Parsing a Perceptual Decision into a Sequence of Moments of Thought

Theoretical, computational, and experimental studies have converged to a model of decision-making in which sensory evidence is stochastically integrated to a threshold, implementing a shift from an analog to a discrete form of computation. Understanding how this process can be chained and sequenced – as virtually all real-life tasks involve a sequence of decisions – remains an open question in neuroscience. We reasoned that incorporating a virtual continuum of possible behavioral outcomes in a simple decision task – a fundamental ingredient of real-life decision-making – should result in a progressive sequential approximation to the correct response. We used real-time tracking of motor action in a decision task, as a measure of cognitive states reflecting an internal decision process. We found that response trajectories were spontaneously segmented into a discrete sequence of explorations separated by brief stops (about 200 ms) – which remained unconscious to the participants. The characteristics of these stops were indicative of a decision process – a “moment of thought”: their duration correlated with the difficulty of the decision and with the efficiency of the subsequent exploration. Our findings suggest that simple navigation in an abstract space involves a discrete sequence of explorations and stops and, moreover, that these stops reveal a fingerprint of moments of thought.