n his naturalistic observations, Milgram noted a high degree of conformity among commuters. In the survey method, he found that almost all subjects claimed they would disobey malevolent authority. Through case studies, he discovered that people differed sharply in response to questions of compliance and obedience, as illustrated by Gretchen Brandt and the others. However, none of this research allowed him to draw conclusions about the causes of conformity, compliance, and obedience. For this purpose, he needed still another method--one that would enable him to collect information about cause-and-effect.

The most promising answers to cause-and-effect questions are found in the experimental method, in which the chief factors in a research problem are manipulated or controlled in precise ways. For this reason, the experimental method is often considered to stand foremost among the various research techniques in psychology.

CLASSICAL EXPERIMENT

In his effort to accomplish morally significant research in social psychology, Milgram placed an announcement in a New Haven newspaper. It offered a reasonable sum of money for participation in a laboratory study of memory. People with all sorts of occupations returned the newspaper coupon and participated in this research. Among them was a 35-year-old drill-press operator named Jack Washington, a pseudonym.

Unknown to Jack and the others, the true purpose of this research was not to study memory but to identify factors that influence obedience and disobedience. To investigate them, Milgram designed a classical laboratory experiment.

In the classical experiment, all potentially influential factors are controlled or held constant except one, which is manipulated. What happens when it is manipulated? One by one, each presumably important factor is tested in this way. In modern multifactor studies, to be considered later, several factors can be examined simultaneously.

Management of Variables. Any changeable factor or element in research is called a variable. It is some condition that the investigator wishes to study. In the classical experiment, the investigator selects for study only one variable at a time. This method follows the rule of one variable, in which just one factor is manipulated, or examined, at any given moment. If an effect is observed--that is, if some specific event takes place--it is regarded as a result of the manipulated factor, assuming that other potentially influential factors are held constant.

Milgram studied obedience this way, using three people and an important deception. The real subject, such as Jack Washington, arrived at the laboratory prepared to engage in a study of memory, unaware of any deceit. In a rigged drawing, Jack drew the role of "teacher'' and was given the task of administering punishment to a person trying to memorize a list of words. This second person, the "learner,'' was an accomplice of the experimenter, and he made many mistakes, for which he had agreed to receive electric shocks as punishment. In fact, he never received any shocks--none were administered. But by complaining and pounding the wall he acted as if he were in pain. The third person, the investigator, served as the authority figure, requesting Jack, as the teacher, to administer stronger and stronger shocks for successive mistakes by the learner. These shocks began at 15 volts and ranged up to 450 volts, each labeled with a verbal description on the shock generator (Figure 2-4).

Figure 2-4 Voltage Scale on the Shock Generator. For each wrong answer by the learner, the teacher was asked to administer the next-higher level of shock. Make an estimate of the highest shock administered by most teachers.

To what extent would Jack Washington and other subjects obey the orders? We have already noted what common sense suggested to most people. On the basis of the questionnaire responses, it seemed that essentially all subjects would disobey the authority.

With the learner strapped to a chair in the next room, Jack began the procedure, shocking the learner for his occasional errors. At the beginning, these errors were infrequent and the shocks mild. At 75 volts, the learner grunted, and at 150 volts his first protests were heard. Jack looked sadly at the authority figure but continued to obey. At 300 volts, the learner hollered: "I absolutely refuse to answer any more. Get me out of here.'' Yet Jack, following orders, continued to administer stronger and stronger punishment, even when the learner no longer answered.

Finally, when he reached the 450-volt level, the maximum shock available, Jack once again turned to the investigator across the room and asked what to do. He was told: "Continue using the 450-volt switch for each wrong answer. Continue, please.'' With a dejected expression on his face, Jack resumed his difficult task (Milgram, 1974).

Jack had been completely obedient. He followed all orders--but he was not alone. In several repetitions of this procedure with various subjects, 60% of all subjects eventually shocked the learner with the maximum voltage available. They obeyed all orders, showing a wide range of emotions, attitudes, and styles. Some were humble, some were self-assured, and most were deeply concerned about the learner. A minority of subjects resisted the orders before reaching the maximum voltage available. One of them was Gretchen Brandt, who disobeyed at 210 volts. At that point, in a calm but resolute manner, she refused to proceed further, thereby placing herself among the first quarter of all subjects in resisting malevolent authority.

Using this situation and many subjects, Milgram then designed a series of experiments to examine the causal factors in this astonishing obedience. One by one, he manipulated these variables: proximity of the learner, closeness of authority, prestige of the setting, and presence of rebellious peers.

Independent and Dependent Variables. The variable to be manipulated is called an independent variable because changes in it are independent of any other aspect of the experiment. It is varied in accordance with the investigator's purpose. If the aim is to discover the influence of the proximity of the learner on the obedience of the teacher, the investigator places the learner at various distances from the teacher. This factor, the distance between learner and teacher, is the independent variable (Table 2-5).

In addition to introducing an independent variable, the experimenter observes and measures the subject's response. This response is referred to as the dependent variable because it is the result of the manipulation; its presence or intensity depends on the independent variable. In most of Milgram's experiments, the dependent variable was the amount of shock the teacher administered. Instead, it might have been the teacher's speed in administering the shock or the duration of the shock.

As a rule, the independent variable is some stimulus; the dependent variable is some response. A word of caution is in order, however. In many experiments, the independent and dependent variables are more complex and cannot be so readily identified in terms of stimuli and responses.

Operational Definitions. In any research, the variables must be clearly defined. As a rule, scientists use operational definitions for this purpose. An operational definition indicates the specific procedures by which something is measured; it depicts the meaning of something in highly explicit, usually quantifiable terms.

To study obedience, for example, Stanley Milgram did not use a dictionary definition: "following orders'' or "carrying out commands.'' He devised an experiment with a fake shock generator, explicitly defining obedience as the highest amount of electric shock administered by the subject. Someone who administered 450 volts was more obedient then someone who administered only 300 volts, and so forth. The merit in this definition lies in its clarity and quantification.

Milgram might have studied obedience in other ways. Using the method of naturalistic observation, he might have observed people on city streets, defining obedience as discarding waste in refuse receptables, following pedestrian signals, or obeying laws about public transportation. He might even have used all of these measures together, employing a composite definition. Using the survey method, he could have consulted public records, defining obedience as timely payment of city and state taxes. In another experiment, he could have requested each subject to engage in an extremely boring task, defining obedience as the length of time the subject persisted in that task (Figure 2-5).

Figure 2-5 Operational Definition of Intelligence. The animal learning this maze begins in the start box, at the upper right end of the maze, and finishes in the goal box, as shown, where it obtains food. Its speed in completing the maze and number of wrong turns are measures of its performance--that is, they may constitute an operational definition of intelligence.

Operational definitions do not necessarily require quantification. For example, how might cohabitation be defined? You may say: "Oh, that's easy--unmarried people living together.'' But how long must they live together? How regularly?

Instead of setting a time limit, some surveys of cohabitation have used marriage applications to create an operational definition. If the partners for a license gave the same address, their living condition was considered cohabitation. If they gave different addresses, they were not considered cohabitants (Reimann- Marcus, 1992). With this definition there may be some false positives, partners considered cohabitants who were not together very much, very long, or in a very loving way. And there may be false negatives, partners who lived together romantically but gave different addresses. However, the definition is precise, which is a requirement of scientific research, and no definition can readily satisfy all potential criteria for a complex concept.

This use of operational definitions restricts the scope of any research and the conclusions that can be drawn from the findings. For example, Milgram's research has been questioned on this basis, though studies using a different operational definition have confirmed the overall outcome (Meeus & Raaijmakers, 1986).

DESIGN OF EXPERIMENTS

Figure 2-6 Proximity of the Learner. As proximity to the learner increased, obedience to the malevolent authority decreased (Milgram, 1974).

This overall plan for different conditions is called the design of the experiment. It includes the choice of subjects and apparatus, as well as the manipulation of the independent variable, but no issue is more important than the control condition. The chief concern here is control of confounding variables. An extraneous or confounding variable is any factor that may exert an unwanted influence on the dependent variable, giving the experiment an uninterpretable result. In Milgram's studies on proximity, suppose the learner in some instances was a child and in others an adult. Then both variables, the learner's age and the learner's proximity, might influence obedience. In this poorly designed experiment, the learner's age would be a confounding variable, producing results which might be confused with those produced by the learner's proximity, the independent variable.

Types of Design. In addition to identifying and manipulating variables, experimenters need to decide how to use their subjects. In one common experimental design, all subjects serve in both conditions, experimental and control. Each subject in the experimental condition is compared with himself or herself in the control condition. In this way, the two conditions should be equal, except for the key factor, the independent variable. This procedure is called the within-subject design or own-control design because comparisons are made within each subject, who serves under different conditions, providing his or her own control condition.

Sometimes the same subjects cannot serve in both conditions. This restriction often occurs when the independent variable extends over a long period of time, as in a program of therapy. If the therapy lasts for two years, the experimental subjects will be two years older and wiser, and they will have completed a program of therapy before beginning the control condition. Hence, the subjects in the two conditions will not be equal. The major method for dealing with this problem involves multiple sets of subjects matched for important characteristics, such as age and sex, each group serving in one condition or the other, experimental or control. This procedure is called the between-groups design because the subjects are randomly assigned to the experimental or control group; then their results are compared.

Milgram debriefed all subjects immediately after the experimental hour, for he did not want them to leave the research thinking that they had really administered pain to someone. Thus, his subjects could not serve as their own control; his method here involved the between-groups design. Using such procedures, he found that the closeness of the authority also made a great deal of difference. When the experimenter sat just a short distance away, 65% of the subjects obeyed all commands. When the experimenter left the laboratory and gave instructions by telephone or a tape recording, obedience diminished sharply. Only 23% of subjects delivered the highest level of shock, and many subjects surreptitiously administered lower shocks than required, assuring the experimenter by telephone that they were proceeding according to the original plan (Milgram, 1974). In the same way, Milgram found that the presence of rebellious peers greatly influenced the subject's obedience. These peers, pretending to be assistant teachers, in fact were accomplices of the experimenter. When they refused to follow orders, the subject refused too. "The effects of peer rebellion are very impressive in undercutting . . . authority,'' Milgram concluded.

Control of Expectations. In many experiments, the subjects have an expectation about what should happen. Controlling expectations is especially important, for example, in testing the physiological effect of drugs, for the investigator wants to know what benefits are available apart from the knowledge that treatment has been received. The experimental group therefore receives the actual drug, and the control group receives a sugar pill, identical in appearance, with no medical properties. This pill, with no active ingredients, is called a placebo, which, freely translated, means "I shall please.'' The aim is to give the control subjects the same set or expectation as that in the experimental group (Figure 2&endash;7).

Figure 2-7 Use of Placebos. One pill has medicinal properties; the other is merely a salt pill. One tape contains a subliminal message; the other makes no such claim. The concept of placebo, derived from drug studies, now applies to any object, event, or other treatment used to control expectations.

The placebo effect also occurs in daily life, as well as in research. One former student had been asked to make a mixed drink each evening for his grandfather, who enjoyed the warm, relaxed feeling it provided. One day he confided that, per the doctor's orders, he never spiked the drink, but the thought of vodka certainly satisfied his grandfather.

There is evidence, nevertheless, that a placebo may not be purely psychological. The thought of receiving a certain medication may prompt activity within the nervous system, releasing neurotransmitters that can influence the subject's reactions in significant ways, especially regarding the perception of pain. Further research on the placebo effect is underway in diverse areas of psychology, ranging from education to health (Adair, Sharpe, & Huynh, 1990; Jensen & Karoly, 1991). Similarly, the experimenter's expectations can be a concern. The experimenter's hopes, habits, and personal characteristics can influence the results of the investigation without his or her knowledge, a condition referred to as experimenter effects. The experimenter unconsciously signals to the subjects the response he or she hopes to obtain from them. This unintentional cuing is also known as the Clever Hans effect, named for the presumably intelligent Berlin horse that tapped out answers by observing subtle cues from people around him (Pfungst, 1911).

When the subjects' expectations are controlled by preventing them from knowing which treatment they have received, the procedure is called a single-blind design. It is illustrated in the use of the placebo. Instead, the experimenter may not know which subjects have received which treatments, again a single-blind design. Sometimes both the experimenter and subjects do not know their treatments, a procedure called a double-blind design. In the latter design, neither the investigator nor the subject knows to which group the subject belongs. A third party decides which subjects receive what treatment and codes them so that the experimenter can evaluate each subject's responses without knowing the treatment received. Under these conditions, expectancy cannot create a bias in the research outcomes. The study of animals offers special possibilities for research design, which is one reason for psychologists' interest in animals. Genetic factors can be manipulated through selective breeding. Environmental conditions can be managed in ways not possible at the human level. Comparisons across species and among different research designs offer a powerful technique for disentangling the diverse determinants of behavior (Timberlake, 1993). Still another advantage is the animals' faster maturation rates. Studies of growth and development often can be completed in a year or two, as opposed to several decades with human beings.

MULTIFACTOR STUDIES

The interest in multifactor studies lies in their greater efficiency and in the capacity to discover what happens when two or more factors are combined. Sometimes the result is an additive effect, meaning that the total influence of all the factors is the sum of their separate influences. No independent variable influences the effect of any other independent variable; the final result is simply their cumulative sum. In Milgram's experiments, the subjects were moderately obedient when the authority figure was nearby and moderately obedient when the learner was in another room. If these effects were additive, both factors together would produce even greater obedience.

In other instances, the result is an interactive effect, meaning that the influence of one variable depends on other variables or that the influence of one variable changes with alterations in other variables. Milgram did not examine interaction effects in these experiments, but a dramatic example can be observed in everyday life with the consumption of alcoholic beverages. Mild consumption of alcohol before going to sleep may have no significant effect on a person's health. Similarly, ingestion of phenobarbital, taken to induce sleep, may have no significant effect. But when taken together, even in moderate amounts, they can be fatal (Figure 2-8).

Figure 2-8 An Interaction Effect. Each factor by itself may have no significant effect. In combination, they can prove fatal.

Outside an experimental situation, these mutual influences among variables are known as the interaction principle, and their subtleties sometimes escape notice in daily life. In supermarkets, for example, baggers and produce clerks sometimes experience an itch or a rash, especially on the hands and forearms. People working with dairy products or stocking shelves are not afflicted. The reason? Chemicals from certain vegetable products, such as celery and parsnips, are deposited on the skin in minute amounts. But that is not all. The problem appears largely in the summer, occurring through a combination of the chemical deposits and exposure to sunlight. The name of this disorder might cause anyone to itch or scratch: phytophotodermatitis. Phyto indicates the plant, photo the light, and dermatitis the resulting rash. To avoid it, one should work with canned goods, wear long gloves, or stay out of the hot sun.