A comparative study on the in vitro and in vivo irritancy of anionic, amphoteric and non-ionic surfactants was performed. In vitro ED50 values of the surfactants were determined by two cytotoxicity assays, the dimethylthiazoldiphenyltetrazoliumbromide (MTT) assay and the neutral red release (NRR) assay on serum-free cultured human foreskin keratinocytes. In vivo human irritancy data were obtained by a 24 hour occlusive patch test in volunteers and the irritant skin response quantified by visual scoring, evaporimetry and colorimetry. A close relationship between the evaluation methods of the patch test was observed (r = 0.92 to r = 0.96), confirming that the 'bioengineering' methods, such as evaporimetry and colorimetry are suitable for measuring skin irritation. For six surfactants evaluated we found a good correlation (r = 0.91) between the ED50 values of the MTT assay and the in vivo irritancy data. The NRR assay yielded less satisfactory correlation coefficients with regard to MTT assay (r = 0.42) and in vivo irritancy data (r = 0.46). This can be mainly attributed to a misinterpretation of the amphoteric and non-ionic surfactants by the NRR assay. While the NRR assay may better evaluate the anionic surfactants, the MTT assay seems to be more suitable when testing a broader range of chemically diverse surfactants. Limitations of cell culture systems are noted, although the potential usefulness of cultured human skin cells for skin irritancy testing has been clearly demonstrated.