Introduction: The relationship between serum galectin-3 and incident cognitive impairment was analyzed in the Reasons for Geographic and Racial Differences in Stroke study. Methods: Baseline galectin-3 was measured in 455 cases of incident cognitive impairment and 546 controls. Galectin-3 was divided into quartiles based on the weighted distribution in the control group, and the first quartile was the referent. Results: There was an increasing odds of cognitive impairment across quartiles of galectin-3 (odds ratios, 1.00 [0.68–1.46], 1.45 [1.01–2.10], and 1.58 [1.10–2.27] relative to the quartile 1; P trend =.003) in an unadjusted model, which persisted after adjusting for age, sex, and race (P =.004). Adjustment for cardiovascular risk factors greatly attenuated this association (odds ratios, 0.97 [0.60–1.57], 1.52 [0.94–2.46], and 1.27 [0.76–2.12]; P =.15). The association differed by diabetes status (P interaction,.007). Among nondiabetics (293 cases, 411 controls), those with galectin-3 in the fourth compared with first quartile had an odds ratio of 1.6 (0.95–2.99; P trend,.02). In diabetics, the odds ratio was 0.23 (0.04–1.33). Discussion: Serum galectin-3 was associated with increased risk of incident cognitive impairment in a large cohort study of blacks and whites but only in nondiabetics.