82

Does unconscious perception really exist? Continuing the ASSC20debate MeganA.K.Peters1,*,†, Robert W. Kentridge2, Ian Phillips3 and Ned Block4 1Psychology Department, University of California Los Angeles, Los Angeles, CA 90095, USA; 2Department of Psychology, Durham University, Durham DH13LE, UK; 3St. Anne’s College, University of Oxford, Oxford OX26HS, UK; 4Department of Philosophy, New York University, New York, NY 10003, USA *Correspondence address. Psychology Department, University of California, Los Angeles, 1285 Franz Hall, Box 951563, Los Angeles, CA 90095-1563, USA. Tel: þ1 323-596-1093; E-mail: meganakpeters@ucla.edu †Megan A. K.Peters, http://orcid.org/0000-0002-0248-0816 In our ASSC20 symposium, “Does unconscious perception really exist?”, the four of us asked some difficult questions about the purported phenomenon of unconscious perception, disagreeing on a number of points. This disagreement reflected the objective of the symposium: not only to come together to discuss a single topic of keen interest to the ASSC community, but to do so in a way that would fairly and comprehensively represent the heterogeneity of ideas, opinions, and evidence that exists concerning this contentious topic. The crux of this controversy rests in no small part on disagreement about what is meant by the terms of the debate and how to determine empirically whether a state is unconscious or not. These are issues that directly concern all of us who study consciousness, so it seems it would be in our best interest to strive for consensus. Given the conversation at ASSC20, we are pleased to have the opportunity to address some of the nuanced topics that arose more formally, and share some of the thinking we have done since the meeting. To reflect the heterogeneity of ideas and opinions surrounding this topic, we have organized this discussion into four distinct contributions. —M.A.K.P. and I.P. Practical and theoretical considerations in seeking the neural correlates of consciousness MeganA.K.Peters Psychology Department, University of California, Los Angeles, Los Angeles, CA 90095, USA. E-mail: meganakpeters@ucla.edu As empirical scientists studying consciousness, we should be concerned with one question above all others: How can we design an experiment that will isolate the “conscious” processing of something from the “unconscious” processing of it, so that we can study the neural processing that underlies awarenessthe neural correlates of consciousness (NCCs)– without inadvertently including a number of other confounds? This is the foundation of the scientific method. Of course, this has always been the goal of studies seeking the NCCs, for example via comparing brain activity in “conscious” and “unconscious” conditions (Baars 1993). But a number of confounds continue to plague our experiments. My goal here, therefore, is to briefly enumerate the current practical concerns in experiments seeking to identify the NCCs, and to discuss how a newly developed paradigm can directly address these practical issues (Peters and Lau 2015). Isolating Awareness: Stimulus Strength and Performance Confounds Two pervasive and closely related potential confounds in the scientific study of consciousness are stimulus signal strength and task performance capacity (Lau 2008; Aru et al. 2012). Each of these needs to be controlled for if we want to isolate awareness to look for its neural correlates. Controlling the external signal strength of a stimulus is relatively easy, by designing an experiment in which stimulus properties do not vary across “conscious” versus “unconscious” conditions. This means the “conscious” condition should not present a large, bright, leisurely stimulus, while the “unconscious” condition uses a small, dim, brief stimulus. Of course neural processing would differ between these two conditions, but the Received: 24 March 2017; Revised: 22 April 2017. Accepted: 15 May 2017 V C The Author 2017. Published by Oxford University Press. This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/ licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact journals.permissions@oup.com Downloaded from https://academic.oup.com/nc/article-abstract/2017/1/nix015/4107416 by New York University user on 28 November 2017 12 | Petersetal. difference would likely have relatively little to do with awareness per se: every region of the visual processing system would likely seem to play some critical role, all the way down to retinal ganglion cells (or even the computer screen!). Although not as obvious as the external signal strength case, an experiment producing unmatched task performance capacity between “conscious” and “unconscious” conditions might imply an internal signal strength difference– especially when one engages in post-hoc data splitting into “seen” versus “unseen” trials (Shanks 2016). As an overly simplified example, imagine a perceptual decision-making experiment in which all stimuli are identical, but the observer blinks on some trials. On “blink” trials the observer does not see the stimulus, and so indicates “unseen” and performs at chance; on “no-blink” trials the observer indicates “seen” and performs well. By post-hoc splitting the trials into “seen” and “unseen” an experimenter has essentially reduced her experiment to the bright versus dim stimulus case, and again has not isolated awareness from the confound of signal strength. Although overly simplified, this example demonstrates the potential for erroneous conclusions if performance is not matched between “conscious” and “unconscious” conditions. Stringent Definitions: What Should Count as “Unconscious” and “Perception”? Assuming we can successfully isolate awareness from signal strength and performance, we must next operationally define unconscious versus conscious perceptioninanexperimentalsetting. First, let us address “unconscious”: for a stimulus to be unconscious, an observer’s subjective experience of the stimulus should be nodifferent from the subjective experience of nothing at all. Although one could define being unconscious of a stimulus in other ways, this conservative definition seems closest to what is targeted when researchers want to distinguish a “conscious” condition or stimulus from an “unconscious” one. A potential objection is that this definition might be too conservative because it fails to capture an important aspect of visual awareness: visual qualia. (For non-visual stimuli, qualia in other modalities should be substituted in this argument.) One might argue that an observer could have an amodal sense or “hunch” that something is present, indicating a lack of true visual awareness of the stimulus. However, this argument requires further definition of just how visual is visual enough to count as visual awareness. Should Type 2 blindsight count, in which some patients with visual cortex damage have residual visual awareness that lacks the usual “particularity, transparency and fine-grainedness” of visual experience (Brogaard 2015, 92; Foley 2015; Foley and Kentridge 2015)? This objection brings up another more important question, however: are visual qualia necessary to study the NCCs of subjective awareness? I would argue no. Whether a subjective experience is visual or not should only affect experiments seeking to identify the neural correlates of visual qualia, not those seeking to uncover the neural correlates of subjective awareness itself. So, in experiments seeking the NCCs we should adopt this definition: awareness of a stimulus is present if the subjective experience of that stimulus is different from the subjective experience of the stimulus’ absence. To define “perception”, I again favor a conservative definition that also will allow us to satisfy the requirement about matched performance. To demonstrate that perception has indeed occurred in a perceptual decision-making task, an observer must be able to make some identification or direct Downloaded from https://academic.oup.com/nc/article-abstract/2017/1/nix015/4107416 by New York University user on 28 November 2017 discrimination decision about a stimulus better than just guessing (Green and Swets 1966; Macmillan and Creelman 2004). Of course, this narrow definition of perception excludes other cognitive processing that may be argued to count as perception, such as priming. A number of studies claim to evince unconscious perception by showing that an unconscious prime (color, word, picture, or even cross-modal event) can influence behavior on a subsequent task (Kouider and Dupoux 2001; Naccache and Dehaene 2001; Kouider et al. 2007; Nakamura et al. 2007; Faivre et al. 2014; Norman et al. 2014). Unconscious primes may also influence objective performance in discriminating a later stimulus without affecting subjective evaluations of that stimulus (Vlassova et al. 2014). However, in priming tasks the unconscious prime does not meet our stringent definition of having been perceived: many studies require that discrimination or detection of the prime be at chance performance for the prime to be considered unconscious (e.g., Kouider et al. 2007; Norman et al. 2014). (Of course, demonstrating true chance performance is a statistical problem in its own right.) Other studies do not ask for any objective or subjective assessment of the prime at all, merely assuming it is unconscious because it has been masked (e.g., Kouider and Dupoux 2001; Naccache and Dehaene 2001; Nakamura et al. 2007). So, while studying unconscious priming can certainly provide insight into visual processing, it does not meet the stringent criteria of “perception” required to isolate awareness from task performance or signal strength. Fortunately, once we have matchedstimulusstrength andperformance, we will have created conditions in which “perception” by this definition is identical as well. Now, the only thing that changes between the two conditions of “conscious” and “unconscious”isawarenessofthestimulus–the“subjectiveawareness” (Giles et al. 2016)– meaning that we have successfully isolated consciousness while controlling for other confounding factors. Exhaustive Measurement: The Criterion Problem Wehavebutonemoreobstacle toovercomeinorder to properly design these experiments: the “criterion problem” (Eriksen 1960; Merikle et al. 2001; Hannula et al. 2005). Just because an observer reports he did not see a stimulus or has low confidence does not mean he had zero subjective experience of it, only that his experience fell below some (potentially very) arbitrary threshold for reporting “seen” or “high confidence”. Unfortunately, all experiments that ask participants to rate stimulus visibility or confidence in a decision on any scale (Ramsøy and Overgaard 2004; Sandberg et al. 2010) are potentially prey to this problem, and until now paradigms developed to avoid the problem (Kolb and Braun 1995; Kunimoto et al. 2001) have met with replicability issues (Morgan et al. 1997; Robichaud and Stelmach 2003)or revealed other theoretical challenges (Galvin et al. 2003; Evans and Azzopardi 2007; Maniscalco and Lau 2012). Although criterion effects can reveal important psychological phenomena in other areas of study (Witt et al. 2015; Peters et al. 2016), in the study of consciousness they represent a very real and fundamental problem by preventing us from exhaustively measuring the presence or absence of awareness (Reingold and Merikle 1988, 1990). Yet these types of scales continue to be used almost exclusively in research seeking the NCCs. Last year, Hakwan Lau and I designed a new paradigm (Peters and Lau 2015) to exhaustively measure awareness in the hopes of demonstrating that it should be possible to achieve “conscious” and “unconscious” conditions that would allow for matched performance. Our task modified the recentThe ASSC20 debate | 3 reapplication (Barthelme´ and Mamassian 2009; de Gardelle and Mamassian 2014) of the classic criterion-free two-interval forced-choice behavioral paradigm (Green and Swets 1966; Macmillan and Creelman 2004). On each trial, only one of two intervals contained a masked, low-contrast target to be discriminated, but we asked subjects to discriminate targets in both intervals (even though there was actually a target in only one) and to bet on which interval they thought they had discriminated correctly. If subjects could discriminate the target when it was present but could not bet on their choices, this would demonstrate the kind of unconscious perception that would allow for matched performance across “unconscious” and “conscious” conditions in subsequent experiments (Peters and Lau 2015). But we found no evidence for unconscious perception. Instead, as soon as participants could discriminate the masked target above chance (i.e., they perceived it) they could tell which interval contained the target (i.e., they were conscious of it); that is, if they were unconscious of the target, they demonstrated no perception (Peters and Lau 2015). A follow-up study asked participants to report which interval contained the more “visible” target, thereby directly asking about visual qualia rather than an amodal subjective sense of “something versus nothing”. However, the results were nearly identical (Peters and Lau 2015). Although it may be argued that the stimulus manipulations necessary for producing unconscious perception can be challenging and fickle, we measured the entire psychometric function and still were unable to identify any point at which the masked stimulus could produce above-chance performance but fail to rise into awareness. WhatThisMeansfortheScientific Study of Consciousness Do these results mean that we can never achieve matched performance in “conscious” versus “unconscious” conditions, rendering the requirements for experiments seeking NCCs impossible to meet? Not necessarily. All we can infer from these results, for now, is that unconscious perception of the type we require seems to be harder to induce than the field may have realized. (We haven’t yet tried all the possible masking or neuromodulation techniques in existence.) Nevertheless, these experimental findings should make us think critically about what has actually been found in studies that do not control for task performance, may be susceptibletothecriterionproblem,orusemaskingorothermanipulations to render a stimulus “unconscious”. The first step toward seeking the NCCs must be for the field to agree on the requirements for experimentally isolating awareness and measuring it exhaustively: tasks in which “conscious” and “unconscious” conditions are performancematched, and which do not depend on arbitrary reporting criteria. Then, we can work to produce an accurate taxonomy of which techniques may produce such matched conscious/unconscious perception, and only then can we use the successful techniques in experiments seeking the NCCs. Of course, this requires that unconscious perception be achievable in the first place. Most of us in the field do believe the unconscious version of perception exists, but many of us don’t believe it has been convincingly demonstrated (Peters and Lau 2015). Despite philosophical objections to the very concept of unconscious perception (Phillips, this commentary), we all need to be convinced we are looking for something that exists– and how to identify it– if we are to use it in experiments that aim to uncover the neural computations, representations, and structures that give rise to our subjective experiences of the world. Downloaded from https://academic.oup.com/nc/article-abstract/2017/1/nix015/4107416 by New York University user on 28 November 2017 Sensation and unconscious perception Robert W. Kentridge Department of Psychology, Durham University, Durham DH13LE, UK. E-mail: robert.kentridge@durham.ac.uk Much has been written about the terms “perception” and “sensation”. Whether either or both might occur unconsciously depends critically upon definitions. If the definition of perception includes a proviso that it is, or gives rise to, experience, it is clear that no empirical evidence will ever show that perception can occur without experience. I therefore want to use as simple an “experience-neutral” definition of perception as is possible. I start with the Oxford English Dictionary (Simpson and Weiner 1989). In the OED the most succinct neutral definition of perception is: “The process of becoming aware of physical objects, phenomena, etc., through the senses”. What we perceive are objects in the world (rather than events in our retina) and the process of perception is separate from that of sensation. I will adopt a working definition of visual perception simply as the process through which we become acquainted with the visual properties of objects in the world (i.e., their distal properties). The OED also has a straightforward definition of sensation: “…the subjective element in any operation of one of the senses, a physical ‘feeling’ considered apart from the resulting ‘perception’ of an object”. So we might define visual sensation as the subjective experience one sometimes has when a stimulus acts on the visual system. In terms of these definitions the question of whether unconscious perception is possible boils down to testing whether representations of the distal stimulus can be constructed when the stimulus being represented does not elicit any sensation. Together with my colleagues I recently published a report of an experiment that addressed this question. Estimation of Distal Stimulus Properties without Awareness: Unconscious Perception? There is much more to color perception than the activation of cones in the retina. Our color experience depends on complex cortical processes that appear to be involved in constructing an estimate of the properties of the distal stimulus– what is known as color constancy (see e.g., Smithson 2005). Having color constancy means that we are able to judge that two identical material samples are the same color even when we see them under different kinds of illumination. As the two samples reflect light in identical ways, but are illuminated by lights with different variations in power across wavelength, the light reflected from them to our eyes will differ. With sufficient visual context from the surroundings we nevertheless judge the color of the materials as the same. We can, however, also make judgments not about the color of materials in the world, but instead about our experiences of color. If we see two samples of a material, one in shade and one in sunlight, we will judge that the materials are the same but we can also say that the color we experience when we look at the material in the shade is duller than the color-experience when looking at the one in direct sunlight. Our percepts of the colors of the objects can be the sameevenwhenwejudgethesensations theyelicit to be different. The question we addressed in Norman et al. (2014) was whether the process of estimating surface color depended upon having color sensations: Was it possible to have color constancy for an unseen surface?4 | Petersetal. HowDifferent Are Seen and Unseen Primes? In our study we used meta-contrast masking to render stimuli invisible. In meta-contrast masking the stimulus, in our case a colored disc, is presented very briefly followed after a short interval by a ring whose inner edge coincides with the location that had been the outer edge of the disc. This arrangement can render the disc quite invisible. It can still, however, be shown that the color of the unseen disc is processed by the visual system if we ask observers to make a decision about the color of the ring, which might be green on some trials and blue on others. If the disc matches the ring in color then observers’ decisions are speeded even when they did not see the disc. A non-conscious estimate of the color of the disc is clearly influencing behavior. What is the nature of that non-conscious estimate? Is it low-level like the signal generated in the retina or is it something more complex like an estimate of an object’s surface reflectance? We test this by presenting the disc and ring in a rich color context (against a background of many differently colored squares) and changing the apparent illumination of the scene between presentation of the disc and presentation of the ring. It appears as if a shadow is moving rapidly across the display during a trial so that the disc is seen under sunlight but, by the time the ring appears, it is in shadow. The color of the disc is the same on all trials. There are two colors of rings, constructed so that one matches the disc in terms of the light it reflects to the observer’s eyes (a spectral match) whereas the other matches in the disc in terms of their apparent surface reflectance across the illuminant change from sunlight to shadow (a surface match). We also selected these colors so that one of the rings is generally classified as “blue” (the spectral match) and the other “green” (the surface match). We compare reaction times for discrimination of the ring’s color in trials with discs compared to trials without discs and ask whether the spectral-match discs or the surface-match discs are the more effective color primes, that is, speed reaction time more. We found (somewhat to our surprise) that the surface-match disc reliably speeds reaction time more than the spectral-match prime. This remained the case in a number of control experiments. Moreover, in a follow-up phase conducted after the reaction time experiments, observers were unable to guess whether trials contained a disc or not (sensitivity measured as d’ was not different from zero), and certainly could not see the discs. We concluded that color constancy, that is, computation of surface color, did not depend upon color experience. Is it reasonable to take this further and conclude that color perception can be unconscious and so perception does not have to follow from sensation? According to my working definitions, in our experiment the distal property of the disc was represented even though observers could not detect its presence and so the disc was perceived despite being unseen. Is there any reason to doubt this conclusion? The main argument leveled against it revolves around whether perception must be something done by a person, as opposed, say, to a subsystem of the visual system. This is one of Tyler Burge’s (2010) stipulations for perception. He says (I summarize) that perception must be sensory, that its content must be of entities in the outside world, that it requires perceptual constancy, and finally, that it must be by the individual insofar as it can initiate or directly guide action by the individual. There is little doubt that the disc in our experiment satisfies the first three stipulations but what about the last? The answer depends upon how the disc speeded reactions; on how this type of “priming” works. Downloaded from https://academic.oup.com/nc/article-abstract/2017/1/nix015/4107416 by New York University user on 28 November 2017 Ansorge et al. (2014) recently comprehensively reviewed research on mechanisms of masked priming and their relationship with conscious executive control. One of the longest standing models of masked priming is Neumann’s (1990) Direct Parameter Specification (DPS) model. Neumann suggests that primes activate a motor response in line with a pre-existing “task-set” (relations between stimuli and the actions to be made in response to them). This task-set can override more reflexive responses to the prime (e.g., using the left hand to respond to stimuli on the left). In DPS the prime may be contributing to construction of a representation matching that of the target intermediate between the raw sensory input and the motor output. It cannot simply be due to summation between the prime and the target at an early sensory stage (e.g., at the receptor level) stage– if it was then the spectral prime should be more effective than the surface prime rather than vice versa. So, in DPS, “a masked prime that is akin to an action trigger sets off the response that is specified in the task-control representation, although the prime can remain below the threshold of awareness” (Ansorge et al. 2014, 272). DPS is almost certainly behind a large fraction of our priming effects and so, in this sense, they are actions, made by the individual, and initiated in response to the prime. Schubert et al. (2013) report unseen primes also affecting temporal attention, speeding responses to events following them closely in time. This may explain why even spectral primes speeded responses to some extent in our study. If a distal representation of the unseen stimulus in our study is initiating action by the individual, then, even going beyond my simple definition of perception, it still seems that perception (including invocation of action by an individual) can be unconscious and does not have to follow from sensation. The action elicited by the unseen prime in our study and in others reviewed in Ansorge et al. (2014) appears to be automatic– there is no evidence that it is under voluntary control. One interpretation of Burge’s stipulation that perception is done by the individual, insofar as it initiates or guides action, is that these actions must be under conscious voluntary control. One might argue that this is slipping in a requirement equivalent to “perception must be conscious” through the back door (to consciously control something must one inevitably be conscious of the relevant properties of that thing?). Instead, let us ask whether the processing of masked primes is significantly different to that of seen primes. What can unseen primes control? Unseen primes can do much more than elicit motor responses. They can modulate switching between “task-sets” (e.g., Lau and Passingham 2007), they can slow or completely inhibit responses by priming “no-go” signals (e.g., van Gaal et al. 2009) and even modify task goals in masked semantic priming (e.g., Fitzsimons and Bargh 2003). Of course, the masked primes in these studies are not testably representations of a distal stimulus. Nevertheless, the range of processes that can be controlled by these unseen primes is not any different from those controllable by seen primes. Ansorge et al. (2014) conclude the key difference between conscious and unconscious primes is not in what they can do but in their flexibility. Unseen primes control responses by selecting from a limited set of alternatives that have previously been set by the individual. Seen stimuli allow novel responses to be executed and new task plans to be set up (even here, unconscious stimuli can play a dominant role in arriving at new task plans, e.g., Reuss et al. 2014). If unseen primes can be used to modify not only just motor responses to stimuli,The ASSC20 debate | 5 but also so many other aspects of our cognitions about a task is it necessary to also require that these responses also be under conscious voluntary control to qualify as percepts of an individual? Unseen primes can control not just one subsystem used in tackling a task but, apparently, all relevant subsystems, albeit in a less flexible but “fast and maybe more error-resistant” (Ansorge et al. 2011, 282) way than seen primes. The ability to use unseen primes in controlling so many diverse aspects of a task suggests that they are available to the individual insofar as the manner of their use is determined by the individual and their processing benefits the individual. The results of our color constancy study satisfy all of Burge’s criteria, and primes defined by surface color might, in principle, be used to control all of these other processes (of course, we don’t know, more experiments are required). I am tempted to draw a conclusion that if it walks like a duck and quacks like a duck then it is a duck– that unseen primes can be perceived and so that color perception can be unconscious and so perception does not have to follow from sensation.