Arsenic (As) bioaccessibility in contaminated soils (n=50) was assessed using the simplified bioaccessibility extraction test (SBET). Soils used in the study were collected from sites where As was used as an herbicide (railway corridor) or pesticide (cattle dip sites), from former gold mines and from highly mineralised locations containing geogenic As sources (gossans). In all but three soils, As bioaccessibility was less than 50% indicating that a significant proportion of the total As concentration may not be available for absorption in the gastrointestinal tract following incidental soil ingestion. When regression models were developed based on soil properties, the descriptive variables best able to describe As bioaccessibility in railway corridor, dip site and mine site soils were total As and total or dithionite-citrate extractable (free) iron (Fe). While As bioaccessibility could be predicted (r(2)=0.955, n=50) in these contaminated soils, As bioaccessibility for gossan soils was a poor fit using linear or multivariate regression analysis.