The Neural Substrates of Conscious Perception without Performance Confounds

Abstract

How does the brain give rise to consciousness? A common approach to addressing this question in neuroscience research involves analyzing differences in neural activity in experimental conditions where consciousness of a stimulus differs. However, unless careful measures are taken, conditions that differ in awareness typically also differ in perceptual task performance, e.g. stimulus detection and discrimination. A large body of research demonstrates that task performance and awareness can dissociate, indicating that they are separate mental processes with separate underlying mechanisms. Thus, task performance looms as a potential confound in consciousness science: computational and neural processes attributed to differences in consciousness may actually be better attributed to correlated but distinct differences in task performance. Here we present an extended exploration of the issue of task performance confounds in consciousness research. We describe the approach of performance matching (i.e. creating experimental conditions that yield identical task performance yet different levels of awareness) as a solution to the problem of performance confounds, and discuss why it is not appropriate to artificially match performance by post-hoc selection of trials (e.g. analyzing correct trials only). We review a growing literature demonstrating matched-performance / different-awareness effects using a variety of experimental designs and discuss signal detection theory models that can both explain extant results and guide the design of future research. Finally, we consider caveats and nuances for performance matching approaches and propose that future research could pool across multiple experimental designs with disjoint sets of confounds to triangulate on the confound-free neural substrates of consciousness.