on the one hand, “introspective modes of presentation” and “phenomenal concepts” and, on the other hand, phenomenal properties. Neither 1 nor 2 appear to concern that relation. 2 seems to be a claim about phenomenal properties, not concepts. 1 might be about properties and it might be about concepts, but it doesn’t seem to concern the relation between the two. So, that’s 1 reason I don’t find 1 & 2 helpful here.

Also, here’s a comparison that might help (it might hurt too, but I’ll chance it). It’s necessarily true that the positive cube root of 8 is 2. This is a priori and metaphysically necessary. It couldn’t turn out otherwise. And it’s pretty obvious, once you understand the the concepts involved. Is it true by definition? Perhaps in a broad sense of “definition”, but this characterization seems more misleading than helpful. The “true by definition” description encourages the idea that the relevant mathematical truth is a trivial consequence of how we use mathematical terms, and that idea seems implausible to me. Likewise, I say, the accuracy claim is probably a priori and metaphysically necessary. And it’s pretty obvious, once you understand the concepts involve. But to say that it’s true by definition is misleading, for similar reasons.

Perhaps we’re just talking past each other. Do you mean something meatier by “true by definition”?

Left by Torin Alter on February 22nd, 2007

Hi Torin,

I understand what you’re getting at far better, and it’s certainly helped to clarify my own thoughts on the matter quite a bit. Thank you. I’ll try to put my thoughts and questions here a bit better:

However phenomenal is defined, the antiphysicalist can stipulate the KA to run in terms of that definition.

Say that, on some occasion, we do introspect innacurately, as in we misrepresent the property we’re trying to introspect. So, the property being introspected isn’t actually phenomenal. Rather, some property of the introspection is phenomenal (whether a property of what’s doing the introspecting or the act of introspecting itself), and it’s That property that the antiphysicalist will tweak the KA for, claiming that That’s what Mary can’t learn about while confined. (Part of what was meant by the “true by definition” thing, though you’ve dissuaded me some)

Of course, we could then claim that our introspection of That (2nd order) property of introspection might be innacurate, but the innacuracy itself is novel, the antiphysicalist will say it’s phenomenal, target THAT (3rd order) property with a new 3rd order KA… off we go.

This is another thing I’m hung up on. If what’s supposed to be phenomenal is the introspection itself… well, what then? By what means do you challenge introspection but by introspection? If consciousness just is introspection, how do we ever get behind it? Any potential counter-theory that might challenge it would still have to be understood through a lense of introspection. I agree with something Derk said earlier in this post, that “a thorough reflective equilibrium procedure requires us to take into account any implausibilities that accepting the accuracy claim would lead to endorsing” but I wonder that the accuracy claim is already presupposed by the equilibration process. What else would be doing the equilibrating?

Left by Dylan Murray on February 23rd, 2007

Hi Dylan,
Thanks for the further thoughts. I don’t think we disagree about your first point, though I would put it differently. I’d say: as long as genuine phenomenality comes into the picture, fiddling with definitions of “phenomenal concept” and “phenomenal property” won’t substantially affect the knowledge argument. I take it that this is what you mean by, “However phenomenal is defined, the antiphysicalist can stipulate the KA to run in terms of that definition.”

But I can’t accept your formulation: someone might define the phenomenal in functional terms, a la Lewis-Armstrong analytic functionalism, for example. Given that definition, the knowledge argument won’t work (and the response to that move is to reject the functionalist definition for failing to capture genuine phenomenality). That’s just a nitpick about your formulation, I think.

Regarding your second point: I’m not quite sure what you’re driving at. First, I don’t think consciousness can be plausibly identified with introspection. Second, I don’t understand the point about presupposition.

Left by Torin Alter on February 26th, 2007

Hey Torin,

Yes, absolutely that’s a better way of putting the point. Thank you. A little nitpick of my own, though: I’m not sure our formulations come out so different in the end. If (indeed when) someone attempts to define the phenomenal in functional terms, the response is to say they’ve ‘failed to capture genuine phenomenality’, as you rightly say. This looks very similar to the objection I made on Feb. 21st about definitions of “phenomenal” on which the accuracy claim would come out false. Namely, if we employ any such definition, “haven’t we stipulated away precisely what was at issue?” At any rate, that’s all mostly legislation. I do prefer your formulation.

As for the second point… a bit off-topic perhaps, but couldn’t resist. The idea is that phenomenal consciousness is inescapable when cognizing. It seems that you cannot have a thought without it being conscious. (A quick stiff-arm: We might have unconscious thoughts but the only time they ever become at all important, particularly in such a way that they can factor into theorizing, is when we become consciously aware of them. So I think they can safely be ignored relative to the idea here.) In particular, this seems to come into play in relation to the ‘phenomenal concepts strategy’ debate…

We experience understanding. In particular, we experience the understanding of explanations afforded by physical concepts, eg. “aha! That’s how Sodium-Potassium pumps work!” On the other hand, we do not understand phenomenal consciousness. In particular, we do not physically understand phenomenal concepts or experience. In other words, it seems phenomenal concepts can refer to nonphenomenal concepts but not vice-versa; back to the hopefully now better fleshed out comment that phenomenal consciousness seems inescapable when cognizing.

Left by Dylan Murray on February 26th, 2007

Something to say?