Research That Matters (January 17 - 20, 2008)

Background & Purpose: The Eating Attitudes Test (EAT-26) is probably the most widely used standardized measure of symptoms and concerns characteristic of eating disorders. The EAT-26 alone does not yield a specific diagnosis of an eating disorder; Neither the EAT-26, nor any other screening instrument, has been established as highly efficient as the sole means for identifying eating disorders. However, studies have shown that the EAT-26 can be an efficient screening instrument as part of a two-stage screening process in which those who score at or above a cut-off score of 20 are referred for a diagnostic interview. Surveys of adolescents or young adult women indicate that about 15% score at or above 20 on the EAT-26. The original factor analysis study showed three interrelated factors such as dieting behavior, oral control, and bulimia nervosa-food preoccupation and then concluded that the EAT-26 is a reliable, valid, and economical instrument which may be used as an objective measure of the symptoms of anorexia nervosa (Garner et al., 1982). However, subsequent confirmatory studies reported the three-factor model was a poor fit to the data and proposed a shorter version of the scale excluding several deviant items. The purpose of this study was to revisit the psychometric properties of the 26-item Eating Attitude Test using a sample of 422 female college students.

Methods: 422 college women [Age: M=20.21, SD=1.91; African Americans =228 (52%), Caucasians = 182 (43%)] voluntarily participated in a 14-page questionnaire. In addition to the EAT-26 scale, three psychological constructs such as self-esteem, internalized shame, and self-objectification were included to test validity with the EAT-26. Only 5% (n=24) of participants reported to have ever been treated for an eating disorder in this sample.

Results: Preliminary item analysis detected five deviant items affecting overall reliability level. After deleting those weak items, the alpha level was elevated from .89 to .93. The prior three-factor model with 26 items showed a poor model fit (CFI = .85, RMSEA = .08). Subsequent CFA results for the four-factor model with final 21 items showed acceptable fit indices (CFI =.93, RMSEA =.06). This study also indicated that the EAT-26 scale is strongly positively correlated with appearance-related self-objectification (r = .30) and internalized shame (r = .47), but is weakly negatively associated with self-esteem (r = -.181). Additional discriminant analysis result revealed that 78.8% of the non-treated group was correctly classified, while 80% of the ever-treated group was correctly classified.

Implications: Although the four-factor model with 21 items shows promising validity, future research is needed to explore the nature of each factor differently identified by different ethnic groups. We also suggest that findings from the present study should be used to start comprehensive validation of a measure of eating attitude for use in non-clinical young female adults, rather than assume that validation of a 21-item four-factor EAT is complete.

Reference: Garner, D.M., Olmstead, M.P., Bohr, Y., & Garfinkel, P.E. (1982). The eating attitude test: Psychometric features and clinical correlates. Psychological Medicine, 12, 871-8.