Prefrontal neurons expand their representation of space by increase in dimensionality and decrease in noise correlation

Encoding of information by populations of neurons is critical for neural processes, but underlying mechanisms are poorly understood. High-dimensional neural representations and changes in inter-neuronal correlations are two candidates for enhancing population-level information. To examine the contribution of both mechanisms, we applied a combination of decoding and encoding methods to simultaneously recorded neural data. We found that increases in the information content of prefrontal neurons prior to saccadic selection of visual targets were greater for the population of neurons than for individual neurons. This differential increase occurred primarily for regions of space weakly encoded by single-cell responses (encoding expansion), relied on high-dimensional neural representations, and was accompanied by selective reductions in noise correlation. Our results show that population-level information is enhanced both by location-specific increases in dimensionality and decreases in noise correlation, enabling prefrontal ensembles to recruit non-informative individual neurons to encode a larger part of space.