The evolution of fruit characteristics in the Geum clade has inspired previous researchers to come up with several genera. Some are dispersed by wind, having plumose styles, and others by animals, for example with the aid of hooked styles. A preliminary morphological data set based mainly on previously used characters resulted in a reasonable but very weakly supported tree when analysed phylogenetically. Wind dispersed species came out basal which previous analyses of rbcL (Morgan et al., 1994) and ITS (Eriksson et al., in press) had indicated. We selected 19 species (covering Fallugia, Geum in strict sense, Waldsteinia, Coluria, Acomastylis, Erythrocoma, Sieversia, Novosieversia, and Oncostylus, along with the outgroups Filipendula, Rosa, Rubus, and Sanguisorba) and obtained sequences of the nuclear ribosomal ITS region. The aligned data (670 sequence characters plus 17 informative gaps scored for presence or absence) were analysed using parsimony in PAUP. There is some well supported resolution in the strict consensus of the 39 shortest trees. For instance, Fallugia and Sieversia are sister groups to rest of the species. The monophyly of this less inclusive group is well supported but the resolution is low within it, except for four species pairs. Unexpectedly, the species differ markedly in morphology within the pairs, and interpretations of fruit evolution seems unrealistic. Examination of chromosome numbers and ploidy levels reveal possible patterns of allopolyploid speciation with subsequent elimination of one of the parental species ITS copies through concerted evolution. To a large extent these conclusions support hypotheses formulated by Gajewski (1957) based on cytogenetic studies. It is clear that the ITS region may not be a suitable marker for the phylogeny within this group. Further studies taking into account sequences of other DNA data are in progress.

Key words: allopolyploidy, concerted evolution, fruit evolution, Geum, ITS, phylogeny