Representational drift in the mouse visual cortex

Abstract

Neuronal representations in the hippocampus and related structures gradually change over time despite no changes in the environment or behavior. The extent to which such ‘representational drift’ occurs in sensory cortical areas and whether the hierarchy of information flow across areas affects neural-code stability have remained elusive. Here, we address these questions by analyzing large-scale optical and electrophysiological recordings from six visual cortical areas in behaving mice that were repeatedly presented with the same natural movies. We found representational drift over timescales spanning minutes to days across multiple visual areas. The drift was driven mostly by changes in individual cells’ activity rates, while their tuning changed to a lesser extent. Despite these changes, the structure of relationships between the population activity patterns remained stable and stereotypic, allowing robust maintenance of information over time. Such population-level organization may underlie stable visual perception in the face of continuous changes in neuronal responses.

Learning from unexpected events in the neocortical microcircuit

“…Importantly, there is evidence that representations in the brain can drift naturally over time, even in the absence of learning, e.g., from [Deitch et al, 2020; Rule et al, 2019]. As such, our above analyses left open the possibility that the changes we observed in the neural responses were not a result of unexpected event-driven learning, but were simply a result of non–specific representational drift.…”

Stimulus-dependent representational drift in primary visual cortex

“…As L4 is the primary recipient of input from the lateral geniculate nucleus (LGN), we hypothesized that it would be comprised of neurons with relatively stable responses to the MOV stimulus compared to neurons in L2/3 and L5, which receive more processed input [62]. However, a recent study of natural movie responses in mouse visual cortex found higher stability of firing rate and stimulus tuning in L2/3 and L5 neurons compared to L4 neurons over multiple days [47]. To investigate differences between layers, we chronically implanted a subset of mice ( n = 4) with a custom glass microprism that granted optical access to nearly the full cortical column [63] (Figure 3A-B), and allowed us to separate neurons by cortical layer based on ROI density (Figure 3C).…”

“…As a result, there may be considerably more representational drift in response to naturalistic stimuli than to gratings, particularly over long time periods (weeks to months). A recent analysis of multi-day naturalistic movie responses from visual cortex found representational drift in several visual cortical areas, which was mainly attributed to shifts in ensemble firing rate patterns and, to a lesser extent, changes in ensemble stimulus tuning patterns [47]. These results are intriguing and unexpected, and prompt further exploration of the stability of naturalistic stimulus representations in visual cortex.…”

Self-healing codes: how stable neural populations can track continually reconfiguring neural representations

“…However, neural tunings continued to tile the task, indicating stable task information at the population level. These features of drift have been observed throughout the brain [5, 9, 10].…”

“…The cellular and molecular components of the brain change over time. In addition to synaptic turnover [1–3], ongoing reconfiguration of the tuning properties of single neurons has been seen in hippocampus [4, 5] and neocortex, including parietal [6], frontal [7], prefrontal [8], visual [9, 10], and olfactory [11] cortices. Remarkably, the reconfiguration observed in these studies occurs in the absence of any obvious change in behavior, task performance, or perception.…”