Strong Reward Signals, Weak Transfer: Limits of Spatial Priority Map Plasticity Across Task Contexts
Reward learning can bias attentional selection, but whether spatially biased reinforcement produces durable, context-general changes in spatial priority over days, and what neurophysiological signals track such learning, remains uncertain. We combined electroencephalography (EEG) and pupillometry with a multi-session spatial reward-learning paradigm (Chelazzi et al., 2014) in which targets could appear at eight locations and reward probability was systematically biased across locations during two days of training. A separate baseline/test visual detection task was administered before training and again four days after training to assess delayed transfer under cross-target competition. Training produced strong reward signals across measures. Feedback-locked ERPs (FRN and P300) differentiated outcome valence and reward magnitude and varied systematically with time-on-task, while pupil dilation was larger following high- than low-reward feedback and overall task-evoked responses decreased across blocks. Reward history also modulated stimulus-locked target processing: targets at high- versus low-reward locations differed reliably across N1/N2 and a late positivity, indicating multi-stage value-dependent influences on visual processing during active learning. In contrast, transfer was weak in both behavior and ERPs: behavioral indices did not show a reliable advantage for highly rewarded locations at delayed test, and neural evidence for persistence was limited to an N2 modulation in the most diagnostic high-versus-low comparison, which should be interpreted cautiously given low trial counts. Accordingly, we did not replicate the robust long-term spatial priority effect reported in the original study. Together, these findings reveal strong reward-learning signals but weak cross-task transfer, suggesting limits on how readily spatial reward learning consolidates into persistent, task-general spatial priority-map plasticity across contexts.