To behave optimally, animals build models of the world based on structural regularities to allow accurate prediction of the value of actions. Although powerful, the computational expense of updating these models limits their usefulness in ever-changing environments. Theoretical models have suggested that a way around these limitations is to separate the representation of structure from the associated value, allowing values to be rapidly updated when necessary. However, it is still unclear how this is implemented in the brain. To address this, we developed an odour-based task that requires mice to learn about the temporal relationship between sequentially presented odour cues, that allows independent manipulation of both the transition statistics between odours, as well as the associated value. We show that mice can rapidly learn this task and that dopamine release in the nucleus accumbens reflect a value signal consistent with the use of the temporal task structure rather than individual odours. We altered the value associated with individual odour sequences and found that mice rapidly adapt their anticipatory behaviour to reflect this change. Additionally, we found that the transition statistics between odours can be manipulated with little change to performance. Together, this task provides an ideal framework to investigate how value and structure are represented and used to support behaviour. Our ongoing work is using this task with in vivo calcium imaging and optogenetic manipulations to delineate how structure and value is represented in the hippocampus, a key unanswered question relevant to computational models of hippocampal function.