The purpose of this report is to describe the reliability of three clinical measures used to evaluate changes in shoulder subluxation. The three methods include measuring the subacromial space in fingers breadth, using calipers, or a plexiglass jig. Thirty-six patients with shoulder subluxation who had experienced a cerebrovascular accident were the subjects. Four occupational therapists with experience in stroke rehabilitation were divided into two teams of two therapists and rated the subjects independently. Each rater repeated her assessments on nine subjects to test intrarater agreement. Inter-rater agreement was assessed both between members of the same team (27 subjects per rater pair) and members of different teams (18 subjects per rater pair). The measure of reliability was the intraclass correlation coefficient (ICC2, 1) as derived from two-way analysis of variance. The highest intra-rater reliability was displayed by the finger breadth method, the second highest by the caliper method and the lowest by the plexiglass jig. The coefficients in the former two cases were always above .8. Using the jig only one rater achieved this level. Agreement between the two members of the same team were above .75 for the fingers and caliper methods, but less than this for the jig. Between members of different teams however, only the finger breadth method attained reliabilities above .7, and the plexiglass jig, in particular, showed very low reliability. These results demonstrate the difficulty of achieving consistent clinical measurement for a condition like shoulder subluxation.(ABSTRACT TRUNCATED AT 250 WORDS)