Restrained eating: measuring an elusive construct.

As a measure of restrained eating, Herman's Restraint Scale (1978) reliably predicts laboratory food consumption in college students regardless of their weight. However, the generality and psychometric properties of the scale have not been established. In the present study, 136 male and female adults were cross-classified as obese and normal and as dieting or non-dieting. The subjects were administered a single questionnaire containing items of the Lie, Social Desirability, and Restraint scales presented in randomized order. Unlike previous reports by Herman, the three adult groups differed significantly on the Restraint Scale in the following order: Obese dieters greater than Obese non-dieters greater than normals. Also, alpha reliability coefficients varied across groups and corrected item-total correlations also displayed considerable variability with no uniformity apparent for individual item correlations. The factor analysis identified three factors within the ten item scale, and for the obese dieters, the scale was not independent of social desirability. These results indicate that the Restraint Scale has limited usefulness beyond laboratory settings with college students.