We computed the accuracy of decoding the contents of WM as a measure of information conveyed in the neural population about WMs. We used a linear discriminant classifier to decode object identity and upper/lower location from the pattern of spiking activity across all MUAs within each hemisphere at each time point. Cross-validated decoding accuracy was significantly above chance for both contralateral and ipsilateral sample objects (Figure 2B, dots; p < 0.01, sign test), but it was significantly higher for contralateral (Figure 2B, stars; p < 0.01, paired t test). Thus, prefrontal sensory and memory-related spiking showed a clear contralateral bias in strength and information content, as previously reported (see Discussion).
