Preliminary results suggest that spontaneous activity encodes both the animals’ behavioral state and whole-brain dynamics. Facial features, such as pupil size and whisker movement, recorded at the same time as neural activity can predict about 40 percent of variance in spontaneous activity. Total movement of the face — a measure of arousal — accounts for about half of that. Recording electrical activity simultaneously from five brain areas — V1, M1, striatum, hippocampus and thalamus — using a novel electrode technology called neuropixels, shows that activity in other brain areas can account for about 60 percent of the variance in spontaneous activity. (Graduate student Carsen Stringer helped to analyze data from spontaneous recordings, and postdoctoral researcher Nicholas Steinmetz recorded neuropixels data.)
The fact that spontaneous activity was comparatively low dimensional was somewhat surprising to Harris. His work and others’ had suggested that spontaneous activity recapitulates stimulus-driven activity, so the large difference in dimensionality between the two conditions was unexpected. “Back when we thought recording from 50 neurons was a lot, we drew certain extrapolations that aren’t necessarily true now that you can record from 10,000 neurons,” Harris says.
But the more questions we asked about the two spaces, the less surprising it became, says Pachitariu. “Spontaneous activity varies slowly with time, on the scale of seconds and minutes, whereas stimulus-driven activity is well localized in time — a flashed stimulus gets a quick burst of spikes,” he says. The spatial scale was also clearly different. Neuropixels recordings showed that spontaneous activity was correlated across the whole brain, whereas stimulus-driven activity was localized to the visual cortex. “In retrospect, the fact that they have different dimensionality is not surprising because they are different things,” Pachitariu says. He speculates that spontaneous activity might be low dimensional because it’s linked to arousal and behavioral state, such as whisking, sniffing running, which itself is low dimensional.