OBJECTIVES: To determine whether benzodiazepine use in older women increased the risk of decline in physical function. DESIGN: A four-year prospective cohort study. SETTING: The communities of Iowa and Washington counties, Iowa. PARTICIPANTS: Eight hundred eighty-five women aged 70 and older who had completed physical performance tests in 1988 and 1992. MEASUREMENTS: Benzodiazepine use was determined during in-home interviews and classified by dose, duration, indication for use, and half-life. Physical performance tests included an assessment of standing balance, walking speed (8-foot distance), and repeated chair raises. RESULTS: Ninety (10.2%) reported benzodiazepine use at baseline. After adjustment for baseline physical performance score and potential confounders, benzodiazepine use was associated with a greater decline in physical performance over 4 years than nonuse (β = -1.16; standard error (SE) = 0.25; P < .001). The use of higher-than-recommended dose was related to decline (β = -2.26; SE = 0.47; P < .001), and use of lower doses was not (β = -0.53; SE = 0.46; P = .246). Long-term use (≥3 years) was related to decline (β = -1.65; SE = 0.34; P < .001), whereas recent and past use were not. Similar results were obtained when restricting the sample to those without disability at baseline. CONCLUSION: This study provides evidence that older women who used benzodiazepines were at risk for decline in physical performance. Subgroup analyses indicated that risk was greater with use of higher-than-recommended doses or for long duration (≥3 years). These findings highlight the importance of using benzodiazepines at the lowest effective dose for a limited duration in older women.