Phenomenal Impressions 

Eric Lormand
University of Michigan
Forthcoming 2005 in Gendler and Hawthorne (eds) Perceptual Experience  

            In normal perceptual experiences, it is as if we cannot introspect any special phenomenal properties, but only normal environmental properties, such as the colors and shapes of seen objects. Call this the impression—veridical or illusory—of “transparency”. In normal imaginative experiences, it is as if we can introspect special phenomenal objects with normal environmental properties, such as colored and shaped visual likenesses of environmental objects, in our minds. Call this the impression—veridical or illusory—of “images”. After describing the scope of these impressions (part 1), my aim is to give a psychological explanation of them (part 2) and to draw from this explanation a positive theory of phenomenal experience (part 3). 

1     A description of the impressions 

            The strongest impression of images is in normal imagination, while the strongest impression of transparency is in normal perception. Since the relevant issues are most familiar in the visual modality, I introduce images with reference to normal visual imagination (section 1.1) and transparency with reference to normal visual perception (section 1.2). Then I argue that the patterns exemplified there are common to other forms of experience: degraded visual perception, upgraded visual imagination, nonvisual imagination and perception, thought, and bodily sensation (section 1.3). 

1.1     Images in visual imagination 

            Jean‑Paul Sartre claims in The Psychology of Imagination that when we imagine things visually, and attempt to introspect this activity, we are subject to an “illusion of immanence”:

We believed, without giving the matter any thought, that the image was in consciousness …. We pictured consciousness as a place peopled with small likenesses and these likenesses were the images. No doubt but that this misconception arises from our habit of thinking in space and in terms of space. This we shall call: the illusion of immanence. … It is also the point of view of common sense. When I say that “I have an image” of Peter, it is believed that I now have a certain picture of Peter in my consciousness. (1940, 4-6)

Such likenesses, however faint, can seem to be essential to imagination—for instance, they can seem to be what distinguishes imagining a banana from merely conceiving of a banana, and even from merely conceiving of the look of a banana. We do sometimes describe what it is like to have introspectible visual‑imaginative experiences as if in having them we are aware of phenomenal denizens of an inner mental world: we say we “form” images of a banana or a building, and we even accept instructions to “flip” them or “rotate” them. Call such alleged mental likenesses “phenomenal objects.” Why do we talk as if in introspecting some experiences we introspect phenomenal objects? 

            One possibility is that we talk this way because it is accurate—maybe we do introspect what are in fact visual images in a strict and literal sense, mental likenesses of imagined physical objects, entities with some of the same perceptible properties as what’s imagined. On this account banana images are yellow and curved just as bananas are and just as (perhaps faded) pictures of bananas are. But a search for such banana likenesses is unlikely to be fruitful. In “forming an image of a banana” there is nothing obvious in one’s brain or body or (causally relevant) environment that is literally yellow like a banana or curved like a banana. It will not help to appeal to mind/body dualism in locating visual likenesses, since presumably items made of a nonphysical substance cannot literally have color and shape at all. The best bet for a defender of immanence would be to suppose that likenesses are arcane: that they exist in the brain, body, or environment but are unknown to current scientific theory (Jackson, 1977, 101‑104). The only alternative is to be an “eliminativist” about literal visual images, to deny they exist. My explanation of the image impression, in part 2, will be of use to the eliminativist, explaining why we speak of images even if they don’t exist. But even if there are literal mental likenesses, some account must be given of how we are sensitive to them in experience. So my explanation of how and why we (mis)represent likenesses may be of use in either case. 

            We cannot explain the image impression simply by giving the ordinary word ‘image’ a more cautious interpretation than the literal ‘likeness’ one. For example, an eliminativist about mental likenesses may accept that there are mental events or states or objects that represent bananas without being visually like bananas—just as (a use of) the phrase ‘yellow, curved banana’ represents yellowness and curvature without being yellow or curved. Some philosophers propose that visual images are these representational mental entities—in effect, that the phrase ‘a mental image’ refers to an event of imaginatively experiencing (see Tye, 1995, 84-87; also, less clearly, Sartre, 1940, 6-8) or refers to a distinctive “symbol structure” housed in the brain (see Block, 1983, 506-507). On such views we are literally correct when we say we have images; in this paper I take no stand on this issue. My focus is on the residual task of explaining why we speak as if there are mental likenesses, why we speak as if there are mental entities with perceptible properties, such as yellowness and curvature, even if these are in fact merely possessed by the ordinary physical objects we imagine, if anything at all.[1]  

            The most common eliminativist strategy for explaining talk of mental likenesses is to attribute them to a kind of looseness in ordinary practices of reporting experiences. Ned Block pursues this strategy by pointing out that “it is easy to slip into ascribing to representations the properties of what they represent”: the phrase ‘a nude painting’ may be used for a painting of nudity, and ‘a loud oscilloscope reading’ may be used for an oscilloscope reading of loudness (1983, 515‑518). On this view it is taken to be no surprise that people loosely describe images (experiencings, symbol structures) of color and shape as themselves being colored and shaped, and so no surprise that in loose talk they treat them as likenesses. Similarly, Michael Tye says that our talk of colored and shaped visual images is part of a “much broader usage” on which “we save breath by speaking as if the representations themselves have the properties of the things they represent” (1995, 107). 

            This explanation of phenomenal‑object claims is overcharitable, much too tidy. While we are not genuinely tempted to think that a painting is literally nude or literally a tree, we are normally very tempted to think that mental images are literally colored and literally shaped. To explain the latter temptation by comparison with the former alleged temptation is to miss what is distinctively powerful about the latter. We say that banana images look yellow and curved. By contrast, we don’t normally say that a “nude” painting looks nude or that a “loud” oscilloscope reading sounds loud. Another indication that this explanation is too weak is that we do not talk in imagistic ways about arbitrary representational mental phenomena, such as propositional attitudes. We don’t speak as if our beliefs that bananas are curved and yellow are themselves curved or yellow, nor do we talk this way about our desires to eat curved yellow bananas. Yet this is what one should expect on the “loose talk” view. If it saves breath to speak of yellow banana-images rather than images of yellow bananas (relieving the burden of saying ‘of’), wouldn’t it save more breath to speak of yellow banana-beliefs rather than “beliefs that bananas are yellow” (yielding the life-prolonging benefits of avoiding ‘that’ and ‘are’)?[2]  

            It is more plausible to suppose that in speaking of colored and shaped mental images people are trying to express their imaginative experiences—or perhaps their most natural beliefs about their experiences—sincerely, strictly, and literally. On this account the experiences or beliefs represent that there are images with color and shape, even if there are no such things. This takes more seriously Sartre’s idea that we under an illusion (or hallucination), or a mistaken impression. We need to explain how we (mistakenly or not) come to be subject to such an impression. Is the impression a (mis)conception, as Sartre thinks, or is it more like a (mis)perception? Does the root cause involve certain mental habits regarding space, as Sartre also thinks, or something else? What is the exact content of the impression? Does it serve some purpose, or is it merely a side-effect of other processes? These are questions I address in part 2. But first I want to intensify the problem by introducing a second, largely complementary, phenomenal impression, and by arguing that the image impression has far greater extent than visual imagination. 

1.2     Transparency in visual perception 

            G. E. Moore claims in “The Refutation of Idealism” that when we perceive things visually, and attempt to introspect this activity, we are subject to a misleading impression of “diaphanousness”:

[T]he moment we try to fix our attention upon consciousness and to see what, distinctively, it is, it seems to vanish: it seems as if we had before us a mere emptiness. When we try to introspect the sensation of blue, all we can see is the blue: the other element is as if it were diaphanous. Yet it can be distinguished if we look attentively enough, and if we know that there is something to look for. (1903, 450)

Moore is trying to explain why it might be natural for certain idealist philosophers to confuse experiences with environmental objects. His explanation is that when one tries to attend, say, to a visual experience of a blueberry, one normally “sees through” the experience to the blueberry itself; it is as if experience is transparent.[3] When one tries to describe what seeing the berry is like, one typically describes what the berry is like (or looks like). This point is strengthened by the fact that all the seen features—blueness, ellipticity, motion, etc.—are naturally experienced as “stuck on” seen objects and the environment, in at least three respects. The features seem to be:

            Nonintervening: We intuitively “locate” features at the distal objects and places we visually attend to in order best to detect them—colors and shapes do not seem to travel through the air from seen objects to us, and spatial relations among seen objects seem to be out “among” the objects.

            Experience‑free: We do not see features as being relations to seeing. We intuitively seem to be mere spectators of them rather than participants in them.

            Objectively possessed: We also do not see features in any other way as being dependent on seeing. We intuitively seem to discover them rather than to create or maintain them.

Otherwise, presumably, we would naturally describe some properties (or relations) of objects partly in terms of properties (or relations) of experience or intervening entities, and there would not be an impression of transparency.[4] This is in contrast with the impression of images in visual imagination, where the phenomenal objects and their apparent features seem to be subjective, experience‑laden, and intervening between us and imagined objects. Why does it sometimes seem difficult to introspect properties of visual perception or intervening entities, as opposed to visually perceived objects? 

            In fact Moore insists that we have only an initial difficulty, since experience is not quite transparent. We can introspect that experiences have some distinctive properties, namely, that they represent features of objects. We introspect that the experience is “an awareness of blue” (1903, 449), and this representational property distinguishes experiences from blueberries.[5] So according to Moore sensations may be distinguished from physical objects introspectively, by their being awarenesses, despite the initial impression of “mere emptiness.” This point does not render the impression of transparency psychologically uninteresting, although it does complicate discussion of it. In introspective reports about a visual experience we do sometimes talk as if the only properties we notice are (i) objective, experience‑free, nonintervening features of seen objects, together with (ii) relations of representing these objective, experience‑free, nonintervening features.[6] It will be useful to have a term for these two kinds of properties—I will call them “objectual” properties. So the question becomes: why do we talk as if in introspecting visual‑perceptual experience we introspect only objectual properties? 

            One possibility is that we talk this way because it is accurate—maybe we do introspect what are in fact only objective, experience‑free, nonintervening features of seen objects, and our representing these features.[7] On this account the only introspectible properties of seeing a blueberry are the relations of representing stuck‑on features of the blueberry; all other properties “introspected” are stuck‑on properties of the berry itself, not phenomenal properties of the experience, or features intervening between the berry and the experience. But a search for such blueberry‑experience properties is likely to be fruitful. Consider seeing the particular ellipticity of a blueberry from various angles and distances. Though the berry can appear invariantly ellipsoidal in all these experiences, there is also a sense in which the berry appears differently in each case; one is introspectively sensitive, in ordinary terms, to the differences among the multiple “looks” of a particular unchanging shape. Perhaps surprisingly, given the connotations of the term “look”, such shape‑looks are subject to an impression of transparency, for they do seem to be stuck on seen objects, in the three respects mentioned above. As one moves around an object, although one is sensitive to multiple looks of its shape, it does not seem to one that the shape changes, and more importantly it does not seem to one as if by moving one brings about new looks in the way one brings about new spatial relations to the object. Rather, it seems that, from here, one can discover that the object has this objective look—stuck on one side of it, say—while from there, one can discover that the object has that objective look—stuck on another side of it.[8] The game is afoot when we ask: what properties or relations are such looks, and how are we sensitive to them?[9]  

            The differences among looks do not seem to be explainable as obvious objectual differences, since multiple looks can all be of the same invariant objectual shape. The best bet for a defender of transparency would be to suppose that looks are more arcane combinations of properties that are in fact stuck on seen objects, and that in introspection we are aware at best of such objectual combinations, and our representing them.[10] The only alternative is to deny literal transparency, to explain looks partly in terms of introspectible properties that are not in fact stuck on seen objects—such as spatial relations between perceivers and seen objects, or properties of proximal stimuli interposed between perceivers and seen objects, or phenomenal properties of visual experiences. My explanation of the transparency impression, in part 2, will be of use to the defender of introspectible nonobjectual looks, explaining why we speak as if they are objectual, even if they aren’t. But even if looks are all objectual, some account must be given of how we are sensitive to them in experience. So my explanation of how and why we (mis)represent looks as objectual may be of use in either case. 

            Although looks seem to be stuck on objects, rightly or wrongly, they also seem to be monomodal, to be detectable only through a single sense modality—in this case, of course, vision. We intuitively think that a blind perceiver, though perhaps sensitive by touch to the ellipticity of a berry, is insensitive to related stuck‑on features of the berry—its ellipticity‑looks. The same is true for ellipticity‑“feels”: in feeling the ellipticity of a berry, using various body parts and motions, we are sensitive to multiple feels of what is nonetheless experienced as an invariant shape. And we intuitively think that a numb perceiver might be sensitive by sight to a berry’s ellipticity without being sensitive to such ellipticity‑feels. Arguably, our sensitivity to the overall differences among looks and feels of the same shape helps explain the ease with which we can typically determine not only the shape but whether we’re seeing or feeling it, and whether we are seeing it in an unfocused or doubled fashion (see note 6). Introspectively, at least, degree of focus and degree of convergence are treated more naturally as features of one’s experience than as objectual features of seen objects. This applies even to nondegraded vision, when one is introspectively sensitive to the contrasting fact that the seeing is nonblurred and nondouble.[11] 

            I have used experiences of a primary property, shape, in initiating a search for nonobjectual introspection, but the same structure holds for visual experiences of alleged secondary qualities, such as colors. Our visual systems respond to complex spectral‑reflectance features of seen surfaces (also spectral‑radiance features, but I will simplify), in isolation and in comparison with neighboring surfaces. These reflectance features are objective, experience‑free, nonintervening, and in principle detectable not only by vision but also by other modalities; a being could in principle detect spectral reflectance properties by a highly sensitive version of touch, for example. Yet in visual experience the objective world does not seem to be populated only with spectral reflectances, but with fully clothed reflectance‑looks, to which a blind reflectance‑feeler could be insensitive. Compared with shape‑looks, reflectance‑looks do not vary much as we move around objects; however they do vary with certain changes in ambient lighting even though the seen objects do not seem to change color. I take no stand in this paper about whether colors are “primary” spectral reflectances, or “secondary” dispositions to cause reflectance-looks, or (certain preferred) reflectance‑looks themselves, or combinations of the above, or something else entirely. My emphasis is on the fact that in introspecting normal, diaphanous, vision it is not as if we perceive our images or experiences, or in any other way perceive properties as mediating between ourselves and ordinary objects, and so reflectance‑looks—whether or not they are colors—are intuitively taken to be stuck on ordinary perceived objects, without relation to our experiences. 

            In normal vision, then, we speak as if looks are stuck on objects, as if introspectively we find only objectual properties. But there is no direct inference from the fact that we introspect a property as objectual to the conclusion that it is in fact objectual (see note 7). This inference is plausible, but only unless and until there is a good account of why introspection should go wrong on these counts. 

            As with the impression of images, it should not be plausible that Moore’s claims about diaphanousness are grounded only in ordinary loose talk. We would not “save breath” by “slipping into” speaking as if seen objects objectively have the properties and relations we introspect seeings or intervening entities as having. It is more plausible to suppose that in speaking only of objectual properties people are trying to express their perceptual experiences—or perhaps their most natural beliefs about their experiences—sincerely, strictly, and literally. On this account the experiences or beliefs represent properties only as objectual, even if some of these properties are nonobjectual. We are under an impression of transparency, whether due to (mis)conception or (mis)perception or both. The distinctiveness of transparency‑talk is another indication of this: our propositional attitudes do not engender such talk, contrary to what would be expected if we were merely given to speaking loosely. When we are introspectively sensitive to the strength or rationality of a belief about a blueberry, or a desire for a blueberry, we don’t speak as if the blueberry itself is strong or rational.[12] 

1.3     The impressions in other phenomenal experiences 

            Although Sartre and Moore focus almost entirely on normal visual imagination and perception, the impressions of images and transparency seem important to what it is like to have many other kinds of experiences. The pattern constituting the image impression, which generalizes beyond normal visual imagination, is that normal environmental properties and relations are experienced (rightly or wrongly) as if they were possessed by mental or intervening objects. The pattern constituting the transparency impression, which generalizes beyond normal visual perception, is that mental or intervening properties and relations are experienced (rightly or wrongly) as if they were possessed by normal environmental objects. In this section I catalog impressions of images in nonvisual imagination and degraded perception (such as afterimaging), then impressions of transparency in nonvisual perception and upgraded imagination (such as dreaming), and finally I turn to hybrid cases, including bodily sensation and pain experiences. My purpose is to give what introspective support I can to the “data” I seek to explain in part 2, so that it stands independent of the theory presented there, and also to lay some ground for the conjecture in part 3 that the impressions are necessarily present in all (and only) phenomenal experiences.[13]  

            Perhaps the clearest nonvisual home of the image impression is auditory imagination. Daniel Dennett remarks that we are less inclined to “strike up the little band in the brain” for audition than we are to “set up the movie screen” for vision (1969, 133), but the two cases are more parallel than this would allow. In auditory imagination we may not be inclined visually to imagine a band, but equally in visual imagination we are little inclined auditorily to imagine a soundtrack. In auditory imagination there seem to be likenesses of environmental sounds just as in visual imagination there seem to be likenesses of environmental surfaces. When one imagines hearing a foghorn, or when one plays a song in one’s head, one often seems to produce a faint or ghostly “mental sound,” typically in the cranial auditorium between and slightly above one’s ears. A similar auditory impression of word‑likenesses occurs when one “hears oneself think.” Though we are genuinely tempted to think that these likenesses have volume or pitch, this impression is often weak and faint compared to ordinary hearing of genuine sounds, just as visual images only weakly or faintly “look” colored and shaped. Similarly, if one imagines oneself lifting a glass of wine, then smelling and tasting the wine, one seems to produce mental likenesses (nonvisual ones, of course) in the vicinity of one’s fingers, nose, and mouth. The glass likeness seems to feel cool, moist, and even curved, and the wine likenesses seem to smell and taste (say) woody and recessive, although all these impressions are faint compared with their perceptual counterparts. And of course, even if there are no objects in one’s body, brain, environment, or soul with the relevant volume, texture, odor, or flavor, there are image impressions of such entities.  

            Is the image impression restricted to cases of imagination, whether visual or nonvisual? It plays little or no role in normal, seemingly transparent, perception.[14] However, it seems to arise in some cases of degraded perception. Consider lucid impressions of vision such as afterimages, floating spots, and the fireworks displays that begin when we close our eyes. In some sense it looks as though there are such faint or ghostly items with color and shape, even if there are no such phenomenal objects, and as in the case of imagination we are rarely wholly convinced by this impression. Nonvisual lucid impressions, such as ringing in one’s ears, or aftertastes, also provide good examples of the image impression.  

            Switch now to transparency. My “primary” illustration of this impression in the previous section concerns varying looks of an unchanging shape, and their contrast with varying feels of the same shape, which in effect illustrates the impression in both visual and tactile perception of primary properties. Feels as well as looks seem objectively stuck on objects, discovered rather than created or participated in. The structure of my “secondary” visual illustration of the transparency impression, for color‑relevant reflectances and reflectance‑looks, carries over to nonvisual experiences of alleged secondary qualities, such as flavors, odors, noises, and degrees of warmth. In these experiences, our perceptual systems respond to complex objective features of molecules and air waves. Yet intuitively, more features seem to comprise the experienced portions of the objective world—not only the structure and motion of air waves and molecules, but fully clothed appearances: air‑wave‑sounds (for noises), molecular‑feels (for warmth), molecular-tastes (for flavors), and molecular‑smells (for odors). The genuinely objective features of air waves and molecules can in principle be detected using modalities other than the ones we naturally use, but we intuitively suspect that such modalities could well “miss” the sounds, tastes, feels, and smells we “detect” as being stuck in the environment. As with colors, I am not concerned with the metaphysical question of what noises, degrees of warmth, flavors, and odors are—primary features of air waves and molecules, secondary dispositions to cause the relevant appearances, (certain preferred) appearances themselves, combinations of these, or something else entirely. I am interested in the independent psychological claim that even if these appearances to which we are sensitive are partly dependent on the nonobjectual properties and relations of experience, we do not experience them as if they were. 

            As far as the transparency impression is concerned, what goes for normal perception goes for misperception, as in the case of (ordinary) perceptual illusions as well as (nonlucid) hallucinations. Even though in these experiences there may be no relevant perceived object or feature, it is not as if we perceive an image instead—what it is like to misperceive that something is yellow is introspectively no (or little) different from what it is like to perceive that something is yellow, but introspectively quite distinct from what it is like to imagine that something is yellow. All the properties we represent in misperception are represented as objectual, even appearance properties that may depend nonobjectually on experience. Likewise, just as the image impression seeps into degraded perception, the transparency impression reaches into upgraded imaginings, especially nonlucid dreams. At least when we are not half‑aware introspectively that we are dreaming, we do not seem to experience mental objects, and all the dreamed features seem discovered, stuck on objects independently of the experience. 

            The image and transparency impressions are normally complementary—each is strongest in experiences in which the other is weakest.[15] Yet there are experiences in which the two coexist or at least seem to oscillate as if in a futile conflict. When one crosses or presses one’s eyes more than slightly, squints, or tries someone else’s strong corrective lenses (or removes one’s own), it can become difficult to describe, intuitively, what kind of objects one seems to see—environmental objects, or mental images? I think the best introspective description is that sometimes one seems to see both, whether simultaneously or in alternation. Lucid dreams present a similarly mixed case. 

            I think that bodily-sensational experiences yield another unstable hybridization of images and transparency. In this as in other respects, proprioception does not easily fit molds of either perception or imagination. The image impression enters into experiences of pressure, warmth, or limb‑position insofar as we are tempted to take there to be “sensations” that are themselves pressing, warm, or located in our limbs, mental likenesses of pressing, warm, physical objects in our limbs. Similarly, in each tickle or itch experience one represents parts of one’s body as being “rubbed” or “prickled” with very specific intensities, directions, speeds, and contact‑point sizes. Often, there seems to be a mental “tickle” or “itch” that itself seems to be moving and pressing in these ways. Such reification seems automatic and nearly irresistible, at least when the experienced features are restrictively localized or pointlike rather than diffuse or pervasive. When the features are experienced as diffuse, it is easier to attribute them to ordinary body parts, which unlike reified sensation‑objects are not experienced as mental. I think this is why when we feel (diffuse) fatigue we don’t easily speak of fatigues (Dennett, 1978, xix‑xx), although when we feel (pointlike) tingling we easily speak of tingles. Similarly, when we feel all warm and fuzzy it is easier to reify (pointlike) “fuzzies” than (diffuse) “warmies,” or one big body‑shaped “warmy.” I think the presence or absence of reification is not merely verbal but has systematic and robust consequences for what the experience is like. Features attributed to reified mental likenesses most naturally seem to be experience‑dependent, to be incapable of persisting unfelt. They typically seem to be “activated” rather than discovered when we attend in their direction.[16] Features attributed to body parts are easier (though not quite easy) to experience as objective, discoverable:

As I pace back and forth in my room I find that I frequently pause in front of the window. Asking myself why, it suddenly dawns on me that I am quite cold, and that my pauses have been due to the succoring warmth of the sun’s rays. (Hill, 1991, 119)

In such a case when we imagine what has been succoring, we imagine that all along our bodies must have had a kind of warm feel that we can discover; the features we would be sensitive to merely in seeing high molecular agitation would not succor at all. To the limited extent that this “feel” seems to be possessed by our body parts, discoverable as afflicting us from there, proprioception can subject us to an impression of transparency, in addition to the more common image impression.[17] 

            As usual, pain experience presents special difficulties. In pain experience it seems to us that something is going on in various parts of our bodies. We often speak of pains as subjective objects felt in our bodies, which is one aspect of the image impression. But a stronger requirement must be met for the image impression to be present. We must feel the pains as if they were subjective likenesses; we must feel them as having features that are in fact had only by normal physical objects. What features, then, do pain experiences represent? The details matter to the presence or absence of the image impression. On some views, for example, pain experiences represent damagedness or disturbance—but these properties don’t engender image impressions, since we don’t suppose our pains are damaged or disturbed. In fact, however, we never have pain experiences that represent merely an inspecific damage or disturbance. In pain experiences we represent parts of our bodies in specific ways we try to express as “throbbing” or “burning,” or as being “stabbed,” “pounded,” “pinched,” “pulled,” etc. The image impression in such cases makes it also seem that there is a pain‑object that itself throbs, burns, stabs, pounds, or is at the limb.[18] 

            With somewhat more hesitation, I think that pain experiences can also breed the impression of transparency. This claim has been disputed by Paul Boghossian and David Velleman. Although they defend the view that visual appearances are projected onto seen objects, they reject such projectivism about pain feelings, on the grounds that we don’t feel the painfulness of a pin‑prick as existing objectively in the pin (1989, 95). They are of course right about the pin, but the general pattern of projection from the mental to the nonmental is apparent even in some pain experiences. We do sometimes feel aspects of painfulness as belonging to our body parts. Which aspects? Ones we may clumsily try to describe as burningness and throbbingness, for example. Burningness and throbbingness are perfectly objectual properties, like shape open both to being seen and to being felt, but the appearance properties we try to express strike us as burningness‑feels and throbbingness‑feels, not open to being seen. These experience-dependent appearance‑properties, nevertheless, often feel stuck in our arms, in our teeth, discovered as afflicting us from there. 

            For these feels to generate a transparency impression, a stronger requirement must be met: it must be that they introspectively seem stuck on our body parts independently of experience. This would be doubtful if it required there to be achings in the absence of experience, since (rightly or wrongly) many people recoil at that idea. But in a burning‑pain experience the burning (and burningness‑feel) can seem to exist independently of experience, even if when it is unexperienced there is no aching. The aching plausibly includes some sort of aversion to the burning, which requires experience of the burning in addition to the apparently objective burning itself.[19] So the transparency impression does not require that hurting seem independent of experience. It only requires (say) that burning and throbbing (clothed in their multiple feels) seem to belong to body parts objectively, in the same way that warmth (clothed in warmth-feels) can seem discoverable in one’s body. To the extent that people can have pain experiences without experientially or introspectively reifying their pains as the objects with these features, they invite instead the transparency impression. This extent is limited: in burning‑pain experience does one typically feel one’s skin as burning, or does one feel pain sensation‑objects at one’s skin as burning, or both (simultaneously or in alternation)? I vote “both”; some cases introspectively seem mixed in the way degraded vision does.[20]  

2     An explanation of the impressions 

            The eliminativist about phenomenal objects of experiences would do well to explain the allegedly illusory impression of their presence (e.g., in normal imagination), and the noneliminativist about phenomenal properties of experience would do well to explain the allegedly illusory impression of their absence (e.g., in normal perception). While many philosophers of mind lament the impression of images—William Lycan (1987) calls it the “Banana Peel” too often used to slip up materialist theories of consciousness—few existing theories of phenomenal experience attempt to explain how or why we have it. And while many have tried to make use of the impression of transparency—Moore wields it against idealism, and others try to argue from it that we “project” colors and odors onto the dull objective world—I know of no theories that attempt to explain how or why we are subject to it. I propose that the two impressions would be explained if there is a certain kind of introspection that produces inner perceptions, given some widespread and natural assumptions about perception and attention (section 2.1). To play this explanatory role such inner perception should be involved in experience (rather than added to experience or directed at it from outside), and this helps insulate inner perception from some of the main objections it faces (section 2.2). Finally, I argue against some possible rival accounts of the impressions (section 2.3). 

2.1     Binding in perception and introspection 

            The image and transparency impressions are both impressions engendered by introspection and attention, in some sense, as they relate to phenomenal experience. Sartre claims that an image “is describable only by an act of the second degree in which attention is turned away from the object and directed at the manner in which the object is given” (1940, 3), while Moore, as quoted earlier, holds that transparency is noticed when we “try to introspect” a sensation or “try to fix attention upon consciousness” (1903, 450). It is likely that to explain the impressions we will need a substantive understanding of the relations among introspection, attention, and experience. I begin with some fairly uncontroversial assumptions about these relations. 

            There are very many theories of introspection, some of which may be genuine rivals, and some of which may describe compatible processes that should each go by the name of ‘introspection’. I will develop an account of one kind of introspection (perhaps coexisting with others) that I think is especially relevant to explaining our two impressions of phenomenal experience; for now, call this ‘E‑introspection’ (‘E’ for ‘experience’). I assume that E‑introspection is an “indicative” process in the sense that it produces some sort of psychological states (events, data structures, etc.—let’s just use ‘state’ broadly) that have a subject matter and purport to be true or accurate about that subject matter. I do not begin with finer assumptions about whether E‑introspection produces beliefs or else perceptions (or judgments, or hypotheses, etc.), and I do not make assumptions about whether these products are conscious, or whether they are complete or infallible or reliable about their subject matter. I do assume that the subject matter of these introspective products can include features of psychological states involved in phenomenal experiences. Finally, I assume at least some of these introspected states have their own subject matter, typically regarding experienced physical objects.  

            Call the states produced by E‑introspection ‘I‑states’, since they are typically inner‑directed (or about one’s own mental states); similarly, call E‑introspected states ‘O-states’, since they are typically outer‑directed (or about things other than one’s own mental states, such as tables, trees and one’s body). (The ‘O’ may also express their status as representational “objects” of E‑introspection.) For instance, on this view, E‑introspection regarding a perceptual or imaginative experience of a table would involve at least the following structure (where the dashed arrows signify causation that is typically present but not strictly required):[21] 

What features of O-states are represented by I‑states? Here I will be quite noncommittal, since the details will not be crucial to the explanation of the impressions. I‑states may be about intrinsic features of O‑states, such as neural structure, or about extrinsic features of O‑states, such as their causal relations to other mental states (including other O‑states), to sense organs, or to environmental stimuli. What does matter to the impressions is that all such features are “mental” or “mental-like”: they are “experience‑dependent” at least in the sense of being dependent on the existence of O‑states; they are normally imperceptible by others because they are partly or wholly realized in one’s head; yet one is sensitive to them in a kind of introspection.[22] By contrast, I assume O‑states typically represent wholly experience‑free features of environmental stimuli, such as shape, motion, and reflectance.[23] The features represented by I-states serve as the looks, feels, and other varying appearances of these environmental stimuli. 

            Now switch from introspection to attention. Attention is focused on some subject matter when a state representing it is given greater priority than “rival” states in processing—in producing further states about (or behavior toward) the subject matter, or in identifying it in the first place. There is also typically a strengthening of the state representing the subject matter, and a weakening of rival states, in dimensions such as conviction and salience. The process of attention direction is not always completely subject to willful control; one may try to attend to a subject matter but fail because rival states remain strong or prioritized. I also assume that O‑states and I‑states are typically rivals. So when attention is (successfully) directed outwardly, O‑states are typically stronger than I‑states and are processed with greater priority; they have greater influence for identification of objects and inferential and behavioral treatment of them. When attention is (successfully) directed inwardly, the reverse is true—I‑states are stronger than and have priority over O-states. The latter point can be difficult to keep straight. What it is for attention to be focused on an O‑state is not for the O-state itself to be strengthened or given priority, but for an accompanying state about it to be strengthened and prioritized (i.e., an I‑state). This parallels the external case—what it is for attention to be focused on a table is not (absurdly) for the table itself to be strengthened or prioritized, but for an accompanying state about it to be strengthened and given priority (i.e., an O‑state). 

            What does all this have to do with the impressions of images and transparency? The impressions may spring from a certain kind of (con)fusion of O-states and accompanying I‑states. (Think “confusion” if an impression is illusory, and “fusion” if it’s veridical.) O‑states and I-states represent features that may not in fact be had by the same things (typically, O‑states represent features of environmental objects and I-states represent features of O‑states).[24] The (con)fusion I postulate is that these states attribute these features to the same objects, or, in other words, that the states operate as if they are about the same objects. For example, suppose a perceptual or imaginative O-state represents something (say, a table) as being square and as reflecting predominantly long‑wave (red) light. Also, suppose the O-state is itself E‑introspected as having certain structural or functional features Q and R. On the view I propose the O-state and I‑state represent the bearer of these features in a unified way, as if by using the same variable, name, or dummy constant (say, ‘x’):

In a moment I will try to explain why there might well be this (con)fusion involving O‑states and I‑states; for reasons to be explained below, call it the ‘binding (con)fusion’. First I want to indicate how such a (con)fusion would, if it existed, yield a unified explanation of the image and transparency impressions. The parameter that varies between these impressions is whether attention is directed outwardly (at the perceived or imagined object) or inwardly (at the perception or imagination of the object).  

            The transparency impression occurs most easily in introspecting normal perception or nonlucid dreams, when O‑states rather than I‑states are strengthened and given priority in object identification—that is, when attention is focused on the environmental objects that O-states are about. In introspecting a visual perception of a red square one identifies the object of the experience most saliently as something environmental—e.g., as something square and (objectively) red—and adjusts one’s inferential and behavioral dispositions toward it accordingly. Since the salient properties are environmental (and publicly perceptible), the object seems environmental (and publicly perceptible). Given E‑introspection, there is also weak and inattentive representing of mental (or at least normally publicly inaccessible) properties Q and R. What would the binding (con)fusion generate in this context? The weakened and subsidiary I‑states would be treated as applying to the (alleged) environmental objects themselves. We would weakly represent the red square as having Q and R—a “projection” of experience‑dependent properties of the O‑state itself as appearances belonging to the environmental, public object. The transparency impression would be explained as an attentive identification of the (environmental) objects O‑states represent, and an inattentive attribution that they have the (mental) properties I‑states represent. The resulting impression would be of features that are had by experiences, or had by objects in relation to experiences (Q and R, in the example) as being stuck on environmental objects.[25] 

            Conversely, the image impression occurs most easily in introspecting some forms of perceptual imagination or abnormal perception, when I‑states rather than O‑states are strengthened and given priority—that is, when attention is focused on the mental entities that I‑states are about (namely, the O‑states).[26] In introspecting a visual‑imaginative experience one identifies the object of the experience most saliently as something mental—e.g., as something Q and R—and adjusts one’s inferential and behavioral dispositions toward it accordingly (typically, by dampening them). Since the salient properties are mental (and publicly inaccessible), the object seems mental (and publicly inaccessible).[27] Given the weakened, inattentive, O‑states about red‑reflectance and squareness, what would the binding (con)fusion yield? The mental object would weakly be represented as being square and red—as imagistic. The image impression would be explained as an attentive identification of the (mental) objects I‑states represent, and an inattentive attribution that they have the (environmental) properties O‑states represent. The resulting impression would be of features—that are had by environmental objects (squareness and red‑reflectance, in the example)—as being subjectively had by mental objects, such as images or pains. 

            On this account, to the limited extent that it is possible to shift attention between perception and introspection, it should be possible to shift between the transparency impression and the image impression. I think this is what we find. This is why wicked philosophy instructors can convince beginning students that all they ever really see are their images—if they sit still in an uneventful room, attending to their experience rather than to the world. And to the extent that attention cannot normally be divided between perceptions and introspections, the transparency and image impressions should normally be alternatives, as in fact they are. 

            The explanation of the image and transparency impressions contains three main components:

   (1)     There is a binding (con)fusion between states representing objectual features of ordinary perceived objects, and states representing other features.

   (2)     These latter states include I‑states produced by a (perhaps distinctive) kind of introspection, E‑introspection, of the former states (O‑states).

   (3)     There is attentive rivalry for strength and priority between the two kinds of states.

(1) and (2) are meant to explain what is common to the impressions, while (3) is meant to explain how they diverge. I have not given direct reasons to believe (1)-(3). Of course, some indirect reason is provided by the very fact that if they were true they would help explain the impressions. The significance of this will become clearer given the failures of alternative explanations (see section 2.3). But a fully convincing argument that (1)-(3) are true and do explain the impressions would require more direct scientific evidence. As a step in that direction, let me try to explain why there might well be a binding (con)fusion between O‑states and I‑states (even if the resulting impressions are illusory). 

            First one must understand a certain feature of normal outer‑directed perception (whether experiential or subliminal, conscious or unconscious). Various sensory “transducers”—small portions of sense organs such as retinal cells, tactile receptors and auditory follicles—each produce in different areas of the brain what Dennett (1991) calls “multiple drafts” (states, events, data structures, etc.) bearing information or misinformation about many properties of stimuli (e.g., sudden discontinuities of brightness, ratios of spectral reflectance, shapes, motions, etc.). This creates what is known in perception research as the “binding problem”: how does a perceptual system keep track of which properties belong to which perceived objects? As usually raised in the cognitive neuroscience of vision, the problem is a bookkeeping one: given that the visual system produces, in different areas of the brain, drafts representing color, motion, orientation, shape, etc., what neural or functional “stamp” does vision give to these separate drafts when the features are seen as coinstantiated? There are various proposals for this stamp of coinstantiation in vision; the most influential is that in spatially separate visual drafts representing features as coinstantiated, the neurons involved fire, repeatedly, in synchrony.[28] What seems seldom noticed is that finding such a stamp of coinstantiation can solve only half the problem about coinstantiation. The remainder of the problem is this: whatever a perceptual system uses as a stamp of coinstantiation, what does it use as a symptom of coinstantiation? How does a perceptual system determine which features to stamp as coinstantiated? I suggest what seems obvious: whatever other symptoms may be used, one extremely plausible and reliable method for determining coinstantiation is to treat drafts caused by the same transducers (at a moment) as applying to the same object. If a blue‑representing draft and a circle‑representing draft have one set of transducers as a common cause, while a red‑representing draft and a square‑representing draft have a different set of transducers as a common cause, then it is an excellent bet that one is seeing a blue circle and a red square, rather than a red circle and a blue square. Call this ‘common‑connection binding’.[29] 

            Now if some of these perceptual drafts (or O‑states) are E-introspectible, common‑connection binding could also apply to I-states about them. An I‑state caused by an O‑state is of course indirectly caused by the same transducers as the O-state. If common‑connection binding extends to the products of E‑introspection, this would explain the presence of a binding (con)fusion between O‑states and I‑states, which is in turn the proposed key to explaining the phenomenal impressions. In other words, on this account the price of solving the binding problem for perceived properties of physical objects is that E‑introspected properties of perceptual drafts are bound along with them. The impressions of images and transparency are side‑effects of a valuable perceptual strategy understandably overextended to E‑introspection. 

            Pending a more powerful and simple explanation of the impressions (see section 2.3), I believe that this outcome—that common‑connection binding applies to E‑introspection—gives both support and meaning to one version of the traditional doctrine that there are “inner perceptions.” I will end this section by describing why I believe this. 

            Common‑connection binding is generally reliable only for a very restricted range of representational states. For example, the reliance upon retinal transducers in binding should be restricted only to visual states, narrowly delimited, not to judgments (or beliefs) generally, even if they happen to be caused by vision. The judgments caused by a single set of transducers (at a moment) may differ arbitrarily in subject matter: some bit of retinal activity at a moment may (distinctively help) cause me to judge that there’s a bee nearby, or that I am going to be stung, or that the lake would be a good place to hide, or that Mickey Mantle would have been proud of the way I’m swinging this stick, etc. Since all these judgments attribute properties to different objects, common‑connection binding cannot be in use for them. For a given perceptual system, we should expect common‑connection binding to apply only to representational states that belong to that perceptual system, in the sense that their subject matter is fixed solely through the system, rather than through other perceptual modalities or through inferential relations to “central” judgment and belief. So if the impressions are to be explained in the manner of (1)-(3) above, via binding (con)fusions that extend to the products of a kind of E‑introspection, then we should expect this E‑introspection to produce states that themselves belong to perceptual systems. E‑introspection of O-states in the visual system should produce I‑states that function as parts of the visual system, explaining why they would be “(con)fused” with visual O‑states. 

            This gives a strong sense in which the products of E‑introspection are perceptual. On this account, rather than being states of a distinctive inner faculty, I‑states are states of the various outer‑perceptual systems, and so count as visual, auditory or other perceptual states, 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 I‑states are produced, in whatever way, they are processed like outer perceptions, as further states in particular sense modalities. For example, visual I‑states help to produce visual beliefs, help to control visuomotor skills, and are not introspectively distinguished in kind from visual O‑states. They may qualify as perceptual due to their use, even if not due to their origin (since there are no inner eyes and ears). By virtue of producing such perceptual outputs, the process of E‑introspection may rightly be called ‘inner perception’ or ‘inner sense’, however the process is structured internally.[30] 

2.2     Inner (mis)perceptions in phenomenal experience 

            For both scientific and everyday purposes we try roughly to distinguish perceptual impressions from judgments. Typically, when in error the two require different sorts of home remedies: broad rationalistic appeals to evidence or authority or prudence may work against the latter while having little or no effect on the former; narrow animalistic strategies such as squinting or moving around or refocusing attention may work against the former while having little or no effect on the latter. There is also typically a difference in self-control: knowing that a particular judgment is in error typically enables one to cease the judgment, while knowing that a perceptual experience is illusory typically does not enable one to cease the experience. There is much controversy about whether and how to draw a distinction between perception and judgment, and I do not wish to delve into this controversy here. The distinction is perhaps especially difficult to draw in the realm of imagination, which often involves purposeful and cognitive influences on systems normally used for perception. What I wish to argue is that the image and transparency impressions bear the hallmarks of the clearest cases of familiar perceptual impressions, whatever the facts may be about borderline cases. 

            We most naturally speak as if the impressions of image and transparency govern how things appear rather than how they are merely judged to be. In the image impression, there seem to be visual images, afterimages, and closed‑eye fireworks that look purple and round, thoughts and ringings‑in‑the‑ear that sound faint or high‑pitched, and tactile images and pains that feel dull or in motion. This claim is supported by the fact that when we have image impressions about (say) visual experiences, we only have them with regard to visible properties, very strictly delimited: we are tempted to think that our images of yellow bananas are banana‑shaped and yellow, but not slippery or imported, and not genuine bananas. Similarly, in the impression of transparency, we seem to see and not merely to judge circularity‑looks as opposed to circularity‑feels, we seem to hear and not merely to judge air‑wave‑appearances, and so on. Also, like familiar perceptual impressions, image and transparency impressions persist virtually unchanged even when we come to believe they are illusory. The perceptual temptations to believe in colored and shaped images remain even when one is convinced by argument that no such things exist. The perceptual temptations to take shape-appearances and color‑appearances as objective persist even when we know better—even when we resist judging that they are independent of our experiential faculties. The impressions therefore seem to be built into phenomenal experience, due to some components of our perceptual systems—or some strictly perceptual components of our imaginative systems—rather than juxtaposed with experience by our highly “cognitively penetrable” systems of judgment.  

            The binding‑(con)fusion account from the previous section can explain how the impressions are built into perceptual experience itself, on the additional assumption that E‑introspection is built into phenomenal experience.[31] This would explain for example why one seems to experience and not merely to judge that there are subjective images and objective appearances. Nevertheless, even if the image and transparency impressions are illusory they would not render the entire content of experience illusory. On this proposal a perceptual experience involves both outer perceptions (O-states about shapes, reflectances, etc.) and inner perceptions (I‑states about some structural or functional properties of perceptions). A visual‑perceptual experience can be both veridical about objective red‑reflectance and squareness (represented by O‑states) and illusory about objective reflectance‑looks and shape‑looks (represented by I‑states). A visual‑imaginative experience can be both veridical about subjective reflectance‑looks and shape‑looks (represented by I‑states), and illusory about subjective red‑reflectance and squareness (represented by O‑states). Furthermore, in each case the allegedly illusory content is borne by weak and background states, while the strong, prioritized states bear the (potentially) veridical content. So the image and transparency impressions, even if they are illusions, need not be much of a threat to one’s pride or one’s hide. 

            Independently of explaining the impressions, perhaps the main reason for postulating an (E‑)introspective component to experience is that this can provide part of an explanation of the (sharp or vague) distinction between normal, conscious, phenomenal, perceptual experience and unconscious, nonphenomenal, nonexperiential perception. Arguably there are perceptual states without experience, in subliminal perception, “blindsight,” and “early” states in processing in the retina, lateral geniculate nucleus, and (perhaps) primary visual cortex.[32] We conceive of these as perceptual without assuming they are experiential—and so we speak of subliminal perception, blindsight, primary visual cortex, etc. Unconscious perception of a table is like conscious perception of a table in generating mental states about the table—states akin to O‑states—but seems unlike conscious perception precisely in the absence of even primitive introspective awareness of these states—in the absence of I‑states.[33] The claim that there can be perceptions wholly lacking in consciousness and phenomenal properties helps to insulate inner perception (and E‑introspection) from its two most influential philosophical objections. It is ironic that these objections concern images and transparency, which I have used in indirect support of inner perception. 

            The objection concerning images stems from the (wholly proper) denial of “sense data”—immanent phenomenal objects interposed between physical objects and one’s perceptions of physical objects. The worry is that accepting inner perception (especially as part of perceptual experience) would involve accepting that one at best perceives outer objects indirectly through inner perceptions of phenomenal objects in one’s own mind (see for example Harman, 1990). My response is that a properly formulated inner‑perception model of experience is not committed to sense data. Inner perceptions needn’t be directed at entities interposed between objects and one’s perceptions of them—the causal chain in perceiving a table needn’t proceed from the table to an introspection and then to a perception of the table. Rather, on a more natural view, the causal chain goes directly from the table to a perception of the table (an O‑state), and then (in cases in which the table‑perception is not merely subliminal) to an introspection of the perception of the table (an I‑state). Both outer perception and inner perception are “direct” in the sense of not requiring mediation by further perceptions. 

            The objection concerning transparency begins by drawing out an alleged commitment of inner perception: since each outer‑perceptual modality (seeing, hearing, etc.) makes its own distinctive contribution to what experience is like, an additional modality of inner perception should be expected to make its own contribution, to change what it is like. The alleged problem with this commitment, given apparent transparency, is that what it is like to introspect a perceptual experience seems simply borrowed from what it is like to have the experience itself (perhaps the best statement of this problem is in Rosenthal, 1990). When one tries to attend to features of normal experiences, one normally “sees through” the experiences to outer objects. So a fundamental disanalogy between outer perception and alleged inner “perception” is that the former but not the latter has its own phenomenology or perceptual quality. This is reason to think that inner perception cannot explain introspection of ongoing phenomenal experiences. My response is based on the idea that inner perception (E‑introspection) is involved in phenomenal experience from the start. Contrary to the objection, outer perceptual modalities are not in themselves sufficient for phenomenal experience, which is how there can be states of perception it is like nothing to have (perhaps retinal states, wholly subliminal states, blindsight, etc.). Rather, on the present account inner perception helps convert ordinary nonphenomenal outer perceptions into phenomenally conscious “experiences.” Instead of borrowing phenomenal qualities from an outer perception, as the transparency objection alleges, inner perception would help generate these qualities together with (otherwise nonphenomenal) outer perception. This explains why inner perception doesn’t add further qualia to an outer‑perceptual experience: inner perception has already made its phenomenal contribution for there to be an outer experience with phenomenal properties in the first place. 

            What specific phenomenal contributions could E‑introspection make to experience? I remain noncommittal about which properties and relations of O‑states are represented by E‑introspection, but it is worthwhile to explore some possibilities. E‑introspectible intrinsic features of an O‑state may include features specific to its hardware realization—such as the rough number of neurons that realize it, or their rough average rates of firing—or more abstract “syntactic” features. E‑introspectible relations among O‑states—those that can’t be reduced to their intrinsic features—may include certain of their functional relations and perhaps even their spatial relations in the brain. These are the sorts of properties that we might expect E‑introspective processes to detect with some reliability. Let me illustrate how I take such E‑introspections to enter into perceptual experiences. 

            First, consider cases of double vision or blurred vision. Typically we are sensitive to the doubleness or blurriness of such experiences, though this is difficult to explain as mere sensitivity to objectual properties—objective features of environmental surfaces together with the generic relation of representing them (see note 11). My suggestion is that we are sensitive to doubleness or blurriness because we E‑introspect relevant nonobjectual structural or functional properties of our O‑states. In double vision, we may E‑introspect of two O‑states (say, two matching perceptions of an edge) that they are two in number—this is not itself an objectual property, but a kind of relation between the O‑states. In blurred vision, we may E‑introspect of a certain O-state (say, a perception of an edge) that it is in a causal relation with an unusual set of other O-states (say, perceptions that line up poorly in the retinotopic maps in primary visual cortex—see note 29). In normal focal vision, by contrast, we are typically sensitive to the nondoubleness and nonblurriness of our experience. This could be explained by our E‑introspection of related structural features of O‑states. We detect of an O‑state of an edge that it has no distinct matching O‑state, and that it is in a causal relation to a “lined-up” set of other O‑states. 

            Or consider Peacocke’s case of the two trees (see note 10). Peacocke argues that “you simply enjoy an experience which has the feature” of different sizes‑in‑the‑visual‑field. However, it is not simply that the experience “has” this visual-field feature, in the way it might “have” the feature of being realized in, say, molecules. In addition, one is normally sensitive to an experience’s visual-field features (in a way one is not normally sensitive to its being realized in molecules). An E‑introspection account can explain the visual-field differences in Peacocke’s two experiences, as well as one’s sensitivity to these features, as follows. Compared with O‑states about the distant tree, O-states about the nearer tree are realized by (or causally connected to) many more O-states in retinotopic maps in the early visual system. This is just the sort of relation to which E‑introspections may be sensitive. 

2.3     Alternative explanations of the impressions 

            The common root of the impressions of images and transparency, on the account I have offered, consists of the following:

   (1)     There is a binding (con)fusion between states representing objectual features of ordinary perceived objects, and states representing other features.

   (2)     These latter states include I‑states produced by a (perhaps distinctive) kind of introspection, E‑introspection, of the former states (O‑states).

I divide the two claims in this way because it is tempting to maintain the first without the second. The core of the binding‑(con)fusion explanation of the impressions can in principle work independently of the alleged role of E‑introspection (inner perception). The most important feature of E‑introspection, for purposes of explaining the transparency impression, is that it is sensitive to nonobjectual properties: properties (e.g., neural or functional ones) that relate many‑to‑one with perceived objectual features (e.g., shape), in the way that looks and feels and appearances generally do. It is the fact that I‑states represent nonobjectual features that generates an illusion of transparency when I‑states are treated as O‑states through the binding (con)fusion. And the most important feature of E‑introspection, for purposes of explaining why there is a binding (con)fusion in the first place, is that it could plausibly be subject to common‑connection binding—the strategy of treating perceptual states caused by the same transducer (at a moment) as about the same object. But the transparency impression might be explained with a similar binding (con)fusion, also due to common‑connection binding, without I‑states. There are two other natural nonintrospective candidates for states representing nonobjectual properties: (a) states representing transduced proximal stimuli that vary while distal properties remain constant, and that vary across sense modalities, and (b) states representing causal or spatial relations between ourselves and experienced objects. These two possibilities need not be rivals to one another (or to inner perception); some cases of transparency may be of one sort, others of another sort. But they do require separate comment.  

            States early in the perceptual process clearly show sensitivity to transduced proximal stimuli. On virtually all detailed theories of normal vision, for example, cell‑firings in each retina cause (or constitute) states representing the amount of incoming light of various wavelengths at various points near each eye, which cause states representing sudden discontinuities of incoming brightness, which cause further proximally‑representing states, and, eventually, familiarly conscious visual experiences. Some may not wish to say these early states “represent,” but the label doesn’t matter here so much as the states themselves (which might be said to “protorepresent” instead). What matters for explaining transparency is that the properties they respond to are vision-specific, at least given our other sensory modalities. One has no nonretinal perceptual way to detect the proximal properties affecting one’s retina. If these proximal properties are the properties we call looks, this might explain why (we think) we can’t be sensitive to looks in any way other than vision.[34] It is very tempting to try to explain transparency via a binding‑(con)fusion involving these states—to hold that proximal stimulus properties are confusedly bound onto distal objects. For one thing, proximally representing states and distally representing states share transducers as common causes, just as distally representing states and inner perceptions of them would. For another thing, we already know that there are proximally (proto)representing states, and we already understand well why there are. The same cannot be said for inner perceptions!  

            Nevertheless, early‑visual states about proximal stimuli are unlikely to figure directly in phenomenally conscious experiences, and so are unlikely to be the states crucially relevant to the transparency impression. Retinal cells are active even in cases of subliminal visual perception and blindsight. Also, since our familiar visual experiences are not continually like double images, there seems to be nothing phenomenal about our separate left‑eye‑caused and right‑eye‑caused early visual states themselves. Normal visual experience clearly involves stereoptical states representing distal properties, but only dubiously involves monoptical states representing proximal stimulus properties. It is also unclear why, unlike inner perceptions, proximally representing states would be attentive rivals for distally representing states, rather than each of them calling upon independent attentional resources: proximally representing states must have strength and processing priority in order for distally representing states to acquire their own strength and processing priority. Without the rivalry appropriate for generating apparent transparency rather than apparent images, there would be no explanation on this view for why consciously we seem not to see arrays of incoming brightnesses, i.e., for why our intuitively conscious experiences do not represent distant objects as being behind two splotches of light near our eyes (nor, in monocular vision, one splotch). So in the end I do not think there is much prospect for identifying proximal stimulus properties with the “looks” we experience as stuck on distal objects. 

            Although we clearly have perceptual states (proto)representing proximal stimuli, it is less clear that there are many perceptual states (proto)representing the obtaining of relations between ourselves and perceived objects. Our retinal cells may be sensitive to the proximal brightness and wavelength patterns that result from (and vary with) our spatial relations to the moon, but they don’t seem to be about our having relations to the moon. In part this is because they are not about the moon, and in part this is because they are not about us. Psychophysicists have not yet identified any retinal cells whose receptive fields include either the moon or the self. But at some point in the perceptual process distance and direction become viable perceptible candidates: we can see or hear how near we are to something along which line, and these relations can vary while the object does not look or sound like it pivots, travels, changes volume, etc. Despite this, such spatial relations do not seem relevant to explaining “looks” and “feels,” because they are not even apparently restricted by sense modality. One can see, feel, hear, and even sometimes smell how far one (or another) is from an object, and in what direction. Also, like states representing proximal stimuli, states representing spatial relations seem to lack the attentive rivalry relations they would need to explain the impressions. Just as it is not taxing to attend simultaneously to the shape and color of an object, so it is easy to attend simultaneously to the shape and distance of the object. So if the binding‑(con)fusion and attentive rivalry are to be operative, all or most of the work of explaining the transparency impression seems to fall to inner perception. 

            Most attempts to explain the image impression, unlike my own, turn on the role of space in imagery. Sartre’s first step toward a diagnosis of the image impression, quoted in section 1.1, blames “our habit of thinking in space” for our “misconception” of images. Even in advance of the details, this is an unpromising direction for developing a space‑based explanation. To the extent that we have a relevant general “habit” of spatial thinking, we should display this habit even when we think of beliefs and desires about color and shape, and so we should be subject to the impressions that these attitudes are colored and shaped. But we are not. There must be something more specific to imagination that explains why the impression is so tempting in this realm. If space is at the root of the impression, it is likely to be due to spatial perception rather than spatial thinking. Georges Rey pursues an explanation of the image impression along these lines:

[P]eople tend to reify the objects of their thought. In the [movies], there are rapidly moving celluloid images that cause certain illusions … not of an object in real space, but in some peculiar space on the screen; and in the case of qualia, there are … illusions of corresponding properties, properties that appear to exist not in the ordinary world, but in some “internal” world of the mind. … [P]redicates [or predicative states] released by the visual module … [are] parameterized for relative position in at least two dimensions. … [S]ince real length and width and color seem to persist in real space, it is difficult to resist the impression that “phenomenal length, width, color” persist in mental space. … [H]owever, [we have] no reason to take any of these reifications seriously. (1992, 309)

Even if we agree with Rey that we should not accept the tempting reifications of mental objects, we should not agree with him (and with Sartre) that spatial oddities create the temptations, inclining us to view images as “phenomenal” or mental or otherwise quirkily unlike normal perceptible objects.  

            Someone might seek to explain the apparent quirkiness of images by their appearing nonspatial or at least (with Rey) by their appearing in nonphysical space. But visual images do appear spatial—we speak of their shapes, directions, sizes, etc. Perhaps surprisingly, images also appear to be in the same physical space as ordinary physical objects. We represent spatial relations between afterimages and perceived objects; closed‑eye fireworks seem to be on or near the backs of our eyelids; the products of visual imagination seem to float around just inside of our eyes; other kinds of images seem to be in our cranial auditoriums, fingers, noses, and mouths; and ringing‑in‑the‑ears seems to be in the ears. The quirkiness of images cannot be explained by their appearing in our bodies or heads, either. Some objects perceived as being in our heads and bodies lack the quirkiness of images (e.g., teeth, and pieces of apple being swallowed), and some alleged images seem to be outside of our heads and bodies (e.g., some visual afterimages, and the feeling of a surface through a held stick or wand[35]). Regardless of their apparent locations, there seems to be something ghostly about how images appear, which needs describing and explaining. The apparent ghostliness of visual images is also not a matter of apparent flatness, since not all of them seem flat; instead, one seems to be able to rotate them in depth. All of these points weigh against attempts to explain the image impression in spatial terms as Rey and Sartre do.[36]  

            Temporal fleetingness does not explain the apparent ghostliness of images any more than spatial oddities do, since some ghostly images persist, and since some ordinary fleeting objects (e.g., lightning, weak soap bubbles) are not ghostly in any way that tempts us to take them as mental. Although the spatiotemporal quirks require explanation, they do not seem essential to the image impression. Likewise, a feeling of having willed or created an image is not necessary—we have no such feelings in cases of lucid hallucinations (afterimages, ringing-in-the-ears, aftertastes), bodily sensations, or of being haunted by unwanted imagery. 

            Although I do not think weird space is the key to images, I agree with Rey that the image impression stems from deep features of perception. I think the most general way in which images appear quirkily unlike perceived physical objects has to do with monomodality. The features represented by I‑states in a given perceptual system are monomodal in the sense that they are normally detectable only by that system—not by other senses and not by other perceivers. For this reason we have no natural dispositions to “test” I‑states against the verdicts of other senses or other perceivers.[37] This contrasts with our visual O‑states about (say) shape, which we are disposed to test against tactile states, and it even contrasts with our visual O‑states about color‑reflectance, which we are disposed to test against (reports of) the visual states of others. In imagination, O‑states about such polymodal properties are weakened and given low priority in processing. Given this, a perceptual system primarily (mis)identifies images in terms of the monomodal properties represented by stronger, prioritized I‑states. Even though visual images weakly appear to have shape and color‑reflectance, then, we find ourselves without a readiness to investigate visual images by other senses or other perceivers. 

            Consider, by contrast, the sense in which normal perception represents objects as nonghostly. Ordinary physical objects, as opposed to ghosts and visual images, are tangible as well as visible. When we seem to see a table, we expect to be able to touch it, but when we seem to see a ghost or visual image, we expect not to be able to touch it.[38] Apparent intangibility cannot capture the ghostliness of all images, since bodily sensations and tactile images do appear tangible—we seem to feel them in our bodies. Nevertheless, when we seem to feel a pain or a tactile image, we expect not to be able to see it, though we expect visibility when we seem to feel a table. When we undergo auditory image impressions, we do not expect to see, feel, smell, or taste the images, nor do we expect others to hear them; similarly, in the grip of other image impressions, we expect only to feel our alleged tactile images and pains, we expect only to taste our alleged gustatory images, and we expect only to smell our alleged olfactory images. We also take allegedly visible images to be untouchable, untasteable, and unsmellable, like allegedly visible ghosts, and even to be inaudible.[39] The content of the image impression is better explained by general features of inner perception than by general features of spatial perception or judgment. 

3     An explanation of phenomenality? 

            Consider the following conjecture:

IT hypothesis: For there to be something it is like for a creature C to have a state s (event, process, data structure, …), it is necessary and sufficient that s engenders for C (veridical or illusory) impressions of images or transparency.

I cannot properly defend this conjecture here; at a minimum, to do so would require an extended philosophical scouring for and wrangling about counterexamples. What I can do is to indicate why I do not think there are likely to be any clear counterexamples, and more positively, to indicate some of what I take the IT hypothesis to explain. 

            In part 1 I argued that the impressions are engendered in all of the following kinds of experiences, which I take to be the most clear cases of phenomenal experiences, of states it is like something to have:
   (i)      clearly conscious perceptual experiences, such as tastings and visual experiences;
   (ii)     clearly conscious bodily-sensational experiences, such as pain, tickle, and itch experiences;
   (iii)    clearly conscious imaginative experiences, such as those of one’s own actions or perceptions; and
   (iv)    clearly conscious thinking experiences, as in streams (or trains) of thought in words or in images.
Call these the “Qualitative Quartet.” The impression as of objective looks and feels is rife in normal perception, upgraded imagination, and diffuse bodily sensation. The impression as of subjective likenesses is most at home in normal imagination, degraded perception, thought, and nondiffuse bodily sensation. Although I proceeded by giving examples, the examples seem arbitrary: my claim is not simply that some members of each of the four categories display the impressions, but that all members of each of the categories do.  

            If true this result would be very startling given that the Quartet seems to be a hodge‑podge sampling of mental phenomena. Even ignoring the wide variations within each group, there are obvious differences among the groups—perceptual and bodily-sensational experiences largely impinge on us, while imaginative and thinking experiences are largely under our control; perceptual, imaginative, and thinking experiences clearly have representational content (we perceive, imagine, or think about things), while bodily-sensational experiences at least seem not to (we don’t seem to hurt or itch or tickle about anything); perceptual and imaginative experiences seem largely pictorial while thinking experiences seem largely linguistic and bodily-sensational experiences seem neither, and so on. Given this heterogeneity, it is extremely striking that the impressions seem common to all the Quartet states, especially given that the impressions can themselves be explained in a unified fashion. 

            Of course, all Quartet experiences are conscious in some sense, but consciousness is not what they have distinctively in common. There are other conscious mental states that, I believe, can be made out as clearly nonphenomenal—such as conscious moods and conscious propositional attitudes (beliefs, desires, emotions, etc.). While there is often something it is like when we have a conscious mood or attitude, I argue elsewhere (1996), on grounds independent of the present paper, that this is not due to the mood or attitude itself but due to coexisting Quartet experiences. Conscious moods and