ABSTRACT
Neurons in the prefrontal cortex are typically activated by multiple factors when performing a cognitive task, and by different tasks altogether. The selectivity of single neurons for the same stimulus dimension often changes depending on context or task performed, a phenomenon known as nonlinear mixed selectivity. It has been hypothesized that neurons with such mixed selectivity offer a computational advantage for performing cognitive tasks due to high-dimensional neural representations. In this study, we sought to determine how nonlinear mixed selectivity is affected by training to perform a cognitive task by examining the neural responses of monkeys before and after they were trained to perform visual working memory tasks. We also compared nonlinear mixed selectivity in different sub-regions of the prefrontal cortex that play different roles in these tasks. Our findings indicate that a small population of prefrontal neurons exhibit nonlinear mixed selectivity even prior to any training to perform cognitive tasks. Learning to perform working memory tasks induces a modest increase in the proportion of neurons with both linear and non-linear mixed selectivity. However, we saw little evidence that nonlinear mixed selectivity is predictive of task performance. Our results provide insights on the representation of stimulus and task information in neuronal populations.
SIGNIFICANCE STATEMENT Working memory depends on the ability of neurons to represent stimuli in their pattern of discharges when they are no longer present. How neurons represent simultaneously different types of information remains a complex computational problem. It has been hypothesized that nonlinear mixed selectivity emerges as a result of training to perform tasks that require maintenance of stimuli and task parameters in memory. We tested experimentally this hypothesis by examining neuronal responses at different areas of the prefrontal cortex, before and after training to perform cognitive tasks. We reveal the regions of the prefrontal cortex that are most responsible for different types of selectivity, as well as how these types of selectivity vary as a result of training, the context of information represented in working memory tasks, and their modulating factors. These insights are critical to formulating a practical understanding of working memory, and by extension, of memory-related disorders dependent on neural selectivity.
Competing Interest Statement
The authors have declared no competing interest.
Footnotes
The authors report no conflicts of interest during the research conducted.