This paper describes the development of the PAS-ADD, a semistructured clinical interview for use specifically with patients with learning disabilities, based on items drawn from the PSE. The PAS-ADD includes a number of novel features including: parallel interviewing of patient and informant; a three-tier structure to provide a flexible interview appropriate to the patient's intellectual level; use of a memorable 'anchor event' in the patient's life to improve time focus; and simplified wording, improved organisation and lay out. Inter-rater reliability was investigated using an experimental design in which two raters viewed and re-rated videotaped PAS-ADD interviews which had been conducted by an experienced clinician. Reliability results compared favourably with those obtained in a major study of PSE reliability with a sample drawn from non-learning disabled individuals. Mean kappa for all items was 0.72. Other indexes of reliability were also good. In the current phase of development, the PAS-ADD is to be expanded to include further diagnostic categories, including schizophrenia and autism. The new version will be updated for use with ICD-10 criteria.