Brain-wide population activity during reaching integrates action-mediated goal expectation
Peng, Y.; Lindersson, C.; Tinelli, S.; Stedehouder, J.; Shah, R. S.; Lak, A.; Stagg, C. J.; Sharott, A.
Show abstract
Large scale recordings have revealed that neurons encoding motor and non-motor variables are highly distributed across the brain. While these neurons generate population level dynamics during spontaneous behavior, it remains unclear how these latent subspaces relate to the simultaneous motor and cognitive demands during ongoing goal-directed behavior. Here, we show that continuously anticipated action outcome, in addition to movement, drives ubiquitous latent dynamics during goal-directed movements. We used multiple Neuropixels probes to simultaneously record spiking activity from cortical and subcortical regions during a reaching task in head-fixed mice. Task-related population dynamics covaried within a common latent subspace across regions and was conserved across recording days and animals. These latent dynamics preceded movement onset and were modulated by reach distance and reward availability. Furthermore, their temporal progression continuously scaled with the timing of reward consumption and their activity decreased afterwards, despite ongoing stereotypical re-reaches. Our findings thus provide evidence for a brain-wide latent subspace for continuous representation of action-mediated proximity to goal, which could provide the basis for ubiquitous temporal difference learning based on predicted action outcome.
Matching journals
The top 4 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- Response outcome gates the effect of spontaneous cortical state fluctuations on perceptual decisions 98%
- Mixed representations of choice direction and outcome by GABA/glutamate cotransmitting neurons in the entopeduncular nucleus 97%
- Orexin population activity precisely reflects net body movement across behavioral and metabolic states 97%
Similar papers in this journal
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.