(A) Illustration of the population response of congruent neurons in module 1 when both cues are presented. Color indicates firing rate. Right panel is the temporal average firing rates of the neural population during cue presentation, with shaded region indicating the standard deviation (SD). Note that the neuron index refers to the preferred direction with respect to the direct cue conveyed by feedforward inputs. (B) The position of the population activity bump at each instance is interpreted as the network’s estimate of the stimulus, referred to as , which is decoded by using population vector. Right panel is the distribution of the decoded network’s estimate during cue presentation. (C–E) The temporal average population activities of congruent (blue) and opposite (red) neurons in module 1 (top row) and module 2 (bottom row) under three cueing conditions: only cue 1 is presented (C), only cue 2 is presented (D), and both cues are simultaneously presented (E). (F–I) Comparing the estimates from congruent and opposite neurons in module 1 with the theoretical predictions, with varying cue intensity (F), with varying cue disparity (G), and with varying reciprocal connection strength between modules (H and I). Symbols: network results; lines: theoretical prediction. The theoretical predictions for the estimates of congruent and opposite neurons are obtained by Equations 4 and 7. Parameters: (A–E) ; (F) ; (G–I) , and others are the same as those in Figure 4. In (F–H), , and in (I), , .