Wide variations in the reproducibility rate of positive patch test responses have been reported. We hypothesized that a major source of non-reproducibility resides in the methodological pitfalls of routine patch testing. Simultaneous duplicate patch testing on opposite sides of the upper back was performed on 500 consecutive patients, using the TRUE Test system consisting of 2 panels, each one containing 12 standard allergens. A rigorous methodological design was applied and relevance was assessed for all discordant patch test reactions. A total of 435 positive patch test reactions were observed either on one or both sides in 289 patients (58.8%). Of these, 22 (5%) were discordant, i.e. interpreted as positive allergic on one side whilst negative or doubtful on the opposite side. The allergens responsible for discordant reactions were nickel sulfate (4 patients), cobalt chloride (3), lanolin alcohol (3), fragrance mix (2), carba mix (2), thiuram mix (2), colophonium (1), potassium dichromate (1), p-phenylenediamine (1), formaldehyde (1), balsam of Peru (Myroxylon pereirae resin) (1) and thimerosal (1). Of the 19 (4%) patients with discordant patch test reactions, the allergen was deemed to be of definite present or past relevance in 9 patients (1.8% of the total and 3.1% of all patients with positive patch test reactions) and of possible relevance in a further 2 patients. These data suggest that patch testing is a reasonably reproducible procedure as long as methodological error is minimized.