A canonical neural mechanism for behavioral variability

Abstract

The ability to generate variable movements is essential for learning and adjusting complex behaviours. This variability has been linked to the temporal irregularity of neuronal activity in the central nervous system. However, how neuronal irregularity actually translates into behavioural variability is unclear. Here we combine modelling, electrophysiological and behavioural studies to address this issue. We demonstrate that a model circuit comprising topographically organized and strongly recurrent neural networks can autonomously generate irregular motor behaviours. Simultaneous recordings of neurons in singing finches reveal that neural correlations increase across the circuit driving song variability, in agreement with the model predictions. Analysing behavioural data, we find remarkable similarities in the babbling statistics of 5–6-month-old human infants and juveniles from three songbird species and show that our model naturally accounts for these ‘universal' statistics.

Strength of Correlations in Strongly Recurrent Neuronal Networks

“…The fact that in Ref. [50] the correlations were independent of N and K [or were Oð1=KÞ when all neurons shared a common input] does not contradict our theory. In the architecture considered in Ref.…”

“…We recently studied the emergence of correlations in a network consisting of two strongly recurrent E − I subnetworks [50]. We showed that when one group of neurons in the first subnetwork projects to all excitatory neurons in the second subnetwork, the correlations in the second subnetwork are Oð1=KÞ.…”

“…In the architecture considered in Ref. [50], the subsets of neurons in the first subnetwork are projecting to a macroscopic fraction of neurons in the second subnetwork. This organization generates correlations between elements of the connectivity matrix, unlike in the models considered here.…”

“…In a previous paper [50], we presented an example of a network model consisting of strongly recurrent populations in which the activity is temporally irregular and spatially correlated. In the present paper, we derive general theorems that relate the strength of pairwise correlations to features of the network architecture.…”

Thalamostriatal and cerebellothalamic pathways in a songbird, the Bengalese finch

“…If for example single DLM neurons contact multiple medium spiny neurons (MSNs) in Area X, similar to the way that thalamic neurons contact multiple MSNs in the mammalian striatum (Deschenes, Bourassa, Doan, & Parent, 1996;Kuramoto et al, 2009;McFarland & Haber, 2001;Parent & Parent, 2005), this would position them to correlate activity across the multiple neurons with which they form synapses. Modeling studies suggest correlations in activity enable neural networks to produce behavioral variability such as that generated by the AFP (Darshan, Wood, Peters, Leblois, & Hansel, 2017). One way to test for this function would be to reversibly and specifically inactivate thalamostriatal synapses in Area X and measure the effects on vocal variability.…”