Reinforcement schedules

The pattern by which reinforcement is delivered for a behaviour. Different schedules produce different learning outcomes, and the choice of schedule is one of the most useful tools in applied training.

Continuous reinforcement (reinforcement every time the behaviour occurs) is best for teaching new behaviours. The clear, immediate connection between behaviour and consequence allows the animal to form the association quickly. Most behaviours start their learning life on continuous reinforcement.

Variable reinforcement (reinforcement on an unpredictable proportion of correct responses) is best for maintaining established behaviours. Variable schedules produce behaviours that are highly resistant to extinction, which is why the principle underpins everything from slot-machine design (where the goal is to produce highly persistent gambling behaviour) to good clicker training (where the goal is to produce highly persistent trained behaviour).

The shift from continuous to variable reinforcement once a behaviour is established is one of the most useful tools in applied training across species, used in horse work, dog training, zoo husbandry training, and animal behaviour work. The transition is typically gradual, with the proportion of unreinforced responses increasing slowly so the animal does not abandon the behaviour during the transition.

There are several specific variable schedules with technical names from the behavioural science literature: variable ratio (a reinforcer arrives after a variable number of responses), variable interval (a reinforcer arrives after a variable amount of time), fixed ratio (reinforcement after every nth response), and fixed interval (reinforcement after a set time period). The detailed differences between these schedules matter mostly for research and analysis purposes; in practical training, the most important distinction is the basic continuous-versus-variable choice.