||This article needs additional citations for verification. (July 2009)|
Extinction is observed in both operantly conditioned and classically conditioned behavior. When operant behavior that has been previously reinforced no longer produces reinforcing consequences the behavior gradually stops occurring. In classical conditioning, when a conditioned stimulus is presented alone, so that it no longer predicts the coming of the unconditioned stimulus, conditioned responding gradually stops. (For example, after Pavlov's dog was conditioned to salivate at the sound of a bell, it eventually stopped salivating to the bell after the bell had been sounded repeatedly but no food came.)
Fear conditioning 
Extinction is typically studied within the Pavlovian fear conditioning framework in which extinction refers to the reduction in a conditioned response (CR; e.g., fear response/freezing) when a conditioned stimulus (CS; e.g., neutral stimulus/light or tone) is repeatedly presented in the absence of the unconditioned stimulus (US; e.g., foot shock/loud noise) with which it has been previously paired.
The simplest explanation of extinction is that as the CS is presented without the aversive US, the animal gradually "unlearns" the CS–US association which is known as the associative loss theory. However, this explanation is complicated by observations where there is some fear restoration, such as reinstatement (restoration of CR in the context where extinction training occurred but not a different context after aversive US is presented again), renewal (restoration of CR in context A but not in B when learning occurred in context A and extinction in context B), and spontaneous recovery (restoration of CR when the retention test occurs after a long but not a short delay after extinction training) and alternative explanations have been offered.
The dominant account of extinction involves associative models. However, there is debate over whether extinction involves simply "unlearning" the US–CS association (e.g., the Rescorla–Wagner account) or, alternatively, a "new learning" of an inhibitory association that masks the original excitatory association (e.g., Konorski, Pearce and Hall account). A third account concerns non-associative mechanisms such as habituation, modulation and response fatigue. Myers and Davis laboratory work with fear extinction in rodents has suggested that multiple mechanisms may be at work depending on the timing and circumstances in which the extinction occurs.
Given the competing views and difficult observations for the various accounts researchers have turned to investigations at the cellular level (most often in rodents) to tease apart the specific brain mechanisms of extinction, in particular the role of the brain structures (amygdala, hippocampus, the prefontal cortex), and specific neurotransmitter systems (e.g., GABA, NMDA). A recent study in rodents by Amano, Unal and Paré published in Nature Neuroscience found that extinction is correlated with synaptic inhibition in the fear output neurons of the central amygdala that project to the periaqueductal gray that controls freezing behavior. They infer that inhibition derives from the prefrontal cortex and suggest promising targets at the cellular level for new treatments of anxiety.
Operant conditioning 
In the operant conditioning paradigm, extinction refers to the decline of an operant response when it is no longer reinforced in the presence of its discriminative stimulus. Extinction is observed after withholding of reinforcement for a previously reinforced behavior which decreases the future probability of that behavior. For example, a child who climbs under his desk, a response which has been reinforced by attention, is subsequently ignored until the attention-seeking behavior no longer occurs. In his autobiography, B.F. Skinner noted how he accidentally discovered the extinction of an operant response due to the malfunction of his laboratory equipment:
My first extinction curve showed up by accident. A rat was pressing the lever in an experiment on satiation when the pellet dispenser jammed. I was not there at the time, and when I returned I found a beautiful curve. The rat had gone on pressing although no pellets were received. ... The change was more orderly than the extinction of a salivary reflex in Pavlov's setting, and I was terribly excited. It was a Friday afternoon and there was no one in the laboratory who I could tell. All that weekend I crossed streets with particular care and avoided all unnecessary risks to protect my discovery from loss through my accidental death.
When the extinction of a response has occurred, the discriminative stimulus is then known as an extinction stimulus (SΔ or S-delta). When an S-delta is present, the reinforcing consequence which characteristically follows a behavior does not occur. This is the opposite of a discriminative stimulus which is a signal that reinforcement will occur. For instance, in an operant chamber, if food pellets are only delivered when a response is emitted in the presence of a green light, the green light is a discriminative stimulus. If when a red light is present food will not be delivered, then the red light is an extinction stimulus (food here is used as an example of a reinforcer).
Successful extinction procedures 
In order for extinction to work effectively, it must be done consistently. Extinction is considered successful when responding in the presence of an extinction stimulus (a red light or a teacher not giving a bad student attention, for instance) is zero. When a behavior reappears again after it has gone through extinction, it is called resurgence.
Extinction burst 
While extinction, when implemented consistently over time, results in the eventual decrease of the undesired behavior, in the short-term the subject might exhibit what is called an extinction burst. An extinction burst will often occur when the extinction procedure has just begun. This usually consists of a sudden and temporary increase in the response's frequency, followed by the eventual decline and extinction of the behavior targeted for elimination. Novel behavior, or emotional responses or aggressive behavior, may also occur.
Take, as an example, a pigeon that has been reinforced to peck an electronic button. During its training history, every time the pigeon pecked the button, it will have received a small amount of bird seed as a reinforcer. So, whenever the bird is hungry, it will peck the button to receive food. However, if the button were to be turned off, the hungry pigeon will first try pecking the button just as it has in the past. When no food is forthcoming, the bird will likely try again ... and again, and again. After a period of frantic activity, in which their pecking behavior yields no result, the pigeon's pecking will decrease in frequency.
Although not explained by reinforcement theory, the extinction burst can be understood using control theory. In perceptual control theory, the degree of output involved in any action is proportional to the discrepancy between the reference value (desired rate of reward in the operant paradigm) and the current input. Thus, when reward is removed, the discrepancy increases, and the output is increased. In the long term, 'reorganisation', the learning algorithm of control theory, would adapt the control system such that output is reduced.
The evolutionary advantage of this extinction burst is clear. In a natural environment, an animal that persists in a learned behavior, despite not resulting in immediate reinforcement, might still have a chance of producing reinforcing consequences if the animal tries again. This animal would be at an advantage over another animal that gives up too easily.
Despite the name, however, not every explosive reaction to adverse stimuli subsides to extinction. Indeed a small minority of individuals persist in their reaction indefinitely.
Extinction-induced variability 
Extinction-induced variability serves an adaptive role similar to the extinction burst. When extinction begins, subjects can exhibit variations in response topography (the movements involved in the response). Response topography is always somewhat variable due to differences in environment or idiosyncratic causes but normally a subject's history of reinforcement keeps slight variations stable by maintaining successful variations over less successful variations. Extinction can increase these variations significantly as the subject attempts to acquire the reinforcement that previous behaviors produced. If a person attempts to open a door by turning the knob, but is unsuccessful, they may next try jiggling the knob, pushing on the frame, knocking on the door or other behaviors to get the door to open. Extinction-induced variability can be used in shaping to reduce problematic behaviors by reinforcing desirable behaviors produced by extinction-induced variability.
D-Cycloserine (DCS) is being trialed as an adjuvant to conventional exposure-based treatments for anxiety disorders. The psychotropic responses are related to D-Cycloserine's action as a partial agonist of the neuronal NMDA receptor for glutamate and have been examined in implications with sensory-related fear extinction in the amygdala.
See also 
- Miltenberger, R. (2012). Behavior modification, principles and procedures. (5th ed., pp. 87-99). Wadsworth Publishing Company.
- Myers & Davis (2007) Mechanisms of Fear Extinction. Molecular Psychiatry, 12, 120–150.
- Amano T, Unal CT, Paré D. (2010). Synaptic correlates of fear extinction in the amygdala. Nature Neuroscience 13: 489–494 doi:10.1038/nn.2499 PMID 20208529
- B.F. Skinner (1979). The Shaping of a Behaviorist: Part Two of an Autobiography, p. 95.