SELECTING SUBJECTS
    Probability Sampling
    Randomization
    Matching
    Matching or Randomization?

    It is important that experimental and control groups be comparable at the outset of an experiment: they should have the same average levels of prejudice against women, for example.  This can be accomplished in a couple of ways.

    If a large number of subjects are involved in the study, probability sampling or randomization provides a solution.  We've seen in Chapter Seven that a probability sample drawn from a population will be representative of it.  If we divide a group of subjects into two groups at random, then each of those groups should be representative of the whole population of subjects, hence each should be the equivalent of the other.

    With a small number of subjects, sampling error is greater, as you saw in Chapter Seven.  If you had a total of ten subjects, for example, selecting five at random would not necessarily achieve comparable experimental and control groups.  In such cases, matching is a better procedure.  In this, you (1) select one subject for either the experimental or control group, (2) note that person's personal characteristics (e.g., gender, age, race, etc.), especially any that might be related to the dependent variable, and (3) find another subject with characteristics that match those of the first subject as closely as possible; and put the second subject in the other group.  Repeat this procedure until all the subjects have been assigned.