Despite its striking flexibility, working memory is severely limited in capacity. However, little is known about how the number of items to be remembered affects the activity of prefrontal neurons during acquisition and retention of working memory. To address this question, we trained monkeys in a multiple-target foraging task which consisted of two distinct phase of working memory, acquisition and retention. In this task, monkeys were asked to forage for a number of targets in a choice display (acquisition phase), and to use this information in the subsequent repetition trials (retention phase). When a new set of problem with different targets was introduced, the animals were required to learn the new display. The number of targets (set-size) was varied from one to three. Thus, the present task enabled us to analyze the effect of set-size on the neural activity of both learning and maintenance of multiple-target memory.
While monkeys performed this task, we recorded single unit activities in the dorsolateral prefrontal cortex (DLPFC) and analyzed the target selectivity of population activity. Consistent with previous findings, we found a set-size effect among activities of target selective neurons. During the acquisition phase, the first target was encoded more strongly than the second target in the two target trials, with even stronger encoding in single target trials. This suggests that encoding capacities of DLPFC neurons were thinly distributed across multiple targets. At the same time, there was a prominent task-phase effect among the same neural population. The target selectivity in DLPFC was stronger in the acquisition phase than in the retention phase, suggesting that the acquisition of working memory requires more target selectivity and DLPFC dynamically modulates its activity to accommodate it. These results suggest that the set-size effect and dynamic modulation of target encoding in the DLPFC underlie the behavioral performance of the multiple-target foraging task.