Phenomenological model links variability in sensory responses to response time.

<p>(<b>A</b>) Example showing that an increase in variability leads to a decrease in mutual information. Two stimuli (S1,S2) are presented with equal probability. This generates ‘sensory evidence’ () which is used to make the decision (here taken as a continuous one-dimensional variable). The variability in the ‘sensory evidence’ is illustrated by plotting the likelihood functions (black for S1, gray for S2). These functions are taken as normal distributions with a standard deviation of 0.5 (solid lines) or 1 (dotted lines). The inset shows how the mutual information depends on the standard deviation of the likelihood functions. (<b>B</b>) Illustration of a qualitative model of how mean response time depends on mutual information. At stimulus onset a decision variable starts moving towards a decision threshold (dashed line). The response time is the time between stimulus onset and reaching threshold. The mean rate of rise (slope) of the decision variable is a monotonously increasing function of the mutual information. The different slopes for ‘sensory evidence’ with low variance (solid line) compared to that with higher variance (dotted line) leads to different mean response times. (<b>C</b>) Variability at an extrastriate site in monkey T (black dot in in <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0043166#pone-0043166-g004" target="_blank">Figure 4B</a>) can account for how mean response times depend on foreperiod duration. Variability of the VEP was converted to mutual information (see <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0043166#s4" target="_blank">Material and Methods</a>) and subsequently used to fit the mean response times.