102

That was the spatial point. The temporal point is that attentional resources available to the supposed spotlight are to some extent shared among other aspects of perception, for example, other modalities, and with executive control mechanisms (Brand-D’Abrescia & Lavie, 2008) and cognition. Nilli Lavie and her colleagues have demonstrated many kinds of cases in which demanding cognitive or perceptual tasks siphon attention away from other perceptual tasks. One experiment (Cartwright-Finch & Lavie, 2007) contrasted a difficult perceptual task (Which arm of a cross is longer?) with an easy perceptual task (Which arm of a cross is green?) In the harder task, subjects were much less likely to consciously see an unexpected stimulus—so called “inattentional blindness.” To take an example more connected to everyday life, talking on a cell phone can foment “inattentional blindness” (Scholl, Noles, Pasheva, & Sussman, 2003; Strayer, Drews, & Johnston, 2003). It is as if electricity for the spotlight shares limited power with the air conditioning and public address systems, causing brownout for one if the other is making more demands. You may object “OK, so there are borderline cases and the borderline is more uneven and larger and more varying in time than one might have thought. So what?” This response seriously underestimates the problem for the idea that there is any distribution or level of attention that entails either veridicality in normal circumstances or illusion. The problem for this view is that there is no way to pick which distribution of attentional resources engenders veridical perception and which engenders illusion. For concreteness, let us consider the claim that when one attends to the fixation point, one’s percept of the 22% patch is illusory, but when one attends to the patch itself, one’s percept of it is (normally) veridical. Is it the absolute value of attention that is supposed to be crucial to attention making for veridicality? It can’t be absolute value since that would have the consequence that if the subject were speaking on a cell phone, nothing would be veridically perceived! So the claim must be based on relative allocation of attention. But the points about the landscape of attention show that many points in the visual field will be very close in amount of attention allocated to them as the supposed focus of attention. How could these very small differences be—in46 NedBlock Figure 7. Every point on the dashed line is equidistant from the two Gabor patches. normal circumstances—the difference between veridicality and illusion? An arbitrary cutoff—which some have proposed for baldness—would make no sense for veridicality. To avoid misunderstanding: I agree that the terms ‘veridical’ and ‘illusory’ are as vague and context relative as many other terms. And I also agree that we can use the phrase ‘attend to’ in an acceptable but loose and context-relative way. My point is rather that the claim that only the attended patch is seen veridically requires a non-context-relative kind of veridicality that cannot be justified. In my view, these points counts decisively against any illusion proposal, but I will mention one other point briefly. Consider the proposal that the supposed attentional spotlight engenders illusion. One motivation for this idea is that areas that are differentially attended are sometimes seen in an illusory manner as in the Tse Illusion mentioned earlier. Also, veridical comparison depends on the supposed spotlight of attention being in between the items compared, not directly on them. However, this advantage would apply to any point equidistant between the items being compared: in the example of the two patches, attending to any point on the vertical line through the fixation point in Figure 4, as shown in Figure 7. Every point on that line is equidistant between the 22% and the 28% patch and within the zone in which one can still see both (if the diagram is the right distance away from the eyes), so every point on that line is one such that attending to it might be predicted to yield a veridical comparison. Yet these different percepts of the 22% patch itself—the ones in which one is attending to different points on the line—would be different phenomenologically from one another, and so the refutation would be regenerated. Attending to the different points on that line would yield different phenomenal experiences of the 22% patch that are equally veridical. And that is the very situation that led to the problem for views opposed to mental paint in the first place.Attention and Mental Paint 47 If one wants to compare two patches, one does best for the supposed spotlight of attention to be aimed at a point equidistant from the two of them (unless someone has a handy map of one’s landscape of attention), but one cannot conclude that only then do the patches look the way they actually are, since the way they look will depend on which of the points that are equidistant from them one chooses to attend to. The vertical line through the square dot in the middle may contain a myriad of such points. Given the landscape of attention, there may be points on that line or elsewhere that yield correct comparisons but are phenomenally different from one another. The facts about the visual system discussed here are not oddities but rather are closely related to the evolutionary purpose of perceptual systems, which is to get information about the world that is relevant to the organisms’ other purposes, which of course include acting (Churchland, Ramachandran, & Sejnowski, 1994). Attention is attracted by movement or change in the environment. This “exogenous” attention is transient and involuntary and happens much faster than the eyes can move, so the experimental situation described earlier in which attention is directed somewhat differently from fixation is a common occurrence. The evolutionary point of the increased acuity and contrast at the attended location is to get more information about what is at that location. Because the effect of increasing acuity at one point inevitably reduces acuity at another (Montagna, et. al. say: “There was no benefit without a cost”), there is no way of making all perceptual comparisons accurate at once. Still, as Carrasco (2006) shows, the increased contrast sensitivity at the locus of attention improves performance on a variety of tasks. (See also Treue, 2004.) These points are directly relevant to disjunctivism, as can be brought out by comparing the attended 22% patch with the seen but unattended 28% patch. There is at least one well defined mental property in common to these two percepts, perceived contrast. Neither of these individual percepts (as opposed to the comparative percepts) are illusory but that same perceived contrast will be shared by various kinds of illusions as well. (See Burge’s (2005) Proximality Principle.) The disjunctivists who claim “no common mental factor” are flouting what perceptual science tells us about the case. It might be thought that the way to decide whether attended (or alternatively unattended) items are seen illusorily would be to consult behavioral tests of veridical perception. Since the heady and methodologically suspect days of the New Look in perception (Bruner & Goodman, 1947), there have been studies that have been taken to suggest that desired objects look closer. Recently, Emily Balcetis and David Dunning (2010) have used a variety of behavioral tests to argue for this conclusion. Most interestingly, when subjects toss a bean bag at objects that differ in subjects’ conative attitudes towards them they tend to come up shorter for objects to which they have a positive conative attitude.48 NedBlock Attended objects look larger, so one might speculate as follows: (1) desired objects are more likely to be attended; (2) They look larger and therefor closer; so (3) the phenomenon is grounded in illusory perception of attended objects. If so, the argument I gave still stands, since, you will recall, even if the attended object is seen illusorily, the point is reinstated by comparing one unattended object with another. (See the discussion connected with Figure 7.) However, I think more would have to be done to show that the “bean bag” test taps a genuine conscious perceptual phenomenon. In the case of the Carrasco phenomenon, I indicated some of the kinds of evidence that show it is a genuine perceptual effect. (For example, it affects perceptual adaptation.) The behavioral measures just mentioned do not reach that kind of standard of evidence and there have been some dramatic cases recently in which such effects appear to reflect “judgmental biases that result from the social, not physical, demands of the experimental context” (Durgin et al., 2009, p. 964). Finally, to conclude this part of the discussion of illusion, I described a number of the alternatives just discussed as unmotivated (for example the idea of drawing an aribitrary cutoff of illusion, a border between illusion and veridicality). These proposals are unmotivated—unless one of your motivations is saving direct realism! I will be satisfied if my argument convinces a neutral reader. 8. The Direct Realist Account of Illusion The points made in the last section count against an illusion in the attentional phenomena I have been talking about independently of any particular theory of illusion. But the point is strengthened when one sees just how implausible any direct realist theory of illusion would have to be. I find direct realists evasive on the topic of illusion, but Bill Brewer (2004, 2008) has advanced a straightforward positive account of illusion from a direct realist perspective. Brewer’s view (2008, p. 172) is that all perception, including illusory perception is veridical perception of something. According to Brewer, in illusory perception, the something is a similarity to a situation which is not the one the subject is viewing. “Illusions are simply cases in which the direct object of experience has such visually relevant similarities with paradigms of a kind of which it is not in fact an instance.” Consider the M¨uller-Lyer illusion: two lines (one with arrowhead endings, the other with reverse arrowheads) are actually equal in length but look unequal. Brewer’s account is that the subject is directly aware of an actually instantiated property, the similarity between the equal lines and other lines that are not present but are actually unequal. As Brewer puts it (p. 176): “Suppose that someone has the diagram visually presented to her, from head-on, and in good lighting conditions, with eyes open and aAttention and Mental Paint 49 normally functioning visual system. According to OV [Brewer’s version of direct realism the “object view”], the core subjective character of her visual experience is simply constituted by that diagram itself. From that viewpoint, and given the circumstances of perception, it has visually relevant similarities with a paradigm pair of unequal lines at different depths. In this sense, the concept of inequality in length is intelligibly applicable to its main lines: the lines look unequal in length.” But what is that visually relevant similarity or similarities that a pair of equal lines has to a pair of unequal lines? Of course the direct realist cannot say it is that they look the same, since that would be to reintroduce unreduced phenomenology. It might help to see just how empty Brewer’s proposal is to consider Wittgenstein’s (Rhees, 1968; Wittgenstein, 1993) supposition—far short of an inverted spectrum—that someone might wake up and find that red things look blue and vice versa.16 The person Wittgenstein is imagining is having an illusory experience: red things look blue. On Brewer’s account, he is seeing the visually relevant similarity or similarities between red and (paradigms of) blue. What are those similarities? Red and blue are both colors, they are both primary colors, and they are both colors of many flags and of Superman’s costume, but none of these similarities will help Brewer’s case. It is very difficult to see how to spell out that similarity between red and blue without appealing to how things look or to a representational content involving a representation of blue. In my view, direct realism is an unstable position for which there is no good account of illusion, whose motivations better support representationism, a subject to which I now return. 9. Representationism and Vagueness As you will recall, representationism holds that the phenomenal character of a perceptual state is or is determined by its representational content. Representationism has an obvious advantage over direct realism: it has a better account of illusion. If a man has the phenomenology of seeing his wife as a hat (Sacks, 1985), he sees his wife, visually representing her as a hat. Direct realists see representationism as scanting “openness to the world”, but as I mentioned I am doubtful about such theoretical introspections. Representationism has another major advantage over direct realism, an account of the difference between conscious and unconscious perceptual representation that gets to first base. The very “openness to the world” that direct realists like in an account of conscious perceptual representation makes it difficult to see what their account of unconscious perception could be, as noted earlier, whereas for representationism, the difference lies in some aspect of function. My argument against direct realism was based on the fact that the way the 22% patch looks depends on where the subject is attending, so there can50 NedBlock be more than one phenomenal character of experience of the same instantiated properties, even if nothing about the environment or the non-mental relations between the subject and the environment differs. An advocate of representationism might suppose that even if my argument vanquishes direct realism, representationism escapes unharmed, since the representationist can simply say that different distributions of attention to the same layout yield different representational contents: if an attended item looks bigger, faster, higher in contrast, higher in saturation, more stripy or earlier, the representationist can, it may seem, handle that fact simply by adjusting the representational contents. The great strength of representationism after all is its ability to handle appearances in representational terms. A similar point might be made on behalf of Colin McGinn’s (1999) and Mark Johnston’s (2004) views of the phenomenology of perception, views that combine elements of both representationism and direct realism. According to McGinn, the phenomenology of perception is determined by a “cluster of properties” ascribed by the percept. The same role is given to Johnston’s “sensible profiles,” which are structured properties. These views are more like representationism than direct realism in one respect: McGinn and Johnston do not require that the cluster of properties/sensible profiles be actually instantiated. The phenomenal character of the visual experience of the man who mistook his wife for a hat could be captured by uninstantiated properties/profiles involving a hat shape, color, texture, etc. And a hallucinator and a veridical perceiver could be aware of the same cluster of properties/sensible profiles, uninstantiated in one case, instantiated in the other, and thus have identical experiences. So McGinn & Johnston could give a response similar to the representationist response just mentioned: different distributions of attention can involve the same object, the 22% patch, but awareness of different clusters of properties/sensible profiles, and these different items could be said to explain the different phenomenologies.17 However, if I am right that none of the different percepts of the 22% patch are illusory, these proposals fall flat. Representationists are not free to postulate representational contents at will so as to reflect appearances— rather these contents have to be grounded in veridical perception. If the representationist says that changing the distribution of attention changes the representational contents (cluster of properties/sensible profiles) without changing or selecting any different property of the actual layout, the upshot is that at least one of those representational contents is illusory, and if my argument against illusion is right, that claim is wrong. In short, if there is no illusion in either of the two percepts of the 22% patch, the extra degree of freedom buys the representationist nothing. However, the representationist has a resource in addition to content: there is also the mode of the representation (Crane, 2007). A belief and a desire can have the same content—e.g. that world peace obtains, but withAttention and Mental Paint 51 different relations to that content. Similarly there may be different modes of representation in perception. One might, for example, hold that vision and audition are such different modes. As mentioned in a slightly different form with respect to direct realism, the mode of representation could be more or less attentive, more or less determinate, more or less specific, more or less vivid, more or less salient. What is the difference between indeterminacy in mode and indeterminacy in content? We can get some illumination by considering a recent dispute. An objection to representationism that has been much discussed (Block, 1996, 2003; Boghossian & Velleman, 1989; Peacocke, 1993; Tye, 1995; Tye, 2002) is that there is a phenomenal difference between blurry vision as of something, say a movie, whose lines may be clear and crisp, and (the contrasting case) a clear and crisp visual experience of a blurry movie— even when the same colors, shapes and textures are represented with the same degrees of determinacy. The first could be produced by taking off one’s glasses in a normal movie theater, the second by defocusing the projector while the viewers keeps their glasses on. Tye (2002) regards the difference as one between experience indeterminately representing (glasses are off) and having an indeterminate content (glasses are on but the movie is blurred). One could argue about whether these categories are exactly right, but they will do for the analogy. The glasses off case (indeterminately representing) is a matter of indeterminacy in mode, whereas the glasses on case is a matter of indeterminacy in content. The point of the analogy is that in both cases, indeterminacy of mode and indeterminacy of content, there is the same attribution to points of space of light and dark and color. If we were to capture these two different percepts in terms of digital pixel arrays, the pixel arrays appropriate to the two cases could be exactly the same. The phenomenal difference consists in whether the blur captured in the pixels is attributed to the display or to the mode of perception. Now consider a specific proposal for accommodating the experimental results in terms of indeterminacy. One option would be to regard the contents that differ but are equally veridical as having different ranges of indeterminacy that include 22%. So in the case of attending to the fixation point (in which case you will recall the 22% patch looks less contrasty than the 28% patch), the content of the percept of the 22% patch might be, say, 16%–28% contrast, whereas in the case of attending to the 22% patch itself (in which case you will recall, it looks equal to the 28% patch), the content might be a higher range in which 22% is at one end, say, 22%–34% contrast. (Specific numbers are useful for concreteness. Since attention shifts perceived contrast at this level by about 6%, I am imagining that there is an “uncertainty” of plus or minus 6%.) Thus both representational contents would be veridical—since they both attribute a range that includes the true contrast, 22%—and they would also be different. The idea here is that the change of representational content imposed by a change of attention is the equivalent of seeing different52 NedBlock indeterminate contrast levels. In the case of McGinn’s & Johnston’s versions, the clusters/profiles would include 16%–28% contrast in one case, and 22%34% in the other.18 The problem with this proposal is that it if the phenomenology of perception flows from representational content, then indeterminacy in content would have to be reflected in an indeterminacy of look. But there need be no such indeterminacy. The Gabor patches used in the contrast experiments are designed to transition between light and dark in a smooth way, but this is not necessary for attentional effects and is not true of many other stimuli. For example, some size stimuli are shaped like the letter ‘C’, where the relevant parameter is the distance from tip to tip. These are not in any reasonable sense fuzzy stimuli, but the subject sees the gap size as bigger if attended. Attention makes attended items bigger, faster, more contrasty, more saturated and earlier, as I have repeatedly noted. These effects do not depend on—and are not reflections of—any sort of fuzziness. The representationist may retort that the point is not that the contents are fuzzy or represented indeterminately but that they are abstract relative to other contents, as determinables are to determinates, for example as red is to scarlet. But this line of thought runs into the following difficulty: the variation of 6%duetoattention is way above the “just noticeable difference” threshold, which for stimuli at these levels is approximately 2%. (Or so I am told. In any case, just looking at the stimuli in Figure 4 shows that the difference is easily detectable. And you may recall that in the discussion of the tilt aftereffect, there was evidence that at higher levels of contrast, the increase due to attention was as much as 14%.) The point is that there is no single “look” that something has if it is 22% plus or minus 6% in contrast. By analogy, consider the supposition that something looks as follows: rectangular or triangular or circular. That disjunctive predicate does not describe one way that something can look—at least not in normal perceptual circumstances. Now I have been considering an indeterminacy in content rather than mode. But the same point applies to both! Consider the cinema case just flagged. The mode version of the case (glasses off) and the content version (glasses on, but projector defocused) involve the same level of “fuzziness”. Versions of all these points apply equally to direct realism. The direct realist can suppose that we are directly aware of a property of objects that is constituted by a range of contrasts, a higher range when one attends than when one does not attend. My objection to direct realism is that direct awareness of an indeterminate property or indeterminate direct awareness of a non-indeterminate property would have to be realized in perception either in the form of a phenomenal indeterminacy or in the form of a disjunctive awareness that does not correspond to any particular look. To avoid misunderstanding: I like the vagueness proposal as a theory of the representational content of perception. What I insist on is that this representational content does not capture the phenomenology of perception.Attention and Mental Paint 53 I am happy with the vagueness proposal so long as the phenomenology is treated as a mode of presentation of that representational content. Thus the percept of the 22% patch when one attends to it could be said to have a representational content a major component of which is something like this: . (Burge (2010) argues persuasively that the content of perception should be understood demonstratively rather than propositionally.) What I am arguing against is that this content is determinative of the phenomenology of the perception. As I mentioned at the outset of the paper, in previous publications I have argued that the phenomenology of perception goes beyond its representational content, but here I am arguing for something a bit different, that the representational content of perception cannot determine the phenomenology of perception. Of course representationists often emphasize that their view is that one kind of content of perceptual states (e.g. non-conceptual content) is determinative of phenomenology even if there are other types of content that are not so determinative. So the argument could be regarded as a challenge to representationists to put up or shut up: what kind of content of perception could determine the phenomenology of perception yet not have the problem just described?19 Onmystory, one could describe the perceptual content of the percept of the 22% patch when one is attending to it this way: {Phenomenal component P; Referential component: }. That is, the percept has both a phenomenal mode of presentation (Burge, 2003; Chalmers, 2004) and a representational content. (Elsewhere (2003, 2007c), I claim to have shown that the phenomenal mode does not supervene on the representational content or conversely). I said that the 22% patch attended (i.e. with more attention) is the same in perceived contrast as the 28% patch unattended (i.e. with less attention). And on the proposal that I have just been considering, both would be captured by {Phenomenal component P; Referential component: } where P is the same for both percepts. “Wait—how could that be? How could the percept of the 22% patch attended be exactly the same as the 28% patch unattended? Since the two patches are in fact different, at least one of those perceptions must be illusory!” This conclusion does not follow. One reason is the one already mentioned: that the actual contrast of both patches lie within the specified range. One is 22% and the other is 28%, both of which are in the range of 22%–34%. But there is another factor that has not come up yet in my discussion. The change invoked by changing attention does not look like a change in the world—at least not to me. Take a look at Figure 4, fixating at the fixation point and moving your attention around. It does not look as if anything is really changing in contrast. That is, these changes don’t have what Burge (2009) calls the phenomenology of objectivity. The change looks unreal (to me). Its unreality is similar to the unreality in the way an afterimage grows54 NedBlock and shrinks as the surface you project it on moves further away or closer. If you look at a colored shape for two minutes and then look at a piece of paper, you see an afterimage as of the sameshapeinthecomplementarycolor. As you movethe paper further away, the afterimage grows. (See any textbook on Emmert’s Law, which describes this change of size.) But its growth looks somehow unreal or unobjective (cf. Masrour, 2010). The subjective unreality of these changes has not as far as I know received any empirical investigation. Thesubjective unreality of the effect of attention was noted by William James. He said (1890, p. 426) “…whatever changes the feeling of attention may bring we charge, as it were, to the attentions’ account, and still perceive and conceive the object as the same…The intensification which may be brought about by attention seems never to lead us astray” (p. 426). James quotes Gustav Fechner (1882) (in English) saying, similarly that when one increases attention, one “feels the increase as that of his own conscious activity turned upon the thing.”20 The upshot is that it is a mistake to treat the change in phenomenology wrought by the change in attention as equivalent in its effect on phenomenology of a change in contrast in the world. If I am right that the distribution of attention does not produce an illusory percept of any single patch, whether attention is focused on that patch, on the fixation point, or on the other patch, then the visual system must in some way track where attention is focused and this information must be in some way reflected in the phenomenology of perception. Intuitively, it seems that this information is realized visually in the phenomena described, although it could be realized phenomenally in some other way. (Subjects can be wrong about which modality is the source of perceptual information.) If it is visual, then there is something phenomenally different between the way the attended 22% patch looks and the way the unattended 28% patch looks, even if they are the same in perceived contrast. Let us think of the attentional tracking as a little voice that says where attention is directed and also that the range of contrasts attributed to the world should be adjusted to reflect reality because of that distribution of attention. Then we could think of the percept of the 22% patch attended as {Phenomenal component P; Referential component: , Little Voice component: attended, so adjust downward} whereas the percept of the 28% patch unattended might be {Phenomenal component P; Referential component: ,Little Voice component: unattended, so adjust upward}. Further research might lead to evidence that the little voice should be partly absorbed into the phenomenal component or the representational component or both. Manyexperiments on perception in effect encourage the subject to judge appearance. If subjects are asked to judge size in afterimages at different distances, apparent size will be reflected in the judgments. A different kind of experimental paradigm—say betting—might encourage a focus on reality instead of appearance.Attention and Mental Paint 55 At the beginning of the paper I mentioned an apparent paradox: two percepts can differ in respect of perceived contrast, size, etc., yet neither be illusory. I mentioned that although loudness presents sound intensity, loudness is a function not only of intensity but also frequency, bandwidth and duration. Two sounds of different intensities can sound the same in loudness and two sounds of the same intensity can sound different in loudness. One might wonder why there isn’t a dimension of perceived size (or contrast or speed or saturation or earliness) that bears the same relation to actual size as loudness bears to actual intensity? Part of the answer is that it would not be very fruitful to project such an evanescent aspect of experience onto the world. That is a functional explanation but there is a more proximate explanation: the auditory system is built to register loudness and that is revealed in the fact that there is loudness constancy. A sound maintains its loudness as you move closer to the source and as its intensity increases. But the phenomena I have been describing involving attention have a kind of anti-constancy in that the perceived size changes even when the actual size is constant. A similar point applies to the behavior of afterimages as indexed in Emmert’s Law. I would be surprised if this anti-constancy feature were not related to the subject’s sense that these changes are not objective, and this idea fits with Burge’s (2009) view of perceptual objectivity as grounded in constancies. 10. Mental Paint I have argued elsewhere that the inverted spectrum (and other related thought experiments) support mental paint. (These arguments are presented in (1990, 1999, 2003) and in a version I am much happier with, in (2007c).) The view of mental paint argued for here is in one respect much less radical than the one argued for in those papers. If things we both call ‘red’ (and think of as red) look to you the way things we both call ‘green’ (and think of as green) look to me, then we have phenomenal qualities that represent ripe tomatoes as red, but do so very differently: your mental paint represents differently than mine. Further, we can have experiences that are the same in respect of mental paint, but represent differently, for example when I am looking at a red thing and you are looking at a green thing. The relation between mental paint and representational content is many-many. However, the argument presented here does not allege any extensional gap betweenmentalpaint andrepresentational content. I have not exhibited a case of same representational content/different phenomenology or same phenomenology/different representational content. I did present the argument as roughly having that form at the outset, but the final view just presented is that vague representational contents may be right. The focus here has56 NedBlock been on a different kind of gap between mental paint and representational content: that to cope with attention, perceptual representation must be vague in content or mode in a way that the phenomenology of perception need not be. Ironically, what sinks direct realism and representationism is the very aspect of perception that advocates of these views often take as their founding insight, the positive thesis of Moorean diaphanousness. Shifting attention makes items in the world look, non-illusorily, different in contrast, size, saturation, stripiness, speed, and time of occurrence. In order to cope with that fact, anti-mental paint views are forced to postulate kinds of vagueness that are not reflected in ways the world looks to be.