Bayesian analysis of phytochrome data from seed plants strongly supports placing Gnetales with Pinaceae when searches assume a general reversible model and allow rates to vary across sites. However, flat-weighted parsimony analysis places Gnetales outside of well-supported clades uniting cycads, ginkgo, and conifers in three gene subtrees. Weighted parsimony analyses in which sites were weighted according to their evolutionary rates suggest that the most conservative and the fastest evolving sites favor placing Gnetales as the sister to extant gymnosperms while sites evolving at intermediate rates support a sister group relationship between Gnetales and Pinaceae. Cycads, ginkgo, and conifers have three phytochrome genes, PHYP, PHYN, and PHYO, but Gnetales have only two genes. One Gnetalean phytochrome gene is related to PHYP, or to PHYB or E of angiosperms, and one gene is related to PHYN or O, or to PHYA or C of angiosperms. In most parsimonious trees from flat-weighted analyses, the PHYP/B/E-related gene is sister to PHYP of cycads, ginkgo, and conifers but the position of the PHYA/C/N/O-related gene is unresolved. Templeton, winning sites, and Kishino-Hasegawa tests to compare trees from constraint analyses that enforced various hypotheses of seed plant relationships reject placing the PHYA/C/N/O-related lineage as sister either to Pinaceae or conifers. However, they do not reject an anthophyte hypothesis in which the unplaced Gnetalean phytochrome lineage is sister to angiosperm PHYA, and which forces the PHYP/B/E-related gene into a clade with angiosperms. Tests for bias in the data may reveal the source of the conflict between parsimony and Bayesian trees and between these parsimony and other molecular trees.

Key words: Bayesian analyses, gene duplication, parsimony, phytochromes, seed plants