We have tested the irritancy of 10 materials of various chemical composition in three in vitro toxicity assays: the Skintex Dermal Assay system (In Vitro International, Irvine, CA, USA), the silicon microphysiometer (Molecular Devices Inc., Menlo Park, CA), and the Living Skin Equivalent (Organogenesis Inc., Cambridge, MA). The purpose was to discover to what degree the in vitro results predict in vivo skin irritation as seen in nine female volunteers over the course of a 5 day cumulative irritancy patch test. Two in vivo assessments of irritancy were made and compared to the in vitro results, a visual scoring system, and a potentially more objective bioengineering assessment: the chromameter. Measures of sensitivity, specificity, positive predictive value, and negative predictive value were used to compare in vitro with in vivo results. Comparison of results using correlation coefficients was avoided since statistically significant rank orders could not be demonstrated for either in vitro or in vivo data. For these in vitro systems, sensitivity ranged from 60 to 100%, specificity from 80 to 100%, positive predictive value from 75 to 100%, and negative predictive value from 67 to 100%. The assay that exhibited the best measures of sensitivity, specificity, and other parameters was the Living Skin Equivalent. This assay is also the one that most closely simulates human skin. The nonionic surfactant triton X-100 gave a false-positive result in both the Living Skin Equivalent and the silicon microphysiometer. False- positive results with nonionic surfactants are consistent with observations made in other cell-based in vitro systems. We caution that the results not be over-interpreted to imply superiority for any one system.