87

Part 4: Block Phillips is right that the controls in the “breaking-Continuous Flash Suppression” (b-CFS) study I cited were inadequate. I had not read the papers by Timo Stein and Phillip Sterzer and their colleagues that he cites. In b-CFS, differences in breaking through the CFS “cloak of invisibility” are compared to differences in detection of “comparable” visible (i.e., without CFS) stimuli. Some stimuli may be easier to detect than others because of salient low-level features. Salient low-level features—rather than high-level features—could explain breaking CFS, and that possibility must be ruled out by controls. In addition, salient low-level features could trigger a non-perceptual response bias or tendency to respond faster to one of the options. However, this control issue probably does not apply to the study I cited, Mudrik et al. (2011). They showed that the anomalous pictures do not differ from the non-anomalous pictures in standard measures of low-level saliency, and they independently controlled chromaticity and spatial frequency, thus making low-level confounds unlikely. Further, not even Stein & Sterzer think that a response bias is a real option in this case: It is important to note that the possible impact of such non-perceptual factors may be limited to the comparison of upright and inverted faces and does not necessarily apply to other b-CFS studies using different stimuli. For example, it is difficult to imagine that observers would have different response criteria for images of complex scenes that differed only in their semantic content (Mudrik et al. 2011). (Stein et al. 2011, 7) 15031-0091-FullBook.indb 176 5/12/2016 9:23:17 AMDebate on Unconscious Perception • 177 Thus, if it is easier to spot a visible picture of someone shaving with a fork rather than with a razor, that can be explained by the same unconscious perception of anomaly underlying both conscious detection and breaking CFS. Summarizing many studies, Phillips says, “Pace Block, CFS appears to abolish high-level perceptual categorization.” On the contrary, there is strong evidence for weak high-level perceptual categorization, including in one of the studies he cites (Yang et al. 2010) so long as there is spatial attention to the location of the stimulus. This point is emphasized in Stein, Thoma, and Sterzer (2015). And in a review of neuroimaging data, Stein and Sterzer (2014) conclude: In summary, neuroimaging studies investigating the processing of visual information during interocular suppression have shown repeatedly that object- or category-specific neural activity in high-level visual areas of the ventral stream is strongly reduced, but can be retrieved when sufficiently sensitive methods of data analysis are used, such as multi-voxel pattern analysis of fMRI data. In sum, there is substantial evidence for high-level unconscious perception. As Phillips notes, the Jiang et al. (2006) CFS study does not use breaking CFS and so is immune to his criticisms of the b-CFS paradigm. In the Jiang et al. (2006) study, the subject sees a fixation point; then each eye gets a pair of stimuli separated by a fixation point. One eye gets a pair of Mondrians while the other eye receives a nude (male or female) on one side with a texture of fragments of nudes on the other side. In Figure 11.6, the top and bottom in the second panel indicate what is presented to each eye. (Do not make the mistake of supposing that the subject sees a nude with one eye and nude texture with the other eye.) The pair of Mondrians separated by the plus sign presented to one eye suppresses conscious perception of both the nude and nude texture in the other eye. Then if the subject’s attention is attracted or repelled by the unconscious perception of the nude, that is due to unconscious perception of the nude. The attraction or repulsion to one side or Figure 11.6 15031-0091-FullBook.indb 177 5/12/2016 9:23:17 AM178 • Ian Phillips and Ned Block the other is measured by a very brief presentation of a stripy noise patch that can be slightly tilted one way or the other, and the subjects have to say which. Attention is known to increase accuracy in this judgment so the direction of attention can be assessed by measuring the accuracy in the judgment of tilt. Subjects are asked whether they saw any difference between the right and the left. If they report any difference, their data is excluded—on the assumption that some aspect of the nude must have leaked into consciousness since only the nude/texture provides any asymmetry in the stimuli. The conclusion of the experiment is that subjects’ attention is attracted or repelled in a way that conforms to their gender preferences. There have been some reports that high-level unconscious perception under CFS may involve low-level conscious perception, for example of a cloud of color (Hong and Blake 2009; Zadbood, Lee, and Blake 2011; Mudrik et al. 2013; Gelbard-Sagiv et al. under review). But the control in this study provides evidence that whatever low-level conscious perception there might be does not differentiate a nude from a nude texture, so the gender of the stimulus is perceived unconsciously. Phillips says unconscious perception of low-level features associated with gender might explain the result. But even if that is right, the point of appeal to the high level is to justify the conclusion that the perception is by the individual rather than a reaction by a subsystem. And that is accomplished here by the relevance to personal-level gender preferences whether the perception is high or low. Some of these effects are diagrammed in Figure 11.7 where the top graph shows bars of attraction (pointing up) and repulsion (down) for 10 heterosexual males (top) and 10 heterosexual females (bottom). Homosexual males resembled heterosexual females. As you can see, attention in heterosexual males was usually repelled by nude men, but heterosexual females and homosexual males tended to be positive or neutral about both nudes. The upshot is that whether high level or not, this is personal-level unconscious perception. Phillips also says that the Jiang effect may not “implicate central agency” if it is due to “automatic” eye movements, as allegedly alleged by Prinz. I can’t imagine how attraction or repulsion keyed to one’s gender preferences could be automatic in any relevant sense of the term. Prinz claims that the unconsciously perceived nudes attract or repel eye movements rather than attention, but that is not to deny unconscious perception. Contrary to Phillips, Burge and I both deny that involvement in central agency is a necessary condition of unconscious perception. The “paradigmatic” quoted by Phillips is a generic, not a universal (Burge 2010, 370). In any case the attraction and repulsion of attention is an individual-level matter and is not unrelated to agency. In conclusion, though Phillips is right that the controls in the b-CFS studies were inadequate, the overall upshot of the studies described is that there is substantial evidence of unconscious perception. 15031-0091-FullBook.indb 178 5/12/2016 9:23:17 AMDebate on Unconscious Perception • 179 0.4 0.2 Attentional effect Attentional effect 0.0-0.2-0.4 0.4 0.2 0.0-0.2-0.4 Figure 11.7 Part 5: Phillips Same gender Opposite gender 1 2 3 4 5 6 7 8 9 10 Average Same gender Opposite gender 1 2 3 4 5 6 7 8 9 10 Average I previously argued that CFS paradigms fail to demonstrate genuine unconscious perception by the individual. Here I reply to Block’s objections before focusing on what I take to be our more fundamental disagreement. In itself, the differential breakthrough of stimuli from CFS does not establish unconscious perception since it may simply reflect differing conscious 15031-0091-FullBook.indb 179 5/12/2016 9:23:17 AM180 • Ian Phillips and Ned Block detection thresholds. Block accepts Stein et al.’s (2011) critique of extant control conditions designed to rule this out. Nonetheless, he suggests that such concerns probably do not apply to Mudrik et al. (2011). In support, Block quotes Stein et al. expressing scepticism that subjects would adopt differing response criteria in relation to Mudrik et al.’s stimuli. In my earlier reply, I did not mention response criteria partly for this reason and partly because Stein et al. provide evidence that the differential breakthrough of upright versus inverted faces in Jiang et al. (2007) also “cannot be ascribed simply to the influence of differential response criteria”. Instead, they suggest that faster breakthrough results from “a lower detection threshold for upright faces”—crucially one not specific to CFS. Block objects to a differential detectability explanation regarding Mudrik et al.’s stimuli on the grounds that they were matched in respect of various low-level features. However, stimuli matched in the relevant ways may still differ in conscious detectability in a non-CFS specific manner. That is all the objection requires, and why Stein et al., whilst well aware of Mudrik et al.’s results, can reasonably claim that their criticisms are “relevant for and extend to all applications of the b-CFS paradigm” (2011: 4; cf. Stein and Sterzer 2014; and Gayet et al. 2014). Certainly, Mudrik et al.’s findings may be due to unconscious perception of anomaly. Yet given what else we know about flash suppression, and can reasonably extrapolate from studies of binocular rivalry (e.g., Zimba and Blake 1983; see Breitmeyer 2014 for a review), our “default stance should . . . be not to expect much high-level unconscious processing during CFS” (Hesselmann and Moors 2015: 3). Block contests my assessment of the relevant literature, finding in it “substantial evidence for high-level unconscious perception” under CFS. Block is right that Yang et al. (2010) only provide evidence that facial expression “is virtually abolished” (as the authors put it) or abolished “to a very large extent” (as I wrote). However, we should be cautious in relying on Yang et al. as positive evidence of weak high-level categorization outside of consciousness. To ensure unawareness, Yang et al. instructed subjects immediately to press a key if “they perceived anything other than the CFS display”. This terminated the trial. If observers pressed the key on more than 15% of trials, they were excluded from analysis. Yang et al. report that 70% of observers completed the session and that these all “later confirmed that they had not perceived any faces during the CFS period upon questioning” (3–4). This methodology is doubly problematic. First, both key pressing and post-session questioning are subject to obvious response biases and so may easily underestimate true awareness. Second, the practice of post-hoc discarding trials/subjects who show awareness introduces a notorious statistical artifact which Newell and Shanks argue “renders the apparent evidence of unconscious processing almost meaningless” (2014: 50, commenting on Sklar et al. 2012; Carmel 2014 provides a clear exposition of the issue).