Some of the difficulties with obesity research concern inconsistencies in the definition of the construct, the methods by which it has been assessed, and the samples that have been employed. The purpose of the present study was to investigate whether there were significant differences in the nature of the simultaneous relationships between a number of variables previously associated with obesity and body size when assessments were made by body mass index (BMI) or an estimation of fatness from skinfold thicknesses. In addition, comparisons were made when subjects were classified into obese and non-obese categories. Appropriate multiple regression and logistic regression analyses were used to analyze these data for a group of adult men and women sampled from the general population. These procedures allowed a comparison of results when the dependent variables were in their continuous versus their dichotomous form. Approximately 250 adult men and women participated in the study. Percent body fat was associated with less frequent leisure time exercise participation, slower walking speed, reduced levels of trait anxiety, and a greater tendency to diet. Among women, but not men, lower socioeconomic status was also related to fatness. However, several of these relationships were absent when BMI served as the measure of body size, and to a greater extent, when subjects were classified into obesity categories. The findings of this study underscore the importance of measurement techniques and classification procedures in determining the outcome of psychosocial and behavioural data in the area of obesity research.