This study describes anorexia nervosa (AN) patients who use excessive exercise for weight management and how this behaviour relates to the classical Diagnostic and Statistical Manual of Mental Disorders - Fourth Edition (DSM-IV) sub-grouping of AN. The study compared 428 clinical AN patients with 119 age and gender-matched controls. The AN cases were initially dichotomised according to DSM-IV subtype criteria into restricting (RAN; N=198) and binge-purge (BPAN; N=230) anorexia. The psychometric instruments were chosen to reflect key features concerning the diagnosis of eating disorders and characteristics of eating and food behaviour and included the 26-item Eating Attitude Test (EAT-26), Three Factor Eating Questionnaire (TFEQ), Dutch Eating Behaviour Questionnaire (DEBQ), Eysenck Personality Inventory (EPI) and Rosenberg Self-Esteem scale (RSE). Structured clinical interviews (1) were carried out in order to identify the subgroup of patients who use excessive exercise in order to facilitate weight control (EAN). The three groups (RAN, BPAN, EAN) did not differ in measures of current age, current body mass index, age of onset of AN and measures of restrained eating. However, significant differences were observed on EAT-26, DEBQ emotional and external factors, TFEQ disinhibition and hunger factors, EPI extraversion and neuroticism, and self-esteem. The EAN were similar to the RAN on the majority of variables but showed significant differences on extraversion, neuroticism, self-esteem and disease pathology (EAT-26). Compared with BPAN, EAN had lower disease pathology (EAT-26 scores), scored higher on the EPI extraversion scale, lower on the neuroticism scale and had greater self-esteem. The EAN also displayed significantly lower emotional and external eating (DEBQ) than BPAN and significantly lower disinhibition and hunger scores (TFEQ). These data suggest that EAN group display a mixed profile of characteristics resembling both BPAN and RAN. When EAN are defined as a separate group they appear to be phenotypically more similar to RAN than BPAN on this particular profile of variables. However when DSM-IV criteria are applied they are almost exclusively classified as BPAN. This outcome suggests that EAN do constitute an identifiable group that can be distinguished from RAN and BPAN. Consequently it is likely that the behaviour of excessive exercising should be considered as clinically relevant on the outcome of treatment.