From unconscious processing to metacognition: Are we explaining consciousness yet?
Friday, September 30th, 2011 [08:30 - 10:30]
SY_07. From unconscious processing to metacognition: Are we explaining consciousness yet?
Cleeremans, A.
Université Libre de Bruxelles, Bruxelles, Belgium
This symposium, consisting of six presentations by leaders in the field, is dedicated to the complex relationships that exist between unconscious information processing, consciousness, and metacognition. Theoretical perspectives on these relationships are strongly contrasted, with some theories of consciousness assuming that it is constitutive of conscious representations that one be aware of possessing them; while other theories assume that subjective experience has little to do with one’s ability to engage in metacognitive judgements. Two presentations are specifically dedicated to exploring this question. Based on brain imaging studies, Hakwan Lau argues that metacognitive reports reflect subjective sensory awareness and that such reports should thus be preferred to forced-choice discrimination as measures of awareness. Cleeremans reports on computational modeling work aimed at exploring the idea that consciousness results from a cognitive system’s unconsciously learning to redescribe its own activity to itself. The following empirical papers then illustrate the scientific fruitfulness of metacognitive theories of conscious awareness and challenge traditional perspectives. Scott & Dienes suggest that associative learning of visual stimuli is possible without conscious perception. Dienes asks whether unconscious learning of artificial grammars is possible (and concludes that it is). Norman challenges the idea that strategic control requires awareness by showing how people can control the expression of learned knowledge without awareness of the knowledge itself. Jiménez & Mendez, finally, explore the interactions between conscious expectancy and congruency effects to develop a dynamic, interactive view of cognitive control in which conscious and unconscious processes play equally important roles.
TALKS
SY_07.1 - Metacognitive reports reflect subjective sensory awareness
Lau, H.
Columbia University, New York, USA
Metacognitive reports of perception (such as visibility judgments or confidence ratings) are reflected by activity in the prefrontal cortex. Critics of prefrontal theories of consciousness argue that such findings are specific to metacognition, and are irrelevant to the primary aspects of sensory awareness. Here I argue that these metacognitive reports can capture some of our deepest intuitions regarding sensory awareness. Our experiments using psychophysics and brain imaging showed that under the lack of attention, metacognitive judgments were inflated, such that subjects rated their subjective sense of perception to be higher than was warranted by the underlying processing capacity. This explains why we may subjectivily feel that we see vividly the whole visual scene in front of it, even though inattentional blindness and change blindness studies showed that we are only good at processing a few items at a time, and that peripheral vision has relatively poor color sensitivity and spatial resolution. The explanation is that we do not actually see all the items in front of us in colorful details. We only think/judge that we are seeing them because of a biased metacognitive mechanism. Whereas philosopher Ned Block argues that visual phenomenology may be too rich to be captured by self-reports, our findings suggest that self-reports can shed important light on the issue, if we use the right kind of measure - metacognitive reports instead of forced-choice task performance.
SY_07.2 - Metacognitive networks
Cleeremans, A. , Pasquali, A. & Timmermans, B.
Université Libre de Bruxelles, Bruxelles, Belgium
Here I explore the idea that consciousness is something that the brain learns to do rather than an intrinsic property of certain neural states and not others. Starting from the idea that neural activity is inherently unconscious, the question thus becomes: How does the brain learn to be conscious? I suggest that consciousness arises as a result of the brain's continuous attempts at predicting not only the consequences of its actions on the world and on other agents, but also the consequences of activity in one cerebral region on activity in other regions. By this account, the brain continuously and unconsciously learns to redescribe its own activity to itself, so developing systems of meta-representations that characterise and qualify the target first-order representations. Such learned redescriptions, enriched by the emotional value associated with them and informed by our constant interactions with others, with ourselves, and with the world, are constitutive of conscious experience. Learning and plasticity are thus central to consciousness, to the extent that experiences only occur in experiencers that have learned to know they possess certain first-order states and that have learned to care more about certain states than about others. This is what I call the “Radical Plasticity Thesis”. In a sense thus, this is the enactive perspective, but turned both inwards and (further) outwards. Consciousness involves “signal detection on the mind”; the conscious mind is the brain's (non-conceptual, implicit, embodied) theory about itself. I illustrate these ideas through neural network models that simulate the relationships between performance and awareness in different tasks, including Artificial Grammar Learning and the Iowa Gambling Task.
SY_07.3 - Associative learning achieved without conscious perception
Scott, R. & Dienes, Z.
University of Sussex, Brighton, UK
Implicit learning paradigms reliably demonstrate associative learning without conscious awareness of the association being learnt. However, while the relationship between stimuli may be unconscious the stimuli themselves are consciously perceived. We examine if associative learning can be achieved without conscious perception. Face stimuli are employed to capitalise on the large cortical resource dedicated to face processing. Repeated back masking is used to extend unconscious exposure to approximately 2 seconds. Participants’ subjective perceptual thresholds were identified using sample images. Whole face pairs or scrambled face pairs were presented and back masked with a random block pattern. Participants indicated whether the faces were whole or scrambled and rated their confidence. Exposure duration was progressively reduced until zero confidence was reported on ten consecutive trials. Training involved sub-threshold exposure to a randomly selected 20 of 30 face pairs. Exposure consisted of two sets of 20 repeats of the image presented below the identified threshold and masked with a random pattern. Each pair included one male and one female face. Participants reported if the male face was on the left or the right and indicated their confidence. If confidence was above zero the exposure duration was further reduced and the given face pair excluded from the test phase. At test participants rated the correctness of 30 pairs of faces: 10 unchanged from training, 10 with the position of the faces swapped, and 10 new pairs. Judgments regarding the location of the male face during training were at chance (M=.51, SE=.02, t(17)=.31, p=.760, CI95% .47:.55). Nonetheless, the correctness ratings given to swapped face pairs (M=1.76, SE=.10) were significantly lower than those for unchanged pairs (M=1.99, SE=.11, t(20)=3.04, p=.006). Sensitivity to the change in configuration of stimuli only previously seen subliminally provides evidence that associative learning can be achieved both unconsciously and without conscious perception.
SY_07.4 - Does unconscious learning of artificial grammars exist?
Dienes, Z.
University of Sussex, Brighton, UK
One of the key domains investigating the difference between conscious and unconscious processes is implicit learning. Within the implicit learning literature, the key papers defining the “believers’” position that unconscious learning exists, and its nature, were published by Arthur Reber in the 1960s and 1970s. The paper defining the sceptics’ position, and still one of the most closely argued papers sceptical of implicit learning, is Dulany, Carlson and Dewey (1984) (DCD). I will present fresh data which, for the first time in 25 years, attempts to replicate DCD, and test alternative interpretations of their results. Reber exposed people to strings of letters, unbeknownst to subjects generated by a finite state grammar, and then asked people to classify new strings as obeying the rules or not. People could do so at above chance levels despite being unable to describe the rules. Reber argued people had acquired unconscious knowledge. DCD repeated the procedure but asked people to underline the part of the string that made it grammatical or non-grammatical. Treating these underlinings as conscious rules, he showed rule validity predicted correct classification almost perfectly. Thus, DCD argued that all the knowledge was conscious. DCD assumed that forced underlining of part of a string amounted to asserting a rule. But, for example, wondering or completely guessing where to underline is not to assert anything. I repeated DCD’s procedure but in addition asked people to report the basis of their underlining: They completely guessed, they relied on intuition, they used a rule, and they used recollection. People predominantly said they guessed (37%) or used intuition (30%). Further, underlinings only partially accounted for classification. Importantly, the underlinings largely expressed unconscious knowledge, by a metacognitive measure shown to dissociate qualitatively different types of knowledge.
SY_07.5 - Strategic control over unconscious structural knowledge
Norman, E. 1 , Scott, R. 2 , Jones, E. 1 , Price, M. 1 & Dienes, Z. 2
1 Faculty of Psychology, University of Bergen, Norway
2 University of Sussex, Brighton, UK
Strategic control over the application of knowledge is traditionally regarded as indicating conscious access to that knowledge (Jacoby, 1991). The current study challenges this assumption by providing evidence from artificial grammar learning (AGL) indicating that unconscious structural knowledge of two grammars can be strategically controlled. In two AGL experiments all participants were trained on two different grammars. Instructions as to which grammar to apply varied randomly between individual trials of a subsequent test phase. The nature of each grammar was disguised by random variation in irrelevant properties of individual string elements. Whether structural knowledge of the grammars was conscious or unconscious was measured in two different ways. In Experiment 1 (N=72) participants reported their decision strategy after each classification response, and reported their degree of awareness of the nature of the rule in a post-experimental questionnaire. In Experiment 2 (N=72) participants made two judgements after each classification response: They reported (a) their decision strategy, and (b) which stimulus property their decision was related to. In Experiment 1, strategic control was found even among participants who expressed unawareness of the nature of the grammar rule, e.g., who reported that the rule was related to colours when in fact colours were irrelevant to the grammar. These participants showed an advantage for trials attributed to “implicit” decision strategies, i.e., random choice, familiarity, or intuition. In Experiment 2, participants also showed strategic control for "implicit" classifications. This was the case even for implicit classifications attributed to irrelevant stimulus properties, e.g. attributing a judgment to an intuition related to colours when in fact colours were irrelevant to the grammar. Findings are interpreted within the framework of "fringe consciousness" (Norman et al., 2007), and as exemplifying the dissociation between consciousness of judgement knowledge versus consciousness of structural knowledge (Dienes & Scott, 2005).
SY_07.6 - Is it what you expect, or all that you do? Dissociations between conscious expectancies and the control of automatic tendencies
Jiménez, L. & Méndez, A.
Universidad de Santiago, Spain
Congruency effects arise when the processing of an irrelevant dimension facilitates responding to another dimension of the stimulus (e.g. Stroop effect). This effect has been described as automatic, in that it is obtained even when people are trying not to rely on the irrelevant word meaning. However, this congruency effect can be dynamically modulated by context factors such as the nature of the previous trials. The sequential congruency (SC) effect thus refers to the fact that this effect of congruency grows larger after a congruent trial than after a non-congruent trial. The SC effect has been taken as a paradigmatic case of control of automatic tendencies, and it has received considerable attention in the literature on cognitive control. In this talk, we will discuss on the nature of this SC effect, which has been attributed to the effect of overt repetition expectancies, to repetition priming, or to a dynamic regulation of control produced as a result of a continuous monitoring of the conflict produced over the previous trials. We will report on a series of experiments in which we controlled for the effects of repetition priming, and dissociated participants’ expectancies from the SC effect. Specifically, we relied on the gambler’s fallacy to show that expectancies can grow in a direction opposite to that predicted by the SC effect. We also manipulated the learning of sequential contingencies to produce either implicit or explicit “expectancies”, and analyzed the interaction between these learned effects and the effects of SC. The results are discussed in terms of a view of control as arising continuously from the processing of the context at a hierarchy of levels, and in which conscious expectancies are seen more as a single piece of the whole picture, rather than as the top layer playing the starring role of metacognition.