We linked risk estimates from the U.S. Environmental Protection Agency's National Air Toxics Assessment (NATA) to racial and socioeconomic characteristics of census tracts in Maryland (2000 Census) to evaluate disparities in estimated cancer risk from exposure to air toxics by emission source category. In Maryland, the average estimated cancer risk across census tracts was highest from on-road sources (50% of total risk from nonbackground sources), followed by nonroad (25%), area (23%), and major sources (< 1%). Census tracts in the highest quartile defined by the fraction of African-American residents were three times more likely to be high risk (> 90th percentile of risk) than those in the lowest quartile (95% confidence interval, 2.0-5.0). Conversely, risk decreased as the proportion of whites increased (p < 0.001). Census tracts in the lowest quartile of socioeconomic position, as measured by various indicators, were 10-100 times more likely to be high risk than those in the highest quartile. We observed substantial risk disparities for on-road, area, and nonroad sources by socioeconomic measure and on-road and area sources by race. There was considerably less evidence of risk disparities from major source emissions. We found a statistically significant interaction between race and income, suggesting a stronger relationship between race and risk at lower incomes. This research demonstrates the utility of NATA for assessing regional environmental justice, identifies an environmental justice concern in Maryland, and suggests that on-road sources may be appropriate targets for policies intended to reduce the disproportionate environmental health burden among economically disadvantaged and minority populations.