GOAL: To explore the influence of social, psychological, and health factors on self-report of function. SUBJECTS: A convenience sample of 289 community-dwelling elderly aged 65-97 years. METHODS: We compared a measure of function based on observed performance, the Physical Capacity Evaluation (PCE) with a self-reported measure of functional limitations (HAQ), in a cross-sectional study. Stepwise multiple regression identified variables predicting self-reported disability, controlling for observed function. RESULTS: Controlling for PCE, self-reports of greater disability (HAQ) were predicted by current joint pain or stiffness, use of prescription medications, urban dwelling, depression, female gender, lack of memory problems, arthritis and lack of exercise. A final model included recent decline in function, dissatisfaction with function, gender, joint pain or stiffness, and observed function, explaining 85% of the variance in self-reported disability. The hypothesis that aging is associated with declining expectations of functional ability was not supported. However, recent health problems affected participants' reporting of limitations, consistent with a recalibration-type response shift. Perceived decline in function over the past six months, a fall within the last month, illness in the last week and pain or stiffness on the day of the exam all raised self-reports of disability. As suggested by adaptation level theory, subjects with recent problems might have an inflated perception of limitations due to shifts in their internal standards. When administered first, the observed performance test improved correlations between observed and self-reported function, primarily among those who did not report a recent decline in function. This suggests that this group may have benefited more from salient information about their abilities provided by performing the PCE before self-report. CONCLUSION: Our data confirm the importance of social, psychological, and health influences in self-report of disability, and are consistent with the hypothesis that people may recalibrate their self assessments based on recent health problems.