FIGURE 4.

Observer simulations and optimal weights. Top panel: a comparison using the root-mean squared error between empirical data from the centre and surround conditions (left and right plot, respectively) and simulated data with varying combination rules (Integration in blue, Switching in red, and Random choice in grey). Ten datasets were simulated for 9 weight values (from 0.1 to 0.9) and for each rule. In both plots the lowest RMSE values across all weights at the 100 ms interval for all combination rules (and their standard errors across the 10 simulations) are shown on the left, while the mean and standard errors of the lowest RMSE values averaged across the remaining intervals are shown on the right. Bottom panel: pattern of weight values yielding the lowest RMSE at each timing and for Integration Model (in blue) and Switching Model (in red) when compared to empirical data (left: centre task, right: surround task). ***p < 0.001.