Tracking Single V1 Neurons Across Days: Does Repeated Exposure to an Ordered Movie Sequence Reshape Their Responses?

Abstract

Primary visual cortex (V1) responses carry information about temporal context and prior experience. I ask whether four days of exposure to a fixed A → B → C movie-clip sequence (the Allen Brain Observatory training paradigm) changes V1 responses to each clip given its predecessor. Using two-photon calcium imaging from 12 mice, I tracked excitatory V1 neurons across pre- and post-training days via a Hungarian-plus-IoU pipeline (286 matched neurons). For each neuron I compared peak ∆F/F when a clip was preceded by its trained versus untrained predecessor, testing whether this difference grew from Day 0 to Day 4. No training-induced trained-versus-untrained difference emerged on Day 4 (Wilcoxon p = 0.966). Instead, a small Day-0 baseline asymmetry (trained > untrained, +0.010 ∆F/F ) was eliminated by Day 4 (trained − untrained ≈ 0), yielding a negative difference-of-differences with 10/12 mice in the same direction (per-mouse Wilcoxon p = 0.052). The change therefore reflects the disappearance of the pre-training baseline asymmetry rather than the development of a clean Day 4 suppression of trained-context responses; this pattern is consistent with, but does not by itself establish, a predictive-coding account in which repeated exposure causes expected transitions to evoke smaller responses than unexpected ones. Response timing was null on average. At the single-neuron level, however, amplitude and latency changes were coupled: shrunken responses peaked earlier and vice versa (Pearson p = 0.0036). The paradigm thus produces a small cohort-consistent suppression of trained-context responses with coupled amplitude–timing adjustments, consistent with an experience-dependent gain change.