Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 4

Maham Khan What role does contingency play in learning?

In this essay I will argue that, in an associative learning framework, the contingency between a stimulus and an associated behavior has a consequential role to play in so far as the stimulus is informative or predictive with regards to another. In effect, I will be arguing that arranging for simple contiguity between two stimuli need not produce an association between them. Specifically, simple CS-US pairing is neither necessary nor sufficient for associative learning to occur. Rather, conditioning depends on the information provided by the CS about the US. Rescorlas (1968) work on contingency, and Kamins (1968) demonstration of a Blocking Effect in learning both lend support to the idea that a simple contingency is not sufficient for conditioning; while evidence for conditioned inhibition (that dates back to Pavlovs work in the late 19th century) illustrates that such a contingency is not necessary for conditioning to occur. Lastly, I will examine a model that tries to account for what this evidence conveys about learning through predictive calculation of associative strength. This is the RescorlaWagner model. Before I go on to expound these ideas any further, some clarifications must be made. In associative learning literature, the term contingency has been, at times, used rather carelessly. In some cases, contingency is taken to mean the degree of dependency which presentation of the US has upon prior presentation of the CS (Rescorla 1967). This is used in contrast to the term pairing which simply means the co-occurrence or proximity of the CS and US in time. Contingency, in this sense, is more quantitative, whereas pairing is nominative. On the other hand, contingency is sometimes taken as an umbrella term for all CS-US relations of which simple pairing then becomes a subtype. Rescorla himself, in a later paper (1988), is somewhat guilty of this. It is then useful to make the distinction between simple contingency (which means simple CS-US pairing) and contingency in general. For the purpose of parsimony in this essay, I will use pairing and contingency in two separate senses. Early theories proposed the simple pairing or occurrence of the CS and US together in time was enough to produce conditioning. Rescorla (1988) argued that in advocating thus, early theories failed to capture adequately the circumstances in which the learning came about, the content of the learning, and the effects of that learning on the animals behavior and performance. In a paper on proper control procedures for Pavlovian conditioning, Rescorla (1968) argued that the contingency rather than the pairing between the CS and US is the important event in conditioning. He advocated that the random stimulus control procedure is more informative than traditional control procedures (CS-alone, novel CS, US-alone, explicitly unpaired, backward conditioning, and discriminative conditioning) in identifying the role of the CS-US contingency. In this control, the CS and US are presented with no contingency whatsoever between then, such that some pairings of CS and US may occur by chance alone. The random stimulus control procedure was used in a study by Rescorla (1968) where he habituated 10 groups of rats to level pressing in return for a

Maham Khan reward, and then conditioned them to a tone (CS) and shock (US). He measured conditioned suppression of lever pressing, by calculating the suppression ratio (CS/(CS+preCS). Each group received a different combination of the probability of the shock of the shock during the CS P(US|CS), and probability of the shock in the absence of the CS P(no US|CS). Rescorla found that when the probability of the shock was the same in the presence or absence of the CS, no suppression of lever pressing occurred (suppression ratio = .5). The greatest change in conditioned suppression ratio was observed (suppression ratio = 0) when the disparity between the two probabilities was greatest (when P (US|CS) was 0.4 and P (no US|CS) was 0). This indicated complete cessation of lever pressing and was therefore suggestive of strong learning. If simple CS-US pairing had been sufficient for learning, conditioning should have been determined by P (US|CS), independently of P (no US|CS). However, the results were not consistent with this prediction. In effect, conditioning was found to be sensitive to the base rate of the US occurrence against which a CS-US contiguity takes place. Rescorla therefore concluded that conditioning depends not on the pairing between the CS and the US but rather how informative the CS is with regards to predictability and contingency of the US. Another modern conditioning phenomena, known as the blocking effect (Karmin 1968) is also consistent with Rescorlas conclusions. The blocking effect can be demonstrated in a simple experiment where two groups of animals receive a compound CS (such as a light + tone) that signals a US (shock). One group, however, has prior exposure to the light alone, whereas the other does not. For both groups, the CS-US contiguity is exactly the same, and yet the group with prior exposure demonstrates a blocking effect to the light/tone compound. Specifically, the prior training renders the light/tone compound CS redundant. The blocking effect demonstrates, again, that conditioning is governed by the informational relation with which the groups differ, rather than the contiguity that they share. Consistent with Rescorlas conclusions, simple contiguity fails to account for the results. Thus far, we have looked at evidence that demonstrates a simple pairing is insufficient to explain conditioning. We now consider some evidence for the proposition that pairing is also not necessary for an association to be formed. This evidence is a variant on the earlier evidence, where the US (in this case the shock) is omitted in the absence of the CS (the tone) rather than in its presence. In this way, the base rate of US occurrence is maintained, whilst any CS-US contiguity is eliminated. Despite the omission of CS-US contiguity, however, learning does take place. Specifically, the tone becomes a conditioned inhibitor: it now signals a negative US-CS contingency. Research on conditioned inhibition lends further support to the notion that simple pairing of two stimuli cannot be considered fundamental to any explanation of Pavlovian conditioning. Such findings emphasize instead the predominant modern thesis that associative learning involves learning about the relations between events.

Maham Khan Having examined some modern evidence on the role of contingency in associative learning, we now look to some theories of conditioning. In particular, we look at the Rescorla-Wagner (1972) model, and how it attempts to explain and synthesize all the findings discussed thus far. The Rescorla-Wagner model emphasizes the importance of surprise. Karmins blocking effects, on this account, do not demonstrate conditioning because there the element of surprise is eliminated through pre-exposure. The model essentially predicts changes in associative strength V, based on the following equation (where is the asymptote of associative strength): V = ( - V) In the compound trials, because there is no element of surprise V equals 0. This means no change in associative strength, and therefore no learning. The model also provides a precise explanation of conditioned inhibition and negative contingencies, through the added consideration of context effects. It brings to the fore the idea that the CS and US never occur in a vacuum; they always occur in concert with some background stimuli. Evidence suggests that humans and animals alike do associate contextual stimuli with the CS and US during conditioning, leading to the emergent view that context may serve as a single, long-lasting CS that occurs in compound with the experimentally manipulated CS and US. In negative contingency experiments therefore, the stimuli may now be seen as the following two types of trial: 1) CS-alone (which now means CS + context) and 2) US-alone in the intertrial interval (which now means US + context). Under this model, the US can be seen as associated with the context alone, rather than with the absence of the CS. This explains which simple CS-US contingency is not necessary for learning. The Rescorla-Wagner model therefore makes the point that conditioning is hardly as simple as a CS-US pairing. Associative learning involves associations between compound stimuli, and conditioning to any one element is therefore determined (facilitated or inhibited) by what the other elements of the compound predict. Having considered the above evidence in light of the Rescorla-Wagner model, in the words of Rescorla, Pavlovian conditioning is not a stupid process by which the organism willy-nilly forms associations between any two stimuli that happen to co-occur. Rather, we might think about the organism as an information seeker that uses logical and perceptual relations among events, along with its own preconceptions, to form a sophisticated representation of its world.

Maham Khan

Bibliography: Rescorla, R.A. (1988). Pavlovian conditioning Its not what you think it is. American Psychologist, 43, 151-160. Bouton, M.E. (2007). Learning and behavior: A contemporary synthesis. Sinaur Pavlov, I.P. (1927). The conditioned reflexes. Dover Press Baker, A.G. (1976).Learned irrelevance and learned helplessness: Rats learn that stimuli, reinforcers and responses are uncorrelated. Journal of Experimental Psychology: Animal Behavior Processes, 2, 130-141. Papini, M. & Bitterman, M.E. (1990). The role of contingency in classical conditioning, Psychological Review, 97, 396-403. Murphy,R.A. & Baker,A.G. (2004). A role for CS-US contingency in Pavlovian conditioning. Journal of Experimental Psychology: Animal Behavior Processes, 30, 229-239. Rescorla, R.A. (1967). Pavlovian conditioning and its proper control procedures. Psychological Review, 74, 71-80. Skinner, B.F. (1950). Are theories of learning necessary? Psychological Review, 57,193-216. Wagner, A.R., Logan, F.A., Haberlandt, K. & Price, T. (1968). Stimulus selection in animal discrimination learning. Journal of Experimental Psychology, 76, 171- 180.

You might also like