High-level cognitive and neural contributions to conscious experience and metacognition in visual perception.

摘要

Visual processing in humans has both objective and subjective aspects. Objective aspects of visual processing consist in an observer's ability to accurately discern objective properties of visual stimuli. Subjective aspects of visual processing consist in an observer's visual experience of the stimuli and the observer's metacognitive evaluation of the reliability of objective visual processing. What is the nature of the relationship between objective and subjective visual processing? A wide range of views exists in the literature today, but a broad distinction can be drawn between (1) views holding that objective and subjective visual processing are intimately interrelated, such that changes in subjective processing should be associated with changes in objective processing; and (2) views holding that subjective visual processing is a separate, higher-order process, such that it is possible to change subjective processing without changing objective processing. Here we perform a series of psychophysical experiments to arbitrate between these views. To make the data analysis more powerful, we created a novel extension of signal detection theory for analyzing the informational content of subjective ratings of perceptual clarity and confidence (Appendix A). We constructed a wide array of signal detection theoretic models capturing different hypotheses on the relationship between objective and subjective visual processing and performed a formal model comparison analysis in order to discern which model structures best accounted for a data set in which objective stimulus discrimination performance was dissociated from subjective ratings of visual clarity (Chapter 1). Results from this analysis favor a higher-order view of subjective visual processing. If the higher-order view is correct, it should be possible to disrupt the informational content carried by subjective ratings of perceptual clarity and decision confidence without affecting an observer's objective ability to visually discriminate stimuli. We found two lines of novel empirical evidence for such dissociations. We show that when subjects perform a working memory task in which the contents of working memory require extensive manipulation, ratings of confidence in a concurrent perceptual task carry less information about perceptual task performance, even taking the influence of task performance into account (Chapter 2). Similarly, we show that transcranial magnetic stimulation to dorsolateral prefrontal cortex selectively impairs the metacognitive sensitivity of visual clarity ratings without affecting perceptual task performance (Chapter 3). Finally, we show that perceptual and metacognitive performance can dissociate over time as an observer performs a continuous block of trials in a visual discrimination task, contrary to views holding that perceptual discrimination and metacognition are closely intertwined processes (Chapter 4). We show that this dissociation can be partly attributed to individual variability in gray matter volume of regions of anterior prefrontal cortex previously linked to visual metacognition. We interpret these results as suggesting that limited prefrontal resources can be dynamically allocated to support the performance of either perceptual or metacognitive processes. Taken together, these results provide converging evidence supporting a higher-order view of subjective visual processing. Functionally, objective and subjective processing are organized hierarchically, such that downstream subjective processes reflect the properties of objective processing but can be independently manipulated. Anatomically, these high-level subjective processes are linked to regions of prefrontal cortex rather than posterior perceptual areas.