Negative Correlations in Visual Cortical Networks

Abstract

The amount of information encoded by cortical circuits depends critically on the capacity of nearby neurons to exhibit trial-to-trial (noise) correlations in their responses. Depending on their sign and relationship to signal correlations, noise correlations can either increase or decrease the population code accuracy relative to uncorrelated neuronal firing. Whereas positive noise correlations have been extensively studied using experimental and theoretical tools, the functional role of negative correlations in cortical circuits has remained elusive. We addressed this issue by performing multiple-electrode recording in the superficial layers of the primary visual cortex (V1) of alert monkey. Despite the fact that positive noise correlations decayed exponentially with the difference in the orientation preference between cells, negative correlations were uniformly distributed across the population. Using a statistical model for Fisher Information estimation, we found that a mild increase in negative correlations causes a sharp increase in network accuracy even when mean correlations were held constant. To examine the variables controlling the strength of negative correlations, we implemented a recurrent spiking network model of V1. We found that increasing local inhibition and reducing excitation causes a decrease in the firing rates of neurons while increasing the negative noise correlations, which in turn increase the population signal-to-noise ratio and network accuracy. Altogether, these results contribute to our understanding of the neuronal mechanism involved in the generation of negative correlations and their beneficial impact on cortical circuit function.

The spatial structure of correlated neuronal variability

“…Modifying our computational model to capture tuning-dependent correlations produces a non-monotonic dependence of residual correlation on tuning similarity in some parameter regimes, but the relevant parameters have not been measured experimentally (Supplementary Figure 6c–e and Supplementary Note S.4). Nevertheless, the modified theory could explain negative correlations previously observed in computer simulations of networks with tuning-specific connectivity 32 and the finding that negative correlations are more frequent between disparately tuned neurons in V1 43 .…”

Novel Interaction between Prefrontal and Parietal Cortex during Memory Guided Saccades

“…Although centered in the positive range, the distribution of spike-count 437 23 correlations typically extends into the negative range. Significant negative correlations, observed 438 to occur more frequently than expected by chance in studies of V1 (Hansen et al, 2012;Chelaru 439 and Dragoi, 2016), MSTd (Gu et al, 2011), FEF (Cohen et al, 2010), area 8a (Leavitt et al, 440 2013) and PFC (Markowitz et al, 2015), tend to occur under conditions otherwise associated 441 with low positive correlations, for instance between neuron pairs that are far apart (Cohen et al, 442 2010;Leavitt et al, 2013) or that have opposed patterns of spatial selectivity (Cohen et al, 2010;443 Hansen et al, 2012;Leavitt et al, 2013;Chelaru and Dragoi, 2016). The push-pull phenomenon 444 we have described is clearly different from within-area interactions insofar as it involves a 445 competitive interaction between neurons with matched spatial selectivity.…”

“…422 In numerous previous studies of neuron pairs in the same cortical area, the spike-count 423 correlation has always been observed to be positive on average (Cohen and Kohn, 2011). The 424 strength of the positive correlation is, however, lower for pairs that are far apart (Constantinidis 425 and Goldman-Rakic, 2002;Smith and Kohn, 2008;Cohen et al, 2010;Leavitt et al, 2013;Smith 426 and Sommer, 2013;Ecker et al, 2014;Katsuki et al, 2014b) or have discordant patterns of 427 selectivity (Zohary et al, 1994;Bair et al, 2001;Constantinidis and Goldman-Rakic, 2002;428 Cohen and Newsome, 2008;Smith and Kohn, 2008;Cohen et al, 2010;Gu et al, 2011;Hansen 429 et al, 2012;Qi and Constantinidis, 2012;Leavitt et al, 2013;Smith and Sommer, 2013;Ecker et 430 al., 2014;Ruff and Cohen, 2014b;Markowitz et al, 2015;Chelaru and Dragoi, 2016;Leavitt et 431 al., 2017b, a) and may vary as a function of wakefulness (Ecker et al, 2014), effort (Ruff and 432 Cohen, 2014a), attention (Cohen and Maunsell, 2009;Mitchell et al, 2009;Herrero et al, 2013;433 Luo and Maunsell, 2015;Ni et al, 2018), learning (Cohen and Newsome, 2008;Cohen et al, 434 2010;Gu et al, 2011;…”

Non-monotonic spatial structure of inter-neuronal correlations in prefrontal microcircuits

“…The spatial structure of intra-areal functional connectivity is frequently inferred by measuring the trial-by-trial correlated variability of neuronal discharges (spike count correlations) (7). One of the most well-established properties (a canonical feature) of intra-areal, mesoscopic, functional connectivity is a so called limited-range correlation structure, reflecting a monotonic decrease of spike count correlations as a function of spatial distance and tuning similarity (7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17). However, this distance-dependent decrease of correlations has been almost exclusively derived from recordings in primary sensory cortical areas or inferred from recordings in the PFC with various constraints like a rather limited scale (18) (see also Discussion section).…”