Eating Attitudes Test (EAT-26)

1. Abstract

The Eating Attitudes Test (EAT-26) is one of the most widely utilized, empirically validated self-report screening instruments designed to assess disordered eating behaviors, attitudes toward food, body weight preoccupation, and risk profiles for eating disorders in both clinical and non-clinical cohorts. Developed by David M. Garner, Marion P. Olmsted, Yvonne Bohr, and Paul E. Garfinkel in 1982, the EAT-26 represents an abbreviated, psychometrically refined version of the original 40-item Eating Attitudes Test (EAT-40) first published in 1979. The instrument comprises 26 self-report items distributed across three robust principal factor dimensions: Dieting (13 items), Bulimia and Food Preoccupation (6 items), and Oral Control (7 items), complemented by five behavioral symptom questions that assess objective eating disorder pathologies including binge eating, self-induced vomiting, use of laxatives or diuretics, excessive exercise, and extreme rapid weight fluctuations, as well as body mass index (BMI) calculation.

Each attitudinal item is scored on a 6-point forced-choice Likert scale (Always, Very Often, Often, Sometimes, Rarely, Never), typically recoded into a weighted non-linear 0-to-3 scoring rubric where the most extreme symptomatic response receives a score of 3, the adjacent symptomatic responses receive 2 and 1 respectively, and the three non-symptomatic responses receive 0 (with Item 26 reverse-scored). Psychometric evaluations consistently demonstrate outstanding internal consistency, with Cronbach's alpha coefficients regularly exceeding .90 in clinical anorexia nervosa and bulimia nervosa samples and ranging between .80 and .88 in broad non-clinical community and university cohorts. Test-retest reliability ranges from .84 to .89 across short intervals. A cut-off score of 20 or higher indicates elevated risk of an eating disorder and warrants immediate comprehensive psychiatric diagnostic interviewing. The scale possesses high criterion-related, convergent, and discriminant validities, making it an indispensable first-line psychometric mechanism for epidemiological surveillance, clinical triaging, and prevention research worldwide.

2. Keywords

Eating Attitudes Test, EAT-26, Eating Disorders, Anorexia Nervosa, Bulimia Nervosa, Screening Instrument, Psychometrics, Factor Analysis, Dieting, Oral Control, Food Preoccupation, Body Image Distortion

3. Authors

The Eating Attitudes Test (EAT-26) was conceptualized, developed, and standardized by a pioneering team of clinical psychologists and psychiatrists specializing in eating pathology:

• David M. Garner, Ph.D. — Eminent clinical psychologist, founder and administrative director of the River Centre Clinic in Sylvania, Ohio, USA; formerly affiliated with the Department of Psychiatry at the University of Toronto and the Clarke Institute of Psychiatry, Toronto, Ontario, Canada.
• Marion P. Olmsted, Ph.D., C.Psych. — Senior clinical psychologist and former Director of the Ambulatory Care Eating Disorders Program at the University Health Network, Toronto; Associate Professor in the Department of Psychiatry, University of Toronto, Canada.
• Yvonne Bohr, Ph.D., C.Psych. — Clinical psychologist and Professor of Psychology in the Department of Psychology at York University, Toronto, Canada; researcher in developmental psychopathology and cognitive-behavioral processes.
• Paul E. Garfinkel, M.D., FRCPC — Renowned psychiatrist, Emeritus Professor in the Department of Psychiatry at the University of Toronto; former President and Psychiatrist-in-Chief of the Centre for Addiction and Mental Health (CAMH), Toronto, Canada.

4. Purpose

The primary objective of the Eating Attitudes Test (EAT-26) is to provide an efficient, standardized, cost-effective, and highly sensitive screening mechanism capable of identifying abnormal eating attitudes, pathological preoccupations with food and body weight, and behavioral markers indicative of clinically meaningful eating disorders. Designed specifically to detect symptoms associated with anorexia nervosa and bulimia nervosa, as well as atypical variants classified under Other Specified Feeding or Eating Disorders (OSFED) and binge eating disorder, the instrument functions predominantly as a risk detection tool rather than an autonomous diagnostic measure.

In epidemiological and public health contexts, the EAT-26 allows researchers to track prevalence rates, demographic risk profiles, and cross-cultural shifts in eating morbidity across large university, school, athletic, and community populations. Because individuals experiencing eating disorders frequently engage in minimization, denial, or emotional concealment of their symptomatology due to deep-seated feelings of shame or ego-syntonic investment in weight suppression, traditional unstructured clinical interviews can fail to detect subclinical presentations. The standardized, anonymous, or quasi-anonymous self-report format of the EAT-26 bypasses initial interpersonal defensiveness, systematically assessing specific cognitive distortions such as carbo-phobia, excessive guilt post-prandially, pervasive drive for thinness, compulsive caloric tracking, and somatic dissociation.

In clinical practice, the EAT-26 is widely deployed during initial psychiatric, psychological, and primary medical intakes to establish baseline symptom severity and determine whether a formal, comprehensive diagnostic evaluation — utilizing structured clinical interviews such as the Eating Disorder Examination (EDE) — is indicated. Furthermore, the scale serves as an objective outcome measure in psychotherapeutic and pharmacological interventions. Administering the EAT-26 at regular intervals across cognitive-behavioral therapy (CBT-E), family-based treatment (FBT), or inpatient refeeding programs enables multidisciplinary clinical teams to monitor therapeutic progress, measure the attenuation of core cognitive disturbances, detect treatment resistance, and evaluate relapse vulnerability following discharge.

5. Psychological Construct

The overarching psychological construct assessed by the EAT-26 is multidimensional disordered eating symptomatology, encompassing cognitive, affective, and behavioral deviations from normative nutritional engagement and somatic self-concept. Rather than viewing eating pathology as a monolithic entity, Garner and colleagues demonstrated that abnormal eating attitudes coalesce around three distinct yet inter-correlated subdimensions:

Dieting (Factor 1)

Comprising 13 items (Items 1, 6, 7, 10, 11, 12, 14, 16, 17, 22, 23, 24, and 26 [reversed]), the Dieting dimension captures an intense, ego-syntonic drive for thinness, acute dread of adiposity, rigid cognitive avoidance of calorie-dense or carbohydrate-rich foods, and chronic guilt triggered by caloric ingestion. Individuals scoring elevated marks on this factor exhibit obsessive cognitive appraisal of caloric expenditure during physical exertion (Item 12), systematic avoidance of foods containing sugar (Item 16) or high carbohydrate contents (Item 7), and persistent distress regarding bodily fat distribution (Item 14). This dimension reflects the foundational cognitive pathology identified in contemporary cognitive-behavioral formulations of eating disorders: the pervasive over-evaluation of shape and weight, wherein personal worth, emotional security, and self-esteem are judged almost exclusively in terms of body morphology and dietary discipline.

Bulimia and Food Preoccupation (Factor 2)

Consisting of 6 items (Items 3, 4, 9, 18, 21, and 25), the Bulimia and Food Preoccupation subscale assesses intrusive, disruptive thoughts concerning food, loss of inhibitory behavioral control resulting in subjective or objective binge eating episodes, and post-prandial compensatory purging behaviors. Prototypical manifestations include episodes of eating binges characterized by a subjective sense of powerlessness (Item 4), mechanical vomiting following food consumption (Item 9), spontaneous impulses to purge after meals (Item 25), and intrusive rumination wherein the individual feels that food actively controls their life (Item 18). Elevated scores on this factor delineate individuals who alternate between rigid restriction and dysregulated compensatory cycles, signaling elevated clinical urgency and higher risk of severe medical complications, including electrolyte disturbances and esophageal erosion.

Oral Control (Factor 3)

Encompassing 7 items (Items 2, 5, 8, 13, 15, 19, 20), the Oral Control subscale captures the behavioral, social, and interpersonal facets of self-imposed dietary restraint, ritualistic eating practices, and perceived social pressure to gain weight or consume food. Prototypical items assess rigid environmental self-control around food (Item 19), ritualistic meal fragmentation such as cutting food into miniscule pieces (Item 5), prolonged duration of meal ingestion compared to peers (Item 15), intentional avoidance of eating when experiencing somatic hunger cues (Item 2), and heightened sensitivity to interpersonal pressure from parents, peers, or clinicians demanding increased intake (Items 8 and 20). Unlike the pure affective panic of the Dieting subscale, Oral Control measures the operationalized mechanisms of restriction and the friction that emerges between the individual's ascetic self-discipline and their interpersonal environment.

6. Theoretical Framework

The theoretical framework underpinning the EAT-26 originates from the transdiagnostic and cognitive-behavioral paradigms of eating disorders pioneered by David Garner, Paul Garfinkel, and subsequently expanded by Christopher Fairburn and Peter Cooper. Historically, early psychodynamic formulations conceptualized anorexia nervosa as a symbolic rejection of psychosexual maturity or an expression of interpersonal enmeshment within dysfunctional family systems. In contrast, Garner and Garfinkel (1979, 1982) posited that disordered eating represents an overt manifestation of aberrant cognitive schemas concerning weight, body morphology, and nutritional autonomy.

Central to this cognitive model is the “core cognitive pathology”: the systemic overvaluation of eating, shape, and weight, and the perceived ability to control them. In healthy populations, self-worth is derived from a diversified constellation of domains, including interpersonal relationships, academic or professional achievements, recreational skills, and intellectual capabilities. In individuals vulnerable to eating disorders, this self-evaluative system contracts severely until an individual's self-esteem is almost entirely contingent upon their ability to maintain strict dietary restriction and weight suppression. The EAT-26 was explicitly constructed to capture the direct affective and cognitive sequelae of this cognitive bias, such as the catastrophic fear of becoming overweight (Item 1) and the internalization of thinness as an absolute imperative (Item 11).

Furthermore, the scale's framework aligns with self-regulation theory and modern dual-pathway sociocultural models of body image disturbance. Sociocultural theory posits that pervasive socio-environmental pressures favoring an unrealistically thin aesthetic become internalized by susceptible individuals, fostering severe body dissatisfaction. This body dissatisfaction acts as a potent distal trigger, propelling the individual into restrictive dieting behaviors. However, prolonged physiological starvation invariably precipitates neurobiological vulnerabilities, neurochemical adaptations, and homeostatic drives to consume food, ultimately resulting in cognitive preoccupation (Factor 2) and reactive binge eating. In turn, reactive binge eating exacerbates catastrophic fears of weight gain, motivating renewed and increasingly draconian oral control behaviors (Factor 3) and compensatory purging. The 26 items of the EAT-26 mathematically model this self-perpetuating, cybernetic feedback loop of cognitive restriction, neurobiological counter-regulation, and behavioral compensation.

7. Validity

The psychometric validity of the EAT-26 has been investigated across hundreds of diverse clinical, community, athletic, and international studies spanning over four decades, establishing robust criterion-related, convergent, and discriminant validity profiles.

Criterion and Known-Groups Validity

In the seminal validation study conducted by Garner, Olmsted, Bohr, and Garfinkel (1982), the EAT-26 was administered to a primary clinical criterion group of 160 female patients diagnosed with anorexia nervosa according to strict clinical criteria and a non-clinical control cohort of 140 female university students. The mean score for the anorexia nervosa cohort was 36.1 (SD = 17.0), compared to an average score of only 9.9 (SD = 9.1) among the normal female control participants, demonstrating profound discriminant capacity (p < .001). Receiver Operating Characteristic (ROC) analyses demonstrated that using the conventional cut-off score of 20 correctly identified 83.6% of clinical anorexia nervosa patients (true-positive rate / sensitivity) while exhibiting a specificity rate of 90.0% in excluding healthy, non-eating-disordered individuals.

Convergent Validity

Convergent validity is robustly documented via high correlations between the EAT-26 and alternative psychometric measures of eating and psychological pathology. The correlation between the EAT-26 total score and the original 40-item parent scale (EAT-40) reached r = .98 across clinical cohorts, indicating near-perfect construct preservation despite the 35% item reduction (Garner et al., 1982). Furthermore, EAT-26 total scores demonstrate substantial positive correlations with the subscales of the Eating Disorder Inventory (EDI), exhibiting strong associations with Drive for Thinness (r = .81 to .86), Body Dissatisfaction (r = .65 to .72), and Bulimia (r = .68 to .77). Correlations with the Eating Disorder Examination-Questionnaire (EDE-Q) Global Score consistently exceed r = .75 across adolescent and young adult clinical cohorts. Moderate to high positive correlations have also been observed with measures of general psychopathology, including the Beck Depression Inventory (BDI; r = .42 to .54) and the State-Trait Anxiety Inventory (STAI; r = .38 to .49), reflecting the extensive clinical comorbidity between disordered eating patterns, major affective disorders, and generalized anxiety.

Discriminant Validity

Discriminant validity has been demonstrated by the scale's capacity to differentiate between distinct psychological conditions. While individuals diagnosed with major depressive disorder, generalized anxiety disorder, or obsessive-compulsive disorder without eating pathology may exhibit mild elevations due to general distress or secondary appetite disturbances, their mean scores typically fall well below the clinical cut-off of 20 (typically averaging between 6.0 and 11.5). Crucially, the EAT-26 differentiates healthy individuals who diet for cardiovascular health, religious reasons, or athletic weight-class compliance from individuals exhibiting authentic clinical psychopathology; the latter score significantly higher on the Bulimia and Food Preoccupation and Oral Control subdimensions, confirming that the scale detects psychopathological fixation rather than benign health-conscious lifestyle adjustments.

8. Reliability

The Eating Attitudes Test (EAT-26) exhibits high internal consistency, inter-item homogeneity, and temporal stability across a diverse spectrum of demographic cohorts and clinical contexts.

Internal Consistency

In the original psychometric derivation by Garner et al. (1982), the EAT-26 demonstrated a Cronbach's alpha coefficient of .90 within the anorexia nervosa clinical cohort and .83 within the non-clinical female comparison group. Subsequent reliability generalization meta-analyses (e.g., Gleaves, Pearson, Ambwani, & Morey, 2014) evaluating psychometric results across dozens of independent empirical investigations found mean Cronbach's alpha estimates of .89 for the total score in clinical cohorts and .84 in general non-clinical community samples. Reliability coefficients for the individual subscales, while naturally slightly lower due to fewer items, remain psychometrically solid: Dieting regularly produces alpha coefficients between .84 and .91; Bulimia and Food Preoccupation ranges from .75 to .84; and Oral Control typically ranges between .68 and .76.

Temporal Stability (Test-Retest Reliability)

Test-retest reliability evaluations demonstrate robust stability across non-interventional intervals. In a longitudinal evaluation of young women across an eight-week interval, Banasiak et al. (2001) observed an intraclass correlation coefficient (ICC) of .84. Similar evaluations conducted over two- to three-week intervals in non-clinical cohorts reported test-retest coefficients between .86 and .89, demonstrating that the underlying cognitive attitudes tapped by the instrument possess substantial temporal durability in the absence of targeted psychiatric or psychological interventions. Conversely, in longitudinal clinical treatment studies, EAT-26 scores decline significantly following successful cognitive-behavioral or residential interventions (dropping from baseline pre-treatment means above 35 to post-treatment means below 15), confirming that while the instrument demonstrates high short-term measurement stability, it remains acutely sensitive to authentic therapeutic change.

9. Factor Analysis

The factorial structure of the EAT-26 was established through exploratory factor analysis (EFA) with principal components analysis and varimax orthogonal rotation conducted on the original 40 items administered to 160 clinical anorexia nervosa patients (Garner et al., 1982). Fourteen items with low factor loadings (< .40), cross-loadings across multiple dimensions, or inadequate discriminatory power between clinical cases and normal controls were pruned, yielding the streamlined 26-item instrument. The resulting three-factor orthogonal model accounted for 40.2% of the total cumulative variance in the clinical cohort:

• Factor 1: Dieting — Accounted for 25.1% of the total variance, comprising 13 items with high factor loadings ranging from .46 to .78 (e.g., Item 1 “I am terrified about being overweight” loaded at .78; Item 11 “I am occupied with a desire to be thinner” loaded at .75).
• Factor 2: Bulimia and Food Preoccupation — Accounted for 8.7% of the total variance, comprising 6 items exhibiting strong factor loadings ranging from .48 to .76 (e.g., Item 4 “I have gone on eating binges where I feel that I may not be able to stop” loaded at .76; Item 9 “I vomit after I have eaten” loaded at .71).
• Factor 3: Oral Control — Accounted for 6.4% of the total variance, comprising 7 items with factor loadings ranging from .41 to .69 (e.g., Item 20 “I feel that others pressure me to eat” loaded at .69; Item 19 “I display self-control around food” loaded at .56).

Confirmatory factor analytic (CFA) investigations conducted in contemporary non-clinical and international samples have occasionally noted that the original three-factor model demonstrates variable fit. In large non-clinical university samples (e.g., Ocker, Lam, Jensen, & Zhang, 2007; Pereira et al., 2008), the original three-factor structure frequently demonstrates moderate fit indices: Comparative Fit Index (CFI) values ranging between .88 and .93, Root Mean Square Error of Approximation (RMSEA) between .052 and .068, and Standardized Root Mean Square Residual (SRMR) between .050 and .062. Some investigators have proposed alternative four-factor or bi-factor solutions in non-clinical cohorts, separating body dissatisfaction from purely restrictive dieting or dividing oral control into internal restraint versus social pressure. Nonetheless, the classic Garner et al. (1982) three-factor model remains the universally recognized standard in clinical diagnostic research, displaying superior structural invariance and construct validity within bona fide clinical populations.

10. Instrument / Measurement Tool

• Instrument Name: Eating Attitudes Test (EAT-26)
• Target Population: Adolescents (aged 14 and older) and adults across clinical, psychiatric, athletic, collegiate, and general community populations. (For children aged 8 to 13, the specialized Children's Eating Attitudes Test [ch-EAT] is utilized).
• Administration Format: Standardized self-administered paper-and-pencil or computerized self-report questionnaire.
• Completion Time: Approximately 5 to 8 minutes.
• Item Count: 26 core attitudinal statements, accompanied by 5 behavioral frequency questions and self-reported height and weight metrics for BMI calculation.
• Response Scale (Attitudinal Items): 6-point forced-choice frequency scale:
  • 1 = Always
  • 2 = Very Often
  • 3 = Often
  • 4 = Sometimes
  • 5 = Rarely
  • 6 = Never
• Scoring and Transformation Rubric:
  • Standard Non-Linear Symptomatic Recoding (Items 1 through 25):
    • Always = 3 points
    • Very Often = 2 points
    • Often = 1 point
    • Sometimes = 0 points
    • Rarely = 0 points
    • Never = 0 points
  • Reversed Scoring Item (Item 26: “I enjoy trying new rich foods”):
    • Always = 0 points
    • Very Often = 0 points
    • Often = 0 points
    • Sometimes = 1 point
    • Rarely = 2 points
    • Never = 3 points
• Total Score Range: 0 to 78 points.
• Subscale Item Distributions:
  • Dieting Subscale (Items 1, 6, 7, 10, 11, 12, 14, 16, 17, 22, 23, 24, 26): Score range 0 to 39.
  • Bulimia and Food Preoccupation Subscale (Items 3, 4, 9, 18, 21, 25): Score range 0 to 18.
  • Oral Control Subscale (Items 2, 5, 8, 13, 15, 19, 20): Score range 0 to 21.
• Behavioral Symptom Screener: Five ancillary behavioral questions inquiring whether, in the past 6 months, the respondent has engaged in:
  1. Eating binges where they felt unable to stop.
  2. Ever made themselves sick (vomited) to control weight or shape.
  3. Used laxatives, diet pills, or diuretics to control weight or shape.
  4. Exercised more than 60 minutes a day specifically to lose or control weight.
  5. Lost 20 pounds (9 kg) or more in the past 6 months.
• Referral Criteria / Clinical Interpretation: A comprehensive diagnostic evaluation by a specialized mental health professional is strongly recommended if an individual meets ANY of the following three referral criteria:
  • Criterion 1 (EAT-26 Cut-Off): An EAT-26 total score equal to or exceeding 20 points.
  • Criterion 2 (Behavioral Symptoms): An affirmative response to any of the behavioral symptom questions meeting established clinical thresholds (e.g., self-induced vomiting or binge eating occurring once a month or more).
  • Criterion 3 (Low Body Mass Index): A calculated BMI indicating significant underweight status relative to established demographic norms (e.g., adult BMI < 18.5 kg/m² or youth age-adjusted percentiles below the 5th percentile).

11. Permissions & Fee and Test Year

The Eating Attitudes Test (EAT-26) was formally published in 1982 by David M. Garner, Marion P. Olmsted, Yvonne Bohr, and Paul E. Garfinkel as an open-access psychometric contribution to the international medical and psychiatric research community. Under the explicit guidance of its primary copyright holder, Dr. David M. Garner, the EAT-26 is placed in the public domain for research, academic, and non-profit clinical screening purposes.

No licensing fees or formal institutional permissions are required for individual clinicians, non-profit institutions, or academic researchers to reproduce, print, or integrate the instrument into computerized research batteries, provided that appropriate bibliographic attribution is maintained and the items and scoring protocols are not altered without authorization. For clinical programs, epidemiological screening initiatives, and interactive online assessments, official scoring rubrics, educational resources, and multi-language translations are maintained directly through the official EAT-26 portal (www.eat-26.com). Commercial utilization or integration into proprietary fee-for-service software requires direct written permission from Dr. David M. Garner.

12. References

Banasiak, S. J., Wertheim, E. H., Koerner, J., & Voudouris, N. J. (2001). Test-retest reliability and internal consistency of three common measures of eating disorder attitudes and behaviors: EAT-26, EDI, and BULIT. Psychological Assessment, 13(1), 158–164. https://doi.org/10.1037/1040-3590.13.1.158

Garfinkel, P. E., & Newman, A. (2001). The Eating Attitudes Test: Twenty-five years later. Eating and Weight Disorders – Studies on Anorexia, Bulimia and Obesity, 6(1), 1–21. https://doi.org/10.1007/BF03339747

Garner, D. M., & Garfinkel, P. E. (1979). The Eating Attitudes Test: An index of the symptoms of anorexia nervosa. Psychological Medicine, 9(2), 273–279. https://doi.org/10.1017/S0033291700030735

Garner, D. M., Olmsted, M. P., Bohr, Y., & Garfinkel, P. E. (1982). The Eating Attitudes Test: Psychometric features and clinical correlates. Psychological Medicine, 12(4), 871–878. https://doi.org/10.1017/S0033291700049161

Garner, D. M., Olmsted, M. P., & Polivy, J. (1983). Development and validation of a multidimensional eating disorder inventory for anorexia nervosa and bulimia. International Journal of Eating Disorders, 2(2), 15–34. https://doi.org/10.1186/2050-2974-2-6

Maloney, M. J., McGuire, J. B., & Daniels, S. R. (1988). Reliability testing of a children's version of the Eating Attitude Test. Journal of the American Academy of Child & Adolescent Psychiatry, 27(5), 541–543. https://doi.org/10.1097/00004583-198809000-00004

Ocker, L. B., Lam, E. T., Jensen, B. E., & Zhang, J. J. (2007). Psychometric properties of the Eating Attitudes Test. Measurement in Physical Education and Exercise Science, 11(1), 25–48. https://doi.org/10.1080/10913670709337424

Pereira, A. T., Maia, B., Bos, S., Soares, M. J., Marques, M., Valente, J., & Macedo, A. (2008). The Portuguese short form of the Eating Attitudes Test-40. European Eating Disorders Review, 16(4), 319–325. https://doi.org/10.1002/erv.845

Rivas, T., Bersabé, R., Jiménez, M., & Berrocal, C. (2010). The Eating Attitudes Test (EAT-26): Reliability and validity in Spanish female samples. The Spanish Journal of Psychology, 13(2), 1044–1056. https://doi.org/10.1017/S1138741600002687

13. Items of the Scale