The Eating Attitudes Test (EAT) is commonly used in studies of anorexia nervosa. This paper reports some unexpected findings concerning the relationship between the EAT and the social class of the respondent. Our analysis is based on data from a sample of over 3000 English schoolgirls from both state and private schools. Important differences were found between the results from the private and state schools suggesting a social class difference in response to this questionnaire. The mean scores on the EAT were significantly higher in the state schools than in the private schools. However the number of cases of anorexia nervosa identified on the basis of individual interviews was higher in the private schools. A detailed analysis of the responses to the individual questions indicate that this discrepancy was due to systematic differences in the way the girls responded to different sets of questions.