Objective: There is increasing evidence that neuron loss precedes the phenotypic expression of Hun- tington's disease (HD). As genes for late-onset neurodegenerative diseases are identified, the need for accurate assessment of phenoconversion (i.e., the transition from health to the disease phenotype) will be important. Methods: Prospective longitudinal evaluation using the Unified Huntington's Disease Rating Scale (UHDRS) was conducted by Huntington Study Group members from 36 sites. There were 260 persons considered "at risk" for HD who initially did not have manifest disease and had at least one subsequent evaluation. Repeat UHDRS data, obtained an average of 2 years later, showed that 70 persons were given a diagnosis of definite HD based on the quantified neurologic examination. Results: Baseline cognitive performances were consistently worse for the at-risk group who demonstrated conversion to a definitive diagnosis compared with those who did not. Longitudinal change scores showed that the at-risk group who did not demonstrate manifest disease during the follow-up study period demonstrated improvements in all cognitive tests, whereas performances in the at-risk group demonstrating conversion to disease during the study declined across cognitive domains. Conclusions: Neuropsychological measures show impairment 2 years before the development of more manifest motor disease. Findings suggest that these brief cognitive measures administered over time may capture early striatal neural loss in HD.