Sensory areas are spontaneously active, even in the absence of sensory stimuli; however, the function of this spontaneous activity remains largely unknown. Recent technological advances have allowed large-scale neural recordings in awake and behaving animals and, accordingly, have transformed our understanding of spontaneous activity. Studies using such recordings have discovered high-dimensional, spontaneous activity patterns; correlations between spontaneous activity and behavior; and differences between spontaneous and sensory-driven activity patterns. These findings are supported by evidence from developing animals, where a transition toward these characteristics is observed as the circuit matures, as well as by evidence from mature animals across species. These newly revealed characteristics call for the formulation of a new role for spontaneous activity in neural sensory computation.
(A) A transition from low- to high-dimensional activity. (A-i) In vivo spontaneous large scale correlated activity among neurons in the superior colliculus and visual cortex in mice P6–P9 demonstrates low-dimensional nature of neural activity (Ackman et al., 2012). (A-ii) Spontaneous neural activity of 42,568 neurons recorded from the mature mouse visual cortex shows high-dimensional nature with a variety of patterns of correlated activity across neurons, with 128 dimensions accounting for 79% of the shared variance in this recording (recording setup described in Stringer et al., 2021). (A-iii) A summary of the changes in dimensionality from early development to mature animals.
(B) Spontaneous activity in developing animals and mature animals is predictive of behavior. (B-i) Tectal spontaneous patterns are predictive of the larval zebrafish tail flicks (Pietri et al., 2017; Romano et al., 2015). (B-ii) Movement-associated spontaneous activity in the visual cortex and visually-driven thalamus (dorsal lateral geniculate nucleus [dLGN]) of a developing awake rat shows an increase in strength and frequency over development (Murata and Colonnese, 2018). (B-iii) Prediction of spontaneous neural activity in the mature visual cortex (shown in Aii) from facial movements (Stringer et al., 2021).
(C) Spontaneous activity (SA) diverges from sensory-driven activity (SdA) over development. (C-i) Rasters of tectal high coactivity patterns during SA and SdA in 4 and 15 days post-fertilization (dpf) larval zebrafish. Data from Avitan et al. (2021). (C-ii) SA-SdA pattern similarity decreases over development. (C-iii) Schematic of activity patterns geometry: sensory-driven (blue dots and arrows) and spontaneous patterns (orange dots and arrows) in the most diverged case, orthogonal subspaces. These subspaces are geometrically closer early in development (light blue and orange planes), and they grow distant over development (Avitan et al., 2021) reaching orthogonality in the adult animal (Stringer et al., 2019b).
