The use of instructional advance directives assumes that preferences for life-sustaining medical treatment remain stable over time and across changes in life condition. A sample of 332 older adults recorded their preferences for 4 life-sustaining treatments in 9 illness scenarios. These preferences were elicited again 1 and 2 years after the original interview. Overall, preferences for life-sustaining treatment were moderately stable over time, but stability varied significantly across judgments. Preferences were most stable for illness scenarios that were most and least serious and for decisions to refuse treatment. Age, gender, education, and prior completion of an advance directive were all related to preference stability, and evidence indicated that declines in physical or psychological functioning resulted in decreased interest in life-sustaining treatment.