Monday, July 2, 2012

Tutorial 4: Cleeremans, Overgaard, Timmermans, and Scott

Monday, July 2 2012 9:15 - 12:30 @ Old Ship Hotel

Tutorial 4: "Behavioral methods to assess awareness"

Axel Cleeremans (Consciousness Cognition & Computation Group, Universite Libre de Bruxelles, Belgium)
Morten Overgaard (Cognitive Neuroscience Research Unit (CNRU), Denmark)
Bert Timmermans (Neuroimaging Group, University Hospital of Cologne, Germany)
Ryan Scott (School of Psychology, Sackler Centre for Consciousness Science, University of Sussex, UK)


The study of consciousness, and in particular the study of the differences between conscious and unconscious information processing presents unique challenges for it requires that one combines subjective (“first-person”) and objective (“third-person”) data. In this tutorial, we will survey recent developments in the measurement of awareness and discuss their theoretical implications. Over the past few years, many new methods to assess the extent to which a person is aware of some state of affairs have been proposed. Interestingly, these novel proposals range from methods that have a strong focus on phenomenology, such as the PAS scale introduced by Overgaard and colleagues, to methods that have a strong metacognitive focus, such as confidence judgements or post-decision wagering. In parallel, Signal Detection Theory has also been the object of intense scrutiny, with strong debate about the relationships between Type I and Type II performance in discrimination tasks, and the introduction of novel indices of awareness such as meta-d'. A further development worth discussing in this light is the increasing use of what one could call sub-personal measures of performance, such as eye movements or EMG measures, as well as the introduction of methods focused on identifying the source of the knowledge involved in decisions.

Many recent studies have used several such measures on every trial, thus raising the possibility of comparing them in within-subjects design and of tracking their respective dynamics over time or over different experimental conditions. The main goal of the tutorial is to introduce the audience to the large repertoire of such new measures, to discuss best practices in the use of such methods (e.g., what are the best conditions under which d' measures should be collected?), and to reflect upon the theoretical implications of the patterns of associations and dissociations revealed by the combination of such methods. The tutorial will be illustrated with recent experimental data.

A syllabus that includes the slides as well as recent articles relevant to the discussion will be provided to participants. Ample opportunities will be provided to participants for discussion, in particular so as to make it possible to examine together specific problems popping up in participants’ own research.

Tutorial Outline:

The tutorial is structured in three parts, described below:

Part I:

Introduction. [Cleeremans & Overgaard, 50m]: In the first part of the tutorial, we will essentially offer an introduction to the challenge of measuring consciousness. Some of this material will be historical, retracing, for instance, the relevant debates as they took place in the subliminal perception or in the implicit learning literature, and overviewing general but central concepts such as subjective vs. objective methods, the relative sensitivity of different measures of a conscious state, the use of verbal reports, confidence judgments, or betting in assessing the extent to which a subject is aware of a of a particular state of affairs, methods such as signal detection or the process dissociation procedure as they can be deployed both in purely behavioural studies or in imaging experiments. Methodological issues such as the question of bias or the pervasive confound between awareness and performance will be introduced through examples involving different paradigms ranging from perception to memory and learning.

The first part of the tutorial will delineate pending challenges, including the following:

• Can the same methods be deployed in different paradigms (e.g. perception vs. memory paradigms) or should some reporting methods be preferred in certain

• How are different measures best combined? Do different ways of administering a d’ task always yield the same result? Does it matter whether a subjective judgment is produced before or after an objective measure?

• Are there differences between dichotomous and graded measures of some content?

• What are the relationships between measures of sensitivity, awareness, and metacognition? What are the respective dynamics of such measures? What are the patterns of associations and dissociations between them?

At the end of this first part of the tutorial, a short discussion followed by a break will be organized with the goal of collecting specific questions that participants would like to see answered. This will also be supported by interactions with the participants in the weeks that precede the meeting, so that the presenters know which issues will be most relevant to the participants.

Part II:

Sensitivity, awareness, and metacognitive judgments. [80m]: The second part of the tutorial will be dedicated to specific illustrative examples taken from very recent and ongoing research. We envision presentations of about 20m each.

Overgaard will survey subjective methods, asking what constitutes “subjective methods” and how one should in practice make use of them. These questions will be illustrated by a detailed review and analysis of findings using one specific approach to subjective reporting: The Perceptual Awareness Scale (PAS). A number of experiments will be described and discussed with a specific focus on exact methodologies. Furthermore, the implications of the experiments will be discussed actively with the tutorial participants. For instance, some experiments suggest that findings of blindsight and even subliminal perception may at least to some degree rely on confounding factors from inaccurate reporting methods. The talk intends to introduce methods for subjective reporting for a scientific audience without prior experience with their use.

Timmermans will survey subjective measures, specifically asking whether they all subtend the same form of metacognition. According to Sandberg et al., 2010, the best method consists of simply asking participants about the clarity of their visual experience Cleeremans, Overgaard, Timmermans & Scott 3 (as in the Perceptual Awareness Scale, PAS). This method outperforms confidence ratings (in response accuracy), and post-decision wagering, presumably because the latter may depend more on participants' answers, and on the information they consider relevant to that answer (judgment knowledge), and are therefore less exhaustive. PAS, as with confidence ratings as they were originally conceived (pertaining to stimulus clarity instead of response accuracy), has the advantage that it is unrelated to this judgment knowledge. However, in designs where one is specifically interested in whether people know on what knowledge they base their answer (like for instance, letter string classification in an Artificial Grammar Learning task), the question is not about what people believe to have perceived (hence PAS cannot be applied), but about what knowledge they consciously think they possess. Here, the target task becomes relevant, and therefore CR (with respect to response accuracy) and PDW (preferably in a no-loss variant, see Dienes & Seth, 2010) are the subjective measures of choice. The following questions ensue: To what degree can both be compared, to what degree is metacognition in the context of knowing-what-you-saw comparable to metacognition in the context of knowing-what-you-know.

Scott will address the following issues: Judgment vs. Structural knowledge, including coverage of recent methods to assess the basis for participants’ decisions. Confidence vs. Attribution judgments – Based on Scott & Dienes (2008) and on more recent work, the issue of unexpected dissociations between confidence and attribution judgments (such participants reporting zero confidence while simultaneously reporting that they exploited a systematic strategy or rule) will be addressed. Type I vs. Type II decisions, including recently developed novel indices and the relationships between Type I and Type II decisions. Relative familiarity and recalibration. Participants’ recalibrate their confidence thresholds in an intelligent way when given bogus information about their accuracy – showing that they do in fact have conscious access to more information than is typically expressed in their confidence ratings. Relative familiarity can be also used as a potential basis for confidence, which provides a tangible example of how SDT can in principle apply to both first order and metacognitive judgments.

Finally, Cleeremans will report on recent experimental work dedicated to exploring the factors that influence visibility as assessed by means of d’. When investigating unconscious influences in subliminal priming experiments, many have argued that primes are invisible because subsequent d’ (“d prime”) task suggest that people are not able to discriminate between the different primes. However, some problems with this technique have been overlooked. Using different versions of the d’ task, a recent study shows that target presentation, attention on prime stimuli and timing of the response are factors that lead to over- or underestimation of the measured d’. This suggests that the standard d’ task is not a straightforward objective measure of prime visibility and that one has to consider these factors when developing a d’ task in further subliminal perception research.

Part III:

Theoretical implications & general discussion. [50m]: The third part of the tutorial will be dedicated to the theoretical implications of the methods and findings presented in Part II of the tutorial as well as providing an opportunity for ample interaction with the participants.

No comments:

Post a Comment