This is when monkeys, based on learning a few S-R associations, could first start to predict the saccade that would DAPT mw lead to reward. Rise time in STR averaged 130.7 ± 12.9 ms (SEM) across trials of the S-R

association phase. This is in contrast to PFC, where average rise time was significantly later, at 822.1 ± 128.2 ms (p < 5 × 10−4, Figure 3B). Likewise, during the early-trial epoch (exemplar display and the first half of the delay), information about the forthcoming saccade was significantly higher in STR (1.90 ± 0.04) than PFC (1.0 ± 0.04, p < 10−4, Figure 3C, left). In contrast, late in the trial (second half of the delay and during saccade execution), saccade information

was stronger in PFC (2.44 ± 0.05) than STR (0.83 ± 0.05, p < 10−4, Figure 3C, right). These results indicate that STR played a more leading role than PFC when performance relied on specific S-R associations. A comparison of correct and error trials during the S-R phase is shown in Figure 4. In both cases, monkeys execute a right or left saccade. If activity reflects a motor signal per se, information should be equal on both. Yet, early-trial information in STR was greatly reduced on error versus correct trials (0.02 ± 0.04, p < 10−4, Figures 4A and 4B). It was lower when correct and error trials were pooled together and classified according to exemplar (1.38 ± 0.04, p <

10−4, Figure 4C), Lapatinib or saccade (0.70 ± 0.03, p < 10−4, Figure 4D). Fossariinae There was also a decrease in PFC saccade information late in error trials (error trials alone: 0.85 ± 0.04, p < 10−4; correct and error trials by exemplar: 0.70 ± 0.05, p < 10−4; correct and error trials by saccade: 1.68 ± 0.06, p < 10−4). The lower information on error trials indicate that the STR and PFC are not reflecting a saccade motor plan per se (including “guesses”), but rather are involved in learning the correct saccade. The saccadic motor plan might have been generated and maintained elsewhere. During the category acquisition phase, monkeys were confronted with increasingly larger numbers of novel exemplars (Figure 1C) and had to move beyond simple S-R association and associate the right and left saccades with each category rather than individual exemplars. Performance was maintained at a high level and improved, even though with each block an increasing proportion of novel exemplars was introduced (Figure 3A, middle row). During this phase, strong early-trial, saccade-predicting activity in PFC first appeared. This was reflected in the sharp reduction in rise time (Figure 3B) and increase in saccade-direction information in the early-trial PFC activity, relative to S-R association (p < 0.005 for rise time and p < 10−4 for information magnitude, Figure 3C).