To clarify the sources of variation in irritant patch testing, variability in delivery of the test substance from the patch test system was studied. An in vitro model was used to study the penetration of sodium lauryl sulphate (SLS) from a patch test system into the skin. Different formulations of SLS applied to the skin for 24 h were studied (aqueous solution and gels), but irrespective of the vehicle used permeation of SLS into the recipient phase was poor. Results were compared to in vivo patch testing in 12 subjects. Approximately 70% of the applied SLS in aqueous soltion was released from the patch test system. Release from gels was poorer. High concordance between the in vivo and the in vitro model was found. No correlation was found between the amount of SLS left in the filter disc and the strentgh of the clinical reaction in vivo.