Iridaceae are one of the largest families of Lilianae and probably also amo
ng the best studied of monocotyledons. To further evaluate generic, tribal,
and subfamilial relationships we have produced four plastid DNA data sets
for 57 genera of Iridaceae plus outgroups: rsp4, rbcL (both protein-coding
genes), the trnL intron, and the trnL-F intergenic spacer. All four matrice
s produce similar although not identical trees, and we thus analyzed them i
n a combined analysis, which produced a highly resolved and well-supported
topology, in spite of the fact that the partition homogeneity test indicate
d strong incongruence. In each of the individual trees, some genera or grou
ps of genera are misplaced relative to morphological cladistic studies, but
the combined analysis produced a pattern much more similar to these previo
us ideas of relationships. In the combined tree, all subfamilies were resol
ved as monophyletic, except Nivenioideae that formed a grade in which Ixioi
deae were embedded. Achlorophyllous Geosiris (sometimes referred to Geosiri
daceae or Burmanniaceae) fell within the nivenioid grade. Most of the tribe
s were monophyletic, and Isophysis (Tasmanian) was sister to the rest of th
e family; Diplarrhena (Australian) fell in a well-supported position a, sis
ter to Irideae/Sisyrinchieae/Tigridieae/Mariceae (i.e., Iridoideae); Bobart
ia of Sisyrinchieae is supported as a member of Irideae. The paraphyly of N
ivenioideae is Suspicious due to extremely high levels of sequence divergen
ce, and when they were constrained to be monophyletic the resulting trees w
ere only slightly less parsimonious (<1.0%). However, this subfamily also l
acks clear morphological synapomorphies and is highly heterogeneous, so it
is difficult to develop a strong case on nonmolecular grounds for their mon
ophyly.