Animal Training by Operant Conditioning
Using Positive Reinforcement to Train Animal Subjects
May 22, 2009
Classical conditioning is associative learning, also known as Pavlovian learning after the most famous example by Ivan Pavlov (1849-1936) whereby dogs were ‘conditioned’ to salivate when a bell was rung, after the ringing of a bell was paired with the arrival of food. It is ‘involuntary’ behavioral responses that are the cause of classical conditioning.
Operant conditioning is the way in which behavior is shaped by consequences (both positive and negative). Operant conditioning was discovered by Burrhus Frederic Skinner (1904 – 1990). Skinner studied the way in which behavior was shaped when some responses were reinforced, some were punished, and others had no effect i.e. were ignored. He found that responses which were reinforced would become more frequent, whilst those that were punished, or had no effect, would become less frequent and eventually disappear.
Which Technique?
- Positive reinforcement: A positive event occurs, such as the delivery of a treat, when a certain behaviour is performed.
- Negative reinforcement: A negative event ceases, such as an unpleasant sound stopping, when a certain behavior is performed.
- Punishment: A negative event increases, such as a mild electric shock being delivered, when a certain behavior is performed.
Negative reinforcement is often confused with punishment but they are very different:
- The goal of negative reinforcement is to strengthen, or increase, the probability of a behaviour occurring because a negative event stops.
- The goal of punishment however, is to weaken, or decrease, the probability of a behaviour occurring because a negative event is introduced.
Reinforcement techniques shape behaviors as they occur. Punishment stops a behavior at that particular moment with no further guidance, and as a result other undesirable behaviors may occur in its place. Although different, both punishment and negative reinforcement are considered aversive approaches to animal training.
As well as the problem of undesirable behaviors taking the place of the one being stopped with punishment, another problem with punishment is that the subject(s) being trained may start to avoid the ‘punisher’. With respect to these problems and regard for the welfare of animals, positive reinforcement should be used as the prominent method of training animals.
Schedules of Reinforcement
Once a behavior has been established not every response has to be reinforced. There are different schedules that can be used for reinforcing a behavior once established:
- Continuous Reinforcement – every response is continued to be reinforced.
- Fixed Ratio Reinforcement – every nth response is reinforced.
- Fixed Interval Reinforcement – the first response after a set timed interval (t secs/mins/hours) is reinforced.
- Variable Ratio – on average every nth response is reinforced.
- Variable Interval – the first response after an average interval of time (t secs/mins/hours) is reinforced.
Behavior that is rewarded on variable schedules is sometimes more resistant to extinction as the subject will keep performing the behaviour to try and get the reward/reinforcement.
Effective Reinforcement
Reinforcement must be immediate and consistent for to be effective. If it is delivered too late the subject may already have moved on to a different behavior and be reinforced for that behavior instead. It is good to use secondary reinforcers to bridge the gap between the behavior occurring, and the subject receiving the stimulus, or reward. Secondary reinforcers commonly used are clickers, whistles and praise. These are delivered immediately after the behavior has been performed and then the reward (food, temperature change etc.) can be delivered once the target behavior has been marked.
Some final pointers when training:
- Your subject can pick up behaviours you didn’t intend them to.
- Your subject can also train you back… “If I sit here long enough you’ll give me that treat anyway!”
