BACKGROUND: Floral morphology, particularly the angle of lip attachment to the column, has historically been the fundamental character used in establishing generic limits in subtribe Oncidiinae (Orchidaceae), but it has also been long recognized that reliance on this character alone has produced a highly artificial set of genera. In essence, lip/column relationships reflect syndromes associated with pollinator preferences; most genera of Oncidiinae as previously defined have consisted of a single floral type. Here, the degree to which this has influenced generic delimitation in Brazilian members of the largest genus of Oncidiinae, Oncidium, which previous molecular (DNA) studies have demonstrated to be polyphyletic, is evaluated. METHODS: Phylogenetic analyses of the following multiple DNA regions were used: the plastid psbA-trnH intergenic spacer, matK exon and two regions of ycf1 exon and nuclear ribosomal DNA, comprised of the two internal transcribed spacers, ITS1 and ITS2, and the 5·8S gene. Results from all regions analysed separately indicated highly similar relationships, so a combined matrix was analysed. KEY RESULTS: Nearly all species groups of Brazilian Oncidium are only distantly related to the type species of the genus, O. altissimum, from the Caribbean. There are two exceptions to this geographical rule: O. baueri is related to the type group and O. orthostates, an isolated species that lacks the defining tabula infrastigmata of Oncidium, is not exclusively related to any previously described genus in the subtribe. Several well-supported subclades can be observed in these results, but they do not correspond well to sections of Oncidium as previously circumscribed or to segregate genera as defined by several recent authors. In spite of their floral differences, these groups of Oncidium, formerly treated as O. sections Barbata, Concoloria pro parte, Crispa, Ranifera, Rhinocerotes, Rostrata (only O. venustum), Synsepala, Verrucituberculata pro parte and Waluewa, form a well-supported clade with Gomesa (including Rodrigueziella and Rodrigueziopsis) embedded in it. Two often recognized segregate genera, Baptistonia and Ornithophora, and the recently described Carriella are also embedded within the Brazilian clade. The level of variation within major subclades of the Gomesa clade is low and similar to that observed within other genera of Oncidiinae. CONCLUSIONS: Convergence on a stereotypical syndrome of floral traits associated with pollination by oil-collecting bees has resulted in these characters not being reliable for producing monophyletic taxa, and the genus Oncidium, defined by these characters, is grossly polyphyletic. Vegetative and a few floral/inflorescence characters link these taxa with a mainly Brazilian distribution, and they were all transferred to Gomesa on this basis rather than separated from Gomesa based on their floral differences, which we hypothesize to be simple shifts in pollination strategies. Other authors have described a large number of new genera for these former members of Oncidium, but most of these are not supported by the results presented here (i.e. they are not monophyletic). A new genus, Nohawilliamsia, is described for O. orthostates because it does not fit in any currently recognized genus and is only distantly related to any other member of Oncidiinae.