Journal of Abnormal Psychology 1988, Vol. 97, No. 1,19-28 Copyright 1988 by the American Psychological Association, Inc. 0021-843X/88/J00.75 The (Mis)measurement of Restraint: An Analysis of Conceptual and Psychometric Issues Todd F. Heatherton, C. Peter Herman, Janet Polivy, Gillian A. King, and Sheila T. McGree University of Toronto, Toronto, Ontario, Canada In this article we examine alleged conceptual and psychometric deficiencies of the Restraint Scale, an instrument intended to identify chronic dieters. These deficiencies include the confounding of restraint with disinhibition, the inapplicability of the scale to obese samples, problems with the factor structure of the scale, and difficulties in completing the scale. We argue that these alleged deficiencies are in most cases chimerical and that the Restraint Scale remains the most useful tool for examining behavioral and other dieter/nondieter differences. Proposed alternatives to the Re- straint Scale are examined and found to be inadequate as replacements, although they may be useful for certain purposes. Closer attention to the intended purpose of such instruments may serve to dispel controversy and confusion. Research on the dynamics of eating has gradually expanded from the examination of behavioral differences as a function of body weight (e.g., Schachter & Rodin's, 1974, survey of obese/ normal differences) to include the investigation of parallel differences as a function of attempted weight suppression (rela- tive to initial weight or relative to presumptive biologically de- fended levels). The Restraint Scale was initially proposed (Her- man & Mack, 1975) as a simple and relatively straightforward self-report device for identifying chronic dieters. At the time it was assumed that because chronic dieters were likely to be maintaining a body weight below "set point," identification of such dieters would therefore permit tests—in normal-weight people as well as the obese—of hypotheses derived from Nis- bett's (1972) seminal article on the effects of long-term hunger. Almost since its inception, the Restraint Scale has been sub- jected to criticism, both psychometric and conceptual. In this article, we review and discuss some of these criticisms of the Restraint Scale, and then consider the alternative scales that have been proposed recently as improvements. The major prob- lems that have been identified are (a) the Restraint Scale's con- founding of dietary restriction with disinhibited eating, (b) its apparent inadequacy when applied to the overweight, and (c) its factor structure. These major issues are not entirely separa- ble from one another, but we shall attempt to distill them into their essentials and address them sequentially, along with a number of lesser problems. Restraint and Disinhibition Early on, it became clear that the dieters identified by the Restraint Scale were as notable for their lapses of restraint as We thank the Natural Sciences and Engineering Research Council of Canada for their support. Correspondence concerning this article should be addressed to Todd F. Heatherton or C. Peter Herman, Department of Psychology, Univer- sity of Toronto, Toronto, Ontario M5S1A1, Canada. for their restraint per se (Herman & Mack, 1975; Herman & Polivy, 1975). Indeed, almost all of the research examining the eating behavior of people scoring high on the Restraint Scale has contrasted experimental situations in which restraint has remained intact with situations in which restraint is broken, with consequent overeating. As a result, our view (Herman & Polivy, 1980; Polivy & Herman, 1983) of the restraint con- struct—or more precisely, of the dieters identified by the scale—has changed so as to acknowledge that most dieters do not succeed in maintaining uninterrupted restriction of intake. The average dieter is more likely to exhibit periods of restraint punctuated by episodes of disinhibited overeating and, in all likelihood, does not achieve significant weight loss relative to physiologically defended levels. (This acknowledgment ap- peared very soon after the original Restraint Scale experiments; see Hibscher & Herman, 1977.) Discussions of the Restraint Scale's purpose—by ourselves and others—have perhaps been insufficiently explicit about the sort of person identified by the instrument. The dieter who suc- ceeds in achieving significant weight loss—presumably by avoiding the splurges of overeating that we have been investigat- ing in the laboratory—may well exist. This is the person who ought to show the long-term deprivation effects of interest to Nisbett (1972). And although this person may obtain a high restraint score, persons who diet less well—whose caloric re- strictions are canceled by bouts of caloric excess—may score as high or higher. Indeed, Polivy (1978) found that bulimic an- orexia nervosa patients (bingers) score higher on the Restraint Scale than do restricting anorexics (starvers). The fact that the most successful dieters do not necessarily obtain the highest scores has been viewed by some (Brief, Stunkard, & Hirsch, cited in Stunkard, 1981; Van Strien, 1986) as evidence for the invalidity of the Restraint Scale. These claims focus on two related features of the extant scale: First, as we have already mentioned, it does not give preeminence to "truly" restrained eaters (i.e., eaters whose behavior is charac- terized exclusively by restraint or restriction); and second, the 19