Dynamic Encoding of Reward Prediction Error Signals in the Pigeon Ventral Tegmental Area during Reinforcement Learning

Behavioral learning curves and representative VTA spike waveforms. a, Key pecking accuracy of P109, P117, and P121 across sessions. b, Spike waveform samples after preprocessing.

Learning-related redistribution of VTA pooled MUA across task epochs during cue⁺ trials. a, Example cue⁺ rasters (left) and PSTHs (right) for pigeon P109 across sessions, aligned to cue onset (0.5 s) and reward delivery (1.0 s; dashed lines). Background shading indicates sessions assigned to the learning phase (blue) versus the consolidation phase (pink) based on behavioral performance. For each session, PSTHs are shown at two temporal resolutions (upper, 10 ms bins; lower, 100 ms bins with a 5-point moving average for visualization). b, Session-wise spike proportions for cue⁺ trials in the precue reference, cue⁺, and reward epochs for each pigeon (mean ± SD across valid trials). Statistical significance is indicated above bars (see Materials and Methods, Statistical analysis). c, Representative example from pigeon P109 illustrating how the session-wise mean spike proportion evolves across training. Each point corresponds to the same mean value shown for P109 in panel b (computed across valid trials within that session for the cue and reward epochs). The curve is provided as a visualization of the learning-related trajectory rather than a group-level summary. d, Session-to-session change in the same metric for P109, computed as Δ = value (Session n + 1) − value (Session n), where each value is the session-wise mean spike proportion (identical to the corresponding bar height in panel b).