We construct a species-level phylogeny for the Pentaschistis clade based on chloroplast DNA, from the following regions: trnL-F, trnT-L, atpB-rbcL, rpL16, and trnD-psbA. The clade comprises 82 species in three genera, Pentaschistis, Pentameris, and Prionanthium.We demonstrate that Prionanthium is nested in Pentaschistis and that this clade is sister to a clade of Pentameris plus Pentaschistis tysonii. Forty-three of the species in the Pentaschistis clade have multicellular glands and we use ancestral character state reconstruction to show that they have been gained twice or possibly once, and lost several times. We suggest that the maintenance, absence, loss, and gain of glands are correlated with leaf anatomy type, and additionally that there is a difference in the degree of diversification of lineages that have these different character combinations. We propose that both glands and sclerophyllous leaves act as defense systems against herbivory, and build a cost/benefit model in which multicellular glands or sclerophyllous leaves are lost when the alternative defense system evolves. We also investigate the association between leaf anatomy type and soil nutrient type on which species grow. There is little phylogenetic constraint in soil nutrient type on members of the Pentaschistis clade, with numerous transitions between oligotrophic and eutrophic soils. However, only orthophyllous-leaved species diversify on eutrophic soils. We suggest that the presence of these glands enables the persistence of orthophyllous lineages and therefore diversification of the Pentaschistis clade on eutrophic as well as oligotrophic soils.KEY WORDS: Ancestral, correlated evolution, Danthonioideae, gland evolution, Pentaschistis, phylogeny, Poaceae.Interest in the evolution of complex structures and their influence upon the success of organisms has a long history. Classic examples include orchid flowers (Darwin 1862) and the evolution of eyes (Fernald 2004). Part of their fascination lies in the very low probability that such structures should evolve at all (Dawkins 1986). However, their (often) unique evolution makes it difficult to tease apart the adaptive aspects from other factors, including chance. In contrast, the loss of such complex structures presents us with an opportunity to explore their function. Losses tend to be more numerous than gains in many complex structures or traits (e.g., heterostyly [Kohn et al. 1996;Schoen et al. 1997]; secondary xylem in aquatic plants [Sculthorpe 1967]; wings in insects [Whiting et al. 2003]; or the chlorophyll producing function of chloroplasts, in holoparasites [Judd et al. 2002;APG 2003;Bungard 2004]; Nickrent et al. 2004]). This makes them more tractable to investigation than gains. For example, if the losses of a particular structure within a clade are nonrandom with respect to other morphological characters or habitat, these losses can be used to investigate Peculiar multicellular glands (see Fig. 1A, B) exist on the inflorescences and/or leaves in 43 species of Pentaschistis and of P...