Eric Lormand

University of Michigan

Nous, 30: 242-261, June 1996

There is not a uniform kind of consciousness common to all conscious mental states: beliefs, emotions, perceptual experiences, pains, moods, verbal thoughts, and so on. Instead, we need a distinction between phenomenal and nonphenomenal consciousness. As if consciousness simpliciter were not mysterious enough, philosophers have recently focused their worries on phenomenal (or qualitative) consciousness, the kind that explains or constitutes there being "something it's like" to be in certain mental states. But there is a tempting suspicion that talk of phenomenal consciousness, and the "what it's like" idiom, are mere mantras, philosophers' attempts to invoke a sense of special mystery where there is none. A critic might simply deny that talk of "what it's like" points to a distinctive kind of consciousness, and so identify phenomenality with consciousness. Against this deflating maneuver, I argue that there are conscious mental states which lack phenomenal character (section 1). From this I extract and try to solve novel problems for theories of phenomenal consciousness (section 2), and I close by defending the hypothesis that nonphenomenal consciousness depends on phenomenal consciousness (section 3).

Our aim is to test whether phenomenal consciousness ("p-consciousness," for short) can simply be identified with consciousness. We can do this by asking: are there any conscious mental

states which lack phenomenal character? If so, there must be more to being p-conscious than simply being conscious. For many conscious states, there is clearly something it's like for us to have them:

Call these kinds of states the "Qualitative Quartet." They all have "qualia," particular what-it's-like properties we sometimes try to describe, for example, by saying that a given pain is "sharp" or "throbbing" to some degree, or that a given visual image is "blurry" or "moving".

Beyond the Quartet, as I will describe in the remainder of this section, there is likely to be some dispute about the scope of p-consciousness. Some states not explicitly listed in the Quartet can be conscious in some sense: in particular, moods and propositional attitudes (beliefs, desires, emotions, etc.).<1> But I think that at least in normal human beings, moods and propositional attitudes are never p-conscious, even when they are conscious. I concede that normally there seems to be something it's like to have a conscious mood or attitude. My conjecture, however, is that whenever this seems so, "what it's like" is completely accounted for by what it's like to be in Quartet states accompanying the mood or attitude. First I will try to explain away alleged mood p-consciousness in terms of both Quartet p-consciousness and alleged propositional-attitude p-consciousness. Then I will try to explain away the residual propositional-attitude p-consciousness in terms of Quartet p-consciousness.

When one has conscious moods, are there distinctive moody qualia or feelings, distinctive "things" it's like to be depressed, elated, or irritable? Or are the only qualia those of various effects, symptoms, or concomitants of the mood? To answer, we need to specify the difference between moods and these surrounding phenomena. I have argued elsewhere (1985) that moods such as depression, elation, and irritability are nonrepresentational states, to be distinguished from emotions, such as sadness, joy, or anger that p.<2> Instead, moods are certain dispositions to activate or deactivate various desires, emotions, and evaluative beliefs, dispositions to facilitate or hinder these attitudes in their service of action guidance, attention direction, imagination, and other processes. If moods are dispositional in this way, they are distinct from the attitudes they facilitate or hinder, and distinct from the imaginings, streams of thought, and other experiences they thereby indirectly influence. (In particular, one can be depressed even when there are no fidgeting attitudes for the depression to facilitate or hinder--as perhaps in dreamless sleep--and even when, perhaps due to distractions, happy thoughts are predominantly or solely active.)

In searching for the qualia of a conscious mood such as depression, then, we must exclude what it's like to be in the various states it influences, such as active attitudes (e.g., sad emotions, unfulfilled desires, and negative evaluations) and correspondingly negative streams of thought, imaginings, bodily sensations and perhaps altered perceptual experiences. We must also exclude what it's like to have thoughts about the mood, as when one thinks to oneself, "Oh, well, I guess I'm depressed," or visualizes oneself frowning. Once we take account of what it's like to be in the various states influenced by a mood, which we introspectively take as evidence that we are in the mood, there is little reason to believe we have omitted any further aspects of what it's like to be in the mood.<3> This is how there can be something it's like when we are in a mood, even though the mood itself has no phenomenal character.

It might be objected that even if there is no distinctive thing it's like simply to be in a mood, there is something distinctive it's like to be in a conscious mood rather than an unconscious mood. Suppose one is in a state of depression that is unconscious, but that otherwise has normal conscious depressive effects, helping to activate negative attitudes, dark imaginings, etc. Now suppose that one becomes conscious of the depression itself, as opposed to being conscious only of its effects--presumably through more careful introspection or through accepting the observations of others--so that the depression becomes (in some sense) conscious. There seems to be a change in what it's like to be in the mood, at least on some occasions, and at least until one ceases to attend to the mood itself. So we need to leave room for a difference between what it's like when one is in a conscious mood and what it's like when one is in an unconscious mood. But this difference is threatened if moods have no qualia of their own--if the only relevant qualia are those of the effects I have mentioned via the facilitation and hindering of attitudes. Since both conscious and unconscious moods have these effects, qualia of these effects cannot explain the difference. It may be tempting, then, to suppose that conscious moods have their own p-consciousness, while unconscious moods do not, to explain the change in what it's like when a previously unconscious mood becomes conscious. However, there is a more plausible alternative.

What is required for one to become conscious of a mood, as opposed to various mood-effects, at least in the way related to introspection or accepting testimony? Intuitively, to become conscious of a mood, at least in the way in which there is something it's like to do so, is to start thinking consciously about the mood.<4> If so, we do not need to suppose that conscious moods are themselves p-conscious. There is room to explain the incremental change in qualia, when a mood becomes conscious, by appeal to the qualia of the thoughts rather than the mood. I will discuss in a moment whether the relevant kinds of thoughts are p-conscious. But they clearly are if they are elements in conscious streams of verbal or imagistic thought. When in becoming conscious of a mood I think of the mood (e.g., as illustrated above, by thinking "I'm depressed," or by imagining my frowns) the changes in what it's like seem to be constituted by the verbal or imagistic qualia of these thoughts. As best I can tell in my own case, there is no residue of qualia for the mood itself to possess. What it's (allegedly) like to have a conscious mood seems to reduce to independently clearer cases of non-mood p-consciousness: what it's like to have the mood's effects (via facilitation and hindering) or symptoms, together with what it's like consciously to think about one's mood.

I think that the p-consciousness that may seem to be possessed by most conscious propositional attitudes can be explained via a similar strategy, as p-consciousness possessed by their effects, symptoms, or concomitants within the Quartet, especially conscious thoughts and imaginings. To make this a substantial claim, I should first say more about what I am calling "thoughts" in the sense of elements of "streams of thought," since there is a different sense of the word according to which it covers all propositional attitudes, and perhaps even all mental representations or all mental states. (In any sense in which all attitudes are thoughts, attitude p-consciousness is trivially a kind of thought p-consciousness.) By streams of thought, I mean first and foremost the fairly slow, roughly serial, and typically deliberate phenomena we try to describe as "talking to oneself" or "thinking in words." (I also include what we are tempted to describe as "thinking in images," which overlaps with what I have been calling "imaginings.") I assume that we often have attitudes--even active, occurrent ones, rather than latent ones or mere dispositions to have them--without reflections in such streams of thought. One's stream of thought is too narrow and shallow to hold at any given time reflections of all the propositional attitudes that actively affect one's actions and mental processes at that time, so not all occurrent attitudes swim in the stream of thought.

What does swim there? In talking to ourselves we don't normally whisper words to ourselves faintly; by my informal polling, anyway, deaf people--at least, people who become deaf after learning to speak--still claim to be able to "hear themselves think." Nor, presumably, can such episodes be explained as proprioception of faint throat or tongue wigglings; it seems likely that talking to oneself can occur despite extreme paralysis or local anesthesia, or even surgical removal of the tongue and throat. Whatever inner speech is, it is something one does only with one's brain. Should we then postulate an inner mental throat? Not literally, of course; we have as much reason for denying that our brains contain public-linguistic expressions as we have for denying that our brains contain bells and whistles. I don't have any bells and whistles in my head--at best, what I have are representations of bells and whistles. Similarly, contrary to my impression that I can literally think "in English," I don't have any English words in my head--at best, what I have are representations of English words (e.g., auditory or visual "images" of English words). Just as nothing in my head--e.g., clumps of brain cells and blood cells--looks like a bell or sounds like a whistle, so nothing in my head looks like a written English word or sounds like a spoken English word. So whatever things constitute the stream of verbal thought, they would have to be not inner speech acts but some sort of representations of outer speech acts.

What sort of representations are these? Well, whatever "inner speech" is, it is very closely related to the actual production of outer speech: inner speech and outer speech are equally serial and slow (roughly), and they seem to share resources (since it is impossible or distinctly difficult simultaneously to say one thing out loud and another thing to oneself).<5> Furthermore, people frequently inadvertently "think out loud" when they mean to think "to themselves," and I have been with very sleepy or pensive people who are surprised to learn that they have been merely thinking, when they meant to be speaking! We can explain all this if we suppose that talking to oneself is a truncated form of normal talking out loud, so that the p-conscious representations of outer speech are certain verbal commands that normally result in outer speech.

Typically, but probably not always, thinking in words is caused by associated, preexisting, propositional attitudes. One's standing belief that snow is white may cause one to think that snow is white, by causing one to form an auditory image of quickly saying the words "snow is white" (or "I believe snow is white").<6> At least normally, if there is anything it's like for me to have a conscious belief that snow is white, it is exhausted by what it's like for me to have such verbal representations, together with nonverbal imaginings, e.g., of a white expanse of snow, and perhaps visual imaginings of words. The important point is that the propositional attitudes are distinct from such p-conscious imagistic representations, just as moods are distinct from their p-conscious effects. For example, typically there is a difference in content: my belief represents that snow is white, but the associated p-conscious elements of my streams of thought represent (e.g., are auditory or visual images of) the words "snow is white" or "Mon Dieu! La neige! Blanche!" or some particular fictitious white expanse of snow, etc. There may also be a difference in mode: my belief is an indicative representation, but the p-conscious representations in verbal thinking are imperative representations or commands.<7> So in searching for attitude qualia, we must exclude the qualia of concomitant thoughts and imaginings (as well as perceptual experiences and bodily sensations). Excluding what it's like to have accompanying Quartet states, however, typically there seems to be nothing left that it's like for one to have a conscious belief that snow is white. There is some motivation, then, for the general hypothesis that, like our moods, our propositional attitudes are not p-conscious, even when they are conscious.

Alvin Goldman has argued against this hypothesis by appeal to the "tip-of-the-tongue" phenomenon, in which one struggles in vain to recover an attitude (which includes what Goldman calls a conceptual "structure" or "unit") that one is to some degree conscious about:

Goldman attributes this argument to Ray Jackendoff (1987), but I think mistakenly; if anything, Jackendoff provides the seeds of a response to Goldman's argument. Jackendoff uses the tip-of-the-tongue phenomenon to "demonstrate" that "conceptual structure is excluded from [phenomenological] awareness" (1987, p. 290). He distinguishes the aspects of what the experience is like into a soundless "form" and an "affect" of effort, so that "one feels as though one is desperately trying to fill a void" (1987, pp. 290 and 315). Neither of these aspects seem attributable to nonsensory attitudes. Jackendoff attributes soundlessness not to propositional attitudes or conceptual structures but to "an empty set of brackets" in a sensory, phonological representation (1987, p. 291). I think this is roughly right; there is something sensory that having the "void" is like, akin to what hearing silence (as opposed to being deaf or asleep) is like.<8> Although Jackendoff provides no account of the feeling of effort, I also see no reason to describe it as nonsensory or to attribute it to Goldman's propositional attitudes and conceptual structures; there is something sensory that the feeling of effort is like, namely, what feeling physical effort is like.<9> I conclude that our attitudes themselves are never p-conscious, although there may often be something it's like when we have them, due to accompanying Quartet states.<10>

We launched this discussion of moods and attitudes as a way of checking whether phenomenal consciousness can be identified with or deflated into consciousness simpliciter. It might be thought that to reject this deflating maneuver, one would need a full positive theory of the meaning of "what it's like," or at least of the states described in this way. However, with the foregoing discussion we are in a position to reject deflation with less effort. Given that conscious moods and attitudes are not p-conscious, p-consciousness cannot be deflated to consciousness. So talk of "phenomenal" consciousness is substantive; there must be a kind of nonphenomenal consciousness (or n-consciousness) available to attitudes and moods.<11>

If I am right, our conscious moods and attitudes must lack some feature necessary for phenomenal consciousness. One task of a theory of p-consciousness, then, is to specify this lacking feature, and perhaps to help determine whether this lack is special to our moods and attitudes, or endemic to any possible moods and attitudes. I find little in existing theories with which to address these questions. I will consider the most prominent variants on functionalist, antifunctionalist, and inner-representational proposals.<12> Then I will outline a more adequate explanation of the nonphenomenality of attitudes and moods.

On Daniel Dennett's theory, phenomenal consciousness is a matter of the functional relations of a mental state: "there is no reality of conscious experience independent of the effects . . . on subsequent action (and hence, of course, on memory)" (1991, p. 132). Which effects on action and memory are relevant to p-consciousness? He seems to favor a kind of "holism":

I have argued elsewhere (forthcoming-b) that this holism is very implausible. Even leaving this aside, we can see that a view like Dennett's fails to explain why conscious attitudes and moods are not phenomenally conscious. Moods and attitudes engender "idiosyncratic reactive dispositions," which Dennett wants to identify with qualia, so why are they not qualitative?<13>

Georges Rey has also defended a functionalist theory of "sensations," by which he seems to mean the half of the Qualitative Quartet consisting of perceptions and bodily sensations. Though he is not tempted to Dennett's holism (1993, p. 256), he does not commit to a principled limitation on the functional relations constitutive of qualia:

In addition, he holds (1993, p. 247) that for the qualitative experience not to be unconscious--and so, presumably, for there to be something it's like to have it--the relevant perceptual representations must be "reportable" or "avowable"--an "input to a system for making assertions and self-reports" (1993, p. 242; see also Dennett, 1969). I think none of this helps to explain the nonqualitativeness of moods and attitudes. Conscious moods and attitudes are reportable verbally, help produce judgments (in both rational and nonrational ways), and are subject to associations and memory-comparisons. Rey mentions various functional and semantic restrictions on the states he calls "sensations," but these do not seem necessary for p-consciousness. They do not clearly apply to phenomenal thoughts and imaginings, and he does not explain why there should be nothing it's like to be in conscious attitudes and moods that are free of such restrictions.

Antifunctionalist theories explain the difference between p-conscious and p-unconscious mental states by appeal to differences in the intrinsic composition of the states, typically their nonpsychological (e.g., biological or physical) properties. Even thinkers who accept a functionalist theory of nonconscious mental states, or of nonphenomenally conscious states, often consider phenomenal consciousness to be resistant to such explanation. Christopher Hill has recently argued that "we are not justified in ascribing immediate sensations to a being unless the being is similar to ourselves in structure and composition" (1991, p. 224), and that:

Ned Block has also expressed sympathy with such a view (e.g., 1990, p. 53-54). There is an almost total lack of detailed suggestions about which intrinsic properties matter to consciousness, but these are presumably to be discovered by scientific investigation, just as on Rey's account the particular qualitatively relevant functional relations should be. So while Hill and Block can predict that conscious moods and attitudes lack the intrinsic properties (to be discovered by science) allegedly relevant to p-consciousness, they seem to have no way to explain why this should be so.

There is somewhat more explanatory promise from another variant of the view that p-consciousness is an intrinsic property, a view prevalent in the phenomenological literature. On this view, consciousness is "reflexive": conscious states are those that are somehow self-representational. Brentano argues that "the presentation which accompanies a mental act and refers to it is part of the object on which it is directed" (1874, p. 128). Husserl claims that "[i]n the case of an immanently [i.e., mentally] directed . . . perception, perception and perceived essentially constitute an unmediated unity, that of a single concrete cogitatio" (1913, p. 112). Sartre speaks of "[reflection's] absolute unity with the consciousness reflected-on" (1943, p. 212), so that "the reflective consciousness must be the consciousness reflected-on" (1943, p. 213). All we need to extract from these passages for present purposes is the claim that a conscious state is a representation of its own existence (and, presumably, a phenomenally conscious state represents its own qualia).

Although reflexivity is typically meant to characterize consciousness in general, it does not cover conscious moods and attitudes very well. As I described in section 1, irritable moods (as opposed to emotions of anger that p) do not represent anything, least of all themselves. Even when one is in a conscious irritable (or depressed, elated, etc.) mood, one needn't be irritated (or depressed, elated, etc.) about it, so if one does represent the mood, it is likely to be via a separate representation rather than the mood itself. A similar point holds about conscious attitudes such as desires or emotions. Although these are representational, and so potentially reflexive, one may consciously want (fear, hate, be amused by, etc.) something without wanting the want (fearing the fear, hating the hatred, being amused by the amusement, etc.). So any representation about these conscious attitudes seems separate from the attitudes themselves. Even for purely indicative attitudes such as beliefs, it is odd to suppose that conscious beliefs are also beliefs in their own existence. If they were, then although an unconscious belief that the Earth is flat is wholly false, a conscious belief that the Earth is flat would be half-true!

As a result, reflexivity theories seem best restricted to p-conscious states rather than n-conscious moods and attitudes. An attraction of this restriction is that reflexivity views may hope to explain why moods and attitudes are not p-conscious, even when they are conscious in some other way. Moods are p-unconscious because, being nonrepresentational (see section 1), they don't represent themselves. And although attitudes are potentially self-representational, perhaps some notion of reflexivity can be specified according to which they are plausibly never reflexive. This would be difficult to provide, however. At least some attitudes are reflexive without clearly being phenomenal: I can consciously believe that I have this very belief, but what it's like when I do so seems to depend on variable Quartet states separable from the belief, such as verbal imaginings of asserting the words "I have this belief" or "Yep, here it is again."

On David Rosenthal's (1990) view, a creature has a (nonphenomenally or phenomenally) conscious state if and only if the creature is conscious that it has the state, in the following sense: the creature must form an occurrent belief that it has the state.<14> Call such a belief "inner" because it is about other mental states rather than about nonmental entities.<15> As Rosenthal seems to recognize, this account alone will not explain why conscious attitudes and moods don't have p-consciousness--for we do often have occurrent inner beliefs about these states. P-conscious states, he says, are those that both are conscious (via inner beliefs) and have what he calls "sensory qualities," which he describes as "the properties on the basis of which we discriminate among our perceptual and bodily sensations" (manuscript, p. 1).<16> So Rosenthal restricts p-consciousness by fiat to states with sensory qualities. The trouble is that we also introspectively discriminate among our conscious attitudes and moods--e.g., by their various contents, dispositions, forces (indicative, imperative, etc.), and relations to p-conscious "symptoms" in the Quartet. What keeps them from having what-it's-like properties? Of course, the mere fact that we do not apply the word "sensory" to them cannot be the answer! At a minimum, Rosenthal's use of "sensory" would have to be extended in a technical way, so as to include imaginings and thinking in words and images. By what principle is the resulting technical term not to be extended to moods and attitudes? We are left without an explanation of what is special about the Quartet.

It might be suggested that Quartet states are special because we innerly perceive them, as opposed to merely forming inner beliefs about them. In some sense we seem to see rather than posit our visual images, feel our pains, hear ourselves thinking. By contrast, even though we form inner beliefs about our moods and attitudes, perhaps we don't innerly perceive them (as opposed to their concomitants in the Quartet). Might these glimmers yield an acceptable explanation of the nonphenomenality of moods and attitudes? Only if there is an illuminating contrast between inner perception and inner belief, presumably based on a distinctive analogy to outer perception. Unfortunately, such an analogy is hard to come by. David Armstrong (1980), for example, explains inner perception as being, like outer perception, "selective" (not omniscient), "fallible" and "causal." This is not much of an analogy, since probably all indicative mental representations have these features, even our most theoretical (vs. "perceptual") scientific beliefs. Armstrong's position does not improve on Rosenthal's.

So, what could be perceptual about inner perception? Of course, we cannot suppose that inner perceptions are generated in the same ways as outer perceptions--e.g., by literal inner eyes, ears, and their attendant representation-forming processes. On the other hand, if they are produced by some distinctive faculty of inner representation, a theory should justify calling them inner "perceptions" rather than mere inner "reactions" of some other sort. A tempting idea is that inner perception is a noninferential source of evidence about mental states. Moods and attitudes can be conscious even though our access to them involves self-directed theoretical inferences, even confabulatory ones (see note ). Perhaps this inferential access rules them out of the phenomenal realm. But we lack an explanation of why noninferentiality is necessary for phenomena. Furthermore, on many accounts ordinary, outer perception does rest on inference, perhaps of a dumb, inflexible, not-many-things-considered sort. If so, presumably the same would be true of inner perception of Quartet states. We could not explain the nonphenomenality of moods and attitudes by our inferential access to them, since access to Quartet states would also be inferential. Instead we would have to suppose that the cleverness or flexibility of inferential access to attitudes and moods renders them nonphenomenal. But why should cleverness and flexibility matter? At best, this standpoint has things backwards: we need clever and flexible inferential access to our moods and attitudes because there is nothing it is like to have them.

Nevertheless, I think that a more carefully characterized kind of inner perception is relevant to explaining the nonphenomenality of moods and attitudes. Elsewhere (under review), I defend the view that each Quartet state involves one of two correlative inner-perceptual illusions, the "image illusion" and the "appearance illusion." While a proper description and explanation of these illusions is too long to include separately here, a sketch may convey the general idea. In the image illusion, we seem to experience mental objects "subjectively" having properties that are had only by nonmental objects: afterimages that are subjectively round and red, pains that are subjectively sharp and throbbing, thoughts that are subjectively soft and medium-pitched, etc. In the appearance illusion, we seem to experience nonmental objects "objectively" having properties that are had only in relation to mental objects: circular surfaces with objective circularity-looks and circularity-feels, air-motions with objective sound-appearances, etc.

Although every Quartet state engenders either the image or appearance illusion, normal human moods and attitudes are not subject to these illusions. By contrast with the image illusion, we do not typically experience our (conscious or unconscious) moods or propositional attitudes, clothed in properties normally possessed only by nonmental objects. Unlike our images of yellow bananas, we don't take our beliefs that bananas are yellow as themselves banana-like or yellow. Unlike blue images, blue moods do not appear blue. And by contrast with the appearance illusion, we do not "project" properties of moods and attitudes onto their causes or objects. A prolonged depression caused by a friend's death does not make the death itself seem prolonged or depressed. A strong but rational desire for a candy does not make the candy itself seem strong or rational.

On the alternative theory of p-consciousness I propose, the disjunction of the illusions is necessary for p-consciousness (leaving open whether it is also sufficient). One advantage of this view is that with it we can explain why moods and attitudes are not p-conscious, by explaining why they are not subject to the illusions. For this purpose, we should ask: why do we ever undergo the illusions, that is, in Quartet states? Again, I can only give the briefest outline of my answer here; the details require a much longer treatment (under review). The illusions are not due to mistaken theory but instead are built into p-conscious experience in the manner of more familiar perceptual illusions. I argue that they spring from a certain kind of confusion between outer-perceptual and inner-perceptual representations. In normal perception, various sensory "transducers"--small portions of sense organs such as retinal cells, tactile receptors, and auditory follicles--each produce a series of outer-perceptual representations.<17> To speed processing, perceptual systems use a certain computational shortcut: reliably enough, they treat representations caused by the same transducer as applying to the same object. I argue that this shortcut applies also to inner perception. An inner perception caused by an outer perception is of course indirectly caused by the same transducer as the outer perception. So perceptual systems treat outer perceptions and the inner perceptions they cause as being about the same objects. This understandable confusion built into perceptual systems explains both the image and appearance illusions. The appearance illusion occurs when outer rather than inner perceptions are attentive and strengthened, as in normal perception, or nonlucid dreams. Attention is focused on the nonmental objects that outer perceptions are about. The weakened inner perceptions represent mental properties (subjective appearances), but are confusedly applied to the nonmental objects themselves, generating the illusion of objective appearances. Conversely, the image illusion occurs when inner rather than outer perceptions are attentive and strengthened, as in abnormal perception, or some forms of perceptual imagination. Here attention is focused on the mental objects that inner perceptions are about (namely, outer perceptions). The weakened outer perceptions represent nonmental properties, but are confusedly applied to the mental objects, generating the illusion of mental images with nonmental features.<18>

The main upshot for present purposes is that normal human moods and attitudes are not subject to the illusions, because they are not subject to confusions with inner perception. Since moods are not representations and so do not apply to anything at all (see section 1), they are in no danger of being confusedly coapplied with inner representations. And even though propositional attitudes do represent, they are not appropriate for the transducer-based shortcut described in the previous paragraph. This shortcut only appears where there is a systematic relationship between representations and a physical field of sensory transducers, that reliably reflects coinstantiation of properties. This is true of outer perceptions produced by the retinal field, and perhaps similar fields for the other sensory transducers, but not for attitudes (or moods). Visual representations caused by the same retinal cells are likely to be of coinstantiated features, but beliefs caused by the same retinal cells may be about anything. (This relative freedom of belief content from specific stimuli is one of the features that distinguishes beliefs from perceptual representations, strictly conceived.) We can thereby understand why neither moods nor attitudes involve the image and appearance illusions. A theory of phenomenal consciousness based on the inner-perceptual illusions therefore has an advantage over all the other existing theories I catalogued above: it alone explains why conscious moods and attitudes are not p-conscious.

Although I have considered the previous theories to be directed at phenomenal consciousness, theorists rarely explicitly consider whether phenomenal consciousness deserves treatment distinct from nonphenomenal consciousness. The shape of a theory of p-consciousness is likely to depend heavily on the supposed relationship between p-consciousness and n-consciousness. A natural and widespread strategy separates p-consciousness into two components: a more general property of "plain" consciousness, and some special phenomenality property. Call this the "factoring" strategy. Rosenthal adopts it, treating p-consciousness as a special case of the kind of consciousness he describes as due to inner beliefs (manuscript, p. 16). Likewise Rey factors p-conscious experiences into reportability and other ingredients, and suggests that reportability is "a familiar, weak notion of consciousness" (1993, p. 242). In fact, any of the theories of p-consciousness mentioned in the previous section, despite failing to explain phenomenal consciousness, could be recast as theories of nonphenomenal consciousness or a more general kind of consciousness neutral between phenomenal and nonphenomenal consciousness. Perhaps they could explain plain consciousness without explaining phenomenality.

I reject any such factoring strategy, in favor of the view that phenomenal consciousness is the most basic kind of consciousness, necessary for the existence of other kinds. Nonphenomenal consciousness is not so "plain" after all, but should be understood in terms of phenomenal consciousness. We have little difficulty conceiving of beings with absolutely nothing "it's like" inside--like rocks or plants or maybe shrimp or sleepwalkers--but we have difficulty convincing ourselves that such beings could have conscious states in any interesting sense. This is a preliminary sign that there are no n-conscious states without accompanying p-conscious states. In the rest of this paper I will develop and defend this claim (see also forthcoming-a), but even in the absence of further argument it has initial plausibility.

How could phenomenal consciousness be necessary for nonphenomenal consciousness? Given that no moods and attitudes are p-conscious, plausibly the only sense in which a mood or attitude can be n-conscious is that one is (disposed without further observation or theorizing to be) conscious of it, or conscious that one has it. This explains why we may argue that a repressed Freudian attitude or mood, or a confabulatory inference, is unconscious by arguing that its bearer has no conscious thought about it. When Freud's patient Dora comes consciously to think of her desire to punish her father, the desire is no longer unconscious. If the required conscious "thoughts" are phenomenally conscious elements of streams of thought, rather than mere beliefs (see note ), we forge a requirement of p-consciousness for n-consciousness. Consider a cautious formulation of this requirement:

The p-primacy hypothesis allows that states can be phenomenally conscious even if we don't have (dispositions to) thoughts about them. A mood or attitude needs an accompanying thought or symptom to be (nonphenomenally) conscious, but the thought or symptom can be (phenomenally) conscious without a second accompanying thought or symptom.<20> So p-conscious states need not be "plainly" conscious in any more general way shared with conscious moods and attitudes (and independent of p-consciousness). This would undermine the factoring strategy; we could not first account for a general way in which both p-conscious states and n-conscious states are conscious, and then invoke phenomenality to differentiate them.

I know of no clear counterexamples to the p-primacy hypothesis, although there are a few potentially troublesome cases.<21> Since inner speech is relatively slow (see section 1), it takes some moments for there to be enough inner speech to constitute thinking about or even expressing to oneself a complex propositional attitude. But attitudes sometimes seem to become conscious too suddenly for this to be explained by such deliberate inner-speech ceremonies. This typically happens after one struggles to solve a specific problem or recall one's attitude about something. Call these "sudden cases." There are two ways a sudden case can be compatible with the p-primacy hypothesis: (a) if the attitude's suddenly being conscious consists in sudden p-conscious states other than slow inner-speech ceremonies, or (b) if the attitude is not conscious suddenly, but only mistakenly thought to be so in retrospect. I think that (a) and (b) hold in all sudden cases, with (b) the more typical.

In some (a)-type sudden cases, we have sudden visual imaginings of words that express or are about the attitudes (we seem to see them all at once, even though we could not innerly speak them all at once). In other (a)-type cases, we have nonverbal Quartet states that are sufficiently complex to constitute distinctive symptoms of the attitudes--e.g., in some contexts a sudden imagining of a fictitious white patch of snow and a sudden feeling as of nodding one's head (or as of saying "yes," breathing easily, etc.) dispose one to think one believes that snow is white.

But even when we lack sudden p-conscious states that count as symptoms or thoughts of the attitude, we do have other sudden p-conscious states that explain how (b) could hold. For example, after struggling with a problem or a memory search, often we suddenly feel (presumably through bodily-perception or bodily-imagination) relief from the previous struggles, or we have short, sharp episodes of inner speech, images of saying "Aha!" or "Oh!" After this sudden experience, often immediately after, we enter into the slower-paced and more normal episodes of inner speech. First we exclaim to ourselves something like "Eureka!," then we explain to ourselves what we have found. By the time we have gone through the latter, nonsudden, p-conscious states, we are normally conscious of our attitude (and, it is normally conscious). But is it clear that we are conscious of our attitude (or, that it is conscious) before then, when all we have is the sudden relief or "Aha!" experience? This seems doubtful, since we sometimes get the sudden experience without the subsequent nonsudden experiences, and in these cases we are not conscious of the attitude (and, it is not conscious). The sudden experience fools us into thinking we have ended our struggle. Likewise, even in the successful cases, our later consciousness of the attitude fools us into thinking it was conscious from the first sudden experience.<22>

Another challenge to the p-primacy hypothesis is that our attitudes seem to be conscious as we verbally express or report them aloud, without any need for concomitant p-conscious inner speech. As with sudden (a)-type cases, such a case is compatible with the p-primacy hypothesis if the attitude's being conscious consists in concomitant p-conscious states other than inner speech, for example, visual imaginings, or perceptual experiences of the outer speech. But there may well be cases in which an attitude is conscious as we express or report it aloud even though these other p-conscious states are absent (e.g., cases in which one visually imagines nothing relevant to what one is saying, and in which one can't hear oneself speak). Fortunately there is a safer possibility for the p-primacy hypothesis. If as I argued in section 1 inner speech is constituted by outer-speech commands, then in normal outer speech we do have simultaneous inner speech, since outer speech is the exercise of these commands. Could this simultaneous inner speech itself be p-conscious? Perhaps it doesn't seem that we have p-conscious inner speech during normal outer speech. But this may be explained by the fact that we consciously hear the actual outer speech (and consciously feel our facial movements), and the resulting vivid p-conscious perceptual states drown out the faint p-conscious inner speech.<23> If so, the consciousness of attitudes expressed or reported aloud could consist in concomitant p-conscious inner speech.

For states such as moods and attitudes, then, it is plausible that the only way in which they become conscious is by bearing certain relations to other states that are p-conscious, such as having p-conscious symptoms and being represented in p-conscious streams of thought. If this is correct, phenomenal consciousness is the root kind of consciousness, and nonphenomenal consciousness is derivative. A theory of consciousness that lacks an argument against hypotheses of phenomenal primacy should allow for a fair examination of them, by not assuming at the outset that phenomenal consciousness can be factored into phenomenality and consciousness.

<1> In this paper I will use the term "propositional attitudes" only for states that are not in the Quartet. (This leaves open whether they are p-conscious, because I am leaving it open for the moment whether Quartet states are the only p-conscious states.) Later in this section I will distinguish propositional attitudes from elements of streams of thought. But in this paper I will have to rest with an intuitive distinction between attitudes and perceptual representations, which I try to sharpen elsewhere (under review).

<2> When we sometimes seem to speak of representational moods, such as a depression that p, we are characterizing hybrid mood-emotion states, such as a nonrepresentational depressed mood with certain relations to a sad emotion that p.

<3> This is supported by the fact that moods are not representational, while every other plausible candidate for p-consciousness is representational. (Even bodily sensations such as pains and tickles are representational--they represent properties such as burning, throbbing, rubbing, and prickling in various parts of our bodies.) When we consider other nonrepresentational mental entities, such as skills, the processes underlying memory, and character traits, we are not tempted to think they have qualia of their own, independently of the qualia their effects may have.

<4> Consider a creature that is unconsciously depressed, and that has p-conscious effects of this depression, but that lacks the ability to have thoughts about depression. (Perhaps small children and monkeys, who may lack any concept of depression, fit this description.) Introspection and testimony would have no way of causing this creature to become conscious of the depression itself, as opposed to being conscious of the ways it is affected by the depression. I say more about the grounds of introspective access in section 3.

<5> One can easily alternate "syllables" in talking aloud and talking to oneself, but this is not what I mean by "simultaneously." Also, one can easily say some words out loud while thinking in visual images of different words, but this is not what I mean by "inner speech."

<6> There are normally more aspects to the imagery, which help one to determine which kind and strength of attitude (belief, desire, suspicion, etc.) is revealed via the thought. In cases of belief, for example, one may imagine saying the words in an assertive tone of voice, and without any concomitant proprioceptive sensations of suppressed giggling, or auditory images as of appending "NOT!," etc.

<7> The differences between thought-images and beliefs run deep. Thought-images are largely voluntary in ways that beliefs are not: I would bet good money that in the next few minutes you can't come to believe that the earth is flat, but I wouldn't bet that you can't think the words. And beliefs (as well as other attitudes and moods) seem subject to confabulatory access in ways that (current) thought-images (and other phenomenal experiences) don't. In trying to identify our own propositional attitudes we tend to think too highly of ourselves, systematically but sincerely reporting attitudes we think we rationally should have in the circumstances, even if we don't have them (as reasonably determined by methods available to the psychologist; see Nisbett and Wilson, 1977). Presumably we make the same kind of mistakes about other mental states, such as our moods, erring systematically in the direction of the moods that we think appropriate in our circumstances. Could this model of hidden confabulatory inference be extended to all access to qualia? It would have difficulty explaining how we access features of our thoughts and other p-conscious experiences that are irrelevant to commonsense rational explanation, and so involve no standards of rationality or appropriateness for us to use in thinking "too highly" of ourselves. For example, no folk-theoretic principles of rationality suggest that one should feel a stinging pain rather than a throbbing pain when a limb has restricted blood flow, yet untutored subjects offer consistent (and apparently reliable) reports of stinging pain "feelings." What premises could we use, consciously or unconsciously, to draw inferences about how these experiences feel? These apparent differences in our self-knowledge of Quartet states versus attitudes and moods support my position that attitudes and moods lack qualia of their own.

<8> More precisely, what it's like to have the void (in the relevant respects, e.g., ignoring the feeling of effort) is what it's like to have auditory images of silence, which in turn is merely akin to what it is like to hear silence. This does not require p-consciousness in states beyond the Quartet, since such states of imagining are already in the Quartet. Also, of course, what matters to the tip-of-the-tongue phenomenon is not that the "void" be silent, but that it be wordless (or rightwordless): we often have auditory experiences as of rhythm, stress patterns, initial word sounds, filler sounds such as "ummm," and so on.

<9> This suggestion receives confirmation in William James' observations that "the acts of attending, assenting, negating, making an effort, are felt as movements of something in the head" (1890, p. 287):

Contrary to certain appearances, as is clear from the context, James is claiming that felt assent and mental effort depends on feelings as of such bodily activities, not necessarily on the bodily activities themselves. Furthermore, as in the previous note, it would be enough for my present response to Goldman if conscious mental effort were like imagining such bodily activities rather than like feeling them.

<10> The distinction between a state's being p-conscious and its merely being accompanied by other p-conscious states provides a response to Goldman's other main argument that conscious attitudes are phenomenal. It is an adaptation of Frank Jackson's (1982) case of Mary, the color-blind scientist:

Goldman wants to argue (a) that attitudes are phenomenal, and (b) that this presents difficulties for functionalism. Our present interest is in the weaker claim, (a). The trouble with Goldman's argument for (a) is that the argument simply doesn't distinguish between the conclusion Goldman needs--that the person learns what it's like to have the doubt or disappointment--and a rival conclusion based on the hypothesis I am defending--that the person learns at most what it's like to have (in tandem) various separable Quartet states that are effects of the doubt or disappointment: pangs, sighs, queasiness, inner speech in novel "tones of voice," verbal thoughts about doubt or disappointment, novel visual imagery, etc.

<11> My argument for this conclusion would survive even if attitudes and moods are sometimes phenomenal when conscious, so long as they are not always so.

<12> Criticisms in this area must be provisional, in that we could simply reject my claims about attitudes and moods if forced to by a theory of p-consciousness with enough other significant advantages. Since my aim here is to highlight the relevant problems for each theory, I will not detail the considerations which motivate the theories, or their other difficulties. I provide fuller descriptions elsewhere (forthcoming-a).

<13> Sometimes Dennett casts himself as an eliminativist about qualia, but since he "[does] not deny the reality of conscious experience" (1992, p. 1), this is not a rejection of p-conscious states, but only of certain bold claims about their qualitative properties. I respond to several of his criticisms elsewhere (forthcoming-b).

<14> Through further requirements he seeks to rule out cases in which one comes to believe in a state solely through accepting the testimony of a psychiatrist, or solely by observing one's own behavior, rather than by any special access relation to one's mental states.

<15> Rosenthal uses "higher-order" for "inner," and speaks of "thoughts" rather than "occurrent beliefs." I substitute "occurrent beliefs" to emphasize (with Rosenthal) that the higher-order (inner-directed) states need not themselves be conscious. (By an "occurrent" belief I mean a belief that is datable, active, and, for emphasis, more than a mere disposition to believe.) As I described in section 1, colloquially thoughts are occurrent states that p-consciously "occur" to one. It would be circular to appeal to such thoughts in offering a theory of p-consciousness.

<16> He comes closest to discussing p-consciousness under the name of "sensory consciousness," which he characterizes as the kind of consciousness "special to sensory states" (manuscript, p. 2), and which he discusses using the "what it's like" idiom (manuscript, pp. 29ff.).

<17> On virtually all detailed theories of normal vision, for example, cell-firings in each retina cause (or constitute) representations of the amount of incoming light of various wavelengths at various points near each eye, which cause representations of sudden discontinuities of incoming brightness at these points, which cause further proximally-representing states, and, eventually, familiarly conscious visual experiences. Some may not wish to call these early states "representations," but the label doesn't matter here so much as the states themselves (which might be called "protorepresentations" instead).

<18> I think the confusion also explains what is perceptual about inner perceptions. Rather than being states of a distinctive inner faculty, inner perceptions are states of the various outer-perceptual systems themselves, and so count as visual, auditory, or other perceptual representations, with minimal violence to the proper use of these terms. Although there are outer sense organs that produce outer perceptions, there need not be inner sense organs that produce inner perceptions. Once inner perceptions are produced, in whatever way, they are processed like outer perceptions, as further representations in particular sense modalities. For example, visual inner perceptions help to produce visual beliefs, help to control visual skills, and are not distinguished from visual outer perceptions introspectively. They qualify as perceptual due to their use, even if not due to their origin.

<19> The argument of section 1 supports the idea that introspective consciousness about attitudes and moods is grounded on symptoms within the Qualitative Quartet; see especially James' account of our consciousness of attention, assent, and deliberation, in terms of what our bodily-perceptual or bodily-imagistic experiences are like (see note 9).

<20> In this way, a state can be phenomenally conscious even though we do not apply concepts to the state, and so do not form any sort of higher-order thoughts or beliefs about it. This is not to deny that phenomenally conscious states require some reflection in "inner awareness"; indeed, according to the account sketched in section 2, a state is p-conscious only if there is an inner perception of the state. I hold that one can innerly perceive a state without applying concepts to it (i.e., without forming thoughts or beliefs about it), just as one can outerly perceive an apple without applying concepts to it.

<21> I am grateful to David Rosenthal for mentioning examples that fuel the following discussion.

<22> Often in trying to solve problems or recall attitudes we get such false alarms of success. Compare James' description of the tip-of-the-tongue phenomenon, in which we "at moments tingle with the sense of our closeness, and then . . . sink back without the longed-for term" (1890, p. 243). Such false tingles do not seem to cause or constitute consciousness of the term; it seems more accurate to say that at these sudden stages our representation of the term is not conscious.

<23> Perhaps we can test this explanation as follows. Suppose there were some way for an experimenter quickly to toggle a subject's (outer) hearing and (facial) proprioception on and off, at will, perhaps with ear or facial implants akin to hearing aids. If one were able to speak continuously through the toggling, presumably one would be able to report whether one seemed to have images of talking, through the periods when one has no perceptions of talking. If so, this would be evidence that inner-speech images accompany, but are typically ignored because of, normal perceptions of outer speech.

References