The Atherosclerosis Risk in Communities (ARIC) study is a multicenter, long-term epidemiologic study that uses B-mode ultrasound primarily to measure carotid arterial intima-media wall thickness. To assess the reliability of the measurement process of this application of B-mode technology to population-based research, 36 volunteers from four centers were scanned at three visits, 7-14 days apart. Estimates of the components of variation in the B-mode measurements of artery wall thickness from between-person, between-sonographer, within-sonographer, between-reader and within-reader variation are presented, along with estimates of the correlation R between measures made at repeat visits by different sonographers and read by different readers. The estimates of R for mean intima-media thickness are 0.69, 0.60, 0.54 and 0.66 for the carotid bifurcation, internal carotid, common carotid, and three-segment overall carotid mean, respectively. When these are adjusted to account for differences in between-person variance between the ancillary study and the main ARIC study, the estimated site-specific reliability coefficients appropriate to the ARIC study population are 0.77, 0.73 and 0.70 for mean carotid far-wall intima-media thickness at the carotid bifurcation and the internal and common carotid arteries.