Phylogenetic relationships in the genus Nicotiana were investigated using parsimony analyses of the internal transcribed spacer (ITS) regions of nuclear ribosomal DNA (nrDNA). In addition, origins of some amphidiploid taxa in Nicotiana were investigated using the techniques of genomic in situ hybridization (GISH), and the results of both sets of analyses were used to evaluate previous hypotheses about the origins of these taxa. Phylogenetic analyses of the ITS nrDNA data were performed on the entire genus (66 of 77 naturally occurring species, plus three artificial hybrids), comprising both diploid and polyploid taxa, and on the diploid taxa only (35 species) to examine the effects of amphidiploids on estimates of relationships. All taxa, regardless of ploidy, produced clean, single copies of the ITS region, even though some taxa are hybrids. Results are compared with a published plastid (matK) phylogeny using fewer, but many of the same, taxa. The patterns of relationships in Nicotiana, as seen in both analyses, are largely congruent with each other and previous evolutionary ideas based on morphology and cytology, but some important differences are apparent. None of the currently recognized subgenera of Nicotiana is monophyletic and, although most of the currently recognized sections are coherent, others are clearly polyphyletic. Relying solely upon ITS nrDNA analysis to reveal phylogenetic patterns in a complex genus such as Nicotiana is insufficient, and it is clear that conventional analysis of single data sets, such as ITS, is likely to be misleading in at least some respects about evolutionary history. ITS sequences of natural and well-documented amphidiploids are similar or identical to one of their two parents-usually, but not always, the maternal parent-and are not in any sense themselves 'hybrid'. Knowing how ITS evolves in artificial amphidiploids gives insight into what ITS analysis might reveal about naturally occurring amphidiploids of unknown origin, and it is in this perspective that analysis of ITS sequences is highly informative.