Figure 10.

Decoding time courses illustrate fluctuation in the accuracy of decoding spatial category based on population activity patterns in parietal cortex (red) and prefrontal cortex (blue), measured in successive 50 ms time bins. Separate decoding time courses in each panel plot the results obtained using population activity on trials in which the rule was compatible with the category preference of each neuron (red and blue), and trials in which the rule was incompatible (black). Time bins in which the proportion of correctly decoded trials differed significantly between compatible and incompatible rule trials are indicated by black circles and corresponding thicker sections of the decoding time courses. Neurons in this analysis were ranked according to the p value associated with the interaction between rule and category in the ANOVA/ANCOVA, and then varying numbers of the most significant neurons were selected to include in the populations used for the decoding. Decoding results indicate the accuracy obtained when based on the activity of the most significant 70 neurons (solid lines), the most significant 200 neurons (dashed lines), or all significant neurons (dotted lines) in each cortical area and monkey. A, B, Decoding accuracy obtained when using population activity in parietal cortex (A) and prefrontal cortex (B) of monkey 1 on sample–boundary trials. The difference between decoding accuracy on compatible and incompatible rule trials measures the modulation of category signals by the rule. For the decoding analysis including all significant neurons, 55 neurons in parietal cortex and 58 neurons in prefrontal cortex contributed. C, The blue and red time courses indicate the mean posterior probability over time (averaged over trials) associated with the correct spatial category on compatible rule trials, based on neural activity in prefrontal and parietal cortex, respectively (data from monkey 1 on sample–boundary trials). The black time course illustrates the cumulative difference between the two time series. The diagonal dashed lines indicate the upper and lower confidence boundaries established by the sequential trials test. The cumulative difference function crosses the upper confidence boundary, indicating that the posterior probability is significantly larger in prefrontal cortex relative to parietal cortex (p < 0.05). D–F, Corresponding data from monkey 2 on sample–boundary trials. For the decoding analysis including all significant neurons, 62 neurons in parietal cortex and 90 neurons in prefrontal cortex contributed. G–I, Corresponding data from monkey 1 on boundary–sample trials. For the decoding analysis including all significant neurons, 104 neurons in parietal cortex and 120 neurons in prefrontal cortex contributed.