In this study, we evaluated the reproducibility of analysis of basal body temperature graphs under optimized conditions for agreement. A total of 160 recordings were selected from spontaneous cycles of infertile women and analysed by three experienced clinicians using uniform criteria. Agreement at the various stages of analysis was assessed by the 'proportions of agreement' with 95% confidence intervals. Agreement in identification of the thermal nadir was clearly superior to that reported in previous publications. Reproducibility of 'ovulatory' graph features (i.e. biphasic graphs and adequate thermal shifts) was excellent. Agreement in classification of monophasic graphs and inadequate thermal shifts, although lower, was still good. Thus, with experienced observers and uniform criteria, a good agreement can be achieved in analysis of the most important parameters of the basal body temperature graph. We believe that an effort should be made to generalize the use of uniform analysis criteria, because only then can results from different institutions be compared and the remaining clinical evaluation of the method be performed.