Contents of Consciousness Investigated as Integrated Information in Direct Human Brain Recordings

Abstract

Abstract Integrated information theory postulates that the particular way stimuli appear when we consciously experience them arises from integrated information relationships across neural populations. We investigated if such equivalence holds by testing if similar/different percepts map onto similar/different information structures. We computed integrated information structure from intracranial EEGs recorded in 6 neurosurgical patients who had electrodes implanted over posterior cortices. During recording, we dissociated subjective percepts from physical inputs in three distinct stimulus paradigms (passive viewing, continuous flash suppression, and backward masking). Unsupervised classification showed that integrated information within stimulus-selective cortical regions classified visual experiences with significant accuracy (peaking on average around 64% classification accuracy). Classification by other relevant information theoretic measures such as mutual information and entropy was consistently poorer (56% and 54% accuracy). The findings argue that concepts from integrated information theory are empirically testable, promising a potential link between conscious experience and informational structures.