Abstract. Soil associated with Northampton Sand within the area of the Borough Council of Wellingborough has elevated levels of arsenic. Arsenic in such soil at a private dwelling was analysed to determine the arsenic oral bioavailability and uptake by fruit and vegetables grown at the site. This information was used to assess the risk to a present occupant with 22 years' exposure, and to a 0–6-year-old girl, the critical receptor in the standard residential land use as stated in UK policy. The exposure pathways comprised soil and dust ingestion, consumption of home-grown vegetables and ingestion of soil attached to home-grown fruit and vegetables. The soil guideline value (SGV) for arsenic in residential land use with plant uptake derived using the Contaminated Land Exposure Assessment (CLEA) model was deemed inappropriate for use at this site, since the SGV does not include the consumption of home-grown fruit, potentially a major pathway in this site for the current occupant, and also assumes 100% arsenic bioavailability. Hence, it was necessary to derive site-specific assessment criteria for arsenic, for which we used the first edition of the Scotland and Northern Ireland Forum for Environmental Research (SNIFFER) framework, modified to reflect guidance in Contaminated Land Research (CLR) Reports CLR 9 and 10, and to incorporate the consumption of home-grown fruit. The incorporation of site-specific data on arsenic bioavailability and vegetable: soil arsenic concentration factor for the 0–6-year-old receptor gave a site-specific assessment criterion (SSAC) of 139 mg kg−1 soil. For the adult female receptor, incorporating the site-specific data on arsenic bioavailability, the vegetable: soil and the fruit: soil concentration factors gave a SSAC of 251 mg kg−1 soil. In both cases, the vegetable consumption pathway drives the risk. None of the 19 soil sample results for arsenic exceeded the site-specific assessment criteria derived for the current occupiers, but 13 of the soil samples slightly exceeded the site-specific assessment criteria for the putative 0–6-year-old girl. Given the degree of conservatism in the selection of input values for the exposure parameters in the SNIFFER framework, this exceedance was not considered to warrant further regulatory action at the time.