Evidence Accumulation of Moving Visual Information With and Without Running: Impact over Learning and Serial Positioning Bias in Mice

Abstract

Perceptual decision making requires animals to integrate noisy sensory evidence over time, but whether the sensorimotor context in which that evidence is acquired shapes how this information contributes to choice election remains poorly understood. Here, I investigate how the presence or absence of running influences both learning rate and temporal evidence weighting in a head-fixed virtual reality accumulation task. Six double-transgenic GCaMP6f mice were trained in one of two conditions: Active (full running over virtual corridor traversal) or Passive (replayed visual flow without running). Training progression was separated into four stages (Level 3 Sublevel 1 < 80%, Level 3 Sublevel 1 ≥ 80%, Level 3 Sublevels 2+3, and Level 4) to enable comparisons across task types at matched task difficulty. Active mice (n = 4) advanced through the training protocol faster and with steeper psychometric slopes than Passive mice (n = 2). Three of the Active mice reached the Level 4 evidence accumulation stage while no Passive mice currently did. A lick-zone validation test confirmed that both conditions suppressed premature licking during the cue region and committed to choices only in the reward zone. At Level 4, a spatial bin logistic regression revealed that Active mice weighted late-corridor evidence far more heavily than early-corridor evidence (population WDR = 15.20 ± 10.73), producing a recency bias that contradicts the mild primacy reported by Pinto et al. (2018) in a T-maze version of the same task. The variation in bias may reflect a critical design difference: the T-maze allows multidirectional ball movement that lets mice lean toward their eventual choice during the cue region, whereas the lick-spout task constrains the ball to unidirectional forward motion and confines the choice to a single lick in the reward zone. These findings suggest that running and intentional navigation accelerates task learning, but that the observed temporal evidence weighting depends on whether the task demands permit progressive commitment during evidence accumulation.