A probabilistic, distributed, recursive mechanism for decision-making in the brain
(a, b) Mean population firing rate of LIP neurons during correct trials on the reaction-time version of the dot motion task (19 neurons). By convention, inRF trials are those when recorded neurons had the motion-cued target inside their response field (solid lines); outRF trials are those when that target was outside the neuron’s response field (dashed lines). (a) Aligned at stimulus onset, starting at the stereotypical dip, illustrating the “ramp-and-fork” pattern between average inRF and outRF responses. (b) Aligned at saccade onset (vertical dashed line). (c, d) Mean time course of the model sensorimotor cortex in rMSPRT aligned at decision initiation (c; t = 1) and termination (d; Term; dotted line), for correct trials. Initiation and termination are with respect to the time of basal ganglia output. Note the suggested saccade time “Sac?”, close to the peak of inRF computations. Simulations are a single Monte Carlo experiment with 800, 1200 total trials for N = 2, 4, respectively, using parameter set Ωd. For simplicity the (r)MSPRT is simulated in discrete time steps, but these have an interpretation in continous time (see Methods). We include an additional step at −1 determined only by initial priors and baseline, where no inference is carried out (yi(t + δyb) = 0 for all i; see Methods). Conventions as in (a). (e) Same as in (c), but for the standard, non-recursive MSPRT (Eq 10 using only the first case of Eqs 8 and 9). (f) Baseline output of the model basal ganglia increases as a function of the number of alternatives, thus increasing the initial inhibition of thalamus and cortex. For uniform priors, the rMSPRT predicts this function is: −log P(Hi) = −log (1/N). Coloured dots indicate N = 2 (blue) and N = 4 (green).