The study investigated longitudinal change in cognitive function in 87 patients with Huntington's disease (HD), using a range of neuropsychological tests, which tap mental manipulative abilities, memory, and frontal executive skills. Over a 1-year period the largest changes were noted in letter fluency, object recall, and Stroop Test performance, whereas no changes were noted over more than 3 years on the modified Wisconsin Card Sorting Test. Contrary to expectation, greater change was evident over 1 year for tasks with low compared to high cognitive demands. The differential sensitivity of tasks was attributed in part to inherent characteristics of the tests themselves: their capacity to detect minor gradations of change and their vulnerability to practice effects. However, the greater change for relatively automatic, speed-based tasks with low cognitive demands was interpreted as reflecting the evolution of HD, with a greater magnitude of change occurring in basal ganglia than cortical function. One purpose of the study was to identify tasks sensitive to the progression of HD and hence most suitable for the evaluation of therapies. Despite reaching statistical significance by virtue of the large group size, numerical differences in test scores over 1 year were very small, suggesting that the use of such tests to evaluate change in individuals or small groups of subjects would be problematic. The data highlight the slow progression of HD, the limitations of standard cognitive tests in detecting change over short periods, and the need for therapeutic studies that encompass a relatively prolonged time frame.