Data from a population-based longitudinal study were used to compare cross-sectional versus prospective predictors of screening mammography. Although the results of the two analyses were not dramatically different, some important differences emerged. More attitudinal variables were related to future behavior compared with past behavior. Using purely cross-sectional data from this study to design an intervention would result in potentially important variables being deemphasized or omitted, which could have a significant impact on the strength of the intervention. Our findings strongly suggest that it may be unwise to generalize cross-sectionally obtained data to a longitudinal situation.