Figure 4.

Results of Experiment 1. (A) Mean shift away from the penalty region is plotted as a function of the shift predicted by the optimal, MEG model for each condition (three penalty levels, Near and Far configurations, three level of stimulus orientation spread) for subject HB. The conditions are indicated by the symbol size (small symbols: Far penalty, large symbols: Near penalty), shape (squares: low orientation variability, circles: medium variability, diamonds: high variability), and color (red: penalty=0, green: penalty=100, black: penalty=500). Filled symbols indicate conditions in which the shift was significantly different from the MEG prediction. Error bars indicate 95% confidence intervals computed by bootstrap simulations. The diagonal line indicates perfect correspondence of data and prediction. The horizontal and vertical lines at a shift of 11 deg indicate the edge of the reward region. The inset shows the same data with expanded range on the axes to show the outlier. (B) Average number of points won per trial is plotted as a function of the MEG prediction for subject HB using the same conventions as in panel A. (C) Efficiency in the two zero-penalty conditions for five subjects. Error bars indicate the range of performance (95% confidence interval) expected from the MEG model. (D) Efficiency in the four nonzero-penalty conditions. Individual data for the other subjects are shown in Supplementary Figure A1