Taxonomically difficult groups are common in the Rosaceae and are often characterized by interspecific hybridization, polyploidy, agamospermy, vegetative reproduction, and variable morphology. In attempting to improve our understanding of systematically complex groups, especially those where many of the morphological characters lack discrete states, molecular data may be of great value. Even then, multiple independent data sets representing different genomes are usually required to establish a robust organismal phylogeny. Rubus is a classic example of such a group, exhibiting all of the above properties. In this analysis, a phylogeny of Rubus will be reconstructed using DNA sequences from five chloroplast (cp) regions (trnL intron, trnL-trnF intergenic spacer, matK, and its 5' and 3' flanking regions in the trnK intron) and two nuclear regions (ribosomal internal transcribed spacers: ITS and granule-bound starch synthase: GBSSI-1). In considering the relationships within Rubus, I will focus on species and clades where reticulate ancestry has been hypothesized. Taxa of particular interest include polyploid R. ursinus, three 6x species of subgenus Orobatus, and eight species of subg. Rubus (blackberries). ITS sequences support a hybrid origin of 12x R. ursinus involving species in distantly related clades, and cpDNA indicate that its maternal parent is likely a member of subg. Rubus. Using ITS data, subg. Orobatus is monophyletic and closely related to subg. Rubus and a clade of trans-Pacific species. However, cpDNA data place species of subg. Orobatus within the subg. Rubus lineage. Finally, hybridization or incomplete lineage sorting may explain persistent nucleotide polymorphism at phylogenetically informative ITS sites in two North American and six European species. Inclusion of data from all DNA regions noted above will provide further insight into the evolutionary history of Rubus.

Key words: allopolyploidization, hybridization, molecular systematics, multiple genes, Rosaceae, Rubus