An objective means based on the carbon and nitrogen stable isotope analysis of five hairs per individual is presented for distinguishing between individuals with anorexia nervosa and/or bulimia nervosa from non-clinical individuals (i.e. clinically normal controls). Using discriminant analysis, an algorithm has been developed that provides both sensitivity and specificity of 80% in making diagnoses of individuals with these eating disorders. With further refinements, the results suggest that it may be also possible to distinguish between individuals with anorexia or bulimia. Finally, the study shows the value of conducting blind tests and using larger sample sizes of both control and treatment groups. Both groups are needed to validate the diagnostic value of a method and to provide measures of sensitivity and specificity of any diagnostic test. Copyright © 2006 John Wiley & Sons, Ltd.