Intrarater and interrater reliabilities were assessed for upper-extremity goniometric measurements of a 5-yr-old girl with severe spastic quadriplegia. Percentage agreement was the method used to evaluate goniometric reliability of 3 motions bilaterally: shoulder flexion, shoulder abduction, and elbow extension. Intrarater reliability for each of the 2 raters was higher than interrater reliability. The mean differences and the range of differences between raters for each of the 6 motions measured also were recorded. There was wide variability in measurements both within and between raters. These results suggest that a difference of .+-. 10-15.degree. in range of motion over time does not justify conclusions of either significant improvement or significant regression in a child with severe spastic cerebral palsy.