The origins of this method are usually credited to two illustrious English cousins, Charles Darwin and Sir Francis Galton, the latter well known for his practical discoveries in testing the sensory abilities of human beings and animals. Here, too, William James had a warning: "Messrs. Darwin and Galton have set the example of circulars of questions sent out by the hundreds to those supposed able to reply. The custom has spread, and it will be well for us in the next generation if such circulars be not ranked among the common pests of life'' (James, 1890).

This method of collecting information of course depends heavily on the questions that are asked. A questionnaire is a printed form with questions of all sorts, often administered by mail or telephone, sometimes in a direct interview. It is intended to be answered by many people. Some questionnaires are simply a list of reminders to be checked as applicable or not.

Types of Items. These questions or reminders can vary from highly specific to broad and vague. Milgram's questions were specific, for people were asked to indicate how long they would follow orders--that is, how much shock they would administer to someone, under the condition that the strength of the shock would be increased by 15 volts each time the learner performed incorrectly. An exact answer was requested, using a numerical scale on the questionnaire.

In a survey on dental care, specific questions might include: "How many times per day do you clean your teeth?'' and "Do you brush with a vertical, horizontal, or circular motion--or with more than one motion?'' Each of these is a structured item, for there is little leeway in answering. Or the question might be an open-ended item, which can be answered in a wide variety of ways, such as: "What is your approach to dental care?'' On an open-ended item, people are more likely to reveal what is important to them, but they may include endless details of no significant value. In either case, variations in the wording of the item may substantially influence the response (Table 2­2).

These results can serve as norms, for they are the responses from a large number of subjects. In short, norms show how people perform, indicating what is common and rare in a given population. Norms provide a means for interpreting a person's results. We can compare the response of that subject with those of the group to find out whether that subject's response is highly typical, somewhat typical, or rather deviant. Usually, a person's response is most appropriately evaluated with the norms for his or her age, sex, economic status, and so forth.

Random Sampling. The most common means of obtaining a representative sample is to develop a random sample, in which each subject in the population has the same chance of being included. To obtain a random sample of 20 students, for example, each student in the college is designated by a different number. Then 20 numbers are selected by random drawings or random-digit dialing on the telephone.

While there are other methods, random sampling is the basic procedure for obtaining a representative sample. In fact, randomization is a fundamental principle in any research, from naturalistic observation to the experimental method. As any sample is increased in size, the influence of chance factors generally becomes less. In other words, a large random sample is more likely to be representative of the population than is a small one.

Incidental Sample. In predicting the outcome of an election or testing a new drug, a representative sample is indispensable. The investigator must know the extent to which the sample reflects the population. In the early phases of some research, however, or if individual differences are not important, the investigator may use an incidental sample, which includes anyone who happens to be available and willing to respond. It is for this reason that so much psychological research is based on college sophomores and white rats. An incidental sample is relatively easy to obtain; the danger lies in the conclusions drawn from such a sample.

In administering his questionnaire at his lectures, Milgram obtained incidental samples. Anyone who came to a lecture could complete the questionnaire. To ensure representative samples in other instances, he used a telephone book and city records for the general population of New Haven, culling thousands of names and addresses by random methods.

Even when the sample is representative, some people do not return the questionnaire or refuse to give an interview. Unless adequate substitutes are found, these situations may produce a nonresponse bias, in which the findings are invalid or biased due to the lack of returns from the full sample. The people who fail to respond may be against the whole issue, or they may simply have different reactions than those who do reply. These difficulties have prompted considerable research aimed at improving the survey method (Dillman, 1991).

Types of Unobtrusive Measures. Subjects in survey research may try to please the investigator, present themselves in a favorable manner, or complete the task as soon as possible. This problem is called the guinea pig effect, meaning that the subjects know they are "guinea pigs''--participants in a research situation. To deal with this problem, survey researchers sometimes employ unobtrusive measures, rather than questionnaires. In unobtrusive measures, the investigator collects information from people without disturbing them in any significant way--and without even observing them directly. The subjects do not know they have been included in a research project. The investigator simply examines the traces of their behavior.

For example, some unobtrusive procedures are called erosion measures, for they indicate the ways in which people use and wear out the environment. In a library, the dirty, wrinkled, and torn pages tell a story. Those magazines are used. Clean pages and stiff bindings give a very different impression. Investigators also use accretion measures, examining the materials people leave behind in corridors and restrooms, such as wrappers, graffiti, and empty bottles and cans. In short, our refuse can be revealing (Figure 2­2).

Figure 2-2 Accretion Measures An observer might make hypotheses about the background and interests of people who read and post notices on this bulletin board.

The use of archival data offers still another unobtrusive measure, for an investigator can consult institutional files for all sorts of information, ranging from births and marriages to thefts and accidents. Retrieving this information is unobtrusive, although the process of obtaining it originally perhaps was quite intrusive for the subjects, as when the census taker appeared at the front door. Institutional files have been used, for example, to study the outcome of legal proceedings and techniques of control with disruptive children (Himelein, Nietzel, & Dillehay, 1991; Skiba & Raison, 1990). The use of unobtrusive measures stands somewhere between naturalistic observation and the survey. It is not traditional naturalistic observation, for the behavior is never observed. It is not a traditional survey either, for the subjects do not know that they have participated in research. Except for the archival method, it offers little information on the people who serve as subjects.

Using Unobtrusive Measures. Bothered by the guinea pig effect with questionnaires, Milgram employed an unobtrusive measure known as the lost-letter technique, in which many letters are distributed throughout a city, stamped and addressed but unposted. Anyone who finds one must decide what to do. Mail it? Destroy it? Ignore it? By varying the addresses on the envelopes and calculating the proportion mailed, this survey technique can be used to measure attitudes toward the addressees (Milgram, 1992).

In one instance, Milgram distributed 400 letters in parking lots, streets, shops, and phone booths throughout the city of New Haven. Each letter was addressed to a medical research group, a private individual, or the Communist or Nazi party. Each envelope was coded and sealed in a manner to indicate later where it had been dropped and whether it had been opened. For most Americans, common sense would have predicted that the letters addressed to the Nazi and Communist parties would be least often mailed and most often opened. Milgram's returns showed these results (Milgram, 1992; Table 2­3).

Second, the method is not useful for asking subtle questions. The finder simply shows an overall attitude, favorable or unfavorable, by mailing or not mailing the letter. This technique is appropriate for measuring positions on either-or questions, such as the abortion issue and sexual orientation (Kunz & Fernquist, 1989; Levinson, Pesina, & Rienzi, 1993). It is not suitable for determining why people take one stand or the other.

Milgram was much impressed by the extent to which people mailed the letters, even though no letter contained any such request. These results showed a widespread compliance with an unstated invitation-- far greater than he might have anticipated. But again, he could not, through the survey method alone, determine the underlying causal factors. Then he thought again about his questionnaire on electric shocks. When confronted with the actual situation, would these people disobey the order?