To evaluate the capabilities of in vitro assays to predict arsenic (As) relative bioavailability (RBA), we examined the relationship between As bioaccessibility, determined using a number of in vitro bioaccessibility (IVBA) methodologies (SBRC, IVG, PBET, DIN and UBM) and As RBA determined in a mouse assay for nine As-contaminated soils and 1 NIST reference material (2710a). Significant differences (P < 0.05) in As IVBA were observed within and between assays indicating that different IVBA methodologies may not produce congruent data, as a result of variability in the extracting medium constituents and/or differences in the pH of gastric and intestinal phases. When results of in vivo determinations of As RBA were compared with As IVBA results, there was no significant difference in slopes of the relationships (P = 0.49-0.88) when SBRC, IVG, PBET, DIN, and UBM gastric and intestinal phase data were used. A significantly (P < 0.05) smaller y-intercept was, however, determined for the in vivo-SBRC gastric phase correlation compared to SBRC, IVG, PBET, and DIN intestinal phase, a factor that may influence prediction of As RBA, especially for soils with low As RBA. When in vivo-in vitro relationships were compared to previously derived correlations from the literature, some differences were observed. These differences may be attributed to factors affecting both in vivo and in vitro data including physiological differences in animal models (e.g., mouse versus swine), which may influence As absorption, differences in the approach used to estimate As RBA, and variability arising from subtle interoperator differences in performance of in vitro assays.