«PANIC theory and the prospects for a representational theory of phenomenal consciousness URIAH KRIEGEL ABSTRACT Michael Tye has recently argued that ...»
A super-sentient creature is conceivable (and also possible, as far as I can tell), who would possess a concept for every shade of red. Any shade of red the creature can discriminate today it can recognize (among hundreds of samples) tomorrow, as well as next year. It would be odd to deny this creature phenomenal experiences of red on account of its augmented sensory and cognitive abilities. Surely it is not the limitations of our sensory and cognitive abilities that give rise to phenomenality.
The difference between phenomenal states and non-phenomenal states cannot, therefore, reside in non-conceptuality. Although mental states that have phenomenal character typically also have non-conceptual content, it is not in virtue of having non-conceptual content that they have phenomenal character. For they would have phenomenal character even if their content was conceptual.
60 U. KRIEGEL
5. Poise So now we are left with the property of poise as our last resort. What is it for a
content to be poised? Tye writes (Tye, 2000, p. 62):
This condition is essentially a functional role one. The key idea is that experiences and feelings, qua bearers of phenomenal character, play a certain distinctive functional role. They … stand ready and available to make a direct impact on beliefs and/or desires.
That is, a content C is poised iff C plays a distinctive functional role D, a functional role presumably different from that of propositional attitudes. So perhaps the difference between mental states with and without phenomenal character is that the former are poised whereas the latter are not.
The problem is that Tye’s characterization of poise commits a category mistake.
What plays the functional role D is perhaps the mental state that carries C, but not C itself. Functional role is not an attribute of representational contents, but of the vehicles that carry these contents . To talk of “poised content” is to that extent misleading: it is the state that is poised, not its content . I conclude that poise is not really a representational property of phenomenal states, but a vehicular property (if you will).
In fact, the same is probably true, if less obviously, of non-conceptuality.
Indeed, I would argue that non-conceptuality is a functional role property as well.
For what does it mean to say that a state S represents a tree conceptually? It does not mean that the content of S is the concept of tree; for S represents a tree, not a concept. Nor does it mean that S represents the tree, intangibly enveloped by a concept. Rather, it means that S represents the tree in a certain way, a way that involves the mobilization of certain cognitive abilities, namely, abilities associated with possession of the concept of tree (e.g. the ability to recognize trees across times and places). This way of representing the tree will manifest itself in the functional role of S. A subject who does not possess the concept of a tree would not be able to form, on the basis of her tree experiences, such beliefs as “all trees have a trunk”;
a subject who does, would. Thus to say that S’s content is conceptual may be nothing more than to say that S would normally lead to the formation of certain beliefs; and to say that S’s content is non-conceptual may be only to say that S would not typically lead to the formation of these beliefs. It is to comment on S’s functional role.
What I am suggesting is that non-conceptuality, like poise, is not really a representational property, but a functional role property, of phenomenal states. It would be impossible to properly defend this view here, and I am not going to try.
As we saw, non-conceptuality is anyway inadequate for distinguishing phenomenal from non-phenomenal states. Moreover, the corresponding point about poise will suf ce for the main argument of this section, as we shall now see. Even so, let me register my contention that the notion of PANIC is, as it stands, a confused hybrid.
It intermingles properties that belong to the representational content of experiences 61
PANIC THEORYand properties that belong, strictly speaking, to the experiences themselves, the vehicles carrying these contents.
The main argument of this section is basically this: if the only difference between a phenomenal state and a non-phenomenal state is that the former is poised and the latter is not, then given that poise is not a representational property, but a functional role one, the difference between phenomenal and non-phenomenal states is not a difference in representational properties—contrary to (4)—but a difference in functional role properties.
This point holds not only for the difference between phenomenal experiences and propositional attitudes. Tye seems to think that there are other mental states,
beyond propositional attitudes, that have all it takes to be phenomenal except poise:
States with non-conceptual content that are not so poised lack phenomenal character. Consider, for example, states generated in vision that nonconceptually represent changes in light intensity. These states are not appropriately poised.
According to Tye, then, there are states of the visual system—call them V states— that have non-conceptual representational content, but are not poised. So Tye takes them to be examples of non-phenomenal states. But what exactly is the difference between V states and phenomenal states, that the latter do and the former do not have phenomenal character? Tye seems to suggest that the only thing that disquali es V states from being phenomenal states is that they are not poised. So the only difference between V states and phenomenal states is that the latter are poised whereas the former are not . We have argued, however, that this is not a difference in these states’ genuinely representational properties. It is only a difference in their vehicular properties. So the fact that V states have no phenomenal character whereas phenomenal experiences do is not a representational fact.
In the same vein, the property of poise is recruited by Tye to account for the lack of phenomenal character in blindsight: the mental states of the blindsighted subject are not phenomenally conscious (unlike the corresponding states of a normal subject) because they are not poised . But again, this is not a representational difference. In general, (5) Whenever S1 has a full PANIC and the only difference between S1 and S2 is that S2 is not poised (i.e. S2 has only an ANIC), the fact that S1 does and S2 does not have phenomenal character is not a fact about the representational properties of S1 and S2.
Rather, it is a fact about the functional roles of S1 and S2. (5) is precisely the case with V states and blindsight. These cases show that, (6) For some states S1 and S2, the fact that S1 does and S2 does not have phenomenal character is not a fact about the representational properties of S1 and S2.
Now, (6) is the explicit negation of (4). It states that two states can be exactly alike with respect to all their genuinely representational properties and still only one of 62 U. KRIEGEL them will have phenomenal character. A theory that entails something like that cannot be seriously called representationalist.
Our line of criticism is that in PANIC Theory the properties that account for the difference between states with phenomenal character and states without phenomenal character are not representational properties, but functional role properties.
This, too, is a way of dealing with the a priori suspicion raised in Section 2. The absurdity of positing environmental features that can only be consciously represented is avoided by quietly sliding to non-representational lines.
One way Tye might respond to our line of criticism is by arguing that we have been rash to determine that poise (as well as non-conceptuality) is a non-representational property. When a mental state S with content C plays the functional role D characteristic of poise, we have to distinguish two properties: (i) the property of playing D; and (ii) the property of being carried by a state that plays D. The former is a property of S, the latter is a property of C. S itself does not have property (ii), for S is not carried by a state that plays D, since S is not carried by anything. If we de ne poise as (ii), then poise is a property of the representational content of S. So it is a genuine representational property of S.
My rejoinder is that this move trivializes representationalism. Suppose we allow such properties as (ii) to count as genuinely representational. Then for any property F a mental state S has, the fact that S is F could be accounted for in terms of the fact that S has a representational content which is carried by an F state, and the fact that S has a representational content which is carried by an F state would count as a genuinely representational fact. A representational account of the fact that S is F would be correct no matter what F is. So the thesis that all mental states are representational facts would be trivially true. The question, however, is whether there is a non-trivial version of this thesis that is true.
6. Conclusion: PANIC Theory as disguised functionalism; and the prospects for a genuinely representationalist account of phenomenal consciousness Every theory of phenomenal consciousness must account for the difference between mental states that have phenomenal character and mental states that do not have phenomenal character. A representational theory of phenomenal consciousness is a theory that does that by adverting to the representational properties of the mental states in question. A theory that does that by adverting to non-representational properties of the mental states in question is not a representational theory. In particular, a theory that does that by adverting to the functional role properties of the mental states in question is a functionalist theory. I have argued that PANIC Theory, despite its advertisement as a representationalist theory of phenomenal character, adverts to functional role properties in accounting for the difference between mental states that do and mental states that do not have phenomenal character. If so, PANIC Theory is really a functionalist theory of phenomenal consciousness.
To say that PANIC Theory is a functionalist theory is not to say that it is false 63
PANIC THEORY. But it is to say that if it is true, then representationalism is false. In this respect, PANIC Theory defeats the purpose for which it was conceived.
Where does all that leave other versions of the representational theory of consciousness? That PT fails to account in genuinely representational terms for the difference between phenomenal and non-phenomenal states does not entail that other representational theories will fail to do so as well. But any genuinely representational theory would have to face the a priori suspicion. To my mind, the a priori suspicion reveals what is fundamentally wrong about a representational approach to phenomenality: if what makes a mental state have phenomenal character at all is that it represents certain features, then the existence of phenomenal consciousness is conditioned by the existence of the features in question, and conversely, it is the existence of the features in question, rather than the nature of conscious subjects, that introduces phenomenal consciousness into the world.
Acknowledgements The ideas in this paper have been in uenced by discussions of Tye’s latest book in a reading group at Brown University during spring 2000. I am indebted to the other participants: Juan Comesana, Jordi Fernandez, Robert Howell, and Jaegwon Kim.
Notes  In particular, they believe that representational facts are facts about causal-informational and/or teleological relationships between brain states and world states (see Dretske, 1981; Millikan, 1984, for some thorough accounts of representational facts along these lines).
 Some philosophers have argued that conscious propositional attitudes have phenomenal character as well (see e.g. Goldman, 1993). I have no quarrel with these philosophers. But even if the attitudes have phenomenal character, it is surely of a wholly different kind than the specially qualitative phenomenal character of sensory experiences. A theory of phenomenal character has to account for that difference. To make my point, I would have to put it in different terms: not in terms of the difference between mental states that have phenomenal character and those that do not, but in terms of those mental states that have the speci c kind of phenomenal character that sensory experiences have and those that do not. In any event, Michael Tye—whose theory this paper discusses—believes that the attitudes do not have phenomenal character, as we will see later.
 Tye (1995, 2000).
 I am using the term “phenomenal state” to refer to mental states that have phenomenal character;
and “non-phenomenal states” to refer to states without any phenomenal character (e.g. the belief that 2 1 2 5 4).
 To my mind, the requirement of abstractness is also utterly implausible: it entails that even when we correctly perceive a tree, the particular concrete tree is not a constituent of the content of our perceptual experience. This means that the concrete tree is not what the experience represents, since in Tye’s framework contents are what experiences represent. More generally, this requirement entails the absurdity that the concrete world is unrepresentable. The problem of misrepresentation is a problem, but making representational content abstract is not the solution  Tye writes (2000, p. 61): “Color experiences … subjectively vary in ways that far outstrip our color concepts. For example the experience of the determinate shade, red29, is phenomenally different from that of the shade, red32. But I have no such concept as red29. So, I cannot see 64 U. KRIEGEL something as red29 or recognize that speci c shade as such. For example, if I go into a paint store and looks at a chart of reds, I cannot pick out red29.”  Some philosophers have contested the point, arguing that the content of a red17 experience is conceptual: it does not deploy the concept of red17, but it does deploy the (demonstrative) concept “this shade of red”.