Decoding the nonconscious dynamics of thought generation

Abstract

Much of economics, psychology and neuroscience have focused on thought dynamics and how they control our behavior, from individual moral choices to the irrationality of market dynamics. However, how much of our thoughts we actually control when we feel we make deliberate choices remains unknown. Here we show that the content of thoughts can be decoded from activity patterns as early as 11 seconds before individuals report having formed the volitional thought. Participants freely chose which of two differently oriented and colored gratings to think about. Using functional magnetic resonance imaging (fMRI) and pattern classification methods, we consistently classified the contents of thoughts using activity patterns recorded before and after the thought was reported. We found that activity patterns were predictive as far as 11 seconds before the conscious thought, in visual, frontal and subcortical areas. These predictive patterns contained similar information to the responses evoked by unattended perceptual gratings and were evident in individual visual areas. Interestingly, neural information present before the decision was associated with the vividness of future thoughts, suggesting that preceding nonconscious sensory-like representations can impact the content and strength of future conscious thoughts. Our results suggest that thoughts and their strength can be biased by prior spontaneous nonconscious perception-like representations, advancing theories of free will and models of intrusive and repetitive thought production.