Skip to main content
v2026.11,986 entries · CC-BY 4.0

EAT-26 (Eating Attitudes Test): Scoring, Cutoff and Use in Research

EAT-26 scoring key, the cutoff of 20, a worked example, code for Excel, R and Python, APA reporting and permission terms.

Written and maintained by CASRAI Editorial Board

Published

On this page

The EAT-26 (Eating Attitudes Test, 26-item version) is a self-report screening questionnaire for eating-disorder risk, not a diagnostic tool. Researchers use it to flag participants whose responses warrant follow-up. This page explains how it is scored, what the cutoff of 20 means, how to compute and report it, and what the permission terms are. It is general research-methods information, not clinical advice and not a self-assessment.

What is the EAT-26?

The EAT-26 is a 26-item questionnaire shortened from the 40-item Eating Attitudes Test (EAT-40) on the basis of a factor analysis. The official site states that it is not designed to make a diagnosis and is intended primarily for adolescents and adults. The EAT-40 was introduced by Garner and Garfinkel (1979); the 26-item version is attributed to Garner and colleagues (1982). In a study it is usually used as a screen, to decide who is referred for a proper assessment, or as a continuous measure of disordered-eating attitudes in a sample.

It differs from longer instruments that profile eating-disorder psychopathology across subscales; see the EDE-Q guide for that kind of measure.

EAT-26 scoring and the cutoff of 20

Items 1 to 25 are scored Always = 3, Usually = 2, Often = 1, and any other answer = 0. Item 26 is scored in the opposite direction (Never = 3). The total is the sum of the 26 item scores, so with a maximum of 3 per item the possible range is 0 to 78 (26 × 3, our arithmetic from the scoring key).

Item group Scoring rule (official key)
Items 1–25 Always = 3, Usually = 2, Often = 1, other answers = 0
Item 26 Reverse-scored (Never = 3)
Total score Sum of all 26 items
Screening threshold 20 or more: recommended for interview by a qualified professional

The official scoring page also lists behavioral questions (A to E) that are separate from the 26 items; a referral is indicated if any box is checked. The site describes a “referral index” built from three things: the total score, the behavioral questions and body mass index (BMI).

Worked example with the arithmetic shown

This is an illustrative, hypothetical respondent, not real data. Suppose the 25 regular items contribute 17 points in total, and the respondent answers item 26 with “Never”.

  • Items 1–25: 17 points
  • Item 26: Never, reverse-scored = 3 points
  • Total: 17 + 3 = 20, which meets the threshold of 20 or more

Had the same respondent answered item 26 with an answer that earns 0, the total would be 17, below the threshold. The example shows why item 26 must be recoded before summing: leaving it unrecoded changes the total and can move a participant across the cutoff.

Interpreting the result

Result Reading (per the official site)
Total of 20 or more Interview with a qualified professional is recommended; this is a referral flag, not a diagnosis
Total below 20 Does not rule out a serious eating problem
Binge eating disorder The site notes that some people with binge eating disorder score low

Treat the instrument as a screen. A study report should say that scores at or above 20 identify participants for follow-up assessment, and should avoid calling them “cases” or “patients with an eating disorder”.

Assumptions and common mistakes

  • Forgetting to reverse item 26. The most common coding error, shown above.
  • Treating 20 as a diagnosis. It is a referral threshold for interview.
  • Ignoring the behavioral questions and BMI. The official referral index uses all three components; a total-score-only analysis is a narrower use and should be described as such.
  • Assuming a low score means no problem. The site states it does not rule out a serious eating problem.
  • Reusing a published reliability coefficient. Reliability belongs to scores in a given sample, so compute and report internal consistency (see Cronbach’s alpha) for your own data, and examine how the measure relates to other constructs (see convergent and discriminant validity).
  • Not checking score distributions. Screening scores in community samples often cluster at the low end, so inspect floor and ceiling effects before choosing an analysis.

How to report the EAT-26 (APA example)

Name the instrument, cite the source, state the scoring and the threshold, and report reliability for your sample. A template, with placeholders you must fill from your own data:

Disordered-eating attitudes were measured with the 26-item Eating Attitudes Test (EAT-26; Garner et al., 1982). Items are summed to a total score (range 0–78), with higher scores indicating more disordered-eating attitudes; item 26 was reverse-scored. Internal consistency in this sample was α = [value]. Participants scoring 20 or above were classified as meeting the screening threshold for further assessment (n = [n], [%]).

Reference entry: 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.

How to calculate the EAT-26 in Excel, R or Python

Each snippet assumes responses stored as text labels (Always, Usually, Often, Sometimes, Rarely, Never) in columns q1 to q26. Item 26 uses the mirror-image key: Never = 3, Rarely = 2, Sometimes = 1, others 0.

Excel

Items 1-25 (per cell, e.g. A2):  =IF(A2="Always",3,IF(A2="Usually",2,IF(A2="Often",1,0)))
Item 26 (cell AA2):              =IF(AA2="Never",3,IF(AA2="Rarely",2,IF(AA2="Sometimes",1,0)))
Total:                           =SUM(first_score_cell:last_score_cell)
Flag:                            =IF(total>=20,"refer","")

R

fwd <- c(Always = 3, Usually = 2, Often = 1, Sometimes = 0, Rarely = 0, Never = 0)
rev <- c(Never = 3, Rarely = 2, Sometimes = 1, Often = 0, Usually = 0, Always = 0)
score <- function(x, key) unname(key[as.character(x)])
items <- paste0("q", 1:25)
d$eat26 <- rowSums(cbind(sapply(d[items], score, key = fwd), score(d$q26, rev)))
d$eat26_flag <- d$eat26 >= 20

Python (pandas)

fwd = {"Always": 3, "Usually": 2, "Often": 1, "Sometimes": 0, "Rarely": 0, "Never": 0}
rev = {"Never": 3, "Rarely": 2, "Sometimes": 1, "Often": 0, "Usually": 0, "Always": 0}
items = [f"q{i}" for i in range(1, 26)]
df["eat26"] = df[items].apply(lambda c: c.map(fwd)).sum(axis=1) + df["q26"].map(rev)
df["eat26_flag"] = df["eat26"] >= 20

Unmapped or missing labels produce missing values in R and pandas; decide on a missing-data rule before summing.

Permission, licensing and citation

The EAT-26 is copyrighted. According to the official permission page, fees and royalties are waived for individuals, a permission letter can be requested to reproduce or link to the test, and commercial web platforms that charge user fees need a license from Eating Attitudes, LLC. The official site asks users to cite Garner et al. (1982) for the EAT-26 and Garner and Garfinkel (1979) for the original EAT-40. If your study reproduces the items in a protocol, appendix or online survey, request permission first.

Frequently asked questions

What does an EAT-26 score of 20 mean?

A total of 20 or more is the threshold at which the official site recommends an interview with a qualified professional. It is a referral flag, not a diagnosis.

What is the EAT-26 score range?

0 to 78, from 26 items each scoring 0 to 3.

Which item is reverse-scored?

Item 26. Items 1 to 25 use Always = 3, Usually = 2, Often = 1, other = 0.

Can a score below 20 rule out an eating disorder?

No. The official site states a lower score does not rule out a serious eating problem, and that some people with binge eating disorder score low.

Is the EAT-26 free to use in research?

Fees and royalties are waived for individuals, but you should request permission to reproduce or link it, and commercial platforms charging user fees need a license.

How is it different from the EAT-40?

The EAT-26 is the shortened form derived from the EAT-40 through factor analysis.

Sources

  • EAT-26 official site, eat-26.com, with the scoring page (/scoring/) and permission page (/permission/), read 2026-11-07.
  • 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.
  • 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.

Follow CASRAI

Research-administration guidance, standards updates and independent tool reviews.

Ask CASRAI · free to try

Ask about EAT-26 (Eating Attitudes Test): Scoring, Cutoff and Use in Research

2 free questions a day, no account.

An AI assistant for research: funder rules, compliance, methods and publishing. It links the sources behind its answers and labels answers that come from a web search.

Answers draw on CASRAI's guides and dictionary plus the federal and funder documents we index: Federal Register, Grants.gov, Regulations.gov and UKRI.

Works on this site and inside Claude, Cursor and the AI tools you already use.

Everything CASRAI publishes — this page, the dictionary, the guides and the news — stays open to read, with no account and no card.

Referenced across the research world

University of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logoUniversity of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logo
  • University of Cambridge logo
  • Columbia University logo
  • Crossref logo
  • University of Edinburgh logo
  • Harvard University logo
  • University of Oxford logo
  • Princeton University logo
  • Stanford School of Medicine logo
  • University College London logo
  • ORCID logo

View CASRAI adoption →