To quantify the within- and between-reader agreement of carotid B-mode ultrasonography and angiography interpretation, images from 117 patients examined by both modalities were read multiple times. Angiographic measurements were more reproducible than those of B-mode scans for all parameters except lesion width, but variations for B-mode scan measurements were similar to those for angiographic measurements. Within-reader agreement on the presence of ulceration was substantial for both modalities, whereas between-reader agreement was poor for B-mode scan and only moderate for angiography.