Objective. To test for differences in reported age at onset (AAO) of bipolar I affective disorder in clinical samples drawn from Europe and the USA. Methods. Admixture analysis was used to identify the model best fitting the observed AAO distributions of two large samples of bipolar I patients from Europe and USA (n = 3616 and n = 2275, respectively). Theoretical AAO functions were compared between the two samples. Results. The model best fitting the observed distribution of AAO in both samples was a mixture of three Gaussian distributions. The theoretical AAO functions of bipolar I disorder differed significantly between the European and USA populations, with further analyses indicating that (i) the proportion of patients belonging to the early-onset subgroup was higher in the USA sample (63 vs. 25%) and (ii) mean age at onset (±SD) in the early-onset subgroup was lower for the USA sample (14.5 ± 4.9 vs. 19 ± 2.7 years). Conclusions. The models best describing the reported AAO distributions of European and USA bipolar I patients were remarkably stable. The intermediate- and late-onset subgroups had similar characteristics in the two samples. However, the theoretical AAO function differed significantly between the USA and European samples due to the higher proportion of patients in the early-onset subgroup and the lower mean age-at-onset in the USA sample.