Dynamic consensus-building between neocortical areas via long-range connections
The neocortex is organized into functionally specialized areas. While the functions and underlying neural circuitry of individual neocortical areas are well studied, it is unclear how these regions operate collectively to form percepts and implement cognitive processes. In particular, it remains unknown how distributed, potentially conflicting computations can be reconciled. Here we show that the reciprocal excitatory connections between cortical areas orchestrate neural dynamics to facilitate the gradual emergence of a consensus across areas. We investigated the joint neural dynamics of primary (V1) and higher-order lateromedial (LM) visual areas in mice, using simultaneous multi-area electrophysiological recordings along with focal optogenetic perturbations to causally manipulate neural activity. We combined mechanistic circuit modeling with state-of-the-art data-driven nonlinear system identification, to construct biologically-constrained latent circuit models of the data that we could further interrogate. This approach revealed that long-range, reciprocal excitatory connections between V1 and LM implement an approximate line attractor in their joint dynamics, which promotes activity patterns encoding the presence of the stimulus consistently across the two areas. Further theoretical analyses revealed that the emergence of line attractor dynamics is a signature of a more general principle governing multi-area network dynamics: reciprocal inter-area excitatory connections reshape the dynamical landscape of the network, specifically slowing down the decay of activity patterns that encode stimulus features congruently across areas, while accelerating the decay of inconsistent patterns. This selective dynamic amplification leads to the emergence of multi-dimensional consensus between cortical areas about various stimulus features. Our analytical framework further predicted the timescales of specific activity patterns across areas, which we directly verified in our data. Therefore, by linking the anatomical organization of inter-area connections to the features they reconcile across areas, our work introduces a general theory of multi-area computation.