86

When this “CFS” process is working properly, 1. Subjects are at chance in making a choice between alternative pictures that have been projected to the suppressed (non-Mondrian) eye. 2. Subjects give the lowest confidence rating on almost all trials. 3. Subjects often insist they are seeing nothing other than the Mondrian. I have used a low tech version of this “CFS” procedure (using red/green Figure 11.4 Mirror stereoscope setup for “continuous flash suppression”. I am grateful to David Carmel for supplying this figure and for permission to use it. 15031-0091-FullBook.indb 171 5/12/2016 9:23:16 AM172 • Ian Phillips and Ned Block glasses) in many classes and talks. People often comment that nothing is being projected other than the Mondrian. 4. There is no difference in confidence between correct and incorrect choices of the input to the suppressed eye (Raio et al. 2012)—that is, confidence does not predict accuracy. And this suggests that even when subjects think they might have consciously seen something, they did not. In addition, there are often differences in kind between processing under CFS and conscious seeing. Raio et al. (2012) compared conscious and unconscious fear conditioning. The stimuli in the suppressed eye was either a male or a female face, one of which was paired with a shock. Fear conditioning was measured by changes in skin conductance in response to the picture paired with the shock. Fear conditioning in the unconscious case ramped up quickly and died off quickly as compared with the conscious case. More interestingly, fear conditioning in conscious—but not unconscious—perception involved suppression of the response to the face that did not predict shock. These facts certainly show unconscious processing, but does that processing involve perception? Mudrik et al. (2011) used a version of CFS (diagrammed in Figure 11.5a) in which the non-Mondrian eye received a picture that slowly ramped up in contrast. The other eye was shown a Mondrian that decreased in contrast when the first picture got to full contrast. The photos included food being put in the oven, a basketball game, and an archery scene (depicted in Figure 11.5c). Each of the photos had an anomalous twin in which, for example, the basketball was replaced by a watermelon, the arrow was replaced by a tennis racket, or the food was replaced by a chessboard. Subjects were asked to press the right or left key as soon as they saw any indication of a scene on the left or the right. What the experimenters were interested in was whether the anomalous photos would break through the “cloak of invisibility” faster. And that was what they found, revealing perceptual integration with context in unconscious perception. As always with purported unconscious perception, we must consider whether CFS allowed fragmentary conscious perception. If so, the appreciation of anomaly might only occur after partial awareness of the stimulus, with faster conscious processing of anomalous pictures being responsible for the result. Mudrik tested this possibility (Figure 11.5b) by integrating the pictures with the Mondrians. In the condition depicted in B, the blended images that were ramped up slowly were presented to both eyes so that the perception of the pictures—both anomalous and ordinary—was conscious. In this condition—in which there was real manufactured partial awareness of the sort postulated by the alternative hypothesis—there was no difference between the anomalous and ordinary pictures, thus disconfirming the alternative hypothesis. Another experiment with similar methodology shows that fearful faces break through the cloak of invisibility faster than neutral or happy faces (Yang, 15031-0091-FullBook.indb 172 5/12/2016 9:23:16 AMDebate on Unconscious Perception • 173 Figure 11.5a–c From Mudrik et al. (2011). I am grateful to Liad Mudrik for this figure. Zald, and Blake 2007). In another, the subject’s attention was drawn or repelled from unconsciously perceived female or male nudes roughly in accordance with gender preferences (Y. V. Jiang et al. 2006). Thus, unconscious perception can involve high-level perceptual categorization that is relevant to personal-level concerns. Unconscious perception must be both unconscious and perception, but there is a potential conflict between these desiderata. The best evidence for lack of consciousness would be if there was absolutely no effect on the visual system—but that would not be perception. And any effect on vision could be used by opponents to argue that the visual registration was not really conscious. 15031-0091-FullBook.indb 173 5/12/2016 9:23:16 AM174 • Ian Phillips and Ned Block Everything depends on the details, and I believe that the details cited show that CFS experiments can thread this needle. I find the CFS form of unconscious perception more convincing than those involved in blindsight or visuo-spatial neglect, syndromes that involve brain damage. Brain damage creates uncertainty about how to understand the response. Another advantage of CFS is that the unconscious perceptions last many seconds, making subjects’ insistence on having no awareness of any stimulus more convincing. Part 3: Phillips Ned Block’s opening statement describes his “favorite case of unconscious seeing”: continuous flash suppression. Here I argue that CFS confronts the same objections which I raised in relation to subliminal priming. As such, it does not convincingly demonstrate that perception of the same fundamental kind as ordinary conscious perception occurs unconsciously. As with subliminal priming, CFS faces two broad concerns. First, is awareness completely abolished (cf. Yang et al. 2014)? Second, do demonstrable effects establish genuine perception by the individual? Here I focus on this second issue. Block takes the alleged fact that “high-level perceptual categorization that is relevant to personal-level concerns” occurs under CFS to establish individual-level perception. In doing so, Block apparently assumes that whether a representation is individually attributable turns on its content. An alternative view is that it turns, not on its content, but on its role. This is the natural understanding of Burge’s suggestion that individual-level representations are paradigmatically those available to central agency. I appealed to this requirement to argue that evidence of objective representation in subliminal priming fails to establish individual-level perception. Various CFS paradigms merit a similar reply. For example, the acquired skin conductance responses evoked by unconsciously presented faces in Raio et al. (2012) do not demonstrate individual-level perception, since such responses are manifestations of the autonomic nervous system, not of central agency. This case seems to me clear-cut; more often it is a delicate question whether a given response-type constitutes an exercise of central agency. Consider the differential orientating responses made to gendered nudes in Jiang et al. (2006). Do these implicate central agency? Not obviously, if the effect is due to the automatic attraction of saccades (as Prinz 2010: 326 suggests; cf. Burge 2010: 333). Even if the effect is attentional, it is controversial whether all attentional effects involve central agency. Setting individual-attributability aside, is Block right that high-level unconscious perceptual representation occurs under CFS? Block cites Yang et al. (2007) and Mudrik et al. (2011), paradigms in which certain stimuli (fearful faces and anomalous scenes) break free from CFS faster than others (neutral 15031-0091-FullBook.indb 174 5/12/2016 9:23:16 AMDebate on Unconscious Perception • 175 faces and familiar scenes). However, the fact that certain stimuli are consciously perceived faster than others may simply indicate that some stimuli are easier (consciously) to detect than others. This no more demonstrates unconscious perception than does the fact that brightly colored objects are easier to spot than dully colored ones. To rule out this simple, differential detectability explanation, recent studies run a control condition wherein stimuli are presented binocularly and made slowly more visible (see Figure 11.5b for Mudrik et al.’s version). Following Mudrik et al., Block argues that since subjects are equally quick to detect both types of stimuli in this control, the faster breakthrough of one type of stimulus from CFS implies specific high-level unconscious perception under CFS. However, as Stein, Hebart, and Sterzer (2011) (also Stein and Sterzer 2014) forcefully argue in relation to the structurally similar study of Jiang, Costello, and He (2007), the control condition used is inadequate. To see why, notice that, whereas in the CFS condition the time after which a stimulus breaks suppression is highly variable, in the control condition the steady ramp in contrast means that the timing of initial awareness is highly predictable. Since CFS and control conditions are studied in separate trial blocks, this creates a crucial difference in temporal uncertainty between conditions. Stein et al. (2011) show that when CFS and control trials are intermixed within blocks, the stimuli which break suppression faster do exhibit a corresponding detection advantage in control trials. This suggests that faster breakthrough is due to differential detectability. Nothing follows regarding unconscious perception. Jiang et al.’s (2006) paradigm in which differential orientating responses are made following the presentation of female or male nudes under CFS avoids these concerns. However, it does not establish “high-level perceptual categorization” since coarse-grained, low-level features statistically associated with gender could equally mediate the effect. Block (2014) argues that perceptual aftereffects indicate genuine perceptual representation. We can test for gender-specific aftereffects by investigating whether an unambiguously female face presented under CFS biases the classification of a subsequently presented gender-ambiguous face. Recent evidence suggests that perceptual aftereffects from stimuli presented under CFS are absent with respect to gender (Amihai, Deouell, and Bentin 2011) and other high-level features including race (ibid.), holistic face processing (Axelrod and Rees 2014), binocular, higher-level components of face shape (Stein and Sterzer 2011), and (to a very large extent) facial expression (Yang, Hong, and Blake 2010). Pace Block, CFS appears to abolish high-level perceptual categorization. These issues have intrinsic interest. Block suggests a wider significance: unconscious perception threatens naïve realism. Block’s view rests on two contentions: (a) that conscious and unconscious seeing are of the same fundamental kind, and (b) that unconscious seeing “must be a matter of perceptual representation”. Naïve realists will likely reject (a). Must they accept (b)? Naive realism’s core tenet is that perceptual episodes involve (non-representational) relations 15031-0091-FullBook.indb 175 5/12/2016 9:23:17 AM176 • Ian Phillips and Ned Block to mind-independent objects, and so have such objects as constituents (Martin 2006). Could a hypothetical naïve realist think of unconscious perception as involving just such relations? It might be objected that such relations are intended to explain phenomenal character. Yet consider two imperfect analogies. Suppose a good life constitutively involves personal relationships, relationships which partly explain why that life is good. It does not follow that personal relationships of the same kind cannot occur within a bad life. Suppose (with Russellians) that true propositions have mind-independent objects amongst their constituents, objects whose identities partly explain such propositions’ truth-values. It does not follow that false propositions lack mind-independent objects as constituents. The naïve realist does face trouble if she insists that sameness and difference in phenomenal character exclusively turns on sameness and difference in perceptual objects. However, naïve realists standardly reject this claim (e.g., Campbell 2009, Brewer 2011). Questions concerning unconscious perception may nonetheless have wider import. Belief in unconscious perception succours scepticism about the significance of consciousness. If perception is essentially conscious, consciousness may partly be important because seeing is.