Tuning and spontaneous spike time synchrony share a common structure in macaque inferior temporal cortex

Abstract

Investigating the relationship between tuning and spike timing is necessary to understand how neuronal populations in anterior visual cortex process complex stimuli. Are tuning and spontaneous spike time synchrony linked by a common spatial structure (do some cells covary more strongly, even in the absence of visual stimulation?), and what is the object coding capability of this structure? Here, we recorded from spiking populations in macaque inferior temporal (IT) cortex under neurolept anesthesia. We report that, although most nearby IT neurons are weakly correlated, neurons with more similar tuning are also more synchronized during spontaneous activity. This link between tuning and synchrony was not simply due to cell separation distance. Instead, it expands on previous reports that neurons along an IT penetration are tuned to similar but slightly different features. This constraint on possible population firing rate patterns was consistent across stimulus sets, including animate vs. inanimate object categories. A classifier trained on this structure was able to generalize category "read-out" to untrained objects using only a few dimensions (a few patterns of site weightings per electrode array). We suggest that tuning and spike synchrony are linked by a common spatial structure that is highly efficient for object representation.

Correlated activity supports efficient cortical processing

“…Together with our previous report (Lin et al, 2014 ), these tests provide additional support for the hypothesis that correlated activity supports efficient processing and behavior. We report that although the concepts of sparseness and decorrelation are often conflated, correlation strength and sparseness (when measured as tuning sharpness) should be considered as separate factors.…”

“…Second, the dense multi-depth arrays allowed us to examine layer specificity, which can tell us about input-output relationships. We and others (Sato et al, 2009 ; Lin et al, 2014 ; Tamura et al, 2014 ) previously reported that local IT populations have a correlational structure in which most neurons are weakly correlated and few neurons have strong tuning correlation and significant spontaneous coincident spiking (~6% of neuronal pairs in IT, vs. ~50% of pairs in V1; Chu et al, 2014 ). Yet, these rare IT choristers are also highly efficient.…”

“…Yet, reports in V1 slices and in vivo have shown the existence of neural ensembles that fire reliably in concert during spontaneous activity (Sadovsky and Maclean, 2014 ), and the same ensembles are active both without stimulation and in response to stimulation (Chu et al, 2014 ; Miller et al, 2014 ). We recently reported (Lin et al, 2014 ) that in macaque IT, correlated neurons “choristers” (Kenet et al, 2005 ; Carandini, 2014 ), neurons that have similar stimulus tuning and coincident spike timing, even during spontaneous activity, carry more generalizable object information than uncorrelated neurons (“soloists”). This surprising result hints that, counterintuitively, correlation supports efficient coding and that current thinking focused on sparsening, decorrelation, and denoising may be flawed.…”

“…Yet, these rare IT choristers are also highly efficient. Just 4–5 choristers per array (the top 6% as defined by k-means clustering, or 8% as defined by average pairwise tuning correlation) have the same object coding capability, for within-category generalization, as the entire array population (no more are needed given their object coding efficiency; Figure 7C of Lin et al, 2014 ). Based on this correlational structure and the much-better object coding capability of choristers vs. soloists, we previously proposed a “pipe cleaner” model (Figure 1B , a “fiber bundle” in mathematical terminology) in which the choristers (the spine) are the substrate of IT’s output, encoding an invariant representation that supports generalization and recognition, and in which the soloists (the bristles) are IT’s inputs, acting as heterogeneous tensors that fine-tune this high-dimensional representation (in the parlance of DiCarlo et al ( 2012 ), to support “cortically local subspace untangling” and to “flatten object manifolds”).…”

Individual Differences in Cognitive Function in Older Adults Predicted by Neuronal Selectivity at Corresponding Brain Regions

“…In contrast, H corr has the potential to estimate neuronal selectivity across brain regions from a single data set, including resting state scans. More specifically, because previous single unit recording studies (Bair et al, 2001; Jermakowicz et al, 2009; Lin et al, 2014) found that pair-wise correlations between two neurons in the presence of visual stimuli can be detected in the absence of stimuli, thus the voxel-wise correlations and H corr within a given region should be similar with or without a task. Similar neural mechanisms might also underlie the findings from fMRI studies of resting state (Fox and Raichle, 2007; Smith et al, 2009).…”