A survey of recent publications in behavioural biology suggests that research in this field often lacks a formal approach linking theory and empiricism, which implies that there is scope for improvement. Frequently, hypotheses guiding empirical work are merely based on plausibility arguments, which is not necessarily wrong and may be useful especially at the early stages of an investigation. Behavioural studies based on a theoretical framework usually test qualitative predictions from a general model either with data from natural situations or by experimental scrutiny. Rarely, a quantitative approach is used where a model is parameterized by existing data and therefore generates precise, testable predictions. In general, there is a lack of awareness that much more is to be learnt from a mismatch between predictions and data than from accordance. The latter is prone to cause the ‘pseudo-proof fallacy’, which is widespread in behavioural ecology. Behavioural physiology, on the other hand, often suffers a lack of proper theoretical models. Much can be gained in both fields if empiricists and theoreticians collaborate more closely towards the ultimate aim – to unravel the mechanisms of behaviour at both, ultimate and proximate levels.