The nucleus accumbens shell (NAcs) is central to reward-guided behaviour, yet how it encodes reward magnitude across different learning modes and training stages remains unclear. Using fibre photometry, we recorded calcium activity in NAcs neurons of rats during Pavlovian and operant tasks with small and large rewards in early and late training. NAcs activity consistently reflected reward magnitude after reward receipt, but the timing and form of this signal depended on task and experience. In Pavlovian conditioning, cue-evoked responses initially increased but shifted to magnitude-dependent inhibition with extended training. In operant tasks, activity ramped before actions regardless of reward size, while post-reward inhibition was stronger and longer after large rewards. These findings demonstrate that NAcs neurons encode reward magnitude primarily at the point of reward receipt, with limited cue-related encoding, and that this encoding adapts with both learning type and training duration. This work identifies dynamic, context-specific roles of the NAcs in shaping reward-driven behaviour.

Individual differences in cognition have been historically recognized in humans, and recent evidence suggests that such variation is widespread across the animal kingdom. Despite its importance for individuals' behaviour and fitness, the biological roots of cognitive variation remain poorly understood. We hypothesize that variation in brain gene expression is important in determining individual cognitive differences. To test this, we focused on 6 neural plasticity genes and examined fish, which exhibit the largest cognitive variation reported for vertebrates. Zebrafish (Danio rerio) exposed to visual discrimination tasks showed substantial variation in their performance, with some learning over 7 times faster than others. Expression of two genes positively predicted learning performance. However, expression levels of most genes were related at the individual level, suggesting that multi-gene expression patterns may be more relevant than single gene variation. Principal component analysis identified two axes of multi-gene expression variation: the first loaded by all genes except neurotrophin bdnf, the second mainly loaded by bdnf and neurod1 expression. Only the latter component significantly predicted learning performance in a visual discrimination task, indicating that individual variation in bdnf expression and with lesser extend neurod1 are critical for learning. Our study bridges the gap between cognitive differences and molecular mechanisms underlying brain function, providing foundation for new understanding what makes individual unique.