Fig 1. Task description, psychophysics and model structure.
(A) Experimental task procedure [3]. (B) Time course of input. Input to pool R, L and S is represented respectively by orange, dashed blue and green line (colors are consistent with B). A first spontaneous phase is followed by the onset of the target that elicit a strong input with temporal adaptation. At t = 1 s the motion input is turned on: pool L receives λL = λ − Δλ and pools R receives λR = λ + Δλ. After motion input, during the delay phase, the network only receives background noise. In free choice trials, pool S receives input due to the onset of the sure target 500 ms after the offset of the motion input. (C) Schematic representation of the attractor network model. Arrowhead connections represent excitatory projections, dot-head connections represent inhibitory projections. Arrows coming from outside represent external inputs. (D) Psychophysics data measured experimentally (top row) and produced by the model (bottom row). The model reproduces qualitatively the effect of both stimulus duration and coherence on the probability of correct responses in forced choice trials (open circles, right panel), as well as on the probability of choosing the “sure” option (left panel). Moreover the model correctly predicts the increased P(correct) in free choise trials (filled circles, right panel).