While changing on many timescales, the world around us looks rather still at any point in time because how our perception works. To be successful, however, living organisms should be able to adapt their behavioural responses according to timescales of regularities and changes in the environment. Identifying these timescales can be beneficial in at least two ways: (1) to adopt a proper response; for example, foraging at the right time of the day; (2) to use deviation from what is predicted to detect drastic environmental changes to adjust response, for example, by moving to a new location when the environment is no longer rewarding. Importantly, the brain possesses multiple mechanisms for responding to and processing information at different timescales, allowing us to learn and predict regularities of the environment and to detect its changes. But is it possible to benefit from both unboundedly?

In this talk, I argue that this is not possible due to a general tradeoff in learning: precision in learning regularities of the environment limits flexibility in adjusting what is learned and vice versa. On a brighter note, I will share some of our findings about different sources of flexibility in learning and choice behaviour. Specifically, I will use computational methods and experimental data across multiple species to address: (1) how is the amount or speed of learning determined and adjusted? (2) how are learning strategies determined and adjusted? (3) how are different sources of information combined to allow flexibility? and (4) how are such adjustments reflected in timescales of neural responses? Addressing these questions is more important than ever when facing multiple global challenges, because adaptability is what made us a successful species and can save us too.