The extension of a learned response from the specific stimulus it was trained with to similar stimuli the animal encounters.

A horse who has learned to load onto one float, and then loads onto another float without specific retraining, has generalised. A dog who has learned to sit at home, and then sits in a new environment, has generalised. A parrot who has learned to step onto one perch, and then steps onto a different perch on cue, has generalised.

Generalisation is one of the foundational processes within learning theory and applies across species. However, it is not automatic. Animals frequently learn responses that are tightly bound to the original training context and need explicit help to transfer the response to new settings. This is one of the most common training problems with established animals: the behaviour works perfectly in the original training environment and falls apart somewhere else.

The technical term for the opposite of generalisation is context-bound learning. An animal whose recall works perfectly in their own garden but not at the park has context-bound learning, not a recall problem per se. The fix is to systematically train the recall in many different contexts, rebuilding the behaviour from the foundations in each new setting, until the response has generalised.

Generalisation can also work too well, producing responses to stimuli the trainer did not intend. A horse who has been conditioned to flinch at one particular sound may generalise that flinching response to many similar sounds. A dog who has been conditioned to growl at one particular type of person may generalise that growling response to all similar-looking people. This can be a welfare concern when fear-based responses generalise widely.

The skill in producing useful generalisation is to train the behaviour with enough variety that the animal learns the underlying pattern (the behaviour, the cue, the consequence) rather than the specific surface features of any one training context.