Incidental soil ingestion is a common contaminant exposure pathway for humans, notably children. It is widely accepted that the inclusion of total soil metal concentrations greatly overestimates the risk through soil ingestion for people due to contaminant bioavailability constraints. The assumption also assumes that the contaminant distribution and the bioaccessible fraction is consistent across all particle sizes. In this study, we investigated the distribution of arsenic across five particle size fractions as well as arsenic bioaccessibility in the <250-, <100-, <10- and 2.5-microm soil particle fractions in 50 contaminated soils. The distribution of arsenic was generally uniform across the larger particle size fractions but increased markedly in the <2.5-microm soil particle fraction. The marked increase in arsenic concentration in the <2.5-microm fraction was associated with a marked increase in the iron content. Arsenic bioaccessibility, in contrast, increased with decreasing particle size. The mean arsenic bioaccessibility increased from 25 +/- 16% in the <250-microm soil particle fraction to 42 +/- 23% in the <10-microm soil particle fraction. These results indicate that the assumption of static arsenic bioaccessibility values across particle size fractions should be reconsidered if the ingested material is enriched with small particle fractions such as those found in household dust.