Pavlovian conditioning. It's not what you think it is.

Current thinking about Pavlovian conditioning differs substantially from that of 20 years ago. Yet the changes that have taken place remain poorly appreciated by psychologists generally. Traditional descriptions of conditioning as the acquired ability of one stimulus to evoke the original response to another because of their pairing are shown to be inadequate. They fail to characterize adequately the circumstances producing learning, the content of that learning, or the manner in which that learning influences performance. Instead, conditioning is now described as the learning of relations among events so as to allow the organism to represent its environment. Within this framework, the study of Pavlovian conditioning continues to be an intellectually active area, full of new discoveries and information relevant to other areas of psychology. Pavlovian conditioning is one of the oldest and most systematically studied phenomena in psychology. Outside of psychology, it is one of our best known findings. But at the same time, within psychology it is badly misunderstood and misrepresented. In the last 20 years, knowledge of the associative processes underlying Pavlovian conditioning has expanded dramatically. The result is that modern thinking about conditioning is completely different from the views psychologists held 20 years ago. Unfortunately, these changes are very poorly appreciated by psychologists at large. The last time many psychologists read anything about Pavlovian conditioning was before these changes took place. Even those more recently educated often received that education from textbooks and instructors that had largely ignored the dramatic conceptual changes that had taken place. The result is that many think of Pavlovian conditioning as an obsolete technical field that is intellectually stagnant. My intention in this article is to show that this view is incorrect. First, I will review some of the changes that have occurred in Pavlovian conditioning in order to give the flavor of its contemporary form. I will argue that it is an intellectually challenging field, in which substantial and exciting progress has been made. Second, I will argue that conditioning continues to have a central place in psychology generally. I will describe how it touches on and informs several related fields that are currently more in vogue. To begin the discussion, consider how conditioning was described 20 years ago, when those in my generation were students. One popular introductory text put it thus: The essential operation in conditioning is a pairing of two stimuli. One, initially neutral in that it elicits no response, is called the conditioned stimulus (CS); the other, which is one that consistently elicits a response, is called the unconditioned stimulus (US). The response elicited by the unconditioned stimulus is the unconditioned response (UR). As a result of the pairing of the conditioned stimulus (CS) and the unconditioned stimulus (US), the previously neutral conditioned stimulus comes to elicit the response. Then it is called the conditioned response (CR). (Morgan & King, 1966, pp. 79-80) This description is typical of those found in both introductory and advanced textbooks 20 years ago. Unfortunately, it is also typical of what one finds in textbooks today. One popular introductory text published in 1987 describes conditioning in this way: "The originally neutral conditioned stimulus, through repeated pairing with the unconditioned one, acquires the response originally given to the unconditioned stimulus" (Atkinson, Atkinson, Smith, & Hilgard, 1987, p. 658). Students are exposed to similar descriptions in textbooks specializing in allied fields of psychology. In a cognitive textbook, one reads, We start out by taking an unconditioned stimulus (UCS) that produces the desired response without training . . . . We pair the UCS with a conditioned stimulus (CS) . . . . This procedure, when repeated several t i m e s . . , will ultimately result in the occurrence of the response following the CS alone. (Klatsky, 1980, p. 281) A widely used developmental text agrees, calling conditioning a "form of learning in which a neutral stimulus, when paired repeatedly with an unconditioned stimulus, eventually comes to evoke the original response" (Gardner, 1982, p. 594). Similarly, a best-selling textbook of abnormal psychology describes a conditioned stimulus as "a stimulus that, because of its having been paired with another stimulus (unconditioned stimulus) that naturally provokes an unconditioned response, is eventually able to evoke that response" (Rosenhan & Seligman, 1984, p. 669). Of course, textbook descriptions vary widely in their precision and sophistication, but these citations represent a common view. Indeed, these quotations will certainly sound so familiar that many readers may wonder what is wrong with them. I want to suggest that the answer is "almost everything." These descriptions make assertions about what I take to be the primary issues to be addressed in the study of any learning process: What are the circumstances that produce learning? What is the content March 1988 9 American Psychologist COl~ri~at 1988 by the American Psychological Association, Inc. 0003-066X/88/$00.75 Vol. 43, No. 3, 151-160 151 of the learning? How does that learning affect the behavior of the organism? But they are mistaken or misleading in virtually every assertion they make about each of these. These descriptions in fact capture almost nothing of modern data and theory in Pavlovian conditioning. I want to illustrate this claim using some data collected in my own laboratory over the years, but first let me make an orienting comment. Descriptions of conditioning, such as those just cited, come from a long and honorable tradition in physiology, the reflex tradition in which Pavlov worked and within which many early behaviorists thought. This tradition sees conditioning as a kind of low-level mechanical process in which the control over a response is passed from one stimulus to another. Much modern thinking about conditioning instead derives largely from the associative tradition originating in philosophy. It sees conditioning as the learning that results from exposure to relations among events in the environment. Such learning is a primary means by which the organism represents the structure of its world. Consequently, Pavlovian conditioning must have considerable richness, both in the relations it represents and in the ways its representation influences behavior, a richness that was not envisioned within the reflex tradition. Let me now turn to illustrating the difference that this alternative view makes for each of three issues: the circumstances producing learning, the content of learning, and the effects of learning on behavior. Circumstances Producing Pavlovian Conditioning Each of the descriptions given earlier cites one major circumstance as responsible for producing Pavlovian conditioning, the pairing or contiguity of two events. To be sure, contiguity remains a central concept, but a modern view of conditioning as the learning of relations sees contiguity as neither necessary nor sufficient. Rather, that view emphasizes the information that one stimulus gives about another. We now know that arranging for two wellprocessed events to be contiguous need not produce an association between them; nor does the failure to arrange contiguity preclude associative learning. The insufficiency of contiguity for producing Pavlovian conditioning can be illustrated by results that have been available for almost 20 years (e.g., Rescorla, 1968) but that have apparently failed to be integrated into the view of conditioning held by many psychologists. Consider a learning situation in which a rat is exposed to two prominent events, a tone CS that occurs for two-minute periods and a brief, mild electric shock US applied to a grid on which the animal is standing. Suppose that those two events are uncorrelated in time, such that the tone This article is an adaptation of a Presidential Address given to the Eastern Psychological Association in Arlington, VA, in April 1987. The research reported here was generously supported by grants from the National Science Foundation. Correspondence concerning this article should be addressed to Robert A. Rescorla, Department of Psychology, University of Pennsylvania, 3815 Walnut St., Philadelphia, PA, 19104. provides no information about the shock. That relation is schematized in the top of Figure 1. Also schematized in that figure is a variation on that treatment in which only those USs scheduled to occur during the tone are actually applied to the animal. The point to notice about those two treatments is that they share the same contiguity of the tone with the US, but they differ in the amount of information that the tone gives about the US. In the first treatment, the shock is equally likely whether or not the tone is present, and so the tone provides no information; in the second treatment, the shock only occurs during the tone, and so the tone is quite informative about shock occurrence. It turns out that in many conditioning situations learning is determined not by what these treatments share but rather by how they differ. The second group will develop an association between the CS and US, but the first will fail to do so. In effect, conditioning is sensitive to the base rate of US occurrence against which a CS/ US contiguity takes place. Indeed, systematic experiments show that in many situations the amount of conditioning is exquisitely attuned to variations in the base rate of the US. An early illustration of that point is shown in Figure 2, which plots asymptotic levels of fear conditioning (measured by the ability of the CS to interfere with ongoing behavior) as a function of the likelihood of the US during the CS. The parameter in the figure is the base-rate likelihood of the US in the absence of the CS. Each curve shows that conditioning is indeed an increasing function of the likelihood of the shock during the tone. For instance, in the frequently studied case in which the shock likelihood is zero in the absence of the CS, then conditioning is greater the greater the probability of the shock during the tone. This is not a surprising result. What is more interesting is the effect of the base rate of US occurrence in the absence of the CS. At any given likelihood of shock during the CS, conditioning is an inverse function of the base rate. When the CS/US contiguity is held constant, conditioning changes from excellent to negligible simply by increasing the shock base rate. Indeed, when the likelihood of a US is the same in the presence and absence of the CS (as is true of the initial point on each function), there is little evidence of conditioning at all. One description of these results is that conditioning depends not on the contiguity Figure 1 Schematic of Two Conditioned Stimulus/ Unconditioned Stimulus (CS/US) Relations That Share the Same Contiguity but Differ in the Information the CS Gives About the US