Reinforcement is a term in operant conditioning and behavior analysis for a process of strengthening a directly measurable dimension of behavior--such as rate (e.g., pulling a lever more frequently), duration (e.g., pulling a lever for longer periods of time), magnitude (e.g., pulling a lever with greater force), or latency (e.g., pulling a lever more quickly following the onset of an environmental event)--as a function of the delivery of a "valued" stimulus (e.g. money from a slot machine) immediately or shortly after the occurrence of the behavior.
A reinforcer is a temporally contiguous environmental event, or an effect directly produced by a response (e.g., a musician playing a melody), that functions to strengthen or maintain the response that preceded the event. A reinforcer is demonstrated only if the strengthening or maintenance effect occurs.
Response strength is assessed by measuring the frequency, duration, latency, accuracy, and/or persistence of the response after reinforcement stops. Early experimental behavior analysts measured the rate of responses as a primary demonstration of learning and performance in non-humans (e.g., the number of times a pigeon pecks a key in a 10-minute session).
Contents |
B.F. Skinner, the researcher who articulated the major theoretical constructs of reinforcement and behaviorism, defined reinforcers according to the change in response strength rather than to more subjective criteria, such as what is pleasurable or valuable to someone. Accordingly, activities, foods or items considered pleasant or enjoyable may not necessarily be reinforcing (because they produce no increase in the response preceding them). Stimuli, settings, and activities only fit the definition of reinforcers if the behavior that immediately precedes the potential reinforcer increases in similar situations in the future, for example, a child who receives a cookie when he or she asks for one. If the frequency of "cookie-requesting behavior" increases, the cookie can be seen as reinforcing "cookie-requesting behavior." If however, "cookie-requesting behavior" does not increase the cookie cannot be considered reinforcing.
Reinforcement theory is one of the motivation theories; it states that reinforced behavior will be repeated, and behavior that is not reinforced is less likely to be repeated.[1]
The sole criterion that determines if an item, activity, or food is reinforcing is the change in probability of a behavior after administration of that potential reinforcer. Other theories may focus on additional factors such as whether the person expected the strategy to work at some point, but in the behavioral theory, reinforcement is descriptive of an increased probability of a response.
The study of reinforcement has produced an enormous body of reproducible experimental results. Reinforcement is the central concept and procedure in special education, applied behavior analysis, and the experimental analysis of behavior.
As Skinner discussed, positive reinforcement is superior to punishment in altering behavior. He maintained that punishment was not simply the opposite of positive reinforcement; positive reinforcement results in lasting behavioral modification, whereas punishment changes behavior only temporarily and presents many detrimental side effects.[2]
The accepted model of reinforcement began shifting in 1966 when Azrin and Holz contributed a chapter[3] to Honig's volume on operant conditioning. Skinner defined reinforcement as creating situations that a person likes or removing a situation he doesn't like, and punishment as removing a situation a person likes or setting up one he doesn't like.[2] Thus the distinction was based on the appetitive or aversive nature of the stimulus. Azrin and Holz defined punishment "as 'a reduction of the future probability of a specific response as a result of the immediate delivery of a stimulus for that response'."[4] This new definition of punishment encroached on Skinner's definition of reinforcement, but most textbooks now only present examples of the 1966 model summarized below:
Helpful definitions:
A positive reinforcer is a consequence that increases the frequency of a behavior or maintains the frequency. What is reinforcing is defined by what happens to the frequency of the behavior. It has nothing to do with whether the organism finds the reinforcer 'pleasant' or not. For example, if a child gets slapped whenever he/she says a 'naughty' word but the frequency of naughty words increases, the slap is a positive reinforcer.
A 'pleasant' consequence is not necessarily a positive reinforcer. Getting a birthday gift is not a positive reinforcer. There is no behavior that will increase (or be maintained) in frequency. When deciding whether or not something is a reinforcer or not, the basic criteria is whether there is a behavior that is increasing or being maintained in the frequency of its occurrence.
Consequences are not universally reinforcing. For example, happy face stickers may be effective reinforcers for some children. Other children may find them silly.
A negative reinforcer increases or maintains the frequency of a behavior. It is not punishment. These terms are often confused. A negative reinforcer increases or maintains the frequency of the behavior that terminates the negative reinforcer. In this case the negative reinforcer is present before the behavior. The organism performs a behavior that terminates the negative reinforcer. The behavior that terminates the negative reinforcer is likely to increase or be maintained in frequency. Suppose someone has a headache (negative reinforcer). The person takes two aspirin but nothing happens. Then the person takes two Tylenol tablets and the headache goes away. The next time the person has a headache it is likely the person will take Tylenol. That is the behavior that has been reinforced.
Forms of operant conditioning:
The following table shows the relationships between positive/negative reinforcements and increasing/decreasing required behavior.
| decreases likelihood of behavior | increases likelihood of behavior | |
|---|---|---|
| presented | positive punishment | positive reinforcement |
| taken away | negative punishment | negative reinforcement |
Distinguishing "positive" from "negative" can be difficult, especially when there are lots of consequences and the necessity of the distinction is often debated.[5] For example, in a very warm room, a current of external air serves as positive reinforcement because it is pleasantly cool or negative reinforcement because it removes uncomfortably hot air.[6] Some reinforcement can be simultaneously positive and negative, such as a drug addict taking drugs for the added euphoria and eliminating withdrawal symptoms. Many behavioral psychologists simply refer to reinforcement or punishment—without polarity—to cover all consequent environmental changes. Others would disagree with the above examples because there is no behavior that is increasing or decreasing in frequency.
A primary reinforcer, sometimes called an unconditioned reinforcer, is a stimulus that does not require pairing to function as a reinforcer and most likely has obtained this function through the evolution and its role in species' survival.[7] Examples of primary reinforcers include sleep, food, air, water, and sex. Some primary reinforcers, such as certain drugs, may mimic the effects of other primary reinforcers. While these primary reinforcers are fairly stable through life and across individuals, the reinforcing value of different primary reinforcers varies due to multiple factors (e.g., genetics, experience). Thus, one person may prefer one type of food while another abhors it. Or one person may eat lots of food while another eats very little. So even though food is a primary reinforcer for both individuals, the value of food as a reinforcer differs between them.
A secondary reinforcer, sometimes called a conditioned reinforcer, is a stimulus or situation that has acquired its function as a reinforcer after pairing with a stimulus that functions as a reinforcer. This stimulus may be a primary reinforcer or another conditioned reinforcer (such as money). An example of a secondary reinforcer would be the sound from a clicker, as used in clicker training. The sound of the clicker has been associated with praise or treats, and subsequently, the sound of the clicker may function as a reinforcer. As with primary reinforcers, an organism can experience satiation and deprivation with secondary reinforcers.
In his 1967 paper, Arbitrary and Natural Reinforcement, Charles Ferster proposed classifying reinforcement into events that increase frequency of an operant as a natural consequence of the behavior itself, and events that are presumed to affect frequency by their requirement of human mediation, such as in a token economy where subjects are "rewarded" for certain behavior with an arbitrary token of a negotiable value. In 1970, Baer and Wolf created a name for the use of natural reinforcers called "behavior traps."[10] A behavior trap requires only a simple response to enter the trap, yet once entered, the trap cannot be resisted in creating general behavior change. It is the use of a behavioral trap that increases a person's repertoire, by exposing them to the naturally occurring reinforcement of that behavior. Behavior traps have four characteristics:
As can be seen from the above, artificial reinforcement is in fact created to build or develop skills, and to generalize, it is important that either a behavior trap is introduced to "capture" the skill and utilize naturally occurring reinforcement to maintain or increase it. This behavior trap may simply be a social situation that will generally result from a specific behavior once it has met a certain criterion (e.g., if you use edible reinforcers to train a person to say hello and smile at people when they meet them, after that skill has been built up, the natural reinforcer of other people smiling, and having more friendly interactions will naturally reinforce the skill and the edibles can be faded).[12]
When an animal's surroundings are controlled, its behavior patterns after reinforcement become predictable, even for very complex behavior patterns. A schedule of reinforcement is a rule or program that determines how and when the occurrence of a response will be followed by the delivery of the reinforcer, and extinction, in which no response is reinforced. Schedules of reinforcement influence how an instrumental response is learned and how it is maintained by reinforcement. Between these extremes is intermittent or partial reinforcement where only some responses are reinforced.
Specific variations of intermittent reinforcement reliably induce specific patterns of response, irrespective of the species being investigated (including humans in some conditions). The orderliness and predictability of behavior under schedules of reinforcement was evidence for B.F. Skinner's claim that by using operant conditioning he could obtain "control over behavior," in a way that rendered the theoretical disputes of contemporary comparative psychology obsolete. The reliability of schedule control supported the idea that a radical behaviorist experimental analysis of behavior could be the foundation for a psychology that did not refer to mental or cognitive processes. The reliability of schedules also led to the development of applied behavior analysis as a means of controlling or altering behavior.
Many of the simpler possibilities, and some of the more complex ones, were investigated at great length by Skinner using pigeons, but new schedules continue to be defined and investigated.
Simple schedules have a single rule to determine when a single type of reinforcer is delivered for specific response.
Other simple schedules include:
Compound schedules combine two or more different simple schedules in some way using the same reinforcer for the same behavior. There are many possibilities; among those most often used are:
The psychology term, superimposed schedules of reinforcement, refers to a structure of rewards where two or more simple schedules of reinforcement operate simultaneously. Reinforcers can be positive, negative, or both. An example is a person who comes home after a long day at work. The behavior of opening the front door is rewarded by a big kiss on the lips by the person's spouse and a rip in the pants from the family dog jumping enthusiastically. Another example of superimposed schedules of reinforcement is a pigeon in an experimental cage pecking at a button. The pecks deliver a hopper of grain every 20th peck, and access to water after every 200 pecks.
Superimposed schedules of reinforcement are a type of compound schedule that evolved from the initial work on simple schedules of reinforcement by B.F. Skinner and his colleagues (Skinner and Ferster, 1957). They demonstrated that reinforcers could be delivered on schedules, and further that organisms behaved differently under different schedules. Rather than a reinforcer, such as food or water, being delivered every time as a consequence of some behavior, a reinforcer could be delivered after more than one instance of the behavior. For example, a pigeon may be required to peck a button switch ten times before food appears. This is a "ratio schedule." Also, a reinforcer could be delivered after an interval of time passed following a target behavior. An example is a rat that is given a food pellet immediately following the first response that occurs after two minutes has elapsed since the last lever press. This is called an "interval schedule." In addition, ratio schedules can deliver reinforcement following fixed or variable number of behaviors by the individual organism. Likewise, interval schedules can deliver reinforcement following fixed or variable intervals of time following a single response by the organism. Individual behaviors tend to generate response rates that differ based upon how the reinforcement schedule is created. Much subsequent research in many labs examined the effects on behaviors of scheduling reinforcers.
If an organism is offered the opportunity to choose between or among two or more simple schedules of reinforcement at the same time, the reinforcement structure is called a "concurrent schedule of reinforcement." Brechner (1974, 1977) introduced the concept of superimposed schedules of reinforcement in an attempt to create a laboratory analogy of social traps, such as when humans overharvest their fisheries or tear down their rainforests. Brechner created a situation where simple reinforcement schedules were superimposed upon each other. In other words, a single response or group of responses by an organism led to multiple consequences. Concurrent schedules of reinforcement can be thought of as "or" schedules, and superimposed schedules of reinforcement can be thought of as "and" schedules. Brechner and Linder (1981) and Brechner (1987) expanded the concept to describe how superimposed schedules and the social trap analogy could be used to analyze the way energy flows through systems.
Superimposed schedules of reinforcement have many real-world applications in addition to generating social traps. Many different human individual and social situations can be created by superimposing simple reinforcement schedules. For example a human being could have simultaneous tobacco and alcohol addictions. Even more complex situations can be created or simulated by superimposing two or more concurrent schedules. For example, a high school senior could have a choice between going to Stanford University or UCLA, and at the same time have the choice of going into the Army or the Air Force, and simultaneously the choice of taking a job with an internet company or a job with a software company. That is a reinforcement structure of three superimposed concurrent schedules of reinforcement. Superimposed schedules of reinforcement can create the three classic conflict situations (approach–approach conflict, approach–avoidance conflict, and avoidance–avoidance conflict) described by Kurt Lewin (1935) and can operationalize other Lewinian situations analyzed by his force field analysis. Other examples of the use of superimposed schedules of reinforcement as an analytical tool are its application to the contingencies of rent control (Brechner, 2003) and problem of toxic waste dumping in the Los Angeles County storm drain system (Brechner, 2010).
In operant conditioning, concurrent schedules of reinforcement are schedules of reinforcement that are simultaneously available to an animal subject or human participant, so that the subject or participant can respond on either schedule. For example, a pigeon in a Skinner box is faced with two pecking keys; pecking responses can be made on either, and food reinforcement might follow a peck on either. The schedules of reinforcement arranged for pecks on the two keys can be different. They may be independent, or they may be linked so that behavior on one key affects the likelihood of reinforcement on the other.
It is not necessary for responses on the two schedules to be physically distinct. In an alternate way of arranging concurrent schedules, introduced by Findley in 1958, both schedules are arranged on a single key or other response device, and the subject can respond on a second key to change between the schedules. In such a "Findley concurrent" procedure, a stimulus (e.g., the color of the main key) signals which schedule is in effect.
Concurrent schedules often induce rapid alternation between the keys. To prevent this, a "changeover delay" is commonly introduced: each schedule is inactivated for a brief period after the subject switches to it.
When both the concurrent schedules are variable intervals, a quantitative relationship known as the matching law is found between relative response rates in the two schedules and the relative reinforcement rates they deliver; this was first observed by R.J. Herrnstein in 1961. Matching law is a rule for instrumental behavior which states that the relative rate of responding on a particular response alternative equals the relative rate of reinforcement for that response (rate of behavior = rate of reinforcement). Animals and humans have a tendency to prefer choice in schedules.[16]
Shaping is reinforcement of successive approximations to a desired instrumental response. In training a rat to press a lever, for example, simply turning toward the lever is reinforced at first. Then, only turning and stepping toward it is reinforced. As training progresses, the response reinforced becomes progressively more like the desired behavior.
Chaining involves linking discrete behaviors together in a series, such that each result of each behavior is both the reinforcement (or consequence) for the previous behavior, and the stimuli (or antecedent) for the next behavior. There are many ways to teach chaining, such as forward chaining (starting from the first behavior in the chain), backwards chaining (starting from the last behavior) and total task chaining (in which the entire behavior is taught from beginning to end, rather than as a series of steps). An example is opening a locked door. First the key is inserted, then turned, then the door opened. Forward chaining would teach the subject first to insert the key. Once that task is mastered, they are told to insert the key, and taught to turn it. Once that task is mastered, they are told to perform the first two, then taught to open the door. Backwards chaining would involve the teacher first inserting and turning the key, and the subject is taught to open the door. Once that is learned, the teacher inserts the key, and the subject is taught to turn it, then opens the door as the next step. Finally, the subject is taught to insert the key, and they turn and open the door. Once the first step is mastered, the entire task has been taught. Total task chaining would involve teaching the entire task as a single series, prompting through all steps. Prompts are faded (reduced) at each step as they are mastered.
Persuasion is a form of human interaction. It takes place when one individual expects some particular response from one or more other individuals and deliberately sets out to secure the response through the use of communication. The communicator must realize that different groups have different values.[17]
In instrumental learning situations, which involve operant behavior, the persuasive communicator will present his message and then wait for the receiver to make a correct response. As soon as the receiver makes the response, the communicator will attempt to fix the response by some appropriate reward or reinforcement.[18]
In conditional learning situations, where there is respondent behavior, the communicator presents his message so as to elicit the response he wants from the receiver, and the stimulus that originally served to elicit the response then becomes the reinforcing or rewarding element in conditioning.[19]
A lot of work has been done in building a mathematical model of reinforcement. This model is known as MPR, short for mathematical principles of reinforcement. Killeen and Sitomer are among the key researchers in this field.
The standard definition of behavioral reinforcement has been criticized as circular, since it appears to argue that response strength is increased by reinforcement, and defines reinforcement as something that increases response strength (i.e., response strength is increased by things that increase response strength). However, the correct usage[20] of reinforcement is that something is a reinforcer because of its effect on behavior, and not the other way around. It becomes circular if one says that a particular stimulus strengthens behavior because it is a reinforcer, and does not explain why a stimulus is producing that effect on the behavior. Other definitions have been proposed, such as F.D. Sheffield's "consummatory behavior contingent on a response," but these are not broadly used in psychology.[21]
In the 1920s Russian physiologist Ivan Pavlov may have been the first to use the word reinforcement with respect to behavior, but (according to Dinsmoor) he used its approximate Russian cognate sparingly, and even then it referred to strengthening an already-learned but weakening response. He did not use it, as it is today, for selecting and strengthening new behaviors. Pavlov's introduction of the word extinction (in Russian) approximates today's psychological use.
In popular use, positive reinforcement is often used as a synonym for reward, with people (not behavior) thus being "reinforced," but this is contrary to the term's consistent technical usage, as it is a dimension of behavior, and not the person, which is strengthened. Negative reinforcement is often used by laypeople and even social scientists outside psychology as a synonym for punishment. This is contrary to modern technical use, but it was B.F. Skinner who first used it this way in his 1938 book. By 1953, however, he followed others in thus employing the word punishment, and he re-cast negative reinforcement for the removal of aversive stimuli.
There are some within the field of behavior analysis[5] who have suggested that the terms "positive" and "negative" constitute an unnecessary distinction in discussing reinforcement as it is often unclear whether stimuli are being removed or presented. For example, Iwata[6] poses the question: "...is a change in temperature more accurately characterized by the presentation of cold (heat) or the removal of heat (cold)?" (p. 363). Thus, you could conceptualize reinforcement as a pre-change condition replaced by a post-change condition that reinforces the behavior that followed the change in stimulus conditions.