Our senses do not provide a deterministic, one-to-one representation of the external world. On the contrary, sensory information is inherently noisy and frequently ambiguous. How do animals make adaptive perceptual decisions in the face of such uncertainty?
A sensible thing for a decision maker to do is establish a level of confidence–a degree of belief about whether a pending decision is likely to be correct. Confidence is crucial for guiding subsequent decisions that depend on the outcome of an initial decision, such as in foraging (explore or exploit?) and medical diagnosis (undergo a risky surgery?). It also has a powerful influence on learning and adapting to new environments. For instance, when negative feedback follows a choice made with high confidence, it implies that something about the world has changed, triggering a shift in behavioral strategy or increase in learning rate.
Our goal is to understand how the brain generates a sense of confidence in a decision, and how this relates to other outcomes of the decision process, namely the choice itself (i.e., accuracy) and the time needed to decide (reaction time). I use behavioral assays to ask monkeys how confident they are in decisions about visual motion while recording and manipulating activity in visual cortical areas (MT, MST). In addition to conventional causal methods such as electrical microstimulation, I have developed and refined optogenetic approaches for manipulating neural activity with greater spatial and temporal specificity. The results so far support the idea that all three key observables of a perceptual decision–choice accuracy, reaction time, and confidence–can be explained by a common mechanism characterized by bounded accumulation of evidence.