Item

Abstract

OBJECTIVE: Drift-diffusion models have been tremendously successful descriptions of the choice process. The abundant behavioral evidence they capture, however, mainly consists of choices and response times, which provide only limited identification of the underlying choice process. In order to examine when and how a prior modulates the choice process during evidence accumulation tasks, we designed a psychophysical experiment that can reveal the evolving decision variable. METHODS: Using a previously studied static perceptual choice task, we systematically varied the stimulus duration in an unpredictable manner, as well as the prior probability that either of the two choice alternatives was the correct one. In addition to subjects' binary choices, we also elicited their decision confidence in an incentive compatible manner. RESULTS: We found that for short trials the evolution of the log-odds of choice is precisely affine linear in log-time: Their dependence on the prior gradually decreases towards zero as time-in-trial evolves, as predicted by the biased starting point in a Bayesian model with exogenous stopping (i.e. in absence of stopping boundaries). However, we found that given sufficient time, conditional choice probabilities became independent of the prior, which is inconsistent with a biased starting point in a drift-diffusion model with endogenous stopping. While subjects' beliefs correlated strongly with their choices, we found that their beliefs did not incorporate the prior even when choices did. CONCLUSION: The two seemingly contradictory findings can be reconciled by a model in which the evolving decision variable captures only the likelihood of the accumulated evidence, rather than being modulated by the prior like a posterior belief. This accumulated evidence then appears to be combined with the prior only after evidence accumulation has stopped. This interpretation, which implies that subjects accumulate the same amount of information regardless of the prior, is further supported by our observation that decision confidence does not take the prior into account even when choices do, suggesting that the prior is maintained by a cognitive process separate from evidence accumulation that subjects use only when evidence is scarce.