OBJECTIVE: Prior studies indicate that olfactory function may be an early marker for cognitive impairment, but the body of evidence has been largely restricted to white populations. METHODS: We studied 2,428 community-dwelling black and white older adults (baseline age 70-79 years) without dementia enrolled in the Health, Aging, and Body Composition (Health ABC) study. Olfaction was measured as odor identification (OI) with the 12-item Cross Cultural Smell Identification Test in year 3. We defined incident dementia over 12 years on the basis of hospitalization records, prescription for dementia medication, or 1.5-SD decline in race-stratified global cognition score. We assessed dementia risk associated with OI score (by tertile) using Cox proportional hazards models. All analyses were stratified by race. RESULTS: Poorer OI in older adults without dementia was associated with increased risk of dementia. After adjustment for demographics, medical comorbidities, and lifestyle characteristics, white participants in the poor or moderate OI tertile had greater risk of dementia (adjusted hazard ratio [HR] 3.34, 95% confidence interval [CI] 2.45-4.54; and HR 1.84, 95% CI 1.33-2.54, respectively) compared to those in the good tertile of function. Among blacks, worse OI was associated with an increased risk of dementia, but the magnitude of the effect was weaker (p for interaction = 0.04) for the poor OI tertile (adjusted HR 2.03, 95% CI 1.44-2.84) and for the moderate tertile (adjusted HR 1.42, 95% CI 0.97-2.10). There was no interaction between OI and APOE ε4 and risk of dementia. CONCLUSIONS: While the magnitude of the association was stronger in whites, we found that poor OI was associated with increased risk of dementia among both black and white older adults.