(A) Assumed neural population activity giving rise to the DM mean and variance of the momentary evidence, and their dependence on coherence. Each curve represents the activity of a population of neurons with a range of heading preferences, in response to optic flow with a particular coherence and a heading indicated by the dashed vertical line. (B) Expected pattern of reaction times if variance is independent of coherence. If neither the DM bound nor the DM variance depend on coherence, the DM predicts the same decision time for all small headings, regardless of coherence. This is due to the DM drift rate, kvis(c)sin(h) being close to 0 for small headings, h≈0, independent of the DM sensitivity kvis(c). (C) Expected pattern of reaction times when variance scales with coherence. If both DM sensitivity and DM variance scale with coherence while the bound remains constant, the DM predicts different decision times across coherences, even for small headings. Greater coherence causes an increase in variance, which in turn causes the bound to be reached more quickly for higher coherences, even if the heading, and thus the drift rate, is small.