To study the reproducibility of the measurement of shoulder movement, we have examined a series of 64 patients with and without shoulder problems, measuring active elevation, abduction, and external rotation in adduction using an inclinometer. The difference within which readings by different observers were expected to lie for 95% of the pairs of observations ranged from 24 degrees to 33 degrees for different movements in asymptomatic shoulders and from 24 degrees to 41 degrees in those with unilateral shoulder symptoms awaiting surgery.