76

Finessing the Bored Monkey Problem Ned Block1,* By recording from microelectrodes in monkey prefrontal cortex (PFC), researchers have decoded the contents of conscious perception in cognitive areas (lateral prefrontal cortex) in conditions in which perceptions are not determined by the stimulus, binocular rivalry, and flash suppression [1–4]. As I noted in my recent Trends in Cognitive Sciences article [5], such results cannot be taken to support cognitive theories of consciousness because of the ‘bored monkey problem’: the idea that subjects whose only task is f ixating a dot may have thoughts about the noticeably different stimuli, causing prefrontal differences that do not reflect prefrontal consciousness. This was the negative point of my article, and in their commentary Phillips and Morales (P&M) [6] do not dispute it. What they do dispute is my positive point: that Brascamp et al. [7] have evaded the bored monkey problem. One innovation introduced by Brascamp et al. is to use stimuli that have two related useful properties. The first is that the stimuli do not afford any ready-to-hand cognitive categories for characterizing them other than as moving dots. Subjects cannot say to themselves: ‘There is the face again.’ P&M say ‘Nothing in Brascamp et al.’s methodologypreventsobserversengaging in extensive cognitive processing’, both in the rivalry transitions and the similar real (objective) transitions. However, P&M are neglecting the fact that the stimuli do not naturally draw cognitive processing in either the rivalry case or the real case. P&M focus on the distinguishability of the rivalrous transitions from the nonrivalrous real (objective) transitions, emphasizing replay subtraction. However,theaforementioned monkey experiments [1–4] do not use any form of replay subtraction. This research does involve comparisons between perception of the rivalrous stimuli and perception of real stimuli, but the purpose is to ascertain which neurons respondto thepercept rather than to the stimulus. The second useful property of these stimuli is that they are subjectively different from each other without being conspicuously different. P&M [6] say: ‘Yet indiscriminable stimuli look the same.’ However, although the stimuli are not noticeably different, they are subjectively different: they differ from each other in the directions of movement of each dot and in the overall directions of motion of the dots. Indeed, they are sufficiently subjectively different to trigger conscious rivalry. I mentioned [5] that rivalry occurs in fruit f lies and can occur in unconscious perception. P&M conclude that the rivalry in Brascamp et al. might be invisible. However, one cannot generalize in this way from rivalry when subjects do not consciously see the stimuli. Rivalry involves the dominance of one whole neural coalition over another. I know of no evidence that rivalry in the case of consciously seen stimuli can somehow slice off the conscious part of the coalition. The competing stimuli are subjectively but not noticeably different. Not being noticeably different, rivalry transitions are less likely to draw more attention than real transitionsas confirmed by Brascamp et al. Would the differences between the neural representations of such stimuli be decodable in the brain at all given how similar they are? Recall that the explanation of binocular rivalry is that pools of neurons that represent each of the stimuli are mutually inhibitory. In the presence of neural noise, one pool wins out. The dominant pool of neurons then weakens due to adaptation, the other pool taking over in the winner-takes-all process of perception, then the cycle repeats. If this weakening and strengthening of content representations were happening in PFC, it would be detectable,eitherwithfMRIorwithelectrophysiological methods (microelectrodes inserted into the cortex). Not finding differences between the rivalrous changes and the real changes is evidence against PFC differences and hence against cognitive theories of conscious contents. This point interacts with the issue of the not-noticeably different stimuli. With readily characterizable (e.g., face/house) stimuli, weakening and strengthening in visual cortex could have caused a cognitive reflection of that weakening and strengthening in PFC, misleading us as to the role of consciousness in PFC. Returning to the comparison of the rivalrous transitions with the nonrivalrous real transitions, P&M concede that failure to f ind PFC differences between rivalrous and real transitions shows that the causes of rivalrous transitions are not to be found in PFC. However, they go on to say that the conscious contents may nonetheless be in PFC. They use this point to conclude that methodologies that compare real with rivalrous transitions– including subtracting replay from rivalry–‘cannot discriminate rival hypotheses concerning NCC’s’ (neural correlate of consciousness). Note that the neural bases of transitions and contents are linked. In particular, if perceptual contents are based in PFC, the strengthening and weakening of content representations that are intrinsic to binocular rivalry would make rivalrous transitions neurally different from real transitions, as noted above. Contents and differences in content have consequences for transitions. So, failure to find differences between rivalrous transitions and real transitions in cognitive areas of PFC disconfirms cognitive theories of conscious contents.