Predicting contraceptive behavior from attitudes: a comparison of within- versus across-subjects procedures

Within- versus across-subjects procedures for predicting behavior from attitudes were contrasted. Each procedure requires a comparison among attitudes in order to generate a prediction; the comparison is either among the same attitudes held by different people (across subjects) or among different attitudes held by the same person (within subject). It was hypothesized that the within-subject model provides a more adequate explanation of behavior from attitudinal constructs and, hence, more accurate prediction of behavior from attitudes than does the across-subjects model. To test this view, a sample of 349 married couples was administered a questionnaire containing measures of three attitudinal components--affect, cognition, and conation--toward each of four contraceptive methods--oral contraceptives, IUD, diaphragm, and condoms. Contraceptive behavior was assessed 1 year later. In support of the hypothesis, the within-subject predictions bore a significantly stronger relation to the behavioral criteria than did the across-subjects predictions. This effect was consistent for each of the three components of attitude, for the prediction of behavior and behavioral intention, for male and female respondents, and for a variety of contraceptive behaviors. In addition, both the within- and the across-subjects analyses demonstrated a clear rank ordering in the predictive validity of the three attitudinal components: Conation was a better predictor of behavior than was affect, which, in turn, was better than cognition.