The most widely used method for establishing the minimal erythema dose (MED) before narrowband ultraviolet (UV) B phototherapy is time-consuming, inconvenient and may yield inconsistent results. To assess the equivalence of MED assessment using a filtered xenon arc lamp UV source, a semiautomated MED tester, and a UV-opaque template method of MED determination with a panel of TL-01 311-nm UVB fluorescent tubes. Secondly, to gauge the current usage of MED testing, and the method used, in a large sample of U.K. phototherapy units. Thirdly, to assess variation in UV output of the semiautomated skin tester immediately after switching on to identify optimum warm-up time. Finally, to assess reproducibility of MED testing by assessing within-patient variability and interobserver variability of MED test results. Sixty-five patients about to undergo UVB phototherapy had their MED determined using three different methods. Within each patient we compared the values of MED determined by a semiautomated Durham MED tester, a panel of narrowband UVB lamps with a UV-opaque template constructed by a phototherapist, and a 310-nm filtered xenon arc lamp with a liquid light guide. MED test results were assessed by clinical evaluation using a 6500 K colour temperature examination lamp. The output of the semiautomated MED tester was measured by spectroradiometer over a period of 20 min in order to identify the time to steady output. Reproducibility of MED testing with the semiautomated MED tester was carried out in 25 normal volunteers. All MEDs were assessed by at least two independent observers. A postal questionnaire was sent to 78 U.K. phototherapy units to assess routine practice concerning MED testing prior to narrowband UVB phototherapy. The semiautomated MED tester showed consistency with the panel method (r = 0.92, panel MED = -0.57 + 1.14 x Durham MED). The semiautomated MED tester produced a slightly lower MED result than the panel MED. Reproducibility tests showed high interobserver agreement (kappa value = 0.79), and high consistency for successive day testing (kappa value = 0.79). Questionnaires were returned from 67 of 78 phototherapy units (85%) and revealed that 19 units (28%) were routinely using MED testing prior to UVB therapy. This study has shown the Durham MED tester to be a convenient, valid and reproducible method for determining patient MED values prior to narrowband UVB phototherapy when used under carefully controlled lighting, by experienced observers.