Randomization provides a solid ground for experimental analysis. It provides a vital means to handle the twin problems of validity and reliability of the response. Virtually all statistics texts emphasize randomization as a basic principle; virtually none tell how to do it.
How to randomize in practice is the concern of this section. The first three subsections present what may perhaps be called standard randomization procedure for the usual run of psychological experiments. Field studies often present complications, as noted later.
Subject Assignment Sheet. The standard way to randomize subjects across experimental conditions is to make up the subject assignment sheet before running the first subject. In one useful form, this sheet lists the sequence of conditions in a randomized order; this is the order in which they are to be run. The first subject to appear is run under the first listed condition, the second under the second listed condition, and so on down the list. The subjects may appear for the experiment in any arbitrary manner; randomization is accomplished by the construction of the list. a
This subject assignment sheet should ordinarily include a line or two to list subject's I.D., experimental condition, and date. Persons who run each subject should be identified. The completed sheet should be retained in the permanent file for the experiment.
Also included on the sheet, no doubt, should be space for comments on the subject's behavior. Such comments can be helpful for understanding unusual scores and thereby for improving experimental procedure in future work. In addition, this requirement encourages the experimenters to improve their observational skills and gain better appreciation of subjects' phenomenology. This is valuable for all of us, not only when we are graduate students.
With human subjects, the question—answer period at the end of the experiment can provide material for comments. One technique is to present selected experimental stimuli and ask subjects to explain their response. To be effective, these comments should be systematic for all subjects.
Randomizing Conditions. One way to randomize the sequence of conditions is to use Table A1 (page 809) or some other table of random numbers. With four treatments, for example, let the digits 1 to 4 stand for the four treatments. Enter the random number table by choosing a haphazard starting place, and use the chosen digit to determine the treatment for the first subject. Then proceed with successive digits, ignoring digits other than 1 to 4.