Introduction: The excess stroke mortality among African Americans and Southerners is well known. Because a higher proportion of the population living in the 'Stroke Belt' is African American, then a portion of the estimated excess risk of stroke death traditionally associated with African-American race may be attributable to geography (i.e., race and geography are 'confounded'). In this paper we estimate the proportion of the excess stroke mortality among African Americans that is attributable to geography. Methods: The numbers of stroke deaths at the county level are available from the vital statistics system of the US. A total of 1,143 counties with a population of at least 500 whites and 500 African Americans were selected for these analyses. The black-to-white stroke mortality ratio was estimated with and without adjustment for county of residence for those aged 45-64 and for those aged 65 and over. The difference in the stroke mortality ratio before versus after adjustment for county provides an estimate of the proportion of the excess stroke mortality inappropriately attributed to race (that is in fact attributable to geographic region). Results: For ages 45-64, the black-to-white stroke mortality ratio was reduced from 3.41 to 3.04 for men, and from 2.82 to 2.60 for women, suggesting that between 10 and 15% of the excess mortality traditionally attributed to race is rather due to geography. Over the age of 65, the black-to-white stroke mortality ratio was reduced from 1.31 to 1.27 for men, and from 1.097 to 1.095 for women, suggesting that between 2 and 13% of the excess mortality attributed to black race is actually attributable to geography. The reductions of all the four age strata gender groups were highly significant. Discussion: These results suggest that a significant, although relatively small, proportion of the excess mortality traditionally attributed to race is rather a factor of geography. Copyright © 2004 S. Karger AG, Basel.