CHAPTER 2
Classical Conditioning
In their attempts to discover the most basic form of learning, early experimenters discovered two procedures, one called classical or Pavlovian conditioning, the other called instrumental or operant conditioning. Beginning in the 1930s traditional theorists, such as Skinner (1938), maintained that the two basic procedures represent two distinct forms of learning. This view dominated research and theory throughout the century even though hard evidence frequently contradicts the traditional theoretical distinction. The procedural distinction, however, remains useful as a way to describe experiments in this field. Accordingly, chapter 2 describes the experimental operations that define classical conditioning and chapter 3 describes the experimental operations that define instrumental conditioning. These descriptions should prepare readers to make their own critical analysis of evidence and interpretation that they encounter in this field. Later chapters consider the relation between modern evidence and the traditional distinction.
TYPICAL PROCEDURES
This chapter describes the procedure known as classical or Pavlovian conditioning. Experimental examples that used different species, different responses, and different stimuli illustrate the broad generality of this procedure. Experimental examples from the 1930s to the 1990s illustrate how well the basic findings have survived stringent tests of time and replication.
Salivation
Around 1900 when he started his study of conditioning, Pavlov was already a distinguished scientist who had won a Nobel Prize for his studies of the physiology of digestion. These studies depended on his surgical technique for creating a fistula, or tube, leading from specific regions of the digestive tract to the outside of a living and otherwise healthy animal. In this way, experimenters could observe and measure normal secretions at any stage of digestion. During these experiments, Pavlov noted that experienced dogs often began their secretions on route from their cages to the experimental room. Secretions that appeared without food and before the start of experiments were an inconvenient artifact in his early research. When he began to study “psychic secretions,” as he called them, for their own sake he set out on the course of research that made him truly famous.
In a typical experiment on salivary conditioning in Pavlov’s laboratory, a hungry dog stood on a table in a special experimental room. The experimenters prepared the dog by surgically diverting a salivary gland so that saliva flowed through the dog’s cheek to the outside and collected in a measuring tube. The experimenter first adapted the dog until it would stand quietly on the table in a loose restraining harness. From an adjoining room, the experimenter observed the dog through a small window and presented experimental stimuli with remote controls. A tone sounded and after about 5 seconds a plate containing a small amount of dry food appeared near the dog’s mouth. Combinations of tone and food separated by intervals of from 5 to 35 minutes occurred several times during each daily session. From time to time the experimenter presented probe trials of tone alone without food for 30 seconds.
Hungry dogs tend to salivate at any sound—a door closing, a tap on the window. The experimental tone is no exception and there was some salivary response the first time an experimenter sounded the tone. But, familiar food always evoked much more salivation than any other stimulus. After 30 to 60 combinations of tone and food, tone alone evoked salivation in probe trials. The arbitrary tone evoked salivation without food.
Leg Withdrawal
In a later American example, Liddell (1942, p. 194) attached electrodes to the left foreleg of a sheep. A metronome beat once per second for seconds before the sheep received a mild shock from the electrode. At first, the metronome had little effect, but the shock evoked a sharp flexion or withdrawal of the foreleg, accompanied by rapid breathing. After four pairings of the metronome and shock in combination, the metronome evoked clear increases in breathing rate. On the sixth trial the leg flexed, slightly. By the 11th trial, the leg withdrew fully and consistently. Figure 2.1 shows the reactions on the 11th trial.
FIG. 2.1. Conditioned responses of sheep. The diagram is from a kymograph record of the 11 th pairing of a metronome beat with a shock delivered to the left foreleg. The metronome beat once per second for 5 seconds before the shock. Both conditioned leg movements and respiratory changes began with the beginning of the metronome sound, anticipating the shock that came at the end of the metronome sound. From Liddell (1942).
Eyeblink
Human beings frequently serve as subjects in modern studies of conditioning. Human eyeblinking can be measured with great precision by painting a small silver spot on the subject’s eyelid and bouncing a light beam off the spot and onto an electronic recording device. The human eye blinks spontaneously, of course, but it is much more likely to blink when the experimental apparatus blows a slight puff of air against the subject’s eye. In a typical experiment, the experimenter pairs a brief flash of light with a puff of air several times during the course of an experimental session. After several pairings the light alone evokes eyeblinks on probe trials (Grant, 1973).
Knee Jerk
In 1902, after Pavlov began his experiments but before news of his findings had reached psychologists in the United States, a student named Edwin Twitmeyer was finishing the experiments for his PhD at the University of Pennsylvania. At the time, many psychologists and neurophysiologists studied the simplest unit of behavior, which they called the reflex arc. The knee jerk, or patellar reflex, is a good example of a simple reflex. Most readers are already familiar with this reflex as tested in thorough medical examinations. A sharp tap on the kneecap, actually the tendon attached to the kneecap, elicits a kicking movement of the lower leg. Repeating this process for many trials increases the extent of the movement. Twitmeyer asked why repetition should enhance this reflex. Why doesn’t repetition fatigue and weaken the reflex instead? To find out, he administered many trials over a long series of experimental sessions.
Twitmeyer’s experimental apparatus released a hammer that hit the kneecap with a measured amount of force at a precise time. To warn his subjects, the apparatus also struck a bell once just a half-second before the hammer struck the kneecap. One day, the apparatus broke down while a well-practiced subject was in place. When Twitmeyer retested the apparatus, the bell sounded but the hammer failed to operate. To Twitmeyer’s surprise, the subject’s knee jerked as if struck by the hammer. Twitmeyer thought that the subject, a fellow student, was joking or possibly kicking without waiting for the hammer. The subject reported that he was as surprised as Twitmeyer. The knee seemed to jerk by itself.
Interested, but still skeptical, Twitmeyer repeated the situation, bell without hammer, with a few other subjects. The results were mixed. Some kicked without the hammer, some did not. As he continued, Twitmeyer found that the number of bell-hammer trials was critical. A few subjects responded to the bell without the hammer after only 30 paired trials, but nearly all showed conditioned responses after 130 or more paired trials.
Twitmeyer grasped the significance of his results, writing:
The movement of the legs following the tap of the bell, without the blows on the tendons, has the characteristics of a simple, immediate reaction to the stimulus. Upon the unanimous testimony of the subjects, it was not produced voluntarily, i.e., there was no idea of the movement in consciousness, antecedent to the movement itself. It may, therefore be held, tentatively at least, that the movement is a reflex action. The afferent excitation must therefore reach the [spinal] cord at the level of the medulla and then pass down to the second or third lumbar segment in which the cell bodies of the efferent conduction path are located. Here then we have a new and unusual reflex arc.
… The occurrence of the phenomenon, therefore, depends upon the preliminary simultaneous occurrence of the sound of the bell with the kick produced in the usual way, i.e., a blow on the tendon. After a certain number of such trials, the number varying for different subjects, the association of the sound of the bell and the kick becomes so fixed that the bell itself is capable of serving as a stimulus to the movement. (1974, pp. 1063–1065)
Note how Twitmeyer, like so many others, treats a half-second interval between bell and hammer as equivalent to a zero interval.
Twitmeyer reported and discussed these results in his PhD dissertation together with his main findings that repeated trials increased the reflex. The faculty examining committee approved the dissertation and he received his degree. He also reported his discovery at the annual meeting of the American Psychological Association in a paper called, “Knee-Jerks Without Stimulation of the Patellar Tendon.” At that time, the association was so small that there was only one session at any one time, and all members could attend all papers. When he finished, the audience, even the great William James who chaired the session, failed to ask a single question (Dallenbach, 1959). Twitmeyer’s experiment first appeared in a regular scientific journal in 1974, long after his death (Twitmeyer, 1974).
Soon after Twitmeyer’s paper was so poorly received by the American Psychological Association, Pavlov’s work on conditioned reflexes appeared and started a whole new field of research. Pavlov, when he published, was already a Nobel Prize-winning scientist. He published a series of systematic explorations of conditioning and also formulated a new theory of cognitive association to go with his new findings. Twitmeyer had only one experiment to report. He called for further systematic research, but he was only a new PhD without a laboratory of his own. He also titled his dissertation modestly, “A Study of the Knee-Jerk.”
It is too bad that others failed to pursue Twitmeyer’s procedure even after its belated publication in 1974. His simple method uses a common reflex of human subjects who care for themselves, which avoids the problem of humane animal care. Indeed, Twitmeyer’s subjects reported less discomfort than subjects in human eyeblink experiments. The apparatus is also simple and relatively inexpensive. In addition, the neurology of the patellar reflex is well known and its relation to the rest of the nervous system is less elaborate than reflexes that have been popular in laboratories.
BASIC TERMS
Unconditioned Stimulus (UCS)
The UCS is a stimulus that evokes the unconditioned response (UCR) with a high probability at the start of the experiment.
Figure 2.2 shows the onset of the UCS, because it is often very difficult to say just when a stimulus such as food comes to an end. In any case, the onset is usually the most significant temporal feature of any stimulus (Albert, Ricker, Bevins, & Ayres, 1993).
FIG. 2.2. Classical conditioning. Schematic diagram of two recordings that might have come from a typical experiment. The upper part, T1, shows the situation on a typical first trial, and the lower part, T2, shows the situation on a typical late trial. In each part of the diagram, the line labeled R is the response line and the line labeled S is the stimulus line. Copyright © 1997 by R. Allen Gardner.
In the examples in this chapter, the UCS is the food for the hungry dog in the salivary conditioning example, the electric shock for the sheep in the leg withdrawal experiment, the puff of air for the human beings in the eyeblink experiment, and the hammer striking the patellar tendon in Twitmeyer’s knee jerk experiment.
The word unconditioned can be confusing because it seems to say “never conditioned” or “present at birth” or something of that sort. For an adult dog, the sight of familiar food must have become associated with eating before the start of the experiment. There is weak evidence from Pavlov’s laboratory that the sight and smell of milk came to evoke the salivation of milk-fed puppies but not of control puppies. Unfortunately, this experiment was inadequately controlled and never replicated so it gives us only suggestive evidence. Certainly, human beings raised in one culture are attracted by the sight and smell of foods that make members of another culture feel like retching, and vice versa.
In an actual experiment with an adult subject of almost any species, it would be very difficult to prove the absence of any conditioning in a lifetime of experience with a particular UCS. Usually, all we can know about the UCS in any particular experiment is that it evokes the UCR with a high probability at the start of the experiment. The past history of this high probability is a separate question.
Conditioned Stimulus (CS)
The CS is an arbitrary stimulus experimentally paired with the UCS. An experiment demonstrates that the CS is arbitrary by using at least two different stimuli, Sa and Sb and pairing one of these stimuli with the UCS for some subjects and the other with the UCS for the other subjects. The test phase of the experiment presents Sa on some trials and Sb on others for both groups of subjects to test whether there is more response to the stimulus that was paired with the UCS. Since the experimenter flips a coin or uses some other arbitrary device to decide which stimulus to pair with the UCS, the experiment tests whether the response depends on the arbitrary pairing of CS and UCS.
In the examples in this chapter, the CS is the tone for the dog in the salivary example, the metronome for the sheep in the leg withdrawal experiment, the flash of light for the human in the eyeblink experiment, and the bell in Twitmeyer’s knee jerk experiment.
The CS usually fails to evoke the full UCR before pairing with the UCS, but often does evoke an orienting response (OR) at the start of the experiment, that Pavlov called an orientation reflex, what-is-it reflex, or investigatory reflex.
Orienting Response (OR)
The OR is a response to the CS at the outset of the experiment. Traditionally, the OR shows that the subject attended to the CS before conditioning. The OR tends to disappear during the course of the experiment as if attention to the CS gradually fades away. A dotted line in T2 of Fig. 2.2 indicates the place where the OR used to appear.
In most experiments the UCR (or a fractional part of the UCR) is a highly probable response to any stimulus under those conditions. For example, almost any stimulus is likely to evoke some salivation from a hungry dog. Conditioning is impossible if the probability of the conditioned response to the CS is truly zero at the start of the experiment.
Unconditioned Response (UCR)
The UCR is a response that the UCS evokes with a high probability at the start of the experiment.
In many cases, such as salivation to the stimulus of dry food placed in the mouth, the UCR is probably an obligatory, species-specific response to the UCS.
In the examples in this chapter, the UCR is salivation for the dog in the salivary example, leg withdrawal for the sheep in the leg withdrawal experiment, eyeblinking for the human in the eyeblink experiment, knee jerking in Twitmeyer’s knee jerk experiment.
Conditioned Response (CR)
The conditioned response (CR) is a response that the CS evokes after paired presentations of the CS and the UCS. In T2 of Fig. 2.2, the CR appears earlier than the UCS appeared in T1.
A dotted outline in T2 of Fig. 2.2 indicates the place where the UCR used to appear. When the CR anticipates the UCS in this way, it leaves a distinct record on every trial. Sometimes, the CR fails to anticipate the UCS. In those cases, experimenters must use probe trials. Probe trials present the CS alone to see if a CR appears without any UCS.
In the examples in this chapter, the CR is anticipatory salivation for the dog in the salivary example, anticipatory leg withdrawal for the sheep in the leg withdrawal experiment, anticipatory eyeblinking for the human in the eyeblink experiment, and anticipatory knee jerking in Twitmeyer’s knee jerk experiment.
In Fig. 2.2, as in most experiments, the CR is similar to the UCR in form and direction, but the CR can be a decrease in response where the UCR was an increase. D. A. Powell, Gibbs, Maxwell, and Levine-Bryce (1993), for example, used lights, tones, and touch stimuli as CSs and painful shock as the UCS. The UCR that the shock evoked in the rabbit subjects of this experiment was always an increase in heart rate. The CR, however, was a decrease in heart rate for all three types of CS.
Response Measures
The amplitude (amount, magnitude) of a response is measured in different units depending on the experiment and the technical resources of the experimenter. Pavlov measured the amount of saliva by reading graduated lines on a measuring tube. Other experimenters have measured leg withdrawal in centimeters of movement, eyeblinks in millimeters of movement, and respiration in millimeters of chest expansion.
The response lines, R, in Fig. 2.2 are irregular, wavy lines to indicate the fact that the baselines are not flat zero values under the conditions of any practical experiment. There is always some fluctuation in the level of response. Hungry dogs frequently salivate, human beings frequently blink their eyes, and sheep often flex their legs without any experimental intervention. The amplitude of response to the UCS and CS must always be relative to the average level of background fluctuation.
The latency of a response is the time between the onset of a stimulus and the initiation of the response. This is a basic response measure used throughout experimental psychology. The OR, CR, and UCR all have latency as well as amplitude. Figure 2.2 shows the latency of the UCR beginning with the onset of the UCS and ending when the amplitude reaches its maximum level. The level of the amplitude measure that the experimenter uses to define the endpoint of the latency is arbitrary. Individual experimenters can choose some middling value or some value that is just significantly higher than the background fluctuations, whatever seems most useful, just as long as they describe it precisely and stick to the same value throughout the experiment.
The amplitude measure must always refer to the amount of response within some specified time interval. Pavlov and his students would never present a CS and then wait hours or even many minutes before reading the level of saliva collected in the test tube. Because the amplitude measured in any trial is always the amount of response recorded within an arbitrarily defined time interval, and the latency measured in any trial is always the amount of time recorded before an arbitrarily defined amount of response, the two response measures are always interdependent. The experimenter decides which measure to make arbitrarily constant and which to measure as a variable. Usually, the decision depends on the cost and convenience of the technology available to the experimenter.
Pavlov and many other experimenters often report results in terms of the probability of response, defined as the number of trials in which there was a recordable response divided by the total number of trials in which the CS appeared. When experimenters report CRs in this way, of course, they are only counting as responses those trials in which there was a response of more than some arbitrarily chosen amount within some arbitrarily chosen interval. They count responses of lower amplitude or longer latency as zero. Responses of higher amplitude or shorter latency are always counted as one. Thus, responses are counted as all-or-none even though they usually vary in amplitude and latency.
Acquisition
Acquisition is a series of trials in which the CS and UCS are paired.
Extinction
Extinction is a series of trials in which the CS appears without the UCS.
Resistance to Extinction
Usually, the CR weakens or ceases entirely during extinction. The rate of decline is called resistance to extinction. Responses persist for many trials without the UCS when resistance is high, but die out soon when resistance is low. Experimenters use one of the standard response measures—latency, amplitude, or probability—to measure resistance to extinction. Sometimes experimenters run a fixed number of extinction trials and sometimes they continue running extinction trials until responding drops below some criterion.
Many experiments must use performance during extinction trials as a measure of conditioning during acquisition. Suppose, for example, that after 100 acquisition trials all response measures in a salivation experiment have reached a maximum and stay at the same level for the next 100 trials. How can the experimenter find out whether conditioning is stronger after 200 trials than after 100 trials even though all measures of acquisition have leveled off? The question can be answered by extinguishing one group after 100 trials and a second group after 200 trials to see if the groups differ in resistance to extinction. In cases like this, two groups that performed identically during acquisition may show differences in resistance to extinction.
There are other cases where certain theoretical questions must be investigated by measuring resistance to extinction. These appear in later chapters of this book as these questions become relevant.
Spontaneous Recovery
Pavlov discovered that extinction is only temporary. After a rest, extinguished CRs usually recover substantially, sometimes completely. In one salivary experiment with a dog, Pavlov (1927/1960) reported that, after a series of seven extinction trials, amplitude of the CR decreased from 10 drops of saliva to 3 drops and latency increased from 3 seconds to 13 seconds. The experimenters let the dog rest for 23 minutes. On the first trial after the rest without intervening trials, the CR rose to six drops and the latency dropped to 5 seconds.
Spontaneous recovery after a rest can be tested with or without the UCS. With the UCS, it is called reacquisition; without the UCS, it is called reextinction (see Fig. 10.1). Spontaneous recovery is a robust phenomenon that appears in virtually all experiments. Plainly, the CR is neither forgotten nor eliminated by extinction. Chapter 10 discusses the theoretical significance of spontaneous recovery.
Habituation
Figure 2.2 illustrates the weakening and eventual loss of the OR over a series of trials. What happens when the experimenter repeats the CS before presenting the UCS? The result is usually the same weakening and eventual loss of the OR. The response habituates. What happens if the experimenter now begins pairing the CS and UCS? Conditioning, if it occurs at all, is usually much slower and more difficult after habituation of the OR to the CS. The effect is sometimes called latent inhibition (Albert & Ayres, 1989; Lubow, 1989).
Habituation can be stimulus-specific. Whitlow and Wagner (1984) repeatedly presented tones of 530 Hz or 4,000 Hz to rats and found selective weakening of the OR, which they measured by the constriction of surface blood vessels. Habituation was selective in that repetitions of the habituated tone evoked a much weakened OR, while a new, unhabituated tone evoked the full OR. Because habituation can be stimulus-specific, it is a type of conditioning in its own right.
Within an hour of birth human infants begin to fixate objects in their field of view (Prechtl, 1974). Fixation is a kind of orienting response, that appears and then habituates if the object stays the same. The amount of time that newborn babies spend fixating on a particular object, before they habituate and their eyes wander, depends on the amount of structure in the figure. Friedman, Carpenter, and Nagy (1970/1973) studied the habituation of this OR in human neonates when the stimuli consisted of black-and-white checkerboard patterns. Friedman et al. presented either a 2 × 2 or a 12 × 12 checkerboard pattern, repeatedly, for 60 seconds at a time. When the amount of fixation fell below the initial average for an infant, the experimenters switched to the other checkerboard pattern. The amount of fixation time recovered to the initial level almost immediately for most of the subjects and the infants scanned the new target with widening eyes as they had initially scanned the first target. Habituation was specific to the first target.
M. H. Bornstein and Benasich (1986) found very similar results with 5-month-old human infants using color slides of human faces and simple red geometric forms as targets. After habituation with one target, fixation and scanning recovered for a fresh target. Reviewing 14 studies of this phenomenon, Bornstein and Sigman (1986) found that the amount of decrement and recovery of attention in habituation tests of this sort correlated with scores on standardized tests of intelligence and language development 2 to 8 years later.
Sensitization
When the UCS is painful or noxious, experimenters often find sensitization which is the opposite of habituation. In an early experiment, Grether (1938) first established that caged monkeys made little response to the sound of a bell. Then he frightened the monkeys several times with loud, noisy powder flashes. After the powder flashes the same bell evoked fright responses. Because the effect of sensitization resembles a conditioned response to a stimulus that was never paired with the UCS, this phenomenon is often called pseudoconditioning.
Later, Harlow and Toltzien (1940) demonstrated a similar sensitization effect in cats by presenting a series of 10, 20, or 30 shock trials before the first presentation of a CS. The greater the number of shock trials, the greater the fear response to the CS on its first presentation. Tests were postponed for 5 minutes, 3 hours, or 24 hours after the shock trials. The new CS evoked about the same amount of response in each of these conditions, so an explanation in terms of generalized excitement seems unsatisfactory. Presumably, such excitement (fear, anxiety) would subside in 24 hours. An alternative possibility seems more likely: that responses to shock become conditioned to the experimental room, the apparatus, and the experimental situation in general (Bouton, 1993).
If the CS appears for several trials without the UCS, the CR gradually weakens or ceases completely. This is extinction, of course. Following extinction a few pairings of the CS and UCS can completely restore the conditioned response. This is reacquisition. Presenting the UCS alone, without the CS, for several trials can also restore the CR. In an extensive study comparing several procedures, Harris (1941) measured finger withdrawal of human subjects with painful shock as the UCS. The CS was a loud tone that lasted 4.75 seconds; the UCS was a .25-second shock. All groups received a total of 80 shocks. The measure of the effectiveness of the different procedures was the percentage of conditioned responses on a series of 10 trials without shock. Harris’ results actually show more CRs after sensitization (trials with shock alone before the first trial with a CS) than any other procedure that he used. The implication of these and many experiments since (e.g., Servatius & Shors, 1994) is that sensitization may be a part of all conditioning with a painful or noxious UCS.
All good experiments include some control for the effect of sensitization. This is accomplished by exposing one group to the UCS without previous pairing of CS and UCS. Viewed another way, however, if sensitization is specific to the experimental situation and can last as long as 24 hours, it is an elementary form of conditioning, like habituation rather than an annoying artifact.
RELATIONSHIP BETWEEN CONDITIONED AND UNCONDITIONED RESPONSES
The traditional view treats the CR as if it were the same as the UCR that the UCS elicited on the first trial. In this view, the dog that formerly salivated only to food should, after conditioning, salivate in the same way to the metronome. The human that formerly blinked at an air puff should, after conditioning, blink in the same way to the light. This is clearly incorrect. The CR and UCR are rarely, if ever, strictly the same, and the CR is hardly a replica of the UCR (D. A. Powell et al., 1993).
Early experiments tended to use crude measuring instruments and seldom reported more than a single index so that the CR seemed to duplicate the UCR. Thus, Culler, Finch, Girden, and Brogden (1935) and Kellogg (1938) reported that, in the early stages of conditioned leg withdrawal, dogs responded to the CS just as they had to the shock. Similarly Pavlov’s records of salivation are seldom precise enough to make exact comparisons between CRs and UCRs. As early as 1948, however, with improved apparatus Konorski could detect clear differences between CRs and UCRs, and this is the usual finding with modern sensitive instruments.
If the CR and UCR are different, what is the relationship between them? The most popular explanation is that the CR is either a fractional component of the UCR or a preparation for the UCS.
Conditioned Response as a Fractional Response
In many experiments, the CR is weaker than the UCR or it is only a component part of the UCR. The amplitude of the conditioned salivary response, for example, is usually significantly lower than the unconditioned salivary response. Moreover, where the UCR to food consists of lapping at the food, salivating, chewing, and then swallowing, the CR usually includes only one or two of these components. A dog that received shock on the leg right after hearing a buzzer may make a conditioned withdrawal response to the buzzer alone without the vocalization that was a part of the reaction to shock.
Conditioned Response as a Preparatory Response
Some theorists claim that the function of the CR is to prepare the organism for the UCS, which would explain the differences between the CR and the UCR. Detailed descriptions certainly support this view. Zener (1937), for example, conditioned salivary responses to a bell that preceded feeding in the usual manner. Zener recorded complete trials on motion picture film.
Except for the component of salivary secretion the conditioned and unconditioned behavior is not identical. (a) During most of the time in which the bell is reinforced by the presence of food, chewing generally occurs with the head raised out of the food pan but not directed either at the bell or into the food pan, or at any definite environmental object. Yet this posture practically never, even chewing only occasionally, occurs to the conditioned stimulus alone. Despite Pavlov’s assertions, the dog does not appear to be eating an imaginary food. (b) Nor is the behavior that does appear an arrested or partial unconditioned reaction consisting of those response elements not conflicting with other actions. It is a different reaction, anthropomorphically describable as a looking for, expecting, the fall of food with a readiness to perform the eating behavior which will occur when the food falls. The effector pattern is not identical with the unconditioned. (c) Movements frequently occur which do not appear as part of the unconditioned response to food: all the restless behavior of stamping, yawning, panting. (p. 393)
When Zener removed the restraining straps, allowing a much greater range of activity, the preparatory character of the behavior became even more evident. At the conditioned signal for food, the dogs approached the food pan; at another signal, which had been associated with the release of acid into the mouth, the dogs either did nothing (since the acid-delivering tube was not attached), or walked away from the neighborhood of the stimulating devices. Furthermore, when satiated with food, their reactions to the CS not only decreased in amount but qualitatively changed.
Zener’s (1937) detailed description shows the complexity of behavior in the conditioning situation. Plainly, the common expression, “the conditioned response,” is misleading.
GENERALITY
Experimental Subjects
Experimenters have conditioned very simple animals with Pavlov’s procedure. Planaria, for example, are flat worms, about one centimeter long, three millimeters wide, and one tenth of a millimeter deep. They have two light-sensitive eyespots, bilateral symmetry, and a ganglion at their head end. They are so primitive that, if you cut one in two, the head end grows a new tail and the tail end grows a new head. Thompson and McConnell (1955) conditioned planaria in a light-shock situation. The experimental chamber was a plastic basin filled with water. The CS was the onset of two 100-watt lights. The UCS was a 28-milliampere shock passed through the water. The experimenters observed two different conditioned responses, a turning of the head to one side or the other and a longitudinal contraction of the entire body. In 150 trials the probability of one or the other of these responses increased from just under 30% to just over 40% (see also, Carney & Mitchell, 1978; Krasne & Glanzman, 1995; Levison & Gavurin, 1979).
Conditioning is possible at a very early age, even before birth. Hunt (1949), with chick embryos as subjects, paired a bell with electric shock and obtained conditioning of a gross bodily movement on the 15th day of incubation. In some cases CRs continued to appear after hatching. Spelt (1948) reported conditioning in human fetuses 6½ to 8½ months of age. As the UCS, he used a very loud noise, which produced an unconditioned startlelike response. As the CS he used a tactile vibration of the mother’s abdomen. He measured movement of the fetus with tambours attached to the mother’s abdomen. Her ears were covered to prevent her from hearing the CS, of course. After pairings of CS and UCS, movement responses occurred to the CS alone. Dorothy Marquis (1931) conditioned the sucking reactions of newborn human infants to the sound of a buzzer. She also altered the feeding schedule of the neonate to a degree by manipulating the amount of time between nursings (Sameroff & Cavanaugh, 1979, reviewed a body of research that grew out of Marquis’ pioneering work).
Unconditioned Responses
Pavlov and his students only used food and painful shock as UCSs and consumatory responses such as salivation or defensive responses such as leg withdrawal as UCRs. Since Pavlov’s time most experimenters who have used his procedure have also used consumatory responses and defensive responses. Swallowing movements that are a UCR to drinking, pupillary dilation and contraction that are UCRs to sharp changes in illumination, dilation and contraction of surface blood vessels that are UCRs to heating up and cooling down, eye movements that are UCRs to rotation, and even immune reactions to antigens are among the many types of responses that experimenters have successfully used as UCRs in classical conditioning (Turkkan, 1989).
Conditioned Stimuli
From the list of CSs already used in classical conditioning experiments, it looks as though just about any stimulus will serve. Experimenters have used lights of various colors, geometrical forms, and rotating objects; pure tones, horns, buzzers, bubbling water, and metronomes; odors, tastes, and surface pressure on particular spots of the skin. Pavlov tended to prefer continuing stimuli, such as metronomes, electric fans, and rotating disks. Later experimenters tend to prefer stimuli with sharp onset, such as flashes of light, clicks, and touch. An increase or decrease in a stimulus or the termination of a stimulus can also serve as a CS.
Arbitrary Stimuli
If the conditioned result is about the same whether the CS is a tone or a light, a high-pitched tone or a low-pitched tone, a red light or a green light, then we can call it an arbitrary stimulus, Sa. The arbitrariness of stimuli in conditioning experiments is only relative, however.
Chapter 1 described how some colors attract bees more than other colors. It is easier to condition bees to return to those colors. Experiments to see whether noisy flowers are more attractive than silent flowers have not yet appeared. We do know that bees sense noises in the hive because the vibrations made by the dancing scouts are essential to the response of the recruits (Kirchner, Dreller, & Towne, 1991).
An extensive series of experiments with rats that started with Garcia and Koelling (1966) demonstrated that the strength of conditioning can depend critically on the appropriateness of the CS to the UCR. In the typical experiment, innocuous chemicals with novel tastes are added to drinking water. After drinking the novel-tasting water, rats are made ill by injections of poison. After they recover, the rats get a choice between the novel-tasting water and plain tap water. They drink dramatically less of the novel-tasting water. The result is called taste-aversion conditioning.
Instead of giving the water a novel taste, the experimenters also sounded buzzers and lighted lights when the rats drank. After illness and recovery, these rats got to choose between a bottle with the noise and lights or a bottle without noise and lights. For rats, illness has markedly less effect on later drinking if the CS is a novel noise or light than if the CS is a taste. Experiments have also conditioned aversion with electric shock. Garcia and Koelling (1966) gave rats painful shocks when they drank novel-tasting water and also when they drank noisy, lighted water. For shocked rats, noise and light depressed drinking more than novel tastes.
Usually, the sensory dimension has a critical effect on conditioning: Different sensory dimensions are more effective for different kinds of conditioning. If the experimenter can arbitrarily choose stimuli within a dimension (colors, tones, tastes) and get much the same effect with a range of stimuli within that dimension, then we call the stimuli arbitrary, with the understanding that this only means relatively arbitrary.
Ethology of Stimulus/Response Compatibility
While it may be true that virtually any animal can be conditioned, that virtually any stimulus can serve as a CS, and that virtually any response can serve as a CR, conditioning is easy with some CS/CR combinations and difficult with others. Differences in the conditionability of particular stimulus/response combinations are species-specific. Naturally, most experimenters are mostly interested in getting results, so they tend to reuse combinations of stimulus, response, and species that have worked well in the past. In recent times, however, more experimenters have investigated species-specific characteristics that favor certain stimulus/response combinations over others. This leads to a less arbitrary and more ethological and ecological view of conditioning.
Keith-Lucas and Guttman (1975), for example, conditioned aversion in rats with painful shock as UCS. They found dramatic conditioned effects when they used a toy hedgehog as CS and only weak effects, if any, when they used a flat, striped surface as CS. Ayres, Haddad, and Albert (1987) and Van Willigen, Emmett, Cote, and Ayres (1987) also conditioned aversive responses in rats using painful shock as the UCS. They found dramatically different amounts of conditioning depending on which CS they used. Conditioning was strong when they used a brief tone as a CS and weak when they used a flash of light.
TIME AND SEQUENCE
In animal behavior, sequence is critical. Suppose I step on your foot and then say, “Excuse me.” Your response would be very different if I said, “Excuse me,” and then stepped on your foot. The amount of time between events also matters. If “excuse me” and stepping on your foot are separated by an hour, you might fail to associate them at all.
In most experiments in classical conditioning, the CS begins just before the onset of the UCS. It may either overlap with the UCS in time or terminate before the appearance of the UCS. The arrangement illustrated in Fig. 2.2 is by far the most studied and most discussed procedure.
Interstimulus Interval (ISI)
The ISI is measured from the onset of the CS to the onset of the UCS. When the CS starts before the UCS, the ISI is positive and the arrangement is called forward conditioning. When the CS starts after the UCS, the ISI is negative and the arrangement is called backward conditioning. When they start at the same time, the ISI is zero and the arrangement is called simultaneous conditioning.
Length of the Interstimulus Interval
No one would expect much conditioning if we sounded the tone on Monday and gave the dog food on Tuesday. Sounding the tone at noon and giving the food at 1:00 p.m. also seems like a poor arrangement. In general, the shorter the ISI the better the conditioning, as most people would expect. Simultaneous appearance of CS and UCS is the shortest possible interval because in that case the ISI is zero. This extreme value is usually unfavorable for conditioning, however.
The importance of some small amount of offset between CS and UCS suggests that temporal offset plays a significant role in the conditioning process. Many experiments have looked for the shortest or the most favorable ISI in the hope of discovering a critical factor in the process. The shortest possible ISI must depend on the speed of bioelectrical conduction. Conduction in the nervous system is a bit slower than conduction in household wiring, but not much slower and the distances are quite short. Of course, there is a good deal of switching circuitry in the nervous system. There are six synapses between the retina and the optical cortex and a few more at least before an impulse can get to the motor cortex and a few more synapses still before an actual muscle moves. Transmission through all of that neural circuitry must cost some time.
The amount of time it takes for an animal to respond to a stimulus such as a light or a tone with a leg or finger movement is called the reaction time and psychologists have studied this interval thoroughly in many animals with many stimuli and many responses. Under favorable experimental conditions simple reaction time (without complex discrimination) takes between one sixth and one eighth of a second from the moment that a light or tone appears to the moment of an overt response such as a leg or finger movement (Wickens, 1984, p. 338). Therefore, the minimum time required for one complete sequence of stimulus in and response out should be between one sixth and one eighth of a second.
Experimenters have studied the ISI in forward classical conditioning (CS then UCS) very thoroughly, and the most favorable interval found in a very large number of forward conditioning studies is roughly one half of a second, depending on experimental particulars, such as subjects, responses, and stimuli (J. F. Hall, 1976, pp. 110–119; Kehoe & Napier, 1991, pp. 195–196). Half a second is at least three to four times as long as the reaction time. And, that is only the most favorable interval; significant amounts of conditioning occur with even longer intervals, under some conditions with intervals of several seconds between CS and UCS. Favorable ISIs that are so much longer than necessary for simple neural circuits to fire indicate strongly that there is more to the process of conditioning than simple association of the CS with the UCS. Chapter 4 discusses the implications of this finding in more detail.
Much of the advance of modern science depends on viewing nature in terms of continuous variables rather than all-or-none categories. Amount of offset between CS and UCS is a variable. The crudeness of Pavlov’s early experiments constrained him in his theoretical interpretations. Pavlov had to divide ISIs into crude all-or-none categories of forward, backward, and simultaneous because the early instruments available to him were too crude for precise measurement.
FIG. 2.3. Results of varying the ISI. From Spooner and Kellogg (1947), from American Journal of Psychology. Copyright © 1947 by the Board of Trustees of the University of Illinois. Adapted with permission of the University of Illinois Press.
After Pavlov, however, better apparatus permitted experimenters to explore the ISI as a variable (J. F. Hall, 1976, pp. 110–119). Figure 2.3 illustrates a few typical examples. Modern experiments show a continuous curve rather than any sharp breaks that divide intervals into three types: forward, simultaneous, and backward. Yes, a zero ISI is unfavorable, but relatively long positive intervals can be even more unfavorable than zero. Under the experimental conditions reported in Fig. 2.3, forward intervals of 1.5 seconds or more produced less conditioning than a zero interval.
Experimenters who treat all forward intervals as equivalent members of a single all-or-none category, rather than differing values of a continuous variable, may choose a single value for the positive ISI and compare this with a zero interval. The result, of course, depends entirely on the single interval that the experimenter chooses. Thus, in three of four separate experiments, Rescorla (1980) used a forward offset of 5 or 10 minutes and compared that with zero offset in each case. In the fourth experiment, the forward offset that Rescorla compared with a zero ISI was 30 seconds. It is hardly surprising that such unfavorably long positive intervals would produce less conditioning than a zero interval. This example appears at length here because it illustrates the distortion that results when theorists divide a continuous variable into a few all-or-none categories. Experiments designed to support a particular theoretical interpretation often fall into this trap. Much more can be learned from experiments designed to explore an interesting variable for its own sake.
Chapter 4 discusses the theoretical significance of the ISI as a variable in classical conditioning.
CONDITIONED EMOTIONAL RESPONSE
Human anxiety often interferes with normal, productive behavior. People who suffer from anxiety report physiological reactions such as heart palpitations, shortness of breath, and sweaty palms during attacks of anxiety. Many clinical psychologists interpret these symptoms as conditioned fear. If they are correct, then experiments on classically conditioned defensive behavior should be critically relevant to human anxiety.
Physiological responses to fear, such as heart rate, blood pressure, and breathing rate, are relatively difficult and expensive to measure, but instrumental responses, such as lever-pressing in a Skinner box (chap. 3), are easy and inexpensive to measure. The conditioned emotional response (CER) takes advantage of the finding that ongoing instrumental behavior, such as lever-pressing in a Skinner box, is sensitive to the effects of defensive conditioning. Meanwhile, anxiety is debilitating, precisely because it interferes with normal, productive behavior, so conditioned fear that interferes with instrumental responses should be highly relevant to human welfare.
Typically, in a CER procedure, experimenters first train hungry or thirsty rats to press a lever for food or water. Next, they remove the lever from the Skinner box and pair a CS such as a light with a painful UCS such as electric shock. Finally, the experimenters replace the lever in the box and observe the amount of lever-pressing under extinction conditions—that is, no food and no shock. During this test phase of the experiment, however, the experimenters probe by presenting the CS alone to see how much the CS suppresses lever-pressing. If the CS suppresses lever-pressing, this is attributed to an emotional response to the CS that interferes with lever-pressing. The amount of suppression should depend on the amount of emotional response that was conditioned to the CS. It is an indirect way of measuring classical conditioning, but it has the advantage of convenience as well as relevance to the intrusive effects of conditioned fear.
In an important early experiment, Mowrer and Aiken (1954) first trained five groups of rats to press a lever in a Skinner box. Then, after removing the lever from the box, they paired a 3-second flashing light with a 10-second painful electric shock in four different ways as shown in the left-hand panel of Fig. 2.4. Group I received the CS immediately before the onset of the UCS; Group II received the CS immediately after the onset of the UCS; Group III received the CS immediately before the offset of the UCS; and Group IV received the CS immediately after the offset of the UCS. Group V received the same number of shocks and lights as the other four groups, but shocks and lights were separated by at least 2 minutes.
FIG. 2.4. Conditioned emotional response (CER). A flashing light CS was paired with a painful shock UCS for Groups I-IV as shown in the left-hand panel. Group V was a control group with the same number of shocks and lights without pairing. From Mowrer and Aiken (1954), from American Journal of Psychology. Copyright © 1954 by the Board of Trustees of the University of Illinois. Used with permission of the University of Illinois Press.
After those differential treatments, Mowrer and Aiken replaced the lever in the box. The results, shown in the right-hand panel of Fig. 2.4, show that the onset of the shock was the most critical point in time for the suppression of food-motivated lever-pressing. The closer the CS to the onset of the shock the greater the suppression. The most effective time was just before the onset of the shock. This is exactly what we would expect if the conditioned emotional response in this experiment was an anticipatory defensive reaction to the flashing light.
SOCIAL CONDITIONING
Following Pavlov, most experiments in classical conditioning have used a rather narrow range of consumatory and defensive responses, such as salivation, leg withdrawal, or eyeblinking. Primitive consumatory and defensive responses are directly involved in emotional conditioning, which is why so many psychologists search for the fundamental laws of classical conditioning. From a broader and more ethological point of view, there is a much wider range of adaptive and maladaptive learning that depends on classical conditioning and that has even more significant implications for human behavior.
Imprinting
A brood of young ducklings following their mother is a common sight in the spring. Ducklings that fail to follow their mother die. How do newly hatched ducklings recognize their own mother? For that matter, how do they recognize an adult duck? How do they tell the difference between adult ducks and other objects anyway? The answer is that newly hatched ducklings follow the first moving object that they see. Following the same object for a while conditions them so that they follow only that object and no others.
Experimenters have exposed ducklings to an arbitrary object such as a red ball or a green shoebox moving back and forth in an enclosure. If the first object is a green shoebox, then after following the shoebox for a while, a duckling always follows that green shoebox, ignoring other objects such as a red ball. Meanwhile, a duckling that has been imprinted in this way on a red ball follows the red ball and ignores a green shoebox. Under natural conditions, of course, the mother duck is the first moving object that a duckling sees. Many kinds of social learning in a wide variety of animals show a similar pattern (Suboski, 1990, 1994).
Conditioned Mobbing
Owls, hawks, and other meat-eating birds prey on blackbirds, starlings, and other smaller birds. The smaller birds can often fight off the predator, if they stick together and fly at their enemy from all directions. They usually cannot kill or injure an owl or a hawk, but they can often chase it right out of the area, and possibly condition it to stay away. This pattern of behavior is called mobbing.
The sooner they start the mobbing defense the more effective it is, so the defending birds need to recognize potential attackers as soon as possible. The first bird to spot the attacker sounds an alarm call and starts the mobbing. All of the defenders in the area join in when they hear or see the first bird begin the mobbing.
Curio, Ernst, and Vieth (1978) demonstrated that inexperienced blackbirds can learn to mob arbitrary targets by a process that looks very much like classical conditioning. The experimenters housed groups of wild and cage-reared blackbirds in adjacent cages in such a way that they could show a stuffed owl to the wild birds while preventing the cage-reared birds from seeing the owl. Instead of an owl, the cage-reared birds saw an arbitrary object such as a plastic bottle or a stuffed specimen of a harmless honey-eating bird. When the experienced blackbirds sounded their alarm calls and attempted to mob the owl, the inexperienced blackbirds followed them with alarm calls and attempts to mob the plastic bottle or the honey-eating bird.
Soon the inexperienced birds started mobbing the arbitrary object as soon as they saw it, without waiting for the experienced birds to lead the attack. The experimenters then used the now trained, cage-reared blackbirds to train a fresh group of inexperienced blackbirds to mob an arbitrary object. They then let this second generation condition a third generation of inexperienced birds, and so on up to a sixth generation of blackbirds that learned to mob a plastic bottle by following other blackbirds.
Chemical Alarms
When they are injured, many social fish release a chemical that alarms other fish in the school. Experimenters can observe the alarm reactions of a school of fish confined to a laboratory fishtank by stimulating the live fish with the body fluids of a fish killed outside of the tank. The alarm reaction of zebra fish is fairly elaborate and includes swimming toward the source of the alarm chemical, infusion of the substance with tasting movements, followed by rapid darting movements as if trying to disperse and escape from the tank, followed by forming a compact group, dropping to the bottom of the tank, and swimming back and forth as close to the bottom as possible. In their natural habitat, that last maneuver would stir up mud and possibly hide the group from a predator.
Suboski, Bain, Carty, McQuoid, Seelen, and Seifert (1990) first demonstrated that a small school of zebra fish would carry out the full alarm reaction when their only stimulus was the sight of another school of zebra fish in an adjacent tank that was executing the alarm reaction after receiving a dose of the alarm chemical. The experimenters then conditioned a small school of zebra fish to react to an innocuous substance, called morpholine, by pairing morpholine with the alarm chemical. After conditioning, morpholine alone induced the complete alarm reaction.
The experimenters next placed a second group of zebra fish in the same tank as the conditioned group and stimulated both with morpholine alone. When first stimulated with morpholine, the second group continued to swim normally, but they soon followed the conditioned fish in going through the alarm response. In this way the experimenters conditioned the second group to make the alarm response to morpholine. When the second group was responding by themselves without the first group in the tank, Suboski et al. (1990) used these fish (which had never experienced morpholine paired with the alarm chemical) to condition a third group to make the alarm response to the innocuous morpholine.
In both birds and fish, fairly elaborate and sustained patterns of response can be conditioned by classical conditioning. This powerful technique is not restricted to simple responses, such as salivation and leg withdrawal.
Emotional Words
Human language is often described as the highest cognitive achievement of biological evolution. Psychotherapy entirely based on verbal treatment is called cognitive therapy. Human beings have ethological reactions to words. Reading sad stories and sad poetry in a book makes many people weep. Reading horror stories can evoke physiological responses of fear. These are powerful emotional responses to printed words on a page, to totally arbitrary stimuli.
Human beings respond emotionally to printed descriptions of sadness or horror without having any personal experience with the events described, and without personal experience with anything nearly as sad or as horrible. They can only learn these emotional responses to printed words from other human beings by a process very like those used in social learning experiments conducted with other animals.
SUMMARY
This chapter reviews a small, representative sample of the hundreds and hundreds of experiments that have used the procedures of classical conditioning. Throughout the 20th century, thorough replication has established the basic findings with a wide range of different species, different responses, and different stimuli. Chapter 3 is a similar review of instrumental conditioning. Chapter 4 outlines the implications of these findings.