This article reports the development and psychometric properties of the Alzheimer's Disease Cooperative Study-Clinical Global Impression of Change (ADCS-CGIC). At present, a number of unvalidated CGIC scales are used in clinical trials, with various methods for making ratings. The ADCS-CGIC was designed on the basis of a survey of ADCS clinicians and by adapting existing instruments. It includes an organized but unstructured format, with which a clinician can address clinically relevant change. The instrument's reliability and validity were assessed in a prospective trial of Alzheimer's disease (AD) and healthy subjects over a 12-month period. It showed good short-term reliability at 1 and 2 months, with 90 and 94% of AD subjects, respectively, rated as having changed not at all or only minimally. The ADCS-CGIC's face validity was demonstrated by untreated AD subjects rated as having worsened over time at both 6 months (56% rated as having worsened) and 12 months (81% rated as having worsened), whereas only 2% of control subjects showed minimal worsening. As a measure of predictive validity, ADCS-CGIC ratings at 12 months were significantly associated with change on four severity scales. As with other measures, change ratings were sensitive to dementia severity. Moderately impaired subjects showed greater worsening than other subjects. ADCS-CGIC ratings of greater worsening were made after the informant interview, regardless of whether informants or subjects were interviewed first. The ADCS-CGIC is a valid and reliable instrument for use in clinical trials.