Learning Through Classical Conditioning

Does the name “Pavlov” ring a bell? If you understand this really bad joke, then you are already somewhat familiar with classical conditioning—learning that one stimulus signals the arrival of another stimulus. A stimulus (plural: stimuli) is any sight, sound, smell, taste, or body sensation that a human or animal can perceive. You know from Chapter 1 that Pavlov’s dogs salivated to the sound of a tone (the first stimulus) because they learned to expect food in their mouth (the second stimulus) after hearing the tone. This may not sound like the research of a Nobel Prize–winning scientist, but it was important work that had a great impact on our understanding of how we learn. Classical conditioning is sometimes referred to as Pavlovian conditioning because Pavlov was the first researcher to study this type of learning systematically. So, let’s take a closer look at Pavlov’s research to gain an understanding of classical conditioning and its importance in learning.

The Elements and Procedures of Classical Conditioning

This is a photograph of Pavlov’s experimental laboratory. Look at the dog’s left cheek. Pavlov devised a surgical insert that diverted the dog’s saliva to exit its cheek so the saliva could be measured accurately.

Science Source

Ivan Pavlov was a Russian physiologist who won a Nobel Prize in 1904 for his earlier work on the physiology of digestion. Pavlov studied the digestive processes of dogs. During these experiments, the dogs were strapped into harnesses and had tubes inserted in their cheeks to measure salivation, the initial step in the digestive process. Pavlov had cleverly found a way to divert the saliva so that it was excreted through the dogs’ cheeks. As part of this research, the dogs were given food (meat powder), and then the amount of saliva excreted was measured. During this research, Pavlov made an accidental discovery. He noticed that the dogs started to salivate before the meat powder was even put in their mouths. For example, the dogs salivated when they heard the footsteps of his assistants bringing the meat powder. Pavlov wanted to know why, and this became the focus of his research for the remainder of his career. So what exactly did he do?

Unconditioned stimulus (UCS) and unconditioned response (UCR).

First, let’s consider why the dogs salivated when the meat powder was put in their mouths. This is a reflexive response—when food is put in your mouth, you salivate. Dogs do it, I do it, you do it. This is called a reflex—a stimulus-response pair (food in the mouth and salivation) in which the stimulus automatically elicits the response. The reflexive stimulus (food in the mouth) that elicits the automatic response (salivation) is referred to as the unconditioned stimulus (UCS), and the response automatically elicited by the UCS is referred to as the unconditioned response (UCR). The key word is “unconditioned.” This means that no learning was necessary for this stimulus (the food, in our example) to elicit the response (salivation, in our example). It’s a naturally occurring reflex. Now let’s examine how such reflexes are used to achieve conditioning.

Conditioned stimulus (CS) and conditioned response (CR).

Pavlov began with a neutral stimulus, a stimulus that does not naturally elicit the to-be-conditioned response. Pavlov used various neutral stimuli, such as a tone generated by a tuning fork, a light, and a touch on the leg, in his conditioning research on dogs (Pavlov, 1927/1960). Before conditioning, his dogs did not automatically salivate to these stimuli. They had to learn (be conditioned) to do this. To achieve such conditioning, the neutral stimulus (for example, a tone) is presented just before the UCS (the food). What does “just before” mean? This can mean a few seconds, but the optimal time interval between the onsets of the two stimuli is usually very brief, only a half-second to a full second (Mazur, 1998). In addition, the two stimuli usually need to be paired together for several trials. Nevertheless, there are some exceptions in which classical conditioning can be obtained without a close temporal pairing of the CS and UCS and in only a small number of trials, sometimes only one. The primary exception is taste aversion, which we will discuss in the last section of this chapter when we discuss biological constraints on learning.

Mark Stivers

Once the conditioning occurs (signaled in Pavlov’s research by the dog salivating to the sound of the tone before the meat powder is put into its mouth), the neutral stimulus is referred to as the conditioned stimulus (CS). This previously neutral stimulus (the tone) comes to elicit a new response (salivating) after repeated pairings with the unconditioned stimulus (meat powder in the dog’s mouth). The learned response (salivating) to the conditioned stimulus (the tone) is called the conditioned response (CR). The conditioned response (salivating) is a preparatory response for the impending UCS (meat powder in the dog’s mouth).

To firm up your understanding of the elements of classical conditioning, let’s consider another example, conditioning of the eyeblink response, which is often used in classical conditioning research with humans. A neutral stimulus, such as a tone, is presented just before a mild puff of air to a person’s eye (the UCS). Initially, an eyeblink (the UCR) occurs reflexively in response to the air puff (the UCS), but with repeated pairings of the tone followed shortly thereafter by the air puff, the previously neutral stimulus (the tone) becomes the conditioned stimulus (the CS) and now elicits a new response, an eyeblink (the CR), in advance of the air puff. In this case, the eyeblink serves as an adaptive, defensive response to the air puff. Instead of an air puff to a person’s eye as the UCS, Clark Hull, a major behavioral psychologist in the first half of the twentieth century, used a slap to the face as the UCS and his graduate students at Yale University as his subjects (Gluck, Mercado, & Myers, 2011). Thus, Hull trained his students to blink in anticipation of a slap to the face. For obvious ethical reasons, researchers have not followed in Hull’s footsteps and do not use the face slap as the UCS in human classical conditioning.

Delayed and trace conditioning.

The timing of the relationship of the CS and UCS is a critical factor in classical conditioning. The conditioned stimulus (the tone) is presented just before the UCS (the meat powder in Pavlov’s experiments) because the conditioning involves learning that this stimulus is a reliable predictor for the arrival of the UCS (Rescorla, 1988). Thus, presenting the UCS before the CS (called backward conditioning) or presenting the UCS and CS at the same time (called simultaneous conditioning) typically leads to no conditioning or poor conditioning because the CS in these two cases would not be a good predictor of the UCS (Powell, Symbaluk, & MacDonald, 2002). The CS has to clearly predict the UCS by being presented first or conditioning usually will not occur. There are two ways to present the CS first: delayed and trace conditioning.

Clark Hull (standing with visor) supervising a classical conditioning experiment at Yale University in the 1920s. The seated graduate student is being conditioned to blink at the sound of a tone in anticipation of a slap to the face. Given that there is a brief interval between the tone and the slap to the face, this is an example of trace conditioning. Actually this particular graduate student, Ernest Hilgard, went on to become one of the most famous psychologists of the twentieth century.

Courtesy of Elizabeth Jecker (daughter of Ernest Hilgard)

In delayed conditioning, the CS remains on until after the UCS is presented, so that the two stimuli occur at the same time. The tone would be turned on and not turned off until after the meat powder was placed in the dog’s mouth. This procedure is called “delayed” because turning off (taking away) the CS is delayed until after the UCS starts. In trace conditioning, there is a period of time between turning off the CS and the onset of the UCS (called the trace interval), when neither stimulus is present. In Pavlov’s case, this would mean turning the tone on and then off, waiting a brief period of time, and then putting the meat powder in the dog’s mouth. For the association between stimuli to be learned, the animal or human (dog in Pavlov’s case) must maintain a “memory trace” of the CS (the tone) to pair with the later-occurring UCS (the meat powder). This is why it is called trace conditioning.

Delayed conditioning is the most effective procedure for classical conditioning, but trace conditioning can be almost as effective if the trace interval between stimuli is very short (Powell, Symbaluk, & MacDonald, 2002). It also appears that for trace conditioning, processing by both the cerebellum and hippocampus is essential for successful conditioning whereas for delayed conditioning, only processing by the cerebellum is necessary (Clark & Squire, 1998). In trace conditioning, processing by the hippocampus is necessary to establish a memory trace of the CS so that it can be associated with the UCS following the trace interval (Bangasser, Waxler, Santollo, & Shors, 2006). In addition, it appears that delayed conditioning can be accomplished without conscious awareness of the temporal relationship between the CS and UCS, but trace conditioning cannot (Clark & Squire, 1998).

The entire classical conditioning process is diagrammed in Figure 4.1. Study the diagram to make sure you understand the four elements that are involved in classical conditioning—the UCS, UCR, CS (the former neutral stimulus), and CR. To ensure that you understand, next we’ll consider another famous example of classical conditioning, the Little Albert study.

Figure 4.1 The Elements of Classical Conditioning In classical conditioning, you start with a reflex—an unconditioned stimulus (UCS) that automatically elicits an unconditioned response (UCR). To condition a new response to a neutral stimulus, the neutral stimulus is presented just before the unconditioned stimulus. After several pairings, the neutral stimulus elicits a new response called the conditioned response (CR). The neutral stimulus is now referred to as the conditioned stimulus (CS).

The Little Albert study.

John Broadus Watson, an American psychologist in the early part of the twentieth century, was the founder of behavioral psychology (Watson, 1913, 1919). He was very impressed with classical conditioning and its potential use to make psychology an objective science of behavior. To examine the possible role of such conditioning in human emotional responses such as fear, Watson and his research assistant, Rosalie Rayner, conducted a study on an infant named Albert (Watson & Rayner, 1920). When the conditioning phase of the study began, Albert was 11 months old.

Let’s think about the Little Albert study in terms of the four elements of classical conditioning—UCS, UCR, CS, and CR. As a neutral stimulus, the researchers used a white laboratory rat. Albert was not afraid of the rat. In fact, when he saw the rat before the conditioning, he moved toward it and touched it. What was the reflex that Watson and Rayner used? While Albert was looking at the white rat, Watson quietly sneaked behind him with an iron bar and a hammer and clanged them together. Albert’s reflexive response, the UCR, was the fear-avoidance response (crying and trying to crawl away) to this unexpected loud noise, the UCS. Watson and Rayner paired the sight of the white rat (the neutral stimulus that became the CS) with the unexpected loud noise (the UCS) several times until the sight of the white rat started eliciting the fear-avoidance response (crying and trying to crawl away from the rat). It took only seven pairings.

In the photograph on the left, Little Albert did not display any fear of the white rat before the classical conditioning trials but rather attempted to play with it. However, following the conditioning trials in which the sight of the white rat was paired with an unexpected loud noise, he acquired a fear response to the rat. In the second photograph, he is generalizing this conditioned fear response to a rabbit.

Archives of the History of American Psychology
Courtesy of Professor Benjamin Harris, University of New Hampshire

This study was unethical, of course. Psychologists would not be allowed to conduct such a study today. Ironically, however, at that time Watson was criticized more by animal rights activists for his research with rats than he was for the Little Albert study (Buckley, 1989). Are you wondering what became of Albert? Sadly, no one knows (Harris, 1979). We do know that Watson and Rayner did not decondition Little Albert’s fear-avoidance response.

One of Watson’s former students, Mary Cover Jones, demonstrated that such a fear could be deconditioned (Jones, 1924). She deconditioned a preexisting fear of rabbits in a 3-year-old boy named Peter. In Chapter 10, we will discuss a behavioral therapy for treating fear disorders (called phobias) that is based upon Jones’s work. In addition, Elsie Bregman in 1934 showed that there could be limits to conditioning such fear responses. In her study, she was unable to condition infants to fear inanimate objects such as wooden blocks and cloth curtains. These results suggest possible biological predispositions to learn certain fears more easily than others. We will return to this idea in the last section of this chapter in our discussion of biological constraints on learning.

It is important to realize that classical conditioning is not only involved in conditioning negative emotional responses such as fear, but can also be used to develop positive emotional responses to stimuli. For example, classical conditioning is used in advertising in order to condition positive attitudes and feelings toward products (Allen & Shimp, 1990; Grossman & Till, 1998; Olson & Fazio, 2001). Think about advertisements that pair celebrities with products. Pairing popular professional basketball players like Michael Jordan and LeBron James with Nike products in advertisements is a good example. Both of these athletes were paid millions of dollars for their Nike endorsements. Why? The advertisers attempt to use our positive feelings about these celebrities to condition positive responses to their products. The celebrity serves as the UCS, and the product as the CS. Advertising campaigns using this evaluative classical conditioning technique have proved very effective (Hofmann, De Houwer, Perugini, Baeyens, & Crombez, 2010). John Watson pioneered this type of advertising when he worked as a psychology consultant for the J. Walter Thompson advertising agency in New York following his dismissal from academics (Buckley, 1989). Watson was forced to leave the academic life after he had an extramarital affair with his research assistant on the Little Albert study, Rosalie Rayner.

General Learning Processes in Classical Conditioning

We have seen how classical conditioning works: A new, learned response (the CR) is elicited by the CS in preparation for the arrival of the UCS. This response learning is also referred to as acquisition, acquiring a new response—the CR to the CS. But what happens after acquisition? Will this new response (fear-avoidance in the Little Albert study) be generalized to other stimuli (say, white rabbits) that are similar to the CS (a white rat)? If so, can we learn to discriminate and respond this way only to specific stimuli (white rats)? Will this response (fear-avoidance) continue to occur when the CS (a white rat) is no longer paired with the UCS (the unexpected loud noise)? To answer such questions, we now turn to a discussion of the other general learning processes that follow acquisition in classical conditioning.

Extinction and spontaneous recovery.

What do you think would happen to the CR if the UCS no longer followed the CS? Would the CR now be “unlearned”? Remember that in classical conditioning the CS reliably signals that the UCS is coming. If the CS no longer serves this function, then the CR is no longer necessary so it is no longer made. No preparation is needed because the UCS is no longer presented. For example, in Pavlov’s work the dog would eventually stop salivating to the tone when the tone no longer signaled that the meat powder was coming. In the Little Albert study, Albert’s fear of a white rat would have diminished after it no longer signaled that the loud, unexpected noise was coming. This unlearning process is called extinction, the diminishing of the CR when the UCS no longer follows the CS.

A visual comparison of the acquisition and extinction processes is given in Figure 4.2 (a and b). The strength of the CR increases during acquisition, but it decreases during extinction. Note, however, that during the extinction process, the CR mysteriously increases somewhat in strength following a rest interval. This is called spontaneous recovery, a partial recovery in strength of the CR following a break during extinction trials (see Figure 4.2c). As extinction continues, however, the recovery observed following the rest intervals continues to decrease until it is minimized. For example, the dog would salivate very little or not at all to the tone, and Albert would show little or no fear of the white rat. The spontaneous recovery that occurs during the extinction process indicates, as Pavlov had concluded, that the CR may not be totally lost during extinction, but only greatly weakened or inhibited. Research since Pavlov’s time has demonstrated that this is indeed the case (Bouton, 1994; Rescorla, 1996).

Figure 4.2 Acquisition, Extinction, and Spontaneous Recovery Panel (a) depicts the acquisition of the CR. Note that during acquisition, the UCS follows the CS on each trial. Acquisition of the CR is reflected in the increase in its strength across conditioning trials. Extinction is depicted in panel (b). On extinction trials, the UCS never follows the CS. This process is reflected in the decrease of the strength of the CR across extinction trials. Note, however, that when a break is taken during extinction training, there is a temporary partial recovery in strength of the CR following the break, depicted in panel (c). This partial recovery is called spontaneous recovery. The amount of recovery lessens as extinction training continues.

Stimulus generalization and discrimination.

In addition to acquisition, extinction, and spontaneous recovery, there are two other general learning processes involving the CR—stimulus generalization and discrimination. Let’s consider generalization first. After acquisition of a CR (salivating) to a CS (a tone), we observe generalization. In stimulus generalization, stimuli similar to the CS elicit the CR. The more similar the stimulus is to the CS, the stronger the response will be. Generalization is an adaptive process. Classical conditioning would not be a very useful type of learning if it only allowed us to learn relationships between specific stimuli. If a dog bites you, isn’t it more adaptive to generalize the fear-avoidance response to other dogs (especially dogs similar to the one that bit you)?

To understand generalization, let’s consider Pavlov’s research with dogs. Assume Pavlov classically conditioned a dog to give the salivary response to a tone with a pitch of 1,000 Hz. (Remember from Chapter 3 that this represents a sound wave that cycles 1,000 times per second.) To test for generalization, the CS (the 1,000 Hz tone) and the other stimuli (tones with higher and lower frequencies) are presented, but the UCS (the meat powder) does not follow any of them. The strength of the CR (the amount of salivation) to each of these stimuli is measured. So what would happen if a 1,025 Hz or a 975 Hz tone were presented? The dog would probably give a fairly strong salivary response to these tones because they are very similar to the original CS of 1,000 Hz. But the strength of the CR (the amount of salivation) would go down as the generalization test tones get farther away from the original CS of 1,000 Hz. This generalization function is illustrated in Figure 4.3. Ignore the discrimination function in the figure for now. We will discuss it shortly.

Figure 4.3 Stimulus Generalization and Discrimination These are idealized stimulus generalization and discrimination results following classical conditioning of the salivary response (CR) in dogs to a tone of 1,000 Hz (CS). The strength of the CR to generalization test stimuli (all tones except the 1,000 Hz tone) decreases as the similarity of the test stimulus to the original CS decreases. These generalization responses, however, can be extinguished through discrimination training in which the UCS (food) only follows the 1,000 Hz tone. The CR to each of the other tones will be extinguished except for the tones so similar to the 1,000 Hz tone (975 Hz and 1,025 Hz) that the dog does not perceive them as different.

Notice how symmetric the generalization test results are around the original CS value (1,000 Hz) in Figure 4.3. This is because the generalization test tones could be varied symmetrically in frequency above and below the original CS tone. But think about the Little Albert study by Watson and Rayner. Their original CS was a white rat. What do you think they used to test for generalization? They used other animals—a rabbit, as we’ve seen, and a dog (neither was white)—but they also used some less similar inanimate objects, including a bearded Santa Claus mask, a sealskin coat, human hair, a package of white cotton, and some wooden blocks. The generalization stimuli did not vary symmetrically around the white rat, but the same pattern of generalization results were observed. As the generalization test stimuli became less similar to the CS (the white rat), the strength of the CR (the fear-avoidance response) decreased. The response to the other animals and the coat was stronger than the responses to the mask and hair. Similarly, the response to the white cotton was even less, and it was nonexistent for the toy blocks.

Sometimes we overgeneralize and need to narrow our responding to a specific stimulus or smaller set of stimuli. In the case of the dog bite, we might overgeneralize and fear any dog. This wouldn’t be rational, so we would need to learn to discriminate dogs who might bite us from those who would not. Such discrimination learning can be thought of as the opposite of generalization. Generalization leads to the CR being given to a broader set of stimuli; discrimination leads to the CR being given to a narrower set of stimuli. In classical conditioning, stimulus discrimination is the elicitation of the CR only by the CS or only by a small set of highly similar stimuli that includes the CS. In the dog bite example, this would be learning only to fear potentially dangerous dogs.

Discrimination training is used to teach stimulus discrimination. We will discuss the simplest case—learning to give the CR only to the original CS. This is the same procedure used to determine the sensory abilities of animals. In this type of discrimination training, the UCS follows only the original CS, the one stimulus to which you want the animal or human to respond (give the CR). None of the other stimuli used in the training are followed by the UCS. During discrimination training, you present many different stimuli numerous times, but the UCS only follows the original CS. What do you think happens? What normally happens to the CR when the UCS is removed in classical conditioning? It is extinguished. This is what discrimination training does; it diminishes the responding to the other stimuli.

Idealized results of stimulus discrimination training for a CS tone of 1,000 Hz in our Pavlovian classical conditioning example are shown in Figure 4.3. As you can see, responding is centered at 1,000 Hz, as it is for the stimulus generalization results. However, unlike the generalization results, the responding following discrimination training is to a much narrower range of stimulus frequencies. The animal has learned only to salivate to the 1,000 Hz tone and tones very similar to it (975 Hz and 1,025 Hz). Why does the animal salivate at all to any stimulus other than the 1,000 Hz tone? It wouldn’t if it had perfect pitch perception and could discriminate all frequencies, but it can’t. Thus, the animal salivates to this narrow range of frequencies because it has difficulty differentiating them. This is why discrimination training can be used to learn what nonverbal animals (and human infants) can discriminate. If they respond to stimuli the same way, then it is assumed that they cannot differentiate between the stimuli.

All five of the general learning processes for classical conditioning are summarized in Table 4.1. If any of these processes are not clear to you, restudy the discussions of those processes in the text and their visual depictions in Figure 4.2 and Figure 4.3 to solidify your understanding. You want to ensure that you understand these learning processes for classical conditioning, because they will be redefined for operant conditioning in the next section.