71

With inhibition of return as with object- based attention, there is a gradient of effects within an object, with the strongest effect at the cued location within the object, and weaker effects in the same object but further away from the cue (Klein and Ivanoff, 2008). This shows integration of object- based effects with spatial effects, again providing evidence against the view that there is a difference in kind between object perception (allegedly discursive) and other perception (iconic). Again, object perception is integrated seamlessly with spatial attention, something that would call for explanation if object representations and spatial representations involved representations of different formats as the pluralists claim. A different kind of support for iconicity in object- perception involves visuospatial neglect, a syndrome in which subjects fail to attend to objects on one side of the body. The point of discussing visuospatial neglect is that it reveals that object representations, spatial representations, temporal representations, and numerical representations are tightly coupled in overlapping systems, counting against the claim that object representations have a different kind of format from other perceptual representations. Of course representations of different formats can be linked, but the question is whether postulating such linkages has independent support. In one kind of visuospatial neglect, subjects ignore or fail to consciously see the left side of the visual field. Patients fail to eat the food on the neglected side of their plates, fail to dress the neglected side of the body, and so on. This kind of neglect is based in one form of egocentric perception. When left- sided neglect patients are asked to bisect a horizontal line, they put the bisecting mark to the right of the midpoint. Interestingly, some left- sided neglect patients show the same effect for imagined lines. The size of the rightward drift in bisecting lines is proportional to the length of the line, i.e., larger displacements with larger lines. This dependence on degrees is indicative of iconic representation. These effects are all matters of degree and interact with many spatial visual features, again providing problems for pluralism. Interestingly, for very short lines, there is an effect in the opposite direction, the “crossover effect” (Zorzi et al., 2012), and this crossover effect also appears in other perceptual activities that use the same system. Neglect often involves inattention to, and perhaps lack of perception of, one side of space. But it often applies to one side of individual objects that have salient axes, showing again that the control of attention depends on spatial aspects of perception, showing integration of object representations with spatial representations. That integration would require explanation if the two kinds of representations were of different formats, one iconic, one discursive. Many patients neglect the left sides of objects all over the visual field, showing some influence of allocentric spatial representation (Beschin et al., 1997; Tipper and Behrmann, 1996). And patients often neglect the initial letter or segment of a word, even if the word is presented vertically; neglect the Western Hemisphere even in an upside- down map; or neglect the left side of a face even in an upside- down photograph (Bisiach and Luzzatti, 1978; Caramazza and Hillis, 1990). Again, these effects are matters of degree and interact with spatial features. A classic demonstration of object- based neglect involved barbells, two circles connected by a line. Neglect patients had trouble with detecting targets on the left circle, but when the barbell was rotated so that the left circle had moved to the right, many patients showed flipped results, with more trouble on the right circle. This effect was only observed if the barbell was a single object: if the line between the circles was omitted, 500 0005464032.INDD 500 Ned Block 09-10-2022 11:32:08there was no such effect (Tipper and Behrmann, 1996). Again we see seamless integration of object perception with spatial perception, a surprising result if the two have entirely different formats. Another classic demonstration of object effects in neglect is shown in Figure 27.5. A patient who was asked to copy a picture left out the left side some of the individual objects in the picture. See Walker, 1995 for other examples. Once again we see that the perception of objects is part of a spatial representation system, something that would be in need of explanation if object representations were discursive but spatial representations were iconic. As just mentioned, neglect also extends to “numerical space.” Left- sided neglect patients asked what number is halfway between 2 and 6 skew their answers toward 6. Strikingly, the crossover effect just mentioned also applies to numerical space (Zorzi et al., 2012). Similar results apply to temporal estimation problems for neglect patients (Bonato et al., 2016). A further crossover effect is that normal subjects who were asked to estimate additions or subtractions of dots showed a leftward bias on the number line (that is, they underestimated) for small numbers of dots, but a rightward bias for large numbers (overestimating) (Zorzi et al., 2012). The explanation usually given for this kind of result is that spatial representation is co- opted for the numerosity system. There is a great deal of evidence for this. For example, if subjects are given the task of pressing one button if the number referred to by a presented digit is bigger than 5 and another button if it is less than 5, subjects are faster for 7 than for 6, faster still for 8, and faster still for 9, with similar results obtaining for digits on the other side of 5 (Dehaene, 2011). These results show that object representations, spatial representations, temporal Patient asked to copy this Patient drew this FIGURE 27.5 A VISUO- SPATIAL NEGLECT PATIENT WAS ASKED TO COPY THE TOP PICTURE. THE BOTTOM PICTURE IS THE PATIENT’S ATTEMPT. NOTE THAT THE PATIENT LEAVES OUT SOME OF THE LEFT SIDE OF SOME OF THE INDIVIDUAL OBJECTS IN THE PICTURE. THANKS TO JAMES DANCKERT FOR THIS PICTURE. How Should We Understand the Distinction 0005464032.INDD 501 501 09-10-2022 11:32:09o o o o o o o t=1 o o o o o o o o t=2 o o o o o o o o o t=3 o o o o o o o o t=4 FIGURE 27.6 SEQUENCE OF EVENTS IN A BASIC MULTIPLE OBJECT TRACKING EXPERIMENT. EIGHT DISKS APPEAR ON THE SCREEN; FOUR ARE INDICATED AS THE ONES THE SUBJECT IS SUPPOSED TO TRACK; THE INDICATIONS OF THE DISAPPEAR; AND THE DISKS MOVE RANDOMLY. AT THE END THE SUBJECT IS ASKED TO IDENTIFY THE FOUR THAT WERE FIRST CUED. MOST PEOPLE CAN TRACK FOUR OBJECTS FAIRLY RELIABLY. FROM PYLYSHYN (2007) UNDER A CREATIVE COMMONS LICENSE. representations, and numerical representations are tightly coupled in overlapping systems, putting a burden on those who think that object representations have a different kind of format from other perceptual representations. In the next section, I will be urging caution about drawing conclusions about perceptual representations from evidence about the remnants of perceptual representations in working memory. To the extent that inhibition of return involves working memory, that caution applies to the results just mentioned. The only way in which working memory would add perceptual features rather than subtract them is via the imposition of perceptual imagery, but it would have to be shown that imagery is involved in the experiments I am talking about. Finally, in the multiple object tracking paradigm, not only can subjects track about four disks but also they can track the average position of the disks. See Figure 27.6. A number of disks (eight in the figure) are shown on a screen. Four of the disks blink or are otherwise indicated, and then the disks move randomly. The subject is supposed to track those objects as they move about in a random way. Most subjects can track about four objects if they do not move too quickly. Subjects turned out to be able to track the centroid of the target disks, but they could also track the centroid of the distractor disks. And they could do this even if their attention was drawn off by a difficult secondary task of counting the number of times the disks crossed some lines on the screen (Alvarez and Oliva, 2008). This experiment suggests that perceptual object representations integrate with spatial representations even though working memory is also involved in multiple object tracking. In sum, there is plenty of evidence that perceptual object representations are of a piece with other sorts of spatial perception, putting pressure on the view that object representations are different in format from other sorts of perception. 2. Object Files of Working Memory and Thought The first section of this article argued that perceptual object representations (the so- called object files of perception) are iconic. I now turn to the other main claim of this article, that the so- called object files of singular thought and working memory are 502 0005464032.INDD 502 Ned Block 09-10-2022 11:32:10fundamentally different from those of perception. Indeed, there is reason for doubt that the object files of perception can even ground singular thought. I will be focusing on working memory at the expense of singular thought. A singular thought can be based on a simultaneous perception, but singular thought can also be based on working memory, and it is the latter case that is relevant to the discussion to follow. Since we are going to be talking about working memory representations, we should first understand what they are, and that requires contrasting them with other short- term forms of perceptual memory: iconic memory and fragile visual short term memory. I now turn to that contrast.