In order to measure disability in a randomised controlled trial of different intensities of rehabilitation following stroke, a modification of existing methods has been developed; it makes use of 17 items of activities of daily living (ADL) which are rated on a three-point scale, and it has been subjected to tests of repeatability and validity. While some activities (e.g. use of taps, ascending or descending stairs) are either redundant or repetitive, others provide independent measures of disability. There was no significant inter-observer variability; the assessors disagreed on 78 (3.7%) out of a total of 2,125 paired observations. In a study of short-term within-patient variability, different scores were assigned on 49 (14.4%) out of a total of 340 observations made on two separate occasions. Minor differences (mainly in activities involving equipment e.g. cookers) emerged when scores obtained in hospital were compared with those obtained at home, but the two sets of scores were still highly correlated (r = 0.962. There was, however, considerable discrepancy between the hospital scores and those derived from the patients' own estimates of their activities at home, the latter falling below actual capabilities as indicated by the hospital scores. A clinical validation of the index showed a significant association between the hospital ADL scores and the extent of cerebral lesions determined by the number of neurological deficits. This modified ADL index is sufficiently repeatable and valid for the assessment of patients with moderate disability as a result of stroke and other chronic diseases; in most circumstances, it can probably be simplified to include only five or six items.