Metastable attractors explain the variable timing of stable behavioral action sequences

Abstract

Natural animal behavior displays rich lexical and temporal dynamics, even in a stable environment. This implies that behavioral variability arises from sources within the brain, but the origin and mechanics of these processes remain largely unknown. Here, we focus on the observation that the timing of self-initiated actions shows large variability even when they are executed in stable, well-learned sequences. Could this mix of reliability and stochasticity arise within the same circuit? We trained rats to perform a stereotyped sequence of self-initiated actions and recorded neural ensemble activity in secondary motor cortex (M2), which is known to reflect trial-by-trial action timing fluctuations. Using hidden Markov models we established a robust and accurate dictionary between ensemble activity patterns and actions. We then showed that metastable attractors, representing activity patterns with the requisite combination of reliable sequential structure and high transition timing variability, could be produced by reciprocally coupling a high dimensional recurrent network and a low dimensional feedforward one. Transitions between attractors were generated by correlated variability arising from the feedback loop between the two networks. This mechanism predicted a specific structure of low-dimensional noise correlations that were empirically verified in M2 ensemble dynamics. This work suggests a robust network motif as a novel mechanism to support critical aspects of animal behavior and establishes a framework for investigating its circuit origins via correlated variability.

Slower prefrontal metastable dynamics during deliberation predicts error trials in a distance discrimination task

“…As in previous accounts (Abeles et al, 1995; Jones et al, 2007; Mazzucato et al, 2015, 2019; Ponce-Alvarez et al, 2012) state transitions varied trial-by-trial and were not necessarily locked to external relevant events. This can be illustrated by warping time so that events in different trials occur simultaneously (Recanatesi et al, 2020), while onset and offset times of states are variable, as shown in Fig. 3a for an illustrative session (see Methods for details).…”

State-dependent regulation of cortical processing speed via gain modulation

“…While barrier heights and the network's attractor landscape can be exactly calculated in the simplified two-cluster network, this task is infeasible in large networks with a large number of clusters where the number of attractors is exponential in the number of clusters. On the other hand, the mean field analysis revealed that changes in barrier heights ∆ are equivalent to changes in gain, and the latter can be easily estimated from spiking activity [37,38] (Fig. 5-S4).…”

“…Metastable activity has been ubiquitously observed in a variety of cortical and subcortical areas, across species and tasks [43][44][45][46][47][48][49][50][51]. Metastable activity can be used to predict behavior and was implicated as a neural substrate of cognitive function, such as attention [45], expectation [20], and decision making [38,48,50]. Metastable activity was observed also during ongoing periods, in the absence of sensory stimulation, suggesting that it may be an intrinsic dynamical regime of cortical circuits [25,45].…”

“…To infer the single-cell transfer function in simulations and data, we followed the method originally described in [38] (see also [37,94] for a trial-averaged version). We estimated the transfer function on ongoing periods when no sensory stimulus was present.…”

“…We estimated single-cell transfer functions from their spiking activity during ongoing periods, in the absence of sensory stimuli but in the presence of different perturbations (Fig. 3d, [38,39]). We found that network perturbations strongly modulated single-cell gain in the absence of stimuli, verifying mean field theory predictions in all cases (Fig.…”