92

The results were that for inverted stimuli, F-F, F-T, T-T and T-F were nearly the same, only 8% of the aftereffect was face-specific. Inverted faces are seen almost entirely via low level shape-general attributives. In the case of upright faces, 55% of the aftereffect was face-specific and 45% was low level. So high level face-specific attributives play a slightly larger role than low level attributives. Of course no single result can rule out that the result is due to differences between low level features of right-side-up and upsidedown faces (e.g. the downward rather than upward curve of the eyebrows). These results show us how evidence can bear on what attributives the visual system actually uses. But do these results concern a technical sense of “see?” Tim Williamson asked me in a presentation of this material whether in this sense of “see” one can see New College. It is useful to distinguish primary from secondary seeing. We can take primary seeing to be the application of a visual attributive to a “visual object”, i.e. an object that is itself picked out by a demonstrative element in a percept of the sort described in the 2nd paragraph of this article. Secondary seeing involves hybrids of visual attributives and concepts applied to objects of primary seeing and complexes of them in states that put together perception with perceptual judgment. Such hybrids can represent things that are not visual objects (i.e. not the referents of the demonstrative elements in perceptions) on the basis of visual objects that compose them. Thus the notion of seeing in which we see New College is secondary seeing and it can be reconstructed in terms of the sciences of perception and cognition. Perception vs Cognition The face in the middle in Figure 2 looks first angry then fearful. But can we be sure from introspection that those “looks” are really perceptual, as opposed to primarily the “cognitive phenomenology” of a conceptual overlay on perception, that is, partly or wholly a matter of a conscious episode of perceptual judgment rather than pure perception? Moving to the Susilo experiment, if a subject sees the adaptee as less elongated than the adaptor, how do we know whether this case of seeing-as is purely perceptual as opposed to partly or wholly a matter of perceptual judgment? (One might raise this worry separately for the adaptor and adaptee.) There is reason to think that concepts don’t adapt in the way percepts do. Ambiguous figure-ground stimuli (see Figure 1) engender oscillating percepts that change every few seconds. Both behavioral and neural evidence indicate that the figure representation suppresses the ground representation, with the strongest suppression directed towards the strongest competitors (Peterson and Skow 2008; Salvagio, Cacciamani et al. 2012). SEEING-AS IN THE LIGHT OF VISION SCIENCE 7The suppression is short-lived, putting a switch of figure and ground within the realm of possibility, but the switching itself is a result of adaptation (Alais, O’Shea et al. 2010). Such forms of multistable perception have three fundamental properties (Logothetis and Leopold 1999): exclusivity (the multistable percepts are not present simultaneously, although transitions can have elements of both), inevitability (one interpretation will eventually replace another) and randomness (the duration of one alternation is not a function of previous durations). Of course figure-ground alternation can be modulated somewhat via intentions, and there are ambiguous figures where cognition makes a difference to perception (Potter 1964), but these facts do not show conceptual adaptation. In displays like those in Figure 1, subjects obtained—as figure—the meaningful shapes like the white shape on the left faster and were able to intentionally maintain them longer only when the figures were roughly upright rather than rotated or inverted—and crucially, whether or not they knew what the shapes were (Peterson and Gibson 1994). The orientationdependence shows the adaptation is at least partly perceptual, and the independence of what the subjects know suggests the adaptation is not cognitive. So at least some kinds of figure-ground alternations are perceptual, not conceptual. Further, the point of perception is to register news, and adaptation contributes to that by filtering out old news; old news however, is important to cognition, so one would not expect adaptation in cognitive systems. Other things equal, if there were conceptual adaptation one would expect conceptual-without-perceptual adaptation—e.g. conceptually ambiguous situations would show alternations. For example, a morally ambiguous situation might lead to alternating interpretations that show the same three fundamental properties. No such phenomena have been reported to my knowledge. I conclude that there is evidence that the alternations observed for facial expressions and other aspects of faces (e.g. identity and gender) are perceptual in nature. And there are other types of evidence that can at least in principle be used in deciding the question in a particular case of whether adaptation is perceptual. If the locus of adaptation is in cortical visual areas, that would provide some evidence that it is perceptual, but even then there could be a cognitive element. For example, working memory, an at least partly cognitive capacity, makes use of neural coalitions involving perceptual representations that provide part of the contents of working memory representations (Curtis and D’Esposito 2003). I will describe evidence that in one case adaptation is wholly perceptual. The technique compares adaptation with priming, an effect that works in the opposite direction from adaptation. Adaptation to a feature makes that feature harder to see; priming makes it easier. For example, exposure of a tenth of a second to a stimulus moving to the left biases the viewer towards seeing an 8 NEDBLOCKambiguous stimulus (ambiguous between rightward and leftward motion) as moving leftwards, so long as the ambiguous stimulus is presented immediately afterward (Kanai and Verstraten 2005). Priming of this sort is thought to result from briefly spreading activation in the visual system. Caspar Schwiedrzik and Lucia Melloni investigated adaptation and priming simultaneously in perception of tilt. Subjects first saw oriented grids of the sort of the leftmost or rightmost grid in Figure 4 for 4 seconds, then indicated which direction of tilt they saw. Then they saw a grid with ambiguous tilt (like the middle one) and indicated its direction of tilt. The idea of the experimental design is that an attractive effect of perception of the first grid on perception of the second (ambiguous) grid indicates priming, whereas a repulsive effect indicates adaptation. The first result is unsurprising: the more 0o-ish the first grid was objectively, the more likely subjects were to classify it as 0o. See Figure 5A which maps the probability of reporting 0o on the first grid (vertical axis) against the objective tilt of the first grid (horizontal axis). Priming is exhibited in Figure 5B by the black line being above the gray line: subjects were more likely to classify the second (ambiguous) grid as 0o if they classified the first grid as 0o. Higher on the vertical axis indicates a greater probability of reporting 0o on the second grid and the black line reflects subjects for whom the first grid was reported as zero. This difference between the black line and the gray line shows attraction between the f irst report and the second report: priming. Most importantly and surprisingly: the more 90o-ish the first grid was objectively, the more likely subjects were to classify the second grid as 0o. This is an adaptation effect and is indicated by the upward slope of both lines in Figure 5B as compared with the downward slope of the line in Figure 5A. Crucially this effect held both for those who classified the first grid as 0o and those who classified the first grid as 90o. The black and gray lines are parallel, so the adaptation effect worked in the same way independently of what Figure 4. The leftmost grid is more likely to be seen as oriented to the northeast (stipulated to be 0o) than the northwest whereas the rightmost grid is northwest (90o), and the middle grid is multi-stable as between the two. This f igure is similar to part of Figure 1 in (Schwiedrzik, Ruff et al. 2013). With permission from Oxford University Press, publishers of Cerebral Cortex. SEEING-AS IN THE LIGHT OF VISION SCIENCE 9high probability of reporting 0° probability of reporting 0° or 90° on the 1st stimulus high probability of reporting 90° high probability of reporting 0° probability of reporting 0° or 90° on the 2nd stimulus high probability of reporting 90° objective tilt 0° 1st report was 0° 1st report was 90° no objective tilt objective tilt 90° objective tilt of the 1st stimulus was 0° no objective tilt of the 1st stimulus objective tilt of the 1st stimulus was 90° Figure 5. See text. This figure is similar to Figure 2 in (Schwiedrzik, Ruff et al. 2013). I am grateful to Caspar Schwiedrzik for adapting this figure to the needs of this article. With permission from Oxford University Press. the subjects thought they saw: what was important for adaptation was not how the subjects took themselves to have seen the first grid but the actual objective tilt of the first grid as registered in the visual system. Since adaptation does not depend on what the perceiver judges, it is hard to see how it could be a cognitive effect. The upshot for our discussion of 10 NEDBLOCKface adaptation is that at least one kind of adaptation is probably perceptual and not cognitive. Schwiedrzik and Melloni also scanned the subjects, finding that priming engaged a wide range of brain areas in the front and middle of the brain whereas adaptation engaged only early visual areas (V2 and V3) in the back of the brain. What is significant here—given the earlier remark about working memory—is that the adaptation does not involve the frontal and parietal areas that are thought to underlie concepts and cognition. They conclude that “adaptation is a purely local phenomenon…” What I like about this experiment is that there is a mesh between the psychological and neural evidence and both suggest that—at least for the admittedly simple stimuli used in this experiment—adaptation is perceptual rather than cognitive. Thus we can distinguish empirically among (1) a high level perceptual attributive, (2) the holistic constellation of color, shape and texture attributives that are recognitionally coextensive with that attributive and (3) the conceptualized attributives that play a role in conscious episodes of thought and judgment rather than perception. Direct realists often claim, on a priori grounds, that representations in perception are all sub-personal (Travis 2004, p. 59). This view is a mistaken product of a deeply wrongheaded armchair methodology. I have described—unfortunately only briefly–evidence that we have visual representations of fearfulness of faces and similar experiments show visual representations of other emotions and of gender and race. These representations are perceptual and also play a direct role in relation to personal level interests and values. That makes them personal, not sub-personal. Painting with an infamously broad brush, Wittgenstein (1958, p. 232) complains of the “confusion and barrenness of psychology” saying “…in psychology there are experimental methods and conceptual confusion.” (As with many of Wittgenstein’s less wise remarks, Wittgensteinians (Hacker 2012) have tried to show that this quotation means something other than what it says.) However, the real confusion lies with philosophers who try to settle the nature of seeing-as in ignorance of the sophisticated conceptual apparatus that has been developed in psychology to approach such issues. Is this Philosophy? This article reports the results of a number of psychological experiments and so it may occur to philosophers reading it to wonder whether this article—and the chapters of Burge’s book that appeal to facts about perception—are philosophy rather than psychology. Both are both! Philosophy of mind is a mix of empirical and relatively a priori considerations, both of which can be seen both as philosophy and as theoretical psychology. However, it would be a mistake to think that those who know nothing of the science of the mind can SEEING-AS IN THE LIGHT OF VISION SCIENCE 11just stick to the relatively a priori parts of philosophy of mind, since one needs to understand the empirical facts to even know where there is room for relatively a priori philosophy. Burge’s book is masterful in matching a priori and empirical methods to the issues that are appropriate for them.2