Chapter 2. The Study of Learning and Behavior “Sit down before fact as a little child, be prepared to give up every preconceived notion, follow humbly wherever and to whatever abysses nature leads, or you will learn nothing.” — T. H. Huxley Preview To engage in the scientific study of behavior, you have to go out of your mind. Nearly everyone, it seems, looks for explanations of behavior inside the mind, a mysterious entity that, in popular thought, resides between our ears and snuggles with, but is separate from, the brain. The scientific study of behavior requires adopting a very different, some will say downright alien, way of accounting for behavior: the natural science approach. Learning Objectives After studying this chapter, you will be able to . . . 2-1 Describe the natural science approach to explaining behavior. 2-2 Explain how psychologists measure learning. 2-3 Compare the benefits and drawbacks of the four different sources of data. 2-4 Describe the role of animals in human learning research. 2-1. The Natural Science Approach Learning Objectives To describe the natural science approach to explaining behavior, you can . . . 2-1-1 Define the natural science approach. 2-1-2 Explain the four assumptions underpinning the natural science approach. Poets, educators, and philosophers have long admired learning, sung its praises, and wondered at its power. But learning has only been the subject of scientific analysis for little more than a hundred years. What does scientific analysis mean when it comes to learning? Answers vary, but many learning researchers take the natural science approach. This approach maintains that learning arises from a natural phenomenon and can and must be accounted for like any other natural phenomenon. The following four assumptions underpin the natural science approach: 1. Something causes natural phenomena. Things do not “just happen”; rather they result from other events. In the 1840s, Dr. Ignaz Semmelweis, a German-Hungarian physician, noticed that women who gave birth in a hospital died of childbed fever more than women who gave birth at home, or even on the street. His attempt to find out why rested firmly on the assumption that the difference in death rates wasn’t a random phenomenon, that something about the hospitals made them more dangerous. This led to research that eventually changed medical practices and saved countless lives. We can never prove that all events have a cause, but science rests on the assumption that effects have causes, including the science of behavior. 2. Causes precede their effects. Dr. Semmelweis assumed that whatever caused patients in a hospital to get childbed fever had to occur before they became ill. In most areas of science, such a statement might appear painfully obvious, yet many people seem to assume that future events can change behavior—in other words, that behavior can precede whatever event causes it. For example, people commonly say that a student studies hard because by doing so they will get a good grade. People assume that the future good grade causes the current studying. Some will counter that students study now, not for the future good grade, but because they currently have an expectation that studying will result in a better grade. And that this expectation causes studying. However, this view fails to take into account that our thoughts often coincide with or follow, rather than precede, the overt act they supposedly cause (Bechara et al., 1997; Libet, 2005; Libet et al., 1983; Libet, Sinnott-Armstrong, & Nadel, 2010; Obhi & Haggard, 2004; Soon et al., 2008). Events (including thoughts) cannot reach into the past to cause changes in behavior. Our proverbial student probably studies because in the past they earned better grades after studying than after not studying. 3. The causes of natural events include only natural phenomena. Mind, spirits, psychic energy, and other mysterious forces have no place in the natural science approach. You cannot explain the movement of tectonic plates (and the earthquakes and tsunamis they cause) by attributing them to God’s anger. Similarly, you cannot explain a person’s overeating by attributing it to a lack of willpower or bad karma. Someone may come up with an original idea, but we do not explain this creativity by attributing it to “the unconscious mind.” To explain behavior, which includes thoughts and feelings (refer to Chapter 1), we must identify the natural events that produce it. These include biological or environmental events. Because learning involves the study of the changes in behavior produced by experience, our main concern in scientifically exploring learning focuses on how events in the individual’s environment change behavior. 4. The simplest explanation that fits the data is best. This fundamental tenet of all sciences, the law of parsimony, means, in part, that the fewer assumptions (unverified events) required by an explanation, the better. In the 2nd century A.D., the Egyptian astronomer Claudius Ptolemy proposed the dominant explanation of astronomical events that lasted for about 1,500 years: the geocentric theory. According to the geocentric theory, the earth sits at the center of the universe, and the sun creates day and night by revolving around the earth once every 24 hours. But as astronomers gathered facts about the stellar bodies, the geocentric theory required modification again and again to accommodate facts that didn’t fit it. These modifications resulted in a complicated and inelegant theory. In 1543, the Polish astronomer Nicolaus Copernicus put forward a radically different explanation of astronomical events. His heliocentric theory proposed that the earth rotates on its axis (thereby creating periods of day and night every 24 hours) and revolves around the sun once every year. Both the geocentric and heliocentric theories attempted to explain the known facts about the universe, but astronomers quickly accepted the simpler and far more elegant heliocentric theory. Today, many people, including many psychologists, accept explanations of behavior that rely on hypothetical events that supposedly take place in the mind. One example, Freud’s Thanatos, proposes that people have an unconscious drive toward self-destruction that causes people to engage in risky behavior, such as misusing drugs, picking fights, and driving recklessly. The law of parsimony suggests that if we can account for such behavior by observable natural phenomena, such as heredity and environmental events, then speculations about mysterious and unobservable forces in a conscious or unconscious mind do not improve our understanding of the behavior (Moore, 2010; Palmer, 2003; Schall, 2005). For many people, applying the assumptions of natural science to behavior requires a major shift in their view of human nature. Fear not: The shift will not turn you into a robot. The natural science approach rests on the four assumptions just described. However, studying learning requires more than those assumptions. It also includes methods consistent with those assumptions. Let us begin with ways of measuring learning. Section Review This text takes the natural science approach to behavior. It assumes that all natural phenomena have causes; that causes are natural phenomena and precede their effects; and that the simplest explanation with the fewest assumptions best accounts for behavior. These assumptions cause most people no problem when applied to fields such as physics, chemistry, and biology, but when applied to behavior, particularly human behavior, people, even physicists, chemists, and biologists, often find them difficult to accept. 2-2. Measures of Learning Learning Objectives To explain how psychologists measure learning, you can . . . 2-2-1 Describe seven ways that psychologists often measure learning. 2-2-2 Explain how a cumulative record can be used to measure learning. Measuring learning requires measuring changes in behavior. Many ways of measuring behavior (in ways consistent with the natural science approach) exist; we will consider the most basic ones here. 1. Errors. Researchers commonly measure learning by seeking a reduction in errors. We can say a rat has learned to run a maze to the extent that it goes from start to finish without taking a wrong turn. As training progresses, the rat will make fewer and fewer errors (refer to Figure 2-1). Similarly, we can say a student has learned a spelling list when they can spell all the words without error. A researcher may measure progress in reading by recording the number of times a student stumbles over a word, with each such stumble counting as one error. Figure 2-1 Details Adapted from Tolman & Honzik, 1930. Errors as a measure of learning. A decline in the number of errors (such as entering the wrong alleys of a maze) provides a measure of learning. 2. Topography. Researchers can measure learning as a change in the topography of a behavior, or the form a behavior takes. (You might think of a topographic map, which shows the form of earth’s surface.) Mirror tracing provides one example of how topography can serve as a measure of learning. In a mirrortracing task, a person traces a form while looking at its reflection in a mirror. This task provides quite a challenge, and at first the pencil line meanders wildly. With practice, however, a person can trace the shape rather neatly (refer to Figure 2-2). The change in topography serves as a measure of learning. Computers can now track topographical changes in three-dimensional space. For example, computers can track the movements of a fish through an aquarium (Pear & Chan, 2001). Figure 2-2 Details Adapted from Kingsley & Garry, 1962, p. 304. Topography as a measure of learning. A person attempted to trace between the lines of a star while looking at the figure’s image in a mirror. On the first trial, the participant’s performance was shaky and erratic; by trial 15, the performance was much improved. The change in topography is a measure of learning. 3. Intensity. We can also measure learning by noting changes in the intensity of a behavior. Once a laboratory rat learns to press a lever, a researcher may increase the resistance of the lever so that the rat has to exert greater force to depress it. The increase in pressure exerted by the rat serves as a measure of learning (refer to Figure 2-3). The same sort of phenomenon occurs outside the laboratory. Having taught a child to sing a song, we can then teach them to sing it more softly. I once impressed my neighbors to no end by teaching my dog, Sunny, to “speak” (bark loudly) and to “whisper” (bark softly) on command. Figure 2-3 Details After Hull, 1943, p. 305. Response intensity as a measure of learning. These frequency distributions show variations in the force exerted by a rat in depressing a lever. The first shows the distribution when all lever presses with a force of at least 21 grams produced a food pellet. The second shows the distribution when the requirement raised to 38 grams. The increase in force exerted provides a measure of learning. The various measures of learning can overlap. In the tracing task in Figure 2-2, for example, we might count the number of times the pencil marks fall outside the star. If we count each such point as an error, we can measure learning as a reduction in errors rather than a change in topography. 4. Speed. A change in the speed with which an organism performs a behavior serves as another measure of learning. The rat that learns to run a maze reaches the goal faster than an untrained rat (refer to Figure 2-4). In the same way, a first grader takes a long time to recite the alphabet at the beginning of the school year but runs through it with the speed of an auctioneer by the end of the year. Likewise, a surgeon usually gets faster at operations the more they perform them. A surgeon once told me that when they first started doing a particular operation, it took them nearly an hour, but after having done the procedure a hundred times, it took only about ten minutes. As these examples illustrate, learning often means doing something more quickly. However, learning can also mean a reduction in speed. Hungry children tend to eat quickly; learning good table manners means learning to eat more slowly. Figure 2-4 Details Adapted from Tolman & Honzik, 1930. Speed as a measure of learning. The decline in the average time it takes rats to run a maze indicates learning. 5. Latency. Change in latency, the time that passes before a behavior occurs, can also provide a good measure of behavior. We will find in the next chapter that a dog can learn to salivate to the sound of a ticking metronome. As the training proceeds, the interval between the ticking and the first drop of saliva shortens; this change in latency provides a measure of learning (refer to Figure 2-5). Similarly, a student beginning to learn the multiplication table pauses before answering a question such as “How much is 5 times 7?” With practice, the pauses become shorter, and eventually the student responds without hesitation. This decrease in hesitation, or latency, serves as a measure of learning. Sometimes, however, learning involves an increase in latency. When someone tells us, “Don’t make snap judgments!” they instruct us to delay before judging; that represents an increase in latency. Figure 2-5 Details Compiled from data in Anrep, 1920. Latency as a measure of learning. There is a long delay before the response (in this case, salivating) appears, but this latency gets shorter with more trials. 6. Rate. Researchers often measure learning as a change in the rate at which a behavior occurs. This term refers to the number of occurrences per unit of time. A pigeon may peck a disc at the rate of, say, five to ten times a minute. The experimenter may then attempt to increase or decrease the rate of disc pecking. The resulting change in rate provides a measure of learning. Similarly, a person may practice receiving Morse code by telegraph. If the rate of decoding (the number of letters correctly recorded per minute) increases, we say that they have learned (refer to Figure 2-6). Learning can also mean a decrease in the rate of behavior. A musician may learn to play the notes of a composition more slowly, for example. Figure 2-6 Details Adapted from Bryan & Harter, 1899. Rate as a measure of learning. Number of Morse code letters correctly received. The increase in rate of decoding is a measure of learning. Rate and speed are related, but not identical, measures of learning. Rate allows us to observe subtle changes in behavior and therefore serves as an especially useful measure of learning. In laboratory studies, researchers once tallied behavior rate by means of an electromechanical cumulative recorder. With this device, an inked pen recorded every occurrence of the behavior on a sheet of paper that moved at a steady pace under the pen. So long as the behavior in question did not occur, the pen made a straight line along the length of the paper. When the behavior occurred, the pen moved a short distance at a right angle to the length of the paper (refer to Figure 27a). The higher the rate of behavior, the more pen movements and the steeper the slope of the ink line; the lower the rate, the flatter the line. Each point on the line indicates the total number of times the behavior has occurred as of that moment, so the graph provides a cumulative record (refer to Figure 27b). Because the data line accumulates, it can never fall below the horizontal. Today, computer software has replaced cumulative recorders, which just gather dust in storage rooms. Fundamentally the cumulative record produced by the computer remains essentially the same as the one produced by the older electromechanical devices. Figure 2-7 Details Cumulative record and response rate. In a cumulative recorder, an inked pen moves at right angles each time a response occurs (A), thus yielding a cumulative record of behavior (B). A change in the rate of behavior indicates learning. Concept Check 1 What happens to the slope of the cumulative record if the rate of a behavior is increasing? What does a flat record indicate about behavior? SHOW ANSWER 7. Fluency. Fluency combines errors and rate to measuring learning by counting the number of correct responses per minute. For example, a student who calls out the answers to single-digit addition problems (such as 9 + 4, 7 + 3, and 5 + 9) provided by a teacher may call out 12 answers in one minute. If they correctly answer ten of those questions, then their fluency measure is ten correct per minute. If, after instruction or practice, their fluency rate increases to 22 correct per minute, their change in fluency provides a clear measure of learning. B. F. Skinner first devised a cumulative recorder, but he was not the first to record behavior cumulatively. That honor goes to biologist James Slonaker, who used it to measure lifelong activity in the white rat (Todd, pers. comm., 2001; refer to Slonaker, 1912). Researchers most commonly use these seven measures of learning. The study of behavior requires not only having reliable ways of measuring learning but also trustworthy ways of obtaining data. Section Review Researchers most often measure learning as a change in number of errors, topography, intensity, speed, latency, rate, or fluency. Other ways of measuring learning exist too. We cannot study learning unless we can measure it in some precise way. When you consider them, remember that measured changes in behavior do not occur as the result of learning; they are learning. That represents an important distinction that a lot of people have trouble with. 2-3. Sources of Data Learning Objectives To compare the benefits and drawbacks of the four different sources of data, you can . . . 2-3-1 Describe anecdotes, case studies, descriptive studies, and experiments. 2-3-2 Give an example demonstrating the limitations of anecdotes. 2-3-3 Provide three limitations of case studies. 2-3-4 Provide one way that descriptive studies offer more explanatory power than case studies. 2-3-5 Provide one limitation of descriptive studies. 2-3-6 Describe the roles of independent and dependent variables in experiments. 2-3-7 Explain the differences between between-subjects experiments and within-subjects experiments. 2-3-8 Differentiate between the experimental group and the control group. 2-3-9 Explain the importance of random assignment in interpretation of experimental studies. 2-3-10 Describe how matched sampling and within-subjects experiments aid in interpretation of experimental research. 2-3-11 Identify the advantages of an ABA reversal design. 2-3-12 Identify the advantages of using an experimental design. 2-3-13 Defend why experimental research is often artificial. Those who study learning have various sources of evidence available to them. Each has strengths and weaknesses. We will begin with the simplest, most common, and least reliable: anecdotes. Anecdotes Anecdotes, first- or secondhand reports of personal experiences, can include specific information about measures of learning, such as the number of errors made, but they usually prove less specific. You can often identify anecdotes by phrases such as “In my experience . . .” and “I’ve found that. . . .” Sometimes anecdotal evidence takes on the character of common wisdom: “They say that . . .”; “It’s common knowledge that . . .”; and “Everybody knows that. . . .” Unfortunately, what “everybody knows” is not always correct. Bloodletting persisted as a treatment for medical disorders for about 2,000 years because “everybody knew” that it worked. Actually, it almost certainly killed many more people than it helped. One of its more famous victims may have been George Washington, the first president of the United States. People can point to anecdotal evidence to support all kinds of principles and practices, but it can be difficult to sort out which anecdotes to believe. Perhaps providing an anecdote will prove the best way to demonstrate the trouble with them. I once spoke to a psychologist who complained about the school his child attended. “You won’t believe this,” he said, “but they don’t teach the kids how to read.” He was talking about a method of reading instruction known as Whole Language, in which teachers read to students and expose them to books but don’t teach them to sound out words. Teachers who use this approach assume that students will pick up reading skills through a kind of osmosis, without any formal instruction. Unfortunately, this approach did not work for the psychologist’s child, whom I will call Sam. I asked my friend what he did about it. “We got a how-to book,” he said, “and taught Sam to read.” Although researchers have discredited Whole Language as a way of teaching beginning reading skills (Chall, 1995; Treiman, 2000), this anecdote does not qualify as hard evidence against it: The psychologist may have exaggerated his child’s failure to read; the problem may have arisen due to a poor teacher, an unnoticed illness, or some other variable besides the method of instruction. We simply cannot say from such anecdotes that Whole Language does not work. You should also not trust positive anecdotes. Let us suppose, for example, that you meet a teacher devoted to Whole Language. You ask the teacher if the method works. They reply, “Absolutely! I’ve seen it work. I had one student who picked up reading easily without any formal instruction at all. Their name was Sam. . . .” You notice the problem: The teacher who “sees” an instructional method working may not know about other important variables, such as instruction in the home. Despite their limitations, anecdotes do have some uses. They can provide helpful leads, and they keep us in contact with “popular wisdom,” which, after all, is not always wrong. Still, we require better evidence for a science of learning (Spence, 2001). Case Studies We get a slightly better grade of data with the case study. Whereas anecdotal evidence consists of casual observations, a case study examines a particular individual in considerable detail. Medicine often makes use of the case study method. Doctors may study a patient with a mysterious symptom with great care as they attempt to understand their illness. Economists also do case studies. They may study a company to find out why it failed or succeeded. Similarly, educational researchers might do a detailed study of a teacher or school that gets particularly good results. And researchers often use case studies in attempting to understand abnormal behavior, such as delusions. Case study evidence, however, comes with serious problems. First, doing a case study takes a good deal of time. Because of this, generalizations are often based on very few cases. If those few cases do not represent the larger group, we may draw incorrect conclusions about that group. Second, the case study cannot answer certain questions about behavior. We cannot, for example, use the case study to determine whether falling off a ladder produces a fear of heights. We may interview a person who fell off a ladder and subsequently developed a fear of heights, but this does not establish that the fall caused the fear. For years, many clinical psychologists and psychiatrists insisted that being attracted to someone of the same gender was a neurotic disorder because their clients who identified as gay or lesbian all displayed neurotic behavior. Then, in the 1950s, Evelyn Hooker pointed out that the heterosexual clients of clinicians also demonstrated neuroticism, but no one concluded that heterosexuality arose from a form of neurosis (refer to Chance, 1975). Third, much of the data obtained in case studies comes not by direct observation of the participant’s behavior but from what the participant or other people report about the participant’s behavior. Such reports often prove notoriously unreliable. When appropriate, the case study improves upon the anecdote because at least researchers obtain the data in a fairly systematic way. But a stable science of behavior requires a better foundation than the sandy soil of the case study. Science requires better control. Descriptive studies provide one way of getting better control. Descriptive Studies In a descriptive study, the researcher attempts to describe a group by obtaining data from its members—often by conducting interviews or administering questionnaires. To devoted advocates of the case study, the descriptive study seems superficial. But by gathering data from many cases and analyzing the data statistically, the descriptive study reduces the risk that a few unrepresentative participants will lead to false conclusions. In a typical descriptive study, we might ask people (in interviews or by means of a questionnaire) about their fears and their childhood experiences. We might then compare the experiences of those who have phobias with those who do not. Statistical analysis might then suggest whether any reliable differences between the two groups exist. We might find, for example, that people with phobias had overprotective parents more than those people without phobias. Descriptive studies represent a vast improvement over case studies, but they have their limitations. Although descriptive studies can suggest hypotheses to explain a phenomenon, they cannot test those hypotheses. We might find that phobia victims describe their parents as overprotective twice as often as others, yet overprotective parenting may not play a role in producing phobias. Perhaps, for example, overprotective parenting correlates with some other variable, such as a genetically based high levels of anxiety, and this other variable actually accounts for the higher incidence of phobias. An experiment offers the only way to determine that. Experimental Studies In an experiment a researcher manipulates one or more variables (literally, things that vary) and measures the effects of this manipulation on one or more other variables. We call the variables the researcher manipulates independent variables; we call those that vary freely dependent variables. In learning experiments, researchers typically use some sort of experience (an environmental event) as the independent variable and typically measure some kind of change in behavior as the dependent variable. Researchers can use many different kinds of experiments, but all fall into one of two types: betweensubjects designs and within-subjects designs. In between-subjects experiments, the researcher typically identifies two or more groups of participants. (Some call these experiments between-group or group designs.) The researcher then varies the independent variable across these groups. Suppose you wish to study the role of certain experiences on aggression. You might assign people to one of two groups and expose those in one group to the experiences you think will produce aggression. We refer to the participants exposed to the aggressioninducing experience as belonging to the experimental group and those not exposed to it belong to the control group. (The participants need not actually appear in groups; here, the term group refers only to assignment to experimental or control conditions.) Next you compare the tendency of participants in the two groups to behave in an aggressive manner. If the experimental participants behave more aggressively than the control participants, then the experiences provided may have caused this difference. Although experiments involving two groups commonly occur, researchers often conduct betweensubjects experiments with many groups. In an experiment on aggression, you might have several experimental groups in which you expose each to a different kind or a different amount of aggressioninducing experiences. You would then compare the performance of each group not only with the control group but also with every other experimental group (refer to Figure 2-8). Figure 2-8 Details Compiled from data in Rosenkrans & Hartup, 1967. Between-subjects design experiment. Data are compared from different individuals in different conditions. Essentially, with a between-subjects design, an independent variable differs across participants. Researchers then assume that any differences in the dependent variable occur because of differences in exposure to the independent variable. The validity of this assumption rests on the extent to which the participants being compared share similarities. It would not do if one group were appreciably older than the other, for example, because any differences in the results might occur because of differences in age rather than because of the independent variable. Likewise, the two groups should not differ in health, gender, education, or a host of other variables. Concept Check 3 What is the essential element of a between-subjects design? SHOW ANSWER To minimize such differences, researchers generally assign participants at random to one of the groups. Sometimes they do this by flipping a coin: If the coin comes up heads, the participant goes into the experimental condition; tails, the participant goes into the control group. Through such random assignment, any differences among the participants distribute more or less equally among the groups. However, with small groups even random assignment leaves open the possibility of differences among the groups, so the more participants in each group, the better. Matched sampling provides another way to reduce pretreatment differences among groups. In matched sampling, researchers identify participants with identical features. They may match animals for age and sex quite easily. We can not only match human participants for age and gender identity but also for other features such as IQ, educational level, socioeconomic background, and so on. After matching participants, researchers assign one member of each pair at random to the experimental group and the other to the control group. After compiling the results of a between-subjects experiment, researchers usually subject the data to statistical analysis in an attempt to estimate the likelihood that differences in results occur due to the independent variable. The within-subjects experiment acts as an alternative to the between-subjects design (Kazdin, 1982; Morgan & Morgan, 2008). (You may also hear these experiments called single-subject designs.) In these experiments, researchers observe a participant’s behavior before the experimental treatment and then during or after it. We call the initial period during which researchers observe a participant’s behavior the baseline period because it provides a basis for comparison. In figures depicting within-subjects data, we usually label this period “A.” The treatment period follows the baseline, and we usually label it “B.” If the A and B periods yield different results (e.g., different latencies or different rates of behavior), you can identify this in the data graph (refer to Figure 2-9). Figure 2-9 Details Hypothetical data. Within-subjects design experiment. Number of times a child hits a stuffed animal each minute during a ten-minute baseline (A) and during an experimental intervention (B). Essentially, in a within-subjects design, the independent variable varies within the participants. In other words, each participant serves in the experimental “group” and in the control “group” at different times. Researchers presume any differences in the dependent variable result from differences in experiences at different times. Because the independent variable varies within the same person or animal, concern that the results might arise from differences among participants greatly reduces. However, some extraneous variable could still explain the results. An animal could become ill during an experiment, for example, and this could give the illusion that the experimental treatment had changed the participant’s behavior when, in fact, it had not. To rule out such possibilities, the experimenter may return to the baseline (A) condition in what we call an ABA reversal design (Sidman, 1960/1988). The experimenter may then reinstate the experimental (B) condition (Figure 2-10). In a sense, the researcher repeats the experiment within the same study. Figure 2-10 Details Hypothetical data. Within-subjects design with reversal. A and B conditions can be repeated to verify the effect of the intervention. Concept Check 4 What is the essential element of a within-subjects design? SHOW ANSWER You can think of using an ABA reversal design like turning a light switch on and off to observe whether it controls a given light. By switching back and forth between A and B conditions, the researcher can demonstrate the extent to which a behavior depends on the independent variable. Researchers generally find data all the more convincing if they can replicate the finding with additional participants, but they generally do not require large numbers of participants. An important difference between within-subjects and between-subjects experiments has to do with the way in which researchers control extraneous differences among participants. In between-subjects experiments, researchers control these differences chiefly through random assignment and matching. These make it easier for researcher to assume that important differences among participants will “even out” across the groups. In within-subjects experiments, researchers control extraneous differences among participants by comparing participants against themselves. Here researchers assume that if they observe the same participant under experimental and control conditions, extraneous differences among participants largely do not matter. Although the single-subject reversal design is thought of as a recent development, the first use of it may date back to the Greek physician Galen. He used it to make a diagnosis 1,700 years ago! (Refer to Brown, 2007.) Both between-subjects and within-subjects experiments allow us, within limits, to observe the effects of independent variables on dependent variables. They provide a substantial advance over descriptive studies. However, not even experimental studies are perfect. Limitations of Experiments The great power of the experiment comes from the control it provides over variables. However, this very control has led to the criticism that experiments create an artificial world from which the researcher derives an artificial view of behavior (Schwartz & Lacey, 1982). In many learning experiments, researchers use an extremely simple behavior for the dependent variable: A rat presses a lever, a pigeon pecks a disc, and a person presses a button. Often they also use a simple independent variable: A light goes on or off, a few grains of food fall into a tray, and a person hears the word correct. The experiment may also occur in an extremely sterile, artificial environment: a small experimental chamber, for example, or (in the case of human participants) a room with only a table, a chair, and a box with a toggle switch. Some people have a hard time believing that the artificial world of the experiment can tell us anything important about behavior in natural environments. To some extent the criticism is fair. Experiments do create artificial conditions, and what we find under those artificial conditions may not always correspond with what would occur under more natural conditions. But the control that makes the experiment seem artificial allows researchers to isolate the effects of independent variables. Although researchers do not particularly care about lever pressing, disc pecking, and button pushing, using such simple acts as the dependent variable allows us to better see the impact of the independent variable. More complicated behavior may offer a more realistic view of behavior, but such complex behaviors may reveal less. Researchers do not aim to understand lever pressing and disc pecking, per se; rather, they seek to understand the effect of the environment on behavior. (For more on this point, refer to Berkowitz & Donnerstein, 1982.) Why do researchers create artificial experiments? Simple: for control. When we create more realistic experiments and study more complicated behavior, we almost inevitably reduce control over important variables and find ourselves with harder to interpret data. We can do two kinds of experiments to get around this problem: laboratory experiments and field experiments. Laboratory experiments offer the control that allows the researcher to derive clear-cut principles. Field experiments—those done in natural settings—allow the researcher to test laboratory-derived principles in more realistic ways. For instance, we might study learning in the laboratory by having rats run mazes and then test the principles derived in field experiments of rats foraging in the wild. Or we might test the effects of different lecture rates on student learning in carefully controlled laboratory experiments and then perform an analogous experiment in a classroom. Despite their limitations, experiments provide a degree of power not available through other means. Consequently, most of the evidence we will consider in the pages that follow derives from experimental research. Much of that research involves animals. Can we really learn about human behavior from experiments with rats, pigeons, and other furry and feathered creatures? Section Review Researchers study learning in various ways. Anecdotal and case study evidence, though unreliable, can provide a good source for hypotheses. Descriptive studies can provide useful and reliable information but cannot account for why a phenomenon occurs. Because of these limitations, researchers usually study learning by means of between-subjects and within-subjects experiments. Experiments have their limits as well, but they provide the best tool available for investigating natural phenomena. 2-4. Animal Research and Human Learning Learning Objectives To describe the role of animals in human learning research, you can . . . 2-4-1 Identify three reasons why animal research improves our understanding of human behavior. 2-4-2 Respond to five reasons why people critique the use of animals in research on human behavior. Some have called animals “furry and feathered test tubes” (Donahoe & Palmer, 1994, p. 71). Most people, including most learning researchers, are far more interested in human behavior than animal behavior. If researchers aim to understand human learning and behavior, why study rats and pigeons? For several reasons. Researchers generally believe that animal research improves our understanding of human behavior. Richard McCarty (1998) writes that “many of the most important developments in psychology over the past century have been linked to or informed by research with animals” (p. 18; also refer to Miller, 1985). Why? First, animals make it possible to get control over the influence of heredity. Because scientists purchase experimental animals from research supply companies, we can know their genetic histories. This means we can substantially reduce an important source of variability in behavior: genetic differences from one participant to another. We cannot, of course, have breeding programs to produce people with uniform genetic backgrounds. Theoretically, researchers can reduce genetic variability in humans by studying learning in identical twins, but finding identical twins to participate in studies proves more difficult than many people think (Moore, 2001). Researchers cannot possibly rely on identical twins for all research on behavior. Second, with animals researchers can control a participant’s learning history. Animals can live from birth in environments with far less variability than their natural environments, thus greatly reducing the influence of unintended learning experiences. Again, this sort of control cannot happen with human participants. Humans, particularly adult humans, come to experiments with very different learning histories. Third, researchers can conduct studies with animals that, for ethical reasons, they cannot conduct with people. Perhaps we might find it interesting or useful to know whether a certain kind of experience would make people depressed or induce them to attack their neighbors, but doing such research with people raises serious ethical problems. Concept Check 5 What are two advantages of animal research? SHOW ANSWER Despite the obvious advantages of animals as research participants, many criticize their use. Perhaps people most often object by saying that the results obtained tell us nothing about people (Kazdin & Rotella, 2009). Critics complain, “People are not rats!” or “Just because pigeons behave that way, doesn’t mean that I behave that way.” Researchers, of course, know all too well about the differences among species; indeed, through animal experimentation some of these differences have come to light (e.g., Breland & Breland, 1961). Researchers must exercise caution in generalizing research on one species to another. If researchers establish that a particular kind of experience affects rat behavior in a certain way, they may assume that human behavior will follow suit, but confidence in such an assumption requires other evidence, such as data from descriptive or experimental studies with people. In most instances such evidence corroborates the findings of animal research. People also object to animal research by suggesting that it has no practical value. Unlike the biomedical research that uses animals to determine the health risks of smoking, this argument goes, animal research on behavior merely provides facts that only a theoretician could find useful. Sure, animal research sometimes aims to answer esoteric theoretical questions, but the findings of this research often have great practical value. People use principles derived from animal research in numerous areas every day, including child rearing (Becker, 1971; Sloane, 1979), education (Chance, 2008; Dermer, Lopez, & Messling, 2009; Layng, Twyman, & Strikeleather, 2004), business (Agnew & Daniels, 2010; Carter et al., 1988), and sports (Boyer et al., 2009; Luiselli & Reed, 2011; Martins & Collier, 2011; Ward, 2011). The beneficiaries of animal research include not only people but also animals. In the past, trainers of pets, saddle horses, farm animals, and circus animals relied heavily on harsh measures. The expression, “Don’t beat a dead horse” derives from the widespread physical abuse of horses in the days of horsepowered transportation. Today animal trainers use more humane and effective techniques, thanks largely to procedures discovered or perfected through animal research (see, e.g., Skinner, 1951; Pryor, 1999). These same procedures resulted in improvements in veterinary care and in the quality of life for animals in zoos and similar facilities (Markowitz, 1982; Stewart, Ernstam, & Farmer-Dougan, 2001; also refer to Chapters 4 and 9). Third, some people object to animal research by suggesting that it is intrinsically unethical. This “animal rights” view maintains that people have no more right to experiment on rats than rats have to experiment on people. The controversy deserves serious treatment (Balster, 1992; Burdick, 1991; Miller, 1985; Shapiro, 1991a, 1991b). Ethical problems do arise in the use of animals in experiments, and researchers take them seriously. In 2011, the National Research Council (2011) concluded that while the genetic similarity of chimpanzees to humans (we share 98% of our genes) makes them an ideal subject for behavioral and biomedical research, it also raises ethical concerns. While the NRC stopped short of ending all funding for research on chimpanzees, it severely limited their use to problems that cannot be answered in other ways (Altevogt et al., 2011; National Research Council, 2011). Concept Check 6 What are two benefits of behavioral research with animals? SHOW ANSWER Some people might also say that researchers do not treat animals as well as animals that serve humans in other ways such as working animals or those living in homes as pets. Granted, laboratory animals do typically live in relatively small and uninteresting spaces. However, researchers do not beat them, tease them, leave them to bake in a hot car, subject them to inclement weather or foul living quarters, deny them veterinary care, or abandon them—as frequently happens to working animals and pets. Various state and federal laws and ethical review boards set conditions for the care of laboratory animals. Indeed, accurately interpreting data from animal research rests on the assumption that the animals have received good treatment. For this reason, one group of researchers (Balster, 1992) writes that “the scientific community goes to great lengths to ensure the welfare of their animals; to do otherwise is bad scientific method” (p. 3; emphasis added). In an effort to prevent unnecessary suffering, the American Psychological Association (2012) and other organizations have established guidelines for the conduct of animal research. These require that research teams meet certain standards in the care and handling of animals. Animals may, for instance, “work” for their food; but animals in domestic service or living in the wild would have to work much harder for the same amount of food. The guidelines also set standards for the use of aversives (stimuli the animal would avoid, given the option). If you can answer a question without the use of aversives, you should. When necessary, aversives must be no more severe than required by the nature of the research. Researchers must also justify the use of aversives with reference to the benefits the study could provide. Inspections by health departments and veterinarians help to ensure that researchers meet these standards. Some critics maintain that animal research is unethical because computer simulations can replace animals in research. We need no longer run rats through mazes or train pigeons to peck a disc for food, they argue; silicon animals can replace those with fur and feathers. Computer programs can simulate the behavior of animals. A good example of this is Sniffy the Virtual Rat (Alloway, Wilson, & Graham, 2011). This program provides an image of a laboratory rat in an experimental chamber on the computer screen. The simulated rat moves about doing rat-like things: exploring its surroundings, scratching and licking itself, drinking water, eating food it finds in a food tray, and learning to press a lever. Sniffy provides a useful instructional tool, but simulated animals (or people, for that matter) cannot replace living research participants. We cannot program a computer to simulate the effects of a variable on behavior until we understand those effects. Research helps us do just that: to understand the effects of one variable on another (behavior). I am not arguing that no valid grounds for objecting to animal research exist. Rather, we should keep in mind that we use animals in a variety of ways for our benefit—and for theirs. Some animals help people who are blind find their way, others detect explosives or illegal drugs, yet others pull a plow through a field, and your dog or cat may provide companionship. Others provide answers to questions about learning and behavior that may greatly improve the lives of both animals and people. Concept Check 7 Why can’t computer simulations replace animal research? SHOW ANSWER Section Review Both humans and other animals may serve as participants for experiments on learning. Animals make greater control possible but leave open the possibility that the results do not apply to humans. Often, researchers conduct basic research with animals, and then apply the principles derived from this research to humans. Animal research raises ethical questions, but so do other common uses of animals, and animals as well as people benefit from animal research. An understanding of learning and behavior requires both animal and human research. A Final Word A major theme of this text is that we can view all behavior, including all human behavior, as a natural phenomenon, and therefore we can study all behavior with the methods of natural science. This means that we can account for behavior with empirically derived principles and without recourse to vague and unverifiable hypothetical concepts. We have not yet satisfactorily analyzed all behavior in terms of natural phenomena, but researchers have made considerable progress toward that goal. You must decide for yourself whether the successes thus far achieved justify adopting the natural science approach to learning and behavior.
Purchase answer to see full attachment