A probabilistic, distributed, recursive mechanism for decision-making in the brain
(a) Comparison of the mean reaction time of monkeys for 2 and 4 alternatives (lines) with that predicted by (r)MSPRT (markers), both for correct trials. Red line: assumed 250 ms of non-decision time. Simulation values are means over 100 Monte Carlo experiments each comprising 3200, 4800 total trials for N = 2, 4, correspondingly, under the parameter set Ω extracted from MT recordings. (b) Discrimination information per ISI in MT statistics (red) compared to the (r)MSPRT’s predictions of the discrimination information available to the monkeys (blue, green). Central lines are for a non-decision time of 250 ms; the edges of the correspondingly-coloured shaded regions are for non-decision times of 300 and 200 ms. (c) As per panel (b), but expressed as a percentage of information lost by monkeys with respect to the information available in MT for the three assumed non-decision times (solid lines and shadings). The information lost if the reaction time match is further enhanced is shown as dashed lines (assuming 250 ms of non-decision time; see Methods). (d) Example ISI density functions before (blue) and after (solid blue and dashed red) information depletion; N = 2, 51.2% coherence, and 250 ms of non-decision time. The null distribution was adjusted to become the ‘new null’ by changing its mean and standard deviation to make it more similar to the preferred distribution. Once done throughout and for a non-decision time of 250 ms, this procedure gives ISI distributions bearing a reduced amount of discrimination information (blue or green lines in panel b), rather than the full discrimination information actually produced by MT (red line). That is, after adjustment, the discrimination information between the preferred and ‘new null’ distributions matches that estimated from the monkeys’ performance.