Figure 1. Three general decision-making models.

Each schematic represents the evidence that resulted in a single, positive choice. (A) The evidence integration model. Sequential samples of noisy momentary evidence are integrated over time until a decision-bound is reached, resulting in a positive or negative choice. The momentary evidence is assumed to be sampled from a Gaussian distribution with mean proportional to the stimulus strength. If the stimulus is extinguished before a decision-bound is reached, then the decision is based on the sign of the integrated evidence. (B) The extrema detection model. Momentary evidence is sequentially sampled but not integrated. A decision is made when one of the samples exceeds a detection threshold (i.e. an extremum is detected). Samples that do not exceed a detection threshold are ignored. If the stimulus is extinguished before an extremum is detected then the choice is determined randomly. (C) The snapshot model. Momentary evidence is neither sequentially sampled nor integrated. Instead, a single sample of evidence (i.e. a snapshot) is acquired on each trial and compared to a decision criterion to render a choice. The sampling time is random and determined before the trial begins.