The ability to perceive time and execute timely behaviors is a fundamental cognitive function of the brain. Despite its importance, it remains unclear where and how the brain encodes time, as well as how temporal information is transmitted between brain regions. To address these questions, we first trained mice in a novel time-interval discrimination task (A). We then implanted silicon electrodes in the secondary motor cortex (M2) and dorsolateral striatum (DLS) to perform large-scale electrophysiological recordings. Using support vector machines (SVMs) to decode temporal information from population activity in M2 and DLS, we found that both regions could decode time equally well. Further population activity analysis revealed that, qualitatively, neural activity in both M2 and DLS exhibited neural sequence encoding patterns (B). To quantify the degree of sequential organization, we developed an algorithm to measure sequentiality (C). Applying this algorithm, they discovered that DLS activity exhibited higher sequentiality.
While both DLS and M2 encoded time information equally, DLS’s higher sequentiality raised a key question: What advantage does greater sequentiality confer for temporal processing? To answer this, we proposed—via a feedforward neural network model with all-positive weights—that highly sequential population activity provides an optimal encoding regime for downstream brain regions to read out time information more easily and accurately (D).
Related paper:
Zhou, S., Masmanidis, S. C., & Buonomano, D. V. (2020). Neural sequences as an optimal dynamical regime for the readout of time. Neuron, 108, 1–8.
Address:No.138, medical college road 
Postcode:200032 
Tel/Fax:021-54237056
Emial:itbr@fudan.edu.cn