Ever since Linnaeus, systematists have been baffled by the seemingly intractable variability of Rosa (c. 150 spp.). Treatments have varied so much that as few as 30 or as many as 4266 species have been recognised. Early cytological studies suggested that this variation was largely due to hybridisation and unusual breeding systems; however, the wider impact of cytology on the classification of roses has been limited. This study was undertaken to examine relationships within Rosa, and to provide a maternal (chloroplast) phylogeny for future studies that will explore the origins of the numerous polyploids in Rosa. Results from trnL/trnL-F analyses using species from all four subgenera (Hulthemia, Rosa, Platyrhodon, Hesperhodos) and ten sections of subg. Rosa recognised by Rehder (1940) suggest that subgenera Hulthemia and Platyrhodon are nested within a clade dominated by subgenus Rosa sections. The position of subgenus Hesperhodos is ambiguous. Comparisons with phylogenies derived from complete matK Genbank sequences suggest that species from sections Rosa, Carolinae, Pimpinellifoliae p.p., and Synstylae p.p., and from sections Bracteatae, Indicae, Synstylae p.p., Caninae p.p., and Gallicanae p.p. might represent the core elements of the two largest clades within Rosa. The latter clade contains almost all of the approximately 15 species cited by Wylie (1954) as having made a major or minor genetic contribution to the formation of modern garden roses. Contrary to previous studies that have used much smaller taxonomic samples, matK and trnL/trnL-F analyses do not support the monophyly of any of Rehder's sections. However, trees are not well resolved, and few clades receive strong statistical support. Despite the largely congruent phylogenies provided by matK and trnL/trnL-F sequences, additional markers are currently being assayed to further resolve and strengthen our hypothesis.

Key words: molecular phylogeny, Rosa, Rosaceae, trnL-F