The study explores whether an atypical form of obsessional illness can be delineated and separated from the conventional form of obsessive-compulsive neurosis (OCN). From a group of 45 obsessive patients, 8 were selected on the basis of 3 criteria: presence of a severely debilitating main obsessive symptom; bordering on the delusional; no schizophrenic symptoms. Assessment and outcome measures included the Psychiatric Questionnaire, the Leyton Obsessional Inventory, Fear Survey Schedule, and IPAT Self-Analysis Form. Self-assessment forms allowed patients to make social adjustment and neurotic symptom ratings. In a multimodal approach, patients were assigned to behavioural and pharmacological treatments on the basis of severity. Reassessment took place after 50 sessions of therapy. Results of analysis of variance statistics indicated that the atypical group had a more malignant form of illness, with more varied and severe obsessions. A poorer prognosis for the atypical group was indicated by: greater social maladjustment, poor employment records, illness of longer duration showing no remissions despite more courses of treatment, and poor response to treatment throughout. The atypical group manifested fewer characteristic features of OCN (example: fewer precipitating events). On the other hand, schizophrenia was not imputed, although delusion-like experiences in the atypical group suggest a psychotic form of illness. The term "obsessive psychosis" suggested by Strauss and recently investigated by Weiss et al and Robinson et al is proposed for our atypical group. Results are compared with those of other investigators. It is concluded that the delineation of a subgroup of obsessional illness is desirable for research and therapy since a form of atypical obsessional illness or obsessive psychosis can be differentiated on aetiological, phenomenological and prognostic factors.