Hosted by the Wellcome Neural Dynamics PhD Programme

In the modern era of big data, neuroscience has unprecedented access to brain-wide activity prior to the formulation of any scientific theories. The International Brain Laboratory (IBL) has recorded almost everything, almost everywhere, not at once (N=547 Neuropixels probe insertions; 115 mice; 11 laboratories) as mice perform a perceptual decision-making task.

Mice control the movement of a visual stimulus with a miniature steering wheel. They receive a sucrose reward for bringing the visual stimulus from either the left or right panel to the center. Stimuli are presented in blocks (80% left/20% right and 20% left/80% right). Mice that learn the task have a psychometric shift based on blocks.

Only half the mice (N=358/799; 43%) learn the block structure of the task. We ask: what distinguishes mice who learn from those who don’t? We show that [1] mice use movement to boost the saliency of low-contrast stimuli and [2] mice do false starts biased by blocks.

We use neural data to show that [1] activity is boosted in primary visual cortex (V1) when mice jitter the wheel during low-contrast stimuli and [2] single-cells but not brain areas distinguish between false starts and non-false starts.

This line of work opens an avenue to accelerate animal training: if we know what strategies mice who learn the task use, then we can use this to predict which untrained mice will learn the task.

If joining online: https://bristol-ac-uk.zoom.us/j/94138286231?pwd=MlRURE1SWjR6OTZCR1Fnak9QbGxhUT09, Meeting ID: 941 3828 6231, Passcode: 277162