30 6.2 Changing Behaviour Through Reinforcement and Punishment: Operant Conditioning

In classical conditioning, the organism learns to associate new stimuli with natural biological responses such as salivation or fear. The organism does not learn something new, but it begins to perform an existing behaviour in the presence of a new signal. Operant conditioning, on the other hand, is learning that occurs based on the consequences of behaviour and can involve the learning of new actions. Operant conditioning occurs when a dog rolls over on command because it has been praised for doing so in the past, when a schoolroom bully threatens classmates because doing so allows them to get their way, and when a child gets good grades because their parents threaten to punish them if they do not. In operant conditioning, the organism learns from the consequences of its own actions.

How reinforcement and punishment influence behaviour: The research of Thorndike and Skinner

Psychologist Edward Thorndike (1874–1949) was the first scientist to systematically study operant conditioning. Thorndike (1898) observed cats who had been placed in a “puzzle box” from which they tried to escape. At first, the cats scratched, bit, and swatted haphazardly, without any idea of how to get out. Eventually, and accidentally, they pressed the lever that opened the door and exited to their prize, which was a scrap of fish. The next time the cat was constrained within the box, it attempted fewer of the ineffective responses before carrying out the successful escape, and after several trials, the cat learned to almost immediately make the correct response.

Observing these changes in the cats’ behaviour led Thorndike to develop his law of effect, which is the principle that responses that create a typically pleasant outcome in a particular situation are more likely to occur again in a similar situation, whereas responses that produce a typically unpleasant outcome are less likely to occur again in the situation (Thorndike, 1911). The essence of the law of effect is that successful responses, because they are pleasurable, are “stamped in” by experience and thus occur more frequently. Unsuccessful responses, which produce unpleasant experiences, are “stamped out” and subsequently occur less frequently.

When Thorndike placed his cats in a puzzle box, he found that they learned to engage in the important escape behaviour faster after each trial. Thorndike described the learning that follows reinforcement in terms of the law of effect. The following YouTube link provides an example of this behaviour:

• Video: Thorndike-Puzzle Box (jenningh, 2007)

Influential behavioural psychologist B. F. Skinner (1904–1990) expanded on Thorndike’s ideas to develop a more complete set of principles to explain operant conditioning. Skinner created specially designed environments known as operant chambers — usually called Skinner boxes — to systematically study learning. A Skinner box (see Figure 6.5) is a structure that is big enough to fit a rodent or bird and that contains a bar or key that the organism can press or peck to release food or water. It also contains a device to record the animal’s responses.

The most basic of Skinner’s experiments was quite similar to Thorndike’s research with cats. A rat placed in the chamber reacted as one might expect, scurrying about the box and sniffing and clawing at the floor and walls. Eventually, the rat chanced upon a lever, which it pressed to release pellets of food. The next time around, the rat took a little less time to press the lever, and on successive trials, the time it took to press the lever became shorter and shorter. Soon, the rat was pressing the lever as fast as it could eat the food that appeared. As predicted by the law of effect, the rat had learned to repeat the action that brought about the food and cease the actions that did not.

Skinner studied, in detail, how animals changed their behaviour through reinforcement and punishment. As seen in the table below, Skinner developed terms that explained the processes of operant learning, using the term “reinforcer” to refer to any event that strengthens or increases the likelihood of a behaviour and the term “punisher” to refer to any event that weakens or decreases the likelihood of a behaviour. It is important to note that the terms positive and negative do not refer to good and bad respectively. Instead, positive means that something is added, and negative means that something is removed. Both reinforcement and punishment can be either positive or negative.

Thus, positive reinforcement strengthens a response by presenting something pleasant after the response, and negative reinforcement strengthens a response by reducing or removing something unpleasant. For example, giving a child praise for completing their homework represents positive reinforcement, whereas doing up your car’s seatbelt to eliminate the annoying warning sound represents negative reinforcement. In both cases, the reinforcement makes it more likely that behaviour will occur again in the future.

Reinforcement, either positive or negative, works by increasing the likelihood of a behaviour. Punishment, on the other hand, refers to any event that weakens or reduces the likelihood of a behaviour. Positive punishment weakens a response by presenting something unpleasant after the response, whereas negative punishment weakens a response by reducing or removing something pleasant. A child who is grounded after fighting with a sibling (i.e., positive punishment) or who loses out on the opportunity to go to recess after getting a poor grade (i.e., negative punishment) is less likely to repeat these behaviours.

Although the distinction between reinforcement, which increases behaviour, and punishment, which decreases it, is usually clear, in some cases it is difficult to determine whether a reinforcer is positive or negative. On a hot day, a cool breeze could be seen as a positive reinforcer, because it brings in cool air, or a negative reinforcer, because it removes hot air. In other cases, reinforcement can be both positive and negative. One may smoke a cigarette both because it brings pleasure (i.e., positive reinforcement) and because it eliminates the craving for nicotine (i.e., negative reinforcement).

It is also important to note that reinforcement and punishment are not simply opposites. The use of positive reinforcement in changing behaviour is almost always more effective than using punishment. This is because positive reinforcement makes the person or animal feel better, helping create a positive relationship with the person providing the reinforcement. Types of positive reinforcement that are effective in everyday life include verbal praise or approval, the awarding of status or prestige, and direct financial payment. Punishment, on the other hand, is more likely to create only temporary changes in behaviour because it is based on coercion and typically creates a negative and adversarial relationship with the person providing the reinforcement. When the person who provides the punishment leaves the situation, the unwanted behaviour is likely to return.

Creating complex behaviours through operant conditioning

Perhaps you remember watching a movie or being at a show in which an animal — maybe a dog, a horse, or a dolphin — did some pretty amazing things. The trainer gave a command, and the dolphin swam to the bottom of the pool, picked up a ring on its nose, jumped out of the water through a hoop in the air, dived again to the bottom of the pool, picked up another ring, and then took both of the rings to the trainer at the edge of the pool. The animal was trained to do the trick, and the principles of operant conditioning were used to train it. However, these complex behaviours are a far cry from the simple stimulus-response relationships that we have considered thus far. How can reinforcement be used to create complex behaviours such as these?

One way to expand the use of operant learning is to modify the schedule on which the reinforcement is applied. To this point, we have only discussed a continuous reinforcement schedule, in which the desired response is reinforced every time it occurs; whenever the dog rolls over, for instance, it gets a biscuit. Continuous reinforcement results in relatively fast learning but also rapid extinction of the desired behaviour once the reinforcer disappears. The problem is that because the organism is used to receiving the reinforcement after every behaviour, the responder may give up quickly when it does not appear.

Most real-world reinforcers are not continuous; they occur on a partial (or intermittent) reinforcement schedule, which is a schedule where the responses are sometimes reinforced and sometimes not. In comparison to continuous reinforcement, partial reinforcement schedules lead to slower initial learning, but they also lead to greater resistance to extinction. Because the reinforcement does not appear after every behaviour, it takes longer for the learner to determine that the reward is no longer coming; as such, extinction is slower. The four types of partial reinforcement schedules are summarized in the table below.

Partial reinforcement schedules are determined by whether the reinforcement is presented on the basis of the time that elapses between reinforcement (i.e., interval) or on the basis of the number of responses that the organism engages in (i.e., ratio) and by whether the reinforcement occurs on a regular (i.e., fixed) or unpredictable (i.e., variable) schedule. In a fixed-interval schedule, reinforcement occurs for the first response made after a specific amount of time has passed. For instance, on a one-minute fixed-interval schedule the animal receives a reinforcement every minute, assuming it engages in the behaviour at least once during the minute. Animals under fixed-interval schedules tend to slow down their responding immediately after the reinforcement but then increase the behaviour again as the time of the next reinforcement gets closer (see Figure 6.6). Consider, for example, a student cramming for an exam. On the other hand, in a variable-interval schedule, the reinforcers appear on an interval schedule, but the timing is varied around the average interval, making the actual appearance of the reinforcer unpredictable. An example might be checking your email: you are reinforced by receiving messages that come, on average, say, every 30 minutes, but the reinforcement occurs only at random times. Interval reinforcement schedules tend to produce slow and steady rates of responding.

In a fixed-ratio schedule, a behaviour is reinforced after a specific number of responses. For instance, a rat’s behaviour may be reinforced after it has pressed a key 20 times, or a salesperson may receive a bonus after they have sold 10 products. Once the organism has learned to act in accordance with the fixed-ratio schedule, it will pause only briefly when reinforcement occurs before returning to a high level of responsiveness. A variable-ratio schedule provides reinforcers after a specific but average number of responses. Winning money from slot machines or on a lottery ticket is an example of reinforcement that occurs on a variable-ratio schedule (see Figure 6.7). For instance, a slot machine may be programmed to provide a win every 20 times the user pulls the handle, on average. Ratio schedules tend to produce high rates of responding because reinforcement increases as the number of responses increases.

Complex behaviours are also created through shaping, which is the process of guiding an organism’s behaviour to the desired outcome through the use of successive approximation to a final desired behaviour. Skinner made extensive use of this procedure in his boxes. For instance, he could train a rat to press a bar two times to receive food, by first providing food when the animal moved near the bar. When that behaviour had been learned, Skinner would begin to provide food only when the rat touched the bar. Further shaping limited the reinforcement to only when the rat pressed the bar, to when it pressed the bar and touched it a second time, and finally to only when it pressed the bar twice. Although it can take a long time, in this way operant conditioning can create chains of behaviours that are reinforced only when they are completed.

Reinforcing animals if they correctly discriminate between similar stimuli allows scientists to test the animals’ ability to learn, and the discriminations that they can make are sometimes remarkable. Pigeons have been trained to distinguish between images of Charlie Brown and the other Peanuts characters (Cerella, 1980) and between different styles of music and art (Porter & Neuringer, 1984; Watanabe, Sakamoto & Wakita, 1995).

Behaviours can also be trained through the use of secondary reinforcers. Whereas a primary reinforcer includes stimuli that are naturally preferred or enjoyed by the organism, such as food, water, and relief from pain, a secondary reinforcer — sometimes called conditioned reinforcer — is a neutral event that has become associated with a primary reinforcer through classical conditioning. An example of a secondary reinforcer would be the whistle given by an animal trainer, which has been associated over time with the primary reinforcer (e.g., food). An example of an everyday secondary reinforcer is money. We enjoy having money, not so much for the stimulus itself, but rather for the primary reinforcers with which it is associated, that is to say, for the things that money can buy.

Image Attributions

Figure 6.5. Skinner Box Photo 02 by Benoit Denis is used under a CC BY-SA 3.0 license; Skinner Box Scheme 01 by Andreas1 is used under a CC BY-SA 3.0 license.

Figure 6.6. Used under a CC BY-NC-SA 4.0 license.

Figure 6.7. HardRockCasinoSlotMachines by Ted Murphy is used under a CC BY 2.0 license.

References

Cerella, J. (1980). The pigeon’s analysis of pictures. Pattern Recognition, 12, 1–6.

jenningh. (2007, November 11). Thorndike-puzzle box [Video file]. Retrieved from https://www.youtube.com/watch?v=BDujDOLre-8

Kassin, S. (2003). Essentials of psychology. Upper Saddle River, NJ: Prentice Hall.

Lapic, S. (2009, October 6). Positive reinforcement – The Big Bang Theory [Video file]. Retrieved from https://www.youtube.com/watch?v=JA96Fba-WHk

Porter, D., & Neuringer, A. (1984). Music discriminations by pigeons. Journal of Experimental Psychology: Animal Behavior Processes, 10(2), 138–148.

Thorndike, E. L. (1898). Animal intelligence: An experimental study of the associative processes in animals. Washington, DC: American Psychological Association.

Thorndike, E. L. (1911). Animal intelligence: Experimental studies. New York, NY: Macmillan.

Watanabe, S., Sakamoto, J., & Wakita, M. (1995). Pigeons’ discrimination of painting by Monet and Picasso. Journal of the Experimental Analysis of Behaviour, 63(2), 165–174.