Fig. 5

PMd activity reflected action decisions, not perceptual decisions. Each unit’s firing rate at a given 20 ms time point across trials was regressed on a linear model with the following predictor variables: direction of the chosen target (green), signed checkerboard coherence favoring a particular reach direction independent of its color (dark blue), color of the chosen target independent of its direction (magenta), and signed checkerboard color coherence independent of the direction of the chosen target (turquoise). This regression analysis was repeated for all units and all time points from −200 ms before to +600 ms after the appearance of the first visual cue in each trial (left of subfigure) and of the second visual cue (right). Plotted are the proportions of units with significant regression weights for a given predictor at each 20 ms time point, for each monkey (columns) and task (rows). Very few significant correlations with any regression predictor were seen during the observation period of the first visual cue (left panel of each pair) in either TF (a, b) or CF/CFD (c, d). In particular, significant correlations with chosen target color (magenta) and signed checkerboard color coherence (turquoise) rarely occurred at any time in either task in either monkey. However, following the second visual cue, there were significant correlations with the direction of the chosen target (green) and the signed level of checkerboard evidence favoring a target direction independent of its color (dark blue). Correlates with the direction of chosen target (green) were comparably frequent across both tasks for each monkey, with more units exhibiting selectivity in Z because the targets were placed in each unit’s preferred and non-preferred directions. The incidence of significant correlations with signed checkerboard evidence favoring a particular reach direction (dark blue) was present, but substantially lower, in both monkeys in the CF/CFD tasks compared to in the TF task