We attempted to adapt an RIA method for cortisol in plasma to the analysis of “free cortisol” in urine. 3H-cortisol could be effectively extracted from urine with methylene chloride, but this solvent also extracted substances other than cortisol which were measured as “cortisol” by RIA. Urine samples which had normal cortisol measured with a competitive protein-binding assay had raised values by RIA. In an attempt to obtain the “correct” urine cortisol values for the samples, we analysed the urines using 5 commerical kits. Inspection of these results confirmed that the in-house RIA method was giving falsely high results for normal subjects, but it also showed that one kit, which was available in Canada until recently, gave spuriously high results with some urines due to interference by dexamethasone which the patients had received as part of a suppression test. We suggest that kit manufacturers should thoroughly evaluate their products prior to release, and that the evaluation should include measurements on a reasonably large and varied patient population to determine appropriate reference ranges and the response to stimulation or suppression tests.