There has been increasing interest in a targeted approach to the screening and prevention of lead exposure in children. Targeted screening requires an understanding of variation in lead exposure in individual children or by region. In order to better understand variation by region, we studied Rhode Island lead poisoning screening data, examining average lead exposure to children living in 136 Providence County census tracts (CTs). The study population included 17,956 children aged 59 months and under, who were screened between May 1, 1992, and April 30, 1993. We evaluated the relationship between the percentage of children with blood lead > or = 10 micrograms/dL (pe10) and sociodemographic and housing characteristics, derived from United States 1990 Census data, of these CTs. CT descriptors included population density, percentage of households receiving public assistance income, median per capita income, percentage of households female headed, percentage of houses owner occupied, percentage of houses built before 1950, percentage of houses vacant, percentage of population Black, percentage of recent immigrants, and intraurban mobility. On average, 109 children were screened in each census tract; mean screening rate was 44%. There was wide variation in average lead exposure among census tracts, with pe10 ranging from 3 to 60% of screened children (mean 27%). Individual census variables explained between 24 and 67% of the variance in pe10 among CTs. A multiple regression model including percentage screened, percentage of households receiving public assistance, percentage of houses built before 1950, In (percentage of houses vacant), and percentage of recent immigrants explained 83% of variance in pe10. The percentage of houses built before 1950, a variable which models the presence of lead paint in old houses, displayed the largest adjusted effect on pe10 over the range observed for that variable in RI CTs. The percentage of houses vacant was also a highly significant and robust predictor; we suggest that vacancy is an ecological marker for the deterioration of leadbased paint, with higher vacancy neighborhoods containing houses in poorer condition. In Rhode Island, census tracts with high vacancy rates also have high rates of recent immigration, making immigrant groups vulnerable to lead exposure. Small-areas analysis may be useful in directing resources to high risk areas, explaining the sociocultural forces which produce such exposure and analyzing the effects of housing policy over time in states with high screening penetration.