More Prominent Nonlinear Mixed Selectivity in the Dorsolateral Prefrontal than Posterior Parietal Cortex

Abstract

Neurons in the dorsolateral prefrontal cortex (dlPFC) and posterior parietal cortex (PPC) are activated by different cognitive tasks and respond differently to the same stimuli depending on task. The conjunctive representations of multiple tasks in nonlinear fashion in single neuron activity, is known as nonlinear mixed selectivity (NMS). Here, we compared NMS in a working memory task in areas 8a and 46 of the dlPFC and 7a and lateral intraparietal cortex (LIP) of the PPC in macaque monkeys. NMS neurons were more frequent in dlPFC than in PPC and this was attributed to more cells gaining selectivity in the course of a trial. Additionally, in our task, the subjects’ behavioral performance improved within a behavioral session as they learned the session-specific statistics of the task. The magnitude of NMS in the dlPFC also increased as a function of time within a single session. On the other hand, we observed minimal rotation of population responses and no appreciable differences in NMS between correct and error trials in either area. Our results provide direct evidence demonstrating a specialization in NMS between dlPFC and PPC and reveal mechanisms of neural selectivity in areas recruited in working memory tasks.

New insights on single-neuron selectivity in the era of population-level approaches

“…Zhou et al ( 2021 ) found that PFC neurons exhibited a higher NMS with respect to parietal neurons during a sequential match/non-match task. Similarly, a working memory task modulated a significantly lower amount of NMS units in PPC with respect to the prefrontal cortex (Dang et al, 2022 ). The authors suggest that these differences could be grounded in anatomical dissimilarities: PFC pyramidal neurons are endowed with the most extended dendritic trees of any other cortical neuron, thus they could represent a perfect substrate to implement NMS.…”

“…The authors suggest that these differences could be grounded in anatomical dissimilarities: PFC pyramidal neurons are endowed with the most extended dendritic trees of any other cortical neuron, thus they could represent a perfect substrate to implement NMS. In addition, some task variables (e.g., reward) were less affected by NMS than others both in PFC and PPC, suggesting an influence of the hierarchical rank of the features on their encoding type (Dang et al, 2022 ).…”

“…In fact, it has been shown that the integration of multiple variables is carried out in a non-linear fashion from a mathematical point of view and therefore the abbreviation NMS (Non-linear Mixed Selectivity) is often used. On the contrary, a “linear mixed selectivity” would not guarantee the advantageous properties of NMS (see below) and it has been found only marginally in neurophysiological data (Dang et al, 2022 ). For simplicity, hereafter, we use the term “mixed selectivity” to intend the NMS.…”

“…Moreover, the interaction between multiple features in the framework of mixed selectivity could produce different responses to the same stimulus in different contexts (Rigotti et al, 2013 ). In this regard, for example, parietal and prefrontal neurons were recently found to exhibit spatial tuning for the location of a visual stimulus depending on the context (i.e., as a cue or the match stimulus; Dang et al, 2022 ). Finally, a mixed code provides a better tradeoff of information channeled vs. energy cost, requiring more neurons, but the same number of spikes to be transmitted with respect to a “pure,” selective code (Johnston et al, 2020 ).…”

Visual Perceptual Learning of Feature Conjunctions Leverages Non-linear Mixed Selectivity

“…Mixed selectivity neurons explicitly provide task-relevant information in conjunction learning tasks. They have been described in a number of brain areas across species, including in visual cortex (31)(32)(33)(34)(35)(36)(37)(38)(39). Mixed selectivity neurons are thought to improve decoding of information in downstream areas by increasing the dimensionality of the representational space, thus enabling additional hyperplanes for linear decoders operating in population space that can aid difficult and even non-linear discrimination problems (40)(41)(42).…”