State-dependent modulation of dependencies in a laminar cortical circuit

Abstract

Spatial attention is a quintessential example of adaptive information processing in the brain and is critical for recognizing behaviorally relevant objects in a cluttered environment. Object recognition is mediated by neural encoding along the ventral visual hierarchy. How the deployment of spatial attention aids these hierarchical computations is unclear. Prior studies point to two distinct mechanisms: an improvement in the efficacy of information directed from one encoding stage to another1-5, and/or a suppression of shared information within encoding stages6,7. To test these proposals, it is crucial to estimate the attentional modulation of unique information flow across and shared information within the encoding stages of the visual hierarchy. We investigated this in the multi-stage laminar network of visual area V4, an area strongly modulated by attention8-10. Using network-based dependency estimation from multivariate data, we quantified the modulation of inter-layer information flow during a change detection task and found that deployment of attention indeed strengthened unique dependencies between the input and superficial layers. Using the partial information decomposition framework11,12, we estimated the modulation of shared dependencies and found that they are reduced specifically in the putative excitatory subpopulations within a layer. Surprisingly, we found a strengthening of unique dependencies within the laminar populations, a finding not previously predicted. Crucially, these modulation patterns were also observed during successful behavioral outcomes (hits) that are thought to be mediated by endogenous brain state fluctuations13-15, and not by experimentally imposed attentive states. Finally, phases of endogenous fluctuations that were optimal for hits were associated with reduced neural excitability. A reduction in neural excitability, potentially mediated by diminished shared inputs, suggests a novel mechanism for enhancing unique information transmission during optimal states. By decomposing the modulation of multivariate information, and combined with prior theoretical work16, our results suggest common computations of optimal sensory states that are attained by either task demands or endogenous fluctuations.