Classical conditioning

1. Types of "associative learning in which the "subject learns that one "stimulus predicts another.

2. Types of training procedures in which two stimuli, the conditioned stimulus (CS) and the unconditioned stimulus (US), are paired with each other, so that the CS comes to evoke a conditioned response (CR), which is similar to the unconditioned response (UR) elicited by the US.

Pavlov and his dogs are probably the first association that comes to mind in most people when prompted to contemplate experiments on animal learning. Indeed, a century after its formalization (Pavlov 1906), the *paradigm of classical conditioning1 (alias Pavlov-ian conditioning, conditioned reflex type I, respondent conditioning, type S(timulus) conditioning) is still a cornerstone of learning research. Following the pioneering work of Sechenov (1862), the father figure of Russian physiology, Pavlov (Figure 13) led a systematic attempt to *reduce the study of higher brain function to that of quantifiable atoms of reflexive behaviour. In doing so he relied mainly on gustatory stimuli as the US, gustatory reflexes as the UR, and auditory or visual stimuli as the CS (Pavlov 1927). Since then, Pavlovian conditioning became *classical. In the past three decades, it has accumulated renewed momentum, due to the impressive developments in behavioural, cognitive,

Fig. 13 Pavlov, assistant, dog, and Pavlov. The heavy sticks (left) are thrown at selected targets from a distance in the Russian game 'Gorodki' ('small towns', or 'Russian pyramids'), on which Pavlov was a very proud expert.The dog seems a bit tired of being classically conditioned, but this might merely be an 'anthropomorphic interpretation. (Courtesy of K.Anokhin and P. Balaban, Moscow.)

Fig. 13 Pavlov, assistant, dog, and Pavlov. The heavy sticks (left) are thrown at selected targets from a distance in the Russian game 'Gorodki' ('small towns', or 'Russian pyramids'), on which Pavlov was a very proud expert.The dog seems a bit tired of being classically conditioned, but this might merely be an 'anthropomorphic interpretation. (Courtesy of K.Anokhin and P. Balaban, Moscow.)

"system, and cellular neuroscience (Holland 1993; Bao et al. 1998; Kim et al. 1998; Pearce and Bouton 2001).

In classical conditioning, the subject learns relations among stimuli (definition 1). This is different from "instrumental conditioning, in which the subject learns relations among actions and their outcome.2 The behaviour of the experimenter in classical conditioning (definition 2) is almost as important as the behaviour of the experimental subject: it is the investigator who chooses and manipulates the CS and the US, and selects the CR from the overall behavioural response of the subject.3 In 'true' classical conditioning, the CR is supposed to be a novel response to the CS, but probably more often than realized, the CR is an intensification of a pre-existing (innate, "a priori) response to the CS. The latter case is termed a-conditioning. If random pairing of the CS and US , or presentation of the CS alone, elicit with time a response similar to the CR, then conditioning is by definition nonassociative. Nonassociative elici-tation of a CR to a previously neutral CS is termed pseudoconditioning. Nonassociative a-conditioning is "sensitization. In some protocols of classical conditioning, the CS, after coming to elicit a CR, is used as a US in a subsequent phase of conditioning. This is termed second-order conditioning (Rescorla 1980); higher-order conditioning can be similarly obtained.

In the first "phases of its investigation, ample attention has been devoted to the role in classical conditioning of the timing and the order of the presentation of the CS and US. This emerged from the observation that, whereas certain conditioning protocols culminate in successful learning, others do not (e.g. Pavlov 1927; Konorski 1948; Mackintosh 1983; Figure 14). An effective protocol is delay conditioning, in which the CS is presented first and the US onset precedes the CS offset or coincides with it. Trace conditioning, in which the CS starts and ends before the US starts, is also effective, provided that the interstimulus interval is kept sufficiently short (usually in the subsecond to the second range; but see a striking exception in "conditioned taste aversion). Simultaneous conditioning, in which the CS and US onset coincide in time, is less effective, unless the CS offset precedes the US offset. Backward conditioning, in which the US onset precedes the CS onset, and the US terminates before the CS, is usually ineffective. It can therefore be concluded that order dependency and temporal contiguity are critical factors in ensuring effective conditioning. Contiguity was noted already by Aristotle as important in recollection, and marked as a condition for association in early associative psychology in the seventeenth and eighteenth century (Warren 1921; "associative learning). Classical conditioning permitted, however, for the first time, a systematic, experimental analysis of the temporal parameters involved, in a variety of species. Reductionist analysis has ultimately chased "coincidence

Fig. 14 Protocols of classical conditioning and their effectiveness. (a) Delay conditioning, which is usually successful. (b) Trace conditioning, which is usually successful provided the interval between the offset of the CS and the onset of the US is kept sufficiently short (but see Conditioned taste aversion for a striking exception). (c) Simultaneous conditioning, which is not an effective procedure. (d) Backward conditioning, which is usually unsuccessful.

detectors, that could implement contiguity, down to the molecular "level (Dudai 1985; Abrams and Kandel 1988; Bourne and Nicoll 1993).

As much as order dependency and temporal contiguity are important, more sophisticated conditioning protocols have demonstrated that temporal contiguity per se is insufficient, and the emphasis shifted to the role of contingency (Rescorla 1968,1988; Mackintosh 1983; Wasserman and Miller 1997). Contingency in this context means comparison of the probability of the occurrence of the US in the presence of the CS as contrasted with the probability of the occurrence of the US in the absence of the CS; conditioning is assumed to occur only when the aforesaid probabilities differ (Rescorla 1968, 1988). Still, contingency per se may also not suffice to account for all the facets of conditioning (Mackintosh 1983; Rescorla 1988; Papini and Bitterman 1990; Wasserman and Miller 1997). So far, in spite of a number of sophisticated theories of associations, there is no accepted unified theory covering all manifestations and properties of classical conditioning (Pearce and Bouton 2001). There might never be, because classical conditioning encompasses multiple training protocols, behavioural phenomena, neuronal circuits and possibly mechanisms ("algorithm). Compare, for example, trace conditioning to delay conditioning (Figure 14); whereas in delay conditioning the CS and US overlap on-line part of the time, in trace conditioning, the subject must hold off-line information about the CS before the US onset. Therefore, trace conditioning is expected to engage brain regions that are not required for conditioning that depends on on-line information only. This was indeed found (Moyer et al. 1990; Clark and Squire 1998; "conscious awareness, "declarative learning). This also raises the question whether classical conditioning should be considered as a distinct type of memory system, specifically, a nondeclarative memory system, as is advocated by the current "zeitgeist ("taxonomy).

Classical conditioning is considered by the majority of scholars in the field to involve reconstruction of knowledge about stimuli, their relationships, and their predictability (Dickinson 1980; Holland 1993; Wasserman and Miller 1997; Pearce and Bouton 2001). A major impetus to this 'cognitive revolution' in the portrayal of classical conditioning was contributed by the study of a rich set of phenomena, which show that the ability of a stimulus to enter into association and control behaviour is altered markedly by the history of the subject with this or other stimuli either before, during, or after training. Many of these stimulus-revaluation phenomena could be construed as reflecting interaction of the "internal representations of the stimuli involved. Here are selected examples (for additional ones, see "cue):4

1. In sensory pre-conditioning, two practically 'neutral' sensory stimuli, CSi and CSj, are repeatedly presented together, followed by the conditioning of CSi to a particular response. Sensory pre-conditioning is said to have occurred if CSj also evokes the response on a test trial (Brogden 1939; Kimmel 1977). Sensory pre-conditioning was demonstrated first in Pavlov's laboratory. It provides a demonstration that associations can take place among stimuli in the absence of an overt response.

2. In conditioned inhibition, a CS- that is conditioned to predict the absence of the US, later inhibits the development of the CR to a composed CS-+CS+ stimulus (Pavlov 1927; Zimmer-Hart and Rescorla 1974). This is construed to imply that the subject anticipates no US in the presence of the CS-, and must overcome this anticipation to form the association of the compound stimulus with the US.

3. Uncorrelated presentation of the CS and US is used as a control for classical conditioning (Rescorla

1967). It also retards, however, subsequent associative conditioning of the same CS-US pair (Kremer 1971). This phenomenon came to be known as learned irrelevance (Mackintosh 1973; Baker 1976).

4. Latent inhibition (Lubow and Moore 1959; Lubow 1989) is attenuation of the associability of the CS as a result of its non"reinforced pre-exposure. For example, in conditioned taste aversion, if instead of pairing an unfamiliar taste with malaise, one pre-exposes the subject briefly to that taste a few days before training, the aversion developed after pairing with the US is significantly weaker (e.g. Rosenblum et al. 1997). Several explanations have been proposed to account for latent inhibition. Some of these suggest that pre-exposure reduces subsequent "attention to the CS (Lubow 1989); others propose that the formation of a CS-context association during pre-exposure interferes either with subsequent acquisition of the new CS-US association, or with its expression (e.g. Grahame et al. 1994).

5. Pre-conditioning exposure to the US could also retard the formation of a CR (Randich and LoLordo 1979). The explanations invoke "habituation, or alternatively, again, the formation of stimulus-context associations, which later compete with the formation of new CS-US association (e.g. Cole et al. 1996), or with its expression (Miller et al. 1993).

6. Responding to a CS is also sensitive to post-training alterations in the ability of the US to control behaviour (Rescorla 1973; Holland and Straub 1979). This is termed US devaluation. It is construed to imply that the internal representation of the CS gains access to adjustment in the value of the US at the time of "performance.

7. And, last but not least on our brief tour of the "surprising world of conditioning phenomena: a CS may also come to modulate the response of the conditioned subject even independently of its direct association with the US, in which case this CS is said to 'set the occasion' for responding to another CS. Occasion setting was first described by Skinner (1938) in the context of instrumental conditioning. He noted that animals can use a discriminative cue, which has been present upon the occasion of a previous reinforcement, to decide whether to emit the conditioned response or not; the cue itself, however, did not elicit the response. Occasion setting was subsequently investigated mostly in classical conditioning (Holland 1992). It is argued that occasion setting is easier to fit into 'configural theories' of conditioning (Pearce and Bouton 2001; on what configural theories are see footnote 3).

All in all, the picture that emerges is hence of the "performance of classically conditioned behaviour as the surface structure, and the dynamic interaction of internal representations of stimuli and of the probability of their co-occurrence as the deep structure.

Selected associations: Associative learning, Cerebellum, Coincidence detector, Instrumental conditioning, Taxonomy

'Definition 3 in *paradigm.

2The truth is that the subject in instrumental conditioning learns other types of relations as well; see there.

3In *real-life, the CS rarely appears in isolation (*context). A singlestimulus CS is called 'elemental', and the conditioning protocol 'elemental conditioning'. Otherwise, the CS is 'composite' or 'compound'. Theories of conditioning debate whether elements of composite stimuli associate with the US independently ('elemental theories') or as a unitary *internal representation ('configural theories'; Pearce and Bouton 2001). But this issue already relates to the knowledge acquired in classical conditioning, which is further referred to in the text.

4Most of these revaluation phenomena could be demonstrated, employed as research tools, and used to infer knowledge structures in instrumental conditioning as well.

0 0

Post a comment