The reproducibility in the identification of carotid artery lesions using B-mode ultrasound was studied in a large random sample selected from the Atherosclerosis Risk in Communities (ARIC) Study. Carotid lesions were defined as plaque with or without acoustic shadowing (indicative of mineralization). A weighted kappa (kappa w) statistic was used as a chance-adjusted measure of repeatability. In the ARIC baseline survey, the kappa w values for the assessment of lesions on repeat reading were 0.47, 0.60 and 0.69 in the left common carotid artery, the carotid bifurcation and the internal carotid artery, respectively. In a repeat scanning, the kappa w values ranged from 0.59 to 0.79 in the left carotid segments. The results were similar in the left and right carotid arteries. Covariates (age, race, gender, body mass index, study center) did not influence the reproducibility. Similar results were also found in both the baseline survey and the first follow-up examination. In conclusion, reproducibility in the assessment of carotid lesions by B-mode ultrasound can be achieved in multicenter studies at fair to good levels of agreement.