A phylogenetic hypothesis for the genus Dendrochilum s.l. (including Bracisepalum) is based on nucleotide sequences of the nuclear ribosomal ITS and two plastid regions (matK, ycf1). The trees based on parsimony analysis of the nrITS and plastid partitions, respectively, are largely congruent, the only strongly supported conflict being the exact position of D. arachnites. Maximum likelihood analysis of the combined molecular matrix resulted in a tree with very high topological congruence to the parsimony consensus tree and to a fossil-calibrated Bayesian inference tree that was used for dating. Our results confirm that monophyly of Dendrochilum is only supported if Bracisepalum is included in the former. They also suggest that only two subgenera (Dendrochilum, Platyclinis) should be recognized, whereas all other subgenera and sections previously proposed on a morphological basis should be treated as synonyms of subg. Platyclinis. Indeed, mapping of five vegetative and nine floral characters of alleged systematic value, using DELTRAN and ACCTRAN optimization, demonstrates pronounced morphological homoplasy. Molecular time calibration dates the origin of the genus to the Miocene and contradicts the hypothesis that most extant Dendrochilum species did not evolve until the Holocene. Inferred phylogenetic relationships between extant species, in combination with previously identified hotspots of narrow endemism, suggest that species endemism in Dendrochilum is predominantly centric.