Experimental extinction

1. A decline in the frequency or intensity of a conditioned behaviour following the withdrawal of *reinforcement.

2. The experimental protocol used to obtain the aforementioned phenomenon.

3. A modification in the *internal representation of a conditioned association that leads to suppression of the conditioned response, due to behaviourally meaningful rearrangements in the relationship among previously associated *stimuli or stimuli and *reinforcers.

Pavlov noted that once one of his famous dogs (and there were many) had been conditioned to salivate to the sound of a metronome by associating the sound with food, the sound alone came to elicit salivation, as expected ("classical conditioning). However, when the metronome continued to be played without subsequent reward, with time salivation diminished. Pavlov termed this phenomenon 'experimental extinction' (Pavlov 1927). Ever since, experimental extinction (which can easily be demonstrated in "instrumental conditioning as well) became one of the most fundamental problems in learning theory. A naive, "Ockham's razor type of explanation, might construe the situation as rather simple: the dog in the above example forms an association between sound and food, but when the sound is sounded without reinforcement, the association weakens and the dog ultimately "forgets. A simple explanation? Probably, but absolutely wrong. A selection of four types of observations will suffice to illustrate the case.

1. Spontaneous recovery. After extinction has occurred, the conditioned response may recover with time without any additional training (Pavlov 1927). This immediately suggests that the original information was preserved. Pavlov himself proposed that in extinction, the conditioned reflex undergoes a process of 'internal inhibition'. To put his proposal in a broader context, we should note that he distinguished central excitatory processes, commanding reflexes, from central inhibitory processes, that act to negate the excitatory processes. Furthermore, he distinguished two types of inhibitory processes, one 'direct' or 'internal', the other 'indirect' or 'external'. 'External' inhibition arises when there is a clash with another excitatory processes in the brain, whereas in 'internal inhibition' the excitatory conditioned reflex itself becomes progressively inhibitory for the behaviour in question.

2. Reacquisition and saving. Generally speaking, 'saving' is facilitation of relearning in retraining (in the 'saving method', retention is quantified by comparing learning and relearning scores; Ebbinghaus 1885). In many cases reacquisition of an extinguished behaviour takes fewer trials than the original training (although see Bouton 1993). Such cases imply that the brain saves more information about the original learning than it volunteers to give up at first.

3. Reinstatement. After experimental extinction of conditioning, exposure to the unconditioned stimulus alone can partially restore the response to the conditioned stimulus (see "fear conditioning in Rescorla and Heth 1975). This again indicates that the conditioned association was not abolished.

4. Renewal. Switching out of the extinction "context could result in re-emergence of the seemingly extinguished response (Bouton and Swartzentruber 1991).

Over the years, the phenomenology of experimental extinction has provided lots of excitement on the one hand, and insights on the other, to experimental psychologists and learning theorists. An additional notable example is the so-called 'partial reinforcement extinction effect' (PREE): behaviour that has been reinforced only intermittently extinguishes more slowly than behaviour that has been reinforced consistently (Humphreys 1939; Weinstock 1954; Amsel 1958; Wagner et al. 1964; Rescorla 1999; Figure 28). This seems rather paradoxical, for why should less reinforcement lead to more memory? Several theories have been proposed to account for PREE. Among the most influential ones are the 'Frustration Hypothesis' (Amsel 1958) and the 'Sequential Hypothesis' (Capaldi 1966). The 'Frustration Hypothesis' argues that PREE works because lack of reward creates 'frustration' (note the "anthropomorphism), and as animals trained under a partial reinforcement schedule got accustomed to frustration, they become more stubborn in extinction training. The 'Sequential Hypothesis' proposes that during partial reinforcement regimes, persisting stimulus traces from the nonreinforced trials become conditioned to the next reinforced response provided there is an appropriate nonreinforced-reinforced alternation schedule, and the memory of these associations maintains responding during extinction. It is thus seen that explanations offered for the phenomena of PREE reflect on fundamental issues in learning theory, such as how associations are reinforced and "acquired, retained, and "retrieved.

Indeed, there is no shortage of hypotheses of why brains behave the way they do in experimental extinction. Many of these hypotheses refer to learning and memory processes that transcend issues of extinction (e.g. Bower and Hilgard 1981; Mackintosh 1983; Flaherty 1985; Bouton 1991). In the frame of the present discussion, suffice it to note that experimental extinction can be viewed as a dynamic learning process, in which "internal representations of target and "context stimuli are rearranged (definition 3). Experimental extinction is hence different from forgetting; it involves a "phase of memory "consolidation for the new extinction experience (Braud and Broussard 1973; Berman and Dudai 2001), and could itself be forgotten (Bouton 1994). Seen this way, experimental extinction provides additional support to the "zeitgeist that classical and instrumental conditioning are not at all solely about the contiguity or contingency of pairs of isolated stimuli; rather these types of learning reflect intricate processes

Group I □—o

Group II •—•

Group III o—o







Extinction trials

Fig. 28 An early example of the puzzling 'Partial Reinforcement Extinction Effect' (PREE): college students were divided into three groups and trained to blink their eyelid to a light because this light was followed by air puff to the cornea. Group I received in training 96 reinforced trails, group II 96 trials (only half of which were reinforced), and group III only 48 reinforced trials. All groups were then subjected to an experimental extinction protocol. The graph depicts the frequency of conditioned response vs. the extinction trial. The naive prediction is that group I will show the strongest acquisition and extinguish most slowly, whereas group II will show weaker acquisition. In reality, all groups showed similar acquisition, while group II was definitely the most resistant to extinction. For hypothetical explanations, see text. (From Humphreys 1939.)

of information processing, only limited facets of which are unveiled under a given experimental situation in the laboratory (for examples see "classical conditioning and "cue). Furthermore, 'experimental extinction' tells us that when we relearn, we preserve the information about previous associations (e.g. Rescorla 1996); hence we constantly create mental "palimpsests.

'Experimental extinction' offers additional take-home messages. Being so widespread, it epitomizes the fact that some general principles and "algorithms are shared by many forms of learning. These learning 'universals' are expected to have selective advantages. Consider, for example, PREE; nothing in life is sure, therefore robust "performance under conditions of uncertainty makes sense. Experimental extinction also illustrates how a phenomenon that at first might appear merely of theoretical interest to a small group of scientists, ultimately becomes important in "real-life. For many years now there is much interest in experimental extinction in psychotherapy, in the treatment of post-traumatic stress syndrome and prevention of its relapse (Bouton and Swartzentruber 1991;Charney et al. 1998).1 The interest in brain mechanisms of extinction is on the rise as well (Falls et al. 1992; LaBar et al. 1998; Berman et al. 2000). One of the goals is to understand the neuronal mechanisms that differentiate learning the new (i.e. in acquisition of the "engram) from learning anew (i.e. in experimental extinction). This could tell us in due time how to prevent corruption of important learned information by subsequent experience ("false memory).

Selected associations: Associative learning, Consolidation, Forgetting, Persistence

1As the clinic is mentioned, it would be only fair to note that the term 'extinction' carries also a different meaning in neuropsychology. It refers to the situation in which the simultaneous application of two identical tactile stimuli results in the report of only one of the stimuli, although each would have been reported if they were to be presented independently. This type of 'extinction' may indicate a lesion in somatosensory cortex. It is not 'experimental' and has nothing to do with memory.

0 0


  • Kristin Kruger
    When classical conditioning extinction relearning?
    7 months ago

Post a comment