To PANIC or Not to PANIC? - Reply to Byrne

by Michael Tye e-mail

Department of Philosophy
The University of Texas at Austin
Austin, Texas 78750

1)   Alex Byrne opens his insightful comments by suggesting that the Müller-Lyer illusion provides a straightforward counter-example to my causal co-variational account of the representational content of sensory states. Byrne notes that if optimal conditions for vision involve “the various components of the visual system operating as they were designed to do in the sort of external environment in which they were designed to operate” (as I say they do, for the case of evolved creatures, in Ten Problems), then the Müller-Lyer illusion occurs under optimal conditions.  Yet in this case, one’s visual experience does not represent that the two lines are of the same length, as the account of content as causal covariation under optimal conditions would seem to predict.  Rather, it represents incorrectly that the lines are of different lengths.

In Consciousness, Color, & Content, I say a little more about optimal conditions than I did in the earlier book. On the latest development of the causal covariation proposal, optimal conditions are essentially ‘no interference’ conditions. Given this understanding, my remark quoted above about optimal conditions is best taken to be like the remark that rocks fall to earth when dropped. Typically this is the case, but there are rare exceptions, as when helium balloons are attached to the rocks.

Understood in this way, optimal conditions do not obtain when one views the Müller-Lyer diagram. The standard explanation of the illusion is that, given our experiences with edges and corners in the external environment, the differing arrow heads at the ends of the two parallel vertical lines lead our visual systems to represent, at a subpersonal level, the left hand vertical line as further away than the right (the left hand line is ‘taken’ in three-dimensional projection to be the edge between two surfaces receding from the subject, the right one is ‘taken’ to be the edge between two surfaces sloping towards the subject). But the vertical lines are actually of the same length and so their projections on the retina are of the same length too. The difference in distance away of the lines is compatible with the same retinal projection only if one line is longer than the other. Accordingly, our visual systems infer that tne left hand line is longer, and this is what we experience. Obviously this account has it that there is a kind of interference created by the differing directions of the arrow heads in our processing of the distance away of the lines. So, conditions are not optimal.

2)   My PANIC theory of phenomenal character requires of a state with phenomenal character that its content be suitably poised.  Byrne’s second objection concerns this condition. Before I turn to it, a quick comment on phenomenal character. Byrne says that we should agree at the outset that the phenomenal character of a state is a property of that state. I don’t see why we should so agree. To be sure, we talk of states having phenomenal character, but nothing in ordinary usage or thought commits us to the view that phenomenal characters are properties. After all, we talk of beliefs as having content and of words as having meaning, but we don’t take belief contents or word meanings to be properties of beliefs and words respectively. Why do that from the outset for phenomenal character? Buy into the Cartesian view of experiences as inner ideas or pictures viewed by an inner eye and it may be natural to take the ‘feel’ of an experience as a property of the idea or picture. But that isn’t commonsense. It’s philosophical dogma — precisely the dogma which representationalism opposes.

So, what is the objection about poisedness? Essentially, Byrne’s claim is that the condition isn’t strong enough for my purposes. States can have PANIC without having any phenomenal character. In making this charge, Byrne relies on my characterization of poisedness as being the property of standing ready and available to make a direct difference with respect to beliefs and/or desires.

3)   I plead guilty to failing to develop the poisedness condition fully enough but not guilty to imposing a condition that is inherently too weak. Let me focus first upon the example of visual experience. Intuitively, visual experiences are not themselves beliefs but they are apt for the production of beliefs. Admittedly, some states that might reasonably be classified as visual experiences, for example, seeing that the table is covered with books, already involve beliefs or belief-like states. But such states, in my view, are hybrid, having a visual experience proper and a belief or thought as components.

Visual experiences proper are not apt for the production of any old beliefs, however. Intuitively, each visual experience is the direct basis for the formation of a belief about the perceptible qualities represented by the experience. Each experience is also, in creatures equipped with the capacity to introspect, the direct basis for the formation of beliefs about the experience and its content.

The content of the visual experience proper supplies the input to the relevant belief-forming processes, where the role of the belief-forming processes is to generate beliefs of the sorts just described. But the appropriate beliefs are not always formed, of course; for introspection can malfunction and, at least in the case of external belief formation, other background beliefs can interfere. There is also the possibility that attention is not appropriately directed.

A visual experience has a poised content, then, in my view, so long as it is apt for the production in the right ways of the right beliefs. In the case of bodily experiences, desire is also relevant. The experience of pain, for example, is the direct basis for the the desire to protect oneself, to avoid damage.

4)   We can now see why Byrne’s proposed counter-examples fail. A green, circular after-image experience has a poised content since it is apt to produce the belief that one is having such an experience and further it is apt to produce the belief that something green and circular is present (albeit something filmy and floating in space), since it would generate the latter belief, were there no interference from background beliefs. Going by the experience alone, one is inclined to believe that something green and circular is present. In the waterfall illusion, one has an experience of something moving and not moving. This experience has a poised content too. For again it is the direct basis for the appropriate introspective belief. And even though the experience (in virtue of its content) does not cause the belief that something is moving and not moving, as one directs one’s attention to the rocks by the side of the waterfall, it would do so, if one were to take the experience at face value without being influenced by other backround beliefs about the impossible.

Mestre’s cortically blind patient undergoes visual states whose contents are not poised, however. For he cannot form beliefs about the contents of those states via introspection even though he has the usual human power to introspect and there is nothing wrong with his introspective mechanisms. His visual states are not sufficiently ‘close’, as it were, to some of the relevant belief-forming processes.

The blindsight subject who undergoes an experience of something red, round, and bulgy, as he views a ripe tomato, and who forms no beliefs about what he sees, is in a visual state that triggers the desire to eat (via, we are to suppose, “some quirk of his inner wiring”). But, contra Byrne, that state’s content is not poised, as explained above.

5)   Byrne’s next objection concerns my abstractness requirement on phenomenal content. He says that “there is no evident reason why the content of perception couldn’t be object-dependent” and he takes this to show that for possible creatures whose perceptual states are of an object-dependent sort, on the PANIC theory, there is nothing it is like for them to perceive things. This, he claims, is an unacceptable result.

Consider my hallucinating a pink rat. It should be uncontroversial that my perceptual state is delusive.  But it cannot be delusive, unless it has an inaccurate representational content. Such a content clearly is not object-dependent, for there is no (relevant) object — I am hallucinating. This perceptual content, then, is abstract.

Since arguments from transparency and the intensionality of phenomenal discourse seem to me to provide strong reasons for thinking that phenomenal character is representational content of some sort and since perceptual states have phenomenal character whether or not their subjects are hallucinating, we have, I suggest, strong reasons for thinking that phenomenal character is representational content of an abstract sort, whatever further conditions are necessary.

Of course, this line of argument presupposes that there is no difficulty in holding that abstract perceptual content attaches to states involved in cases of veridical perception. But why shouldn’t we hold that?  Even if veridical perceptual states have object-dependent contents, it does not follow from this that they don’t have abstract contents too. If I see a picture as a duck, then my visual state has a conceptual content, but it doesn’t follow that it lacks any nonconceptual content. There are, it seems to me, many layers of perceptual content [1]; and the possession by a perceptual state of one of these layers does not preclude it from having others.

6)   Byrne notes that I hold what he calls “the content view” of nonconceptual content and not “the state view”. On the content view, nonconceptual content is content of a sort that cannot be had by beliefs. He goes on to argue that, on this view, “the PANIC theory is seriously underdescribed.” But what exactly is the problem? Here is what Byrne says:

Now, although it might be that the PANIC theory supplemented with a well-worked out version of the content view can explain why beliefs lack phenomenal character, and why perceptual experiences have it, the immediate problem is that Tye has supplied no good reason in favor of the content view.

First, some remarks on content. In my view, phenomenal ‘looks’ contexts are intensional to the following extent: in the context ‘looks F’, where ‘looks’ is used phenomenally, ‘G’ can safely be substituted for ‘F’ so long as ‘F’ and ‘G’ pick out the same property. [2] In belief contexts, the intensionality runs deeper. Beliefs use concepts and concepts individuate in a fine-grained way. For example, the concepts WATER and H2O are different concepts as are the concepts BENJAMIN FRANKLIN and THE INVENTOR OF BIFOCALS, even though the members of each pair co-refer. This is why one can believe that water is thirst quenching without believing that H2O is thirst, for example. Admittedly, phenomenal concepts, in my view, are a special case. They refer directly. [3] But they are, I believe, the only concepts that so refer. In saying this, I am not implying that indexical concepts, for example, are tantamount to descriptions. The concept I, for example, contributes both its referent and a first-person mode of presentation to the content of any belief that exercises it. But the first-person mode is not descriptive and it does not fix the referent. For this and many other concepts, a two-factor theory of content seems best.

Since one cannot have a belief that uses only phenomenal concepts — at a minimum other logical concepts must be involved — belief content is always at least partly sensitive to the concepts exercised. Change any nonphenomenal concept or replace a phenomenal concept with a nonphenomenal one and the content of the belief changes. This is not true with content of the sort had by purely phenomenal states. If the picture looks square to Samantha, for example, then it looks the shape of the tiles, and vice versa, assuming square is the shape of the tiles (though, of course, the picture need not look the shape the tiles look). Moreover, the point extends to nonconceptual content generally. So, on my view, nonconceptual content is not content of the sort that belongs to beliefs.

Byrne says that I have supplied no good reason in favor of the content view. But I just did. And that reason is in Consciousness, Color, & Content. Of course, the reason is not one that is likely to persuade Byrne, since he does not accept the orthodox view that belief contexts are intensional in the way I have noted. But the alternative he proposes, under which beliefs have very coarse-grained contents, has well known counter-intuitive consequences.

Beliefs, then, on my view, lack phenomenal character because they lack nonconceptual content. But this is not the only reason for their having no ‘feel’. They also are not poised. Beliefs are not conscious states at all. They are manifested in consciousness in thoughts, just as they are manifested in speech by the production of sentences. Beliefs no more have poised contents than do any of the states in early vision (e.g., states representing changes in light intensity). For a state’s content to be poised, it must be available to the processes that form beliefs and desires (and further not just any old belief/desire formation processes). And that requires that the state be at the interface of the conceptual and nonconceptual domains, on the nonconceptual side. Beliefs themselves, then, do not play the right role, not even phenomenal beliefs.

Perhaps it will be argued that in tailoring the account of poisedness so as to rule out phenomenal beliefs as the bearers of phenomenal character I have made the proposal ad hoc. I disagree. A priori reflection upon the role that experiences qua experiences play with respect to cognition delivers the poisedness condition. Moreover, there are independent reasons for denying that beliefs have phenomenal character. Indeed there are such reasons for supposing that thoughts, the conscious manifestations of beliefs, do not have phenomenal character. [4] Whatever phenomenal character goes with a thought attaches to associated images, most notably linguistic auditory images. And these images are experiences in their own right.


[1].  Thus Chris Peacocke (1992, “Scenarios, Concepts, and Perception,’ in The Contents of Experience, ed. By T. Crane, 105-35, Cambridge, Cambridge University Press) allows scenario contents, protopropositional contents, and conceptual contents. A perceptual state can have an accurate scenario content, for example, but an inaccurate protopropositional one.

[2].  I am categorically not making this claim for the context ‘looks to be F ’ or the context ‘looks as if it is F ’ or indeed for any nonphenomenal ‘looks’ contexts.

[3].  Even so, their identity as phenomenal concepts demands that they play the right functional role. This role ensures that phenomenal thought (belief) types play a different role in rationalizing explanations than nonphenomenal thought (belief) types. See here “A Theory of Phenomenal Concepts,” Philosophy forthcoming.

[4].  For some relevant comments here, see my review of Galen Strawson’s Mental Reality, Journal of Philosophy, 1996, 421-424. For the full argument, see my The Unity of Consciousness, Chapter 3, in preparation.

top | back to symposium index | Byrne's paper