Fig. 9:

Mean field model simulation results of the S1 synaptic activation variable with regards to the prediction error (PE) and stimulus prediction. (A) During pre-stimulus baseline, time-averaged S1 synaptic activation variable increased monotonically with z(0). (B) During post-stimulus presentation, the average S1 synaptic activation variable exhibited a V-shaped profile from varying stimulus amplitude (x = 2.0, 2.5, 3.0), where the minimum occurs when x = z(0) or PE = 0. The minimum shifted rightward with increasing x, indicating that the post-stimulus S1 synaptic activation variable was proportional to |x−z|. 100 Monte Carlo trials were run with random z(0) ∈ [1.0, 5.0]. Mean and SEM for each group are plotted. (C) Comparison of average S1 synaptic activation at different time (pre vs. post-stimulus) and PE: PE = x (i.e., z = 0) and PE = 0 (i.e., z = x). Ten Monte Carlo trials were run with random input amplitude x ∈ [2.0, 2.4]. There was a significant difference in the average S1 synaptic activation between the pre vs. post-stimulus period in both cases. All p-values for pair comparisons marked in the graph were less than 0.0001 (rank-sum test). The pre-stimulus firing was computed from the expectation z onset (from time 0 if no expectation) to the stimulus x onset; the post-stimulus firing was computed from the stimulus onset to withdrawal.