We present a phylogeny of Poaceae subfamily Panicoideae based on analyses of three independent datasets, two chloroplast (ndhF, rpoC2) and one nuclear (phytochrome B). We will discuss both maximum parsimony and Bayesian analyses, with particular emphasis on their respective utility for large datasets. Panicoid grasses form a well-supported monophyletic group. The subfamily has traditionally been partitioned into two tribes. Our results unequivocally support a monophyletic Andropogoneae, but tribe Paniceae is clearly paraphyletic and is divided into two well-supported clades. One clade comprises x=10 taxa; it is sister to the Andropogoneae with which it shares the base chromosome number. The second clade contains taxa with a base chromosome number of x=9. There are three major subtypes of C4 photosynthesis that correspond to mediating enzyme. Most panicoids are of the NADP-ME subtype. Within the x=9 group, however, two well-supported clades comprise taxa using NAD-ME and PCK enzymes, indicating a single origin for each of these subtypes. Optimization of C4 origins is sensitive to taxon sampling. Although Andropogoneae are exclusively C4-NADP-ME, C3 taxa are interspersed throughout the remainder of the Panicoideae and some are basal in each clade of the Paniceae. Preliminary analyses suggest that a single origin with multiple reversals is slightly less parsimonious than multiple independent origins followed by a variable number of reversals.

Key words: Bayesian analysis, C4 photosynthesis, Panicoideae, Poaceae