The use of chemical information in phylogenetic studies is an under-used resource for studying evolutionary relationships. The absence of chemical data in phylogenetic inference is likely due in part to difficulties associated with coding chemical data as characters. Using floral and vegetative fragrance data from twenty species in three southwestern genera of Nyctaginaceae, we have explored several methods for handling fragrance data. Chemical data collected from several individuals within species were summarized to yield a single value per species and optimized onto a phylogeny inferred independently using molecular sequence data. Fragrance data were also coded in a variety of ways including (a) presence/absence of each compound, (b) presence/absence characters according to the biosynthetic pathway by which each compound was synthesized, and (c) multistate characters coded according to amount of floral compounds from each biosynthetic pathway. Phylogenies were inferred using these data sets, and the resultant topologies were compared to the molecular phylogeny. In addition to using fragrance data in phylogeny reconstruction, we also experimented with optimizing these fragrance data sets onto the molecular phylogeny. It appears that there is no one clear approach to handling fragrance data that retains the most phylogenetic information. Rather, the "best" coding method may differ among characters; treating biosynthetic pathways as characters may sometimes be most phylogenetically informative, whereas for other fragrance components coding specific compounds as characters might be most appropriate. Further, despite much homoplasy, results suggest that fragrance data do contain phylogenetic information.

Key words: Acleisanthes, Mirabilis, Nyctaginaceae, phylogenetics, scent, Selinocarpus