85

Debate on Unconscious Perception IAN PHILLIPS AND NED BLOCK Part 1: Phillips To the untutored ear, the idea that perception does not require consciousness may sound as absurd as the idea that thunderstorms do not require anything to be happening in the sky (cf. Wiggins 2001: 218), or that being red does not require being coloured (cf. Moore 1925: 46–7). What then persuades contemporary theorists that unconscious perception is hard, scientific fact? To avoid “getting bogged down in theories of perception”, Prinz (2015) stipulates that “unconscious perception” is to mean the “unconscious transduction of information . . . useable by the organism that transduces it”. But this will not suffice if our interest is in defending the philosophically substantive thesis that episodes of the same fundamental kind as episodes of conscious perception can occur unconsciously (Burge 2010: 374–5; Block 2012: 11–12; cf. Prinz 2010: 310). A traditional idea is that perception essentially involves occupying a subjective perspective on an objective world. Pursuing this idea, Burge characterizes perception as constitutively a matter of “objective sensory representation by the individual” (2010: 368). So characterized, not all useable transduced information constitutes perception. For an informational state to constitute perception it must (a) have objective content, representing how particulars are arranged in the subject’s external environment; and (b) be attributable to the individual, not merely, say, to their visuomotor system. If consciousness were a requirement either for objectivity (e.g., Eilan, forthcoming) or for individual attributability, Burge’s characterization of perception would rule out unconscious perception. However, Burge rejects both 165 15031-0091-FullBook.indb 165 5/12/2016 9:23:15 AM166 • Ian Phillips and Ned Block suggestions. For Burge, perceptual objectivity is achieved by exercise of the perceptual constancies (399): “capacities to represent environmental attributes, or environmental particulars, as the same, despite radically different proximal stimulations” (114). Further, a representation’s being attributable to the individual turns, paradigmatically, on whether the representation’s content is available to central, coordinating agency (333). Does a commitment to unconscious perception ineluctably follow? Only if centrally available, constancy-implicating representations can occur unconsciously. Here I argue that the empirical evidence which Burge and many others cite fails to establish this contention. Thus, even granting Burge’s controversial claims about perception, the existence of unconscious perception remains an open question. (For much more on these issues, see Phillips, forthcoming a, b.) Cases of two types ground much contemporary belief in unconscious perception. First, clinical cases in which perception appears preserved despite loss of consciousness. Second, paradigms in which a stimulus continues to influence responding despite apparently being suppressed from conscious awareness. Consider (type-1) blindsight, the striking phenomenon of preserved visual function (standardly evinced by successful forced-choice responding to a narrow range of stimuli) despite destruction of V1 and in the absence of acknowledged awareness (Weiskrantz 1986/2009, Cowey 2010). According to Burge: “blindsight patients perceive environmental conditions. The perception involves perceptual constancies—including motion, location, and size constancies. The perception guides action. There is strong reason to believe that some of these patients lack phenomenal consciousness in the relevant perceptions” (374). Does blindsight involve the perceptual constancies though? Early work on blindsight neglected this issue. Thus, Weiskrantz (2002: 572) notes “that size constancy, or in fact any of the visual constancies, has never been addressed in any blindsight studies of which I am aware”. More recent work supports a negative answer. For example, Alexander and Cowey (2010) provide evidence that whilst their two patients (MS and GY) retain a capacity to locate and detect stimuli, this capacity is exclusively based on the ability to detect sharp luminance contours and stimulus transients. Neither ability implicates the perceptual constancies, instead being interpretable purely in terms of a sensitivity to proximal stimulation (cf. Burge 2010: 352). In keeping with this, Alexander and Cowey conclude that MS and GY have only the ability to detect “events” varying in “subjective salience” (532). Similarly, Azzopardi and Hock (2011) show that motion detection in GY is limited to detection of “objectless” first-order motion energy (i.e., spatiotemporal changes in luminance) as opposed to detection of changes in position or shape. And Kentridge, Heywood, and Weiskrantz (2007) show that their patient DB matches colored stimuli purely by wavelength and so lacks even rudimentary color constancy. Such evidence suggests that the preserved visual functions of blindsight do not constitute perception proper. 15031-0091-FullBook.indb 166 5/12/2016 9:23:15 AMDebate on Unconscious Perception • 167 Another issue raised by many clinical conditions including blindsight, neglect, and prosopagnosia is how we can be sure that a failure to report awareness reflects a genuinely complete absence of awareness. Signal detection theory (SDT) provides a helpful framework (Green and Swets 1966). The core insight of SDT is that responses in a perceptual task are the joint upshot of two factors: discriminative capacity (d‘) and response criterion (c). In so-called forced-choice tasks in which subjects must select which of two (e.g., spatial or temporal) intervals a target stimulus is presented in, subjects naturally adopt unbiased criteria, simply picking whichever interval produces the largest sensory response. Such tasks provide a direct guide to perceptual sensitivity. However, in many other tasks, the decision space is not symmetric and subjects often exhibit strong biases toward a particular response. For example, in “yes/no” detection tasks in which subjects are asked whether a stimulus is presented or seen, the threshold which a sensory response must meet to elicit a “yes” may be highly variable, exhibiting either “liberal” or “conservative” bias (see Figure 11.1). Many hard-to-control factors encourage bias in ways which are (very plausibly) independent of awareness. For instance, a subject may exhibit conservative bias because they are naturally under-confident, in a low mood state, or have certain pre-conceptions about their own capacities or the experiment’s purpose. Furthermore, subjects with acquired field defects may have standing problems adjusting their criterion from that formerly appropriate, or across their differentially sensitive fields. Such difficulties will yield conservative bias (Azzopardi and Cowey 2001). The upshot is that a failure to report stimulus presence or awareness in the presence of preserved discriminative capacity cannot be assumed to reflect unconscious perception. Instead, we must always take seriously the possibility that it simply reflects conscious perception combined with conservative response bias. This “problem of the criterion” was used to cast doubt on much early psychological work on unconscious Figure 11.1 SDT analysis of a simple “yes/no” task. Only sensory responses above the subject’s variable response criterion elicit a positive, “yes” response. Two possible criteria are shown: the first (c1 ) moderately liberal, the second (c2 ) highly conservative 15031-0091-FullBook.indb 167 5/12/2016 9:23:15 AM168 • Ian Phillips and Ned Block perception (e.g., Eriksen 1960 and Holender 1986). It continues to plague studies of perception in clinical populations as well as many other putative cases of unconscious perception—for example, inattentional blindness (e.g., Mack and Rock 1998; see Dulany 2001) and attentional blink (e.g., Luck, Vogel, and Shapiro 1996). One way of avoiding the problem is to turn to studies in which subjects display no preserved discriminative capacity in respect of some feature (i.e., d‘ = 0) and yet that feature continues to exert a perceptual influence. The classic paradigm is masked priming in which masking renders a prime stimulus indiscriminable (d‘ = 0), and yet the prime still facilitates subsequent responses to congruent, supraliminal targets. However, such paradigms do not provide unproblematic evidence of unconscious perception. First, establishing that task relevant perceptual sensitivity is completely absent, as opposed merely to very low, is a methodological minefield. Second, the assumption that d‘ is a fully adequate measure of phenomenal consciousness is far more controversial than usually supposed (e.g., Schmidt 2007). Third, studies rarely concern themselves with showing that the priming effect implicates objective representation, and so perception proper. Here I highlight a fourth broad concern—namely, whether priming effects suffice to evidence individually attributable representations. Consider an exemplary recent study by Norman, Akins, and Kentridge (2014) in which subjects were asked to identify the colour of a mask. A preceding prime matched the mask either in surface colour or reflected colour (i.e., wavelength) but, because of a shift in illumination between prime and mask, not both (Figure 11.2). A c.25ms response advantage accrued under Figure 11.2 Trial sequence from Norman et al. (2014: 2824). Reproduced with permission. Copyright © 2014 Elsevier Ltd. All rights reserved. 15031-0091-FullBook.indb 168 5/12/2016 9:23:16 AMDebate on Unconscious Perception • 169 surface-matching conditions indicating constancy-based objective colour representation. Moreover, the effect remained even when subjects could not detect the prime (d’ = 0). This is arguably good evidence of unconscious objective representation. However, the natural understanding of the priming effect provides no reason to think such representations are available to central, coordinating agency. Consequently, the paradigm provides no evidence of individual-level perception. The natural understanding of the effect is that the prime activates objective colour representations in the visual system. This activation results in more fluent processing of subsequently presented congruent colours. As a result, the surface-matched mask is seen more quickly and easily, and so responded to faster. Responses are thus facilitated even though representation of the prime itself is entirely restricted to the visual system and so not individually attributable. This opening discussion barely scratches the surface of decades of empirical work. Its purpose is simply to challenge the common conviction that unconscious perception is an incontrovertible empirical datum. To the contrary, the ambition of establishing the existence of individually attributable, objective representations completely outside of phenomenal awareness is fraught with difficulties. We should not assume that it will be achieved. Part 2: Block Is There Unconscious Seeing, and Why Care? Seeing is a single fundamental natural kind of which conscious and unconscious seeing are sub-kinds (Block 2010; Burge 2010). This fact provides difficulties for some of the major theories of perception. For example, naïve realism posits that there are no perceptual representations and that the phenomenology of perception is a matter of direct awareness of things and properties in the world (Travis 2004). But if there is unconscious perception, it must be a matter of perceptual representation that is not available to awareness. And if unconscious seeing is unconsciously representing, there is a strong case that conscious seeing is consciously representing. Introduction There is no well-confirmed theory of the scientific nature of consciousness. Consequently there is no scientific proof of any particular type of state that it is unconscious. And given the lack of any “criterion” for consciousness, any particular experiment will depend on assumptions that themselves have a questionable scientific status. But what holds for unconscious states also applies to conscious states—there is no scientific proof that you are consciously reading these words right now, and this point should put the first one into perspective. With unconscious states as with conscious states we have good reasons for attribution in particular cases—even if they fall short of the status of other scientific truths. 15031-0091-FullBook.indb 169 5/12/2016 9:23:16 AM170 • Ian Phillips and Ned Block Ian Phillips and I have disagreements about whether there is such a thing as unconscious perception. Instead of focusing on the details of those disagreements, what I will do here is describe my favorite case of unconscious seeing. Continuous Flash Suppression (CFS) To understand what continuous flash suppression (CFS) is, let us start with binocular rivalry, a phenomenon diagrammed in Figure 11.3. “A” shows the bottom of the brain of a subject who is wearing red/green glasses and is viewing Figure 11.3 This diagram is similar to diagrams in Tong et al. (1998). I am grateful to Frank Tong for this diagram and permission to use it 15031-0091-FullBook.indb 170 5/12/2016 9:23:16 AMDebate on Unconscious Perception • 171 a red house superimposed on a green face. One eye receives a face stimulus, and the other a house stimulus. The conscious perception—diagrammed in “B”—is not a combined image but rather conscious alternation, with the whole visual field filled first by one image, then the other, then the first again, and so on, every few seconds for as long as the subject is perceiving the stimulus. Subjects report that when they are aware of a face, they are not at all aware of the house (except for occasional brief transitional images). Much of the early visual processing does not change with the changing conscious percept. But when the face fills the whole visual field, the face processing system in the brain (part of which is indicated by the blue dot) is much more active while the areas that process the house (red dots) are suppressed. And the reverse happens when the subject is conscious of the house. The reports suggest that the suppressed representation is not conscious. Nao Tsuchiya discovered that a high-contrast rapidly changing (at 10 Hz) colored image—known as a “Mondrian”—such as the one on the right in Figure 11.4 could make the perception of the stimulus in the other “suppressed” eye unconscious very reliably for up to 10 seconds and pretty reliably unconscious for several minutes.