Laboratory and Field Experiments
All sciences use experiments. In the laboratory experiment, materials .or 'people are brought into the laboratory for study. In
laboratory experiments with people, people a~ recruited, assembled, and perhaps paid for engaging in the experiment. Dollars
famous frustration-aggression studies (1939] were conducted by assembling a number of students as experimental subjects, supposedly to study the effects of fatigue upon task performance. These students were subjected to intense frustration through prolonged boredom, non arrival of promised food and games, and other intentional annoyances, while their aggressive responses were cataloged The field experiment takes research out to people instead of bringing people to the research laboratory. A massive field experiment involving vaccination of several million children established the value of the Salk polio vaccine. A continuing series of field experiments are seeking to find effective ways of promoting birth control in underdeveloped countries and among disadvantaged groups in the United States [Berelson, 1966; Ridker, 1976; Singh, 1979]
The concept of any experiment IS quite simple: Hold all variables constant except one, cause it to vary, and see what happens. One of.the best ways to control all variables is to use control groups. A control group is a group of subjects who are like the experimental group in all respects except the variable(s) which we are studying. As an example, suppose we want to know Whether abolishing grades would increase learning's or increase loafing.
To test this by experiment we would need. a control group of classes which follow the usual teaching and grading procedures and an experimental group of classes using whatever experimental procedure is being tested. To "hold all other variables constant the control and experimental groups would need to be alike In students abilities, subject studied, quality of teaching, student' work load, students' finances, and anything else likely to affect their performance. We would also need a reliable instrument to measure learning outcomes (after reaching agreement upon what learning outcomes were important).
Then the results of the trial could be objectively determined. If the experimental group shows greater or lesser learning gains than the control group and this difference is confirmed by repletion (repetitions of the experiment by other researchers), then significant conclusions. can be drawn. Failure to use suitable control-groups ma destroy a study's usefulness, For example, two psychologists [Miale and Selzer, 1970) examined the Rorschach tests which were given to sixteen Nazi leaders at the 'time uf the Nuremberg war crimes trials and reported . that fifteen of them, were "psychopathic" in various degrees, But Miale and Selzer failed to compare the Nazi leaders' tests with Rorschach tests from a control group of leaders from other countries. Thus, even if we assume that the analyses are correct, we do not know whether these researchers have uncovered personality characteristics of Naw leaders, or characteristics of leaders. Thus, this study is of limited value. There are two common ways of setting up experimental and control groups. One is the matched-pair technique. For each person in the experimental group, another person similar in all important variables (such as age, religion, education, occupation, or anything important,to this research) is found and placed in the control group. Another technique is the random-assignment technique, in which statistically random assignments of persons 'to experimental and control groups are made such as assigning the first person to the experimental group, the next to the control group, and so on. Suppose we wish to measure the effectiveness of an experimental treatment program for delinquents in a reformatory. Using one technique, we should match each delinquent who received the experimental treatment (experimental group) with and other delinquent, matched for other variables thought important, who received the usual treatment (control group). Using the random-assignment technique, every second. (or third, or tenth) delinquent would be assigned to the experimental group upon arrival at the reformatory, with the others becoming the control group.
Wherever the researcher is permitted to make assignments in this way, the random-assignment technique is far easier research situation does not allow this technique, the matched-pair technique may be used. Experiments in sociology face certain difficulties. An ,experiment involving thousands of people may be prohibitively expensive. It may take years, to complete a prospective study. Our values forbid us to use people in any experiments which may injure them. The scientific world reacts strongly in those infrequent instances where human subjects have been used in a hazardous or harmful manner [J. Katz, 1972; Jones, 1981]. When people are unwilling to cooperate in an experiment, we cannot force them to do so (although we may occasionally trick them into unconscious cooperation). Furthermore, when people realize that they are experimental subjects, they begin to act differently, and the experiment may be spoiled. Almost any kind of experimental or observational study upon people work know they art being stolidity will give some interesting findings which may vanish soon after the study is ended. Planned experiments upon human subjects are most reliable when these subjects do not know the true object of the experiment. They may be given a rationale, personable explanation of what the experimenter is doing, but this rationale may be a harmless but necessary deception which conceals the true purpose of the experiment.
For example, McClellan [1971] wished to study the effects of alcohol upon normal people in a party atmosphere but told the subjects that he was studying the effects of a party atmosphere upon fantasy, and had them write imaginative stories about pictures he showed them at intervals. But .as Kelman points out [1966], the use of deception in social research poses the ethical question of distinguishing between harmless deception and intellectual dishonesty and may even produce errors in the outcome (subjects may detect the deception and begin second-guessing the researcher!). Because.of all these limitations, social sciences (excepting psychology) make limited use of planned experiments. We use them wherever practical, but depend more heavily on other techniques
