Learning relationships between cues in our environment enables us to recognise common underlying structures of events and is thought to form the basis of episodic memory. This type of learning - often called relational, hierarchical or structured learning - requires retaining both the stimulus identity as well as its relation to other cues, for example their relative order in time or location in space. Despite being essential for many of our everyday decisions, relational learning cannot be performed by classic reinforcement learning algorithms, and there is limited insight into how it is achieved within the brain. One area implicated in relational learning is the hippocampus. Specifically, neurons in the CA1 area of the hippocampus have been shown to represent variables that are essential for constructing a relational structure of the environment, such as cue configurations, the value of such configurations and their order in space and time. To investigate the neural mechanisms of relational learning, we designed an odour sequence task for mice that requires subjects to learn about both the identity of an odour and its temporal position within a sequence. Importantly, the task design allows for manipulation of the temporal structure and value of cues separately. Using this ability allowed us to probe generalisation to novel cues in the same temporal structure, and the ability to rapidly update the value associated with learnt relational structure. We found that mice quickly learnt this task and, in line with a key role for hippocampal circuitry, optogenetic inactivation of CA1 markedly impaired task performance. In addition, we found that after initial learning, mice could rapidly adapt to manipulations of cue value and identity, suggesting flexible use of previously learnt relational structures. Ongoing work aims to understand how the role of CA1 in these computations differs along the dorso-ventral axis of the hippocampus.