MrBayes 3.1.2 (Huelsenbeck and Ronquist, 2001; Ronquist
and Huelsenbeck, 2003) was used to infer a phylogenetic hy-
pothesis using a Bayesian approach. Two different ways to
partition the data were tried: one where all the data were allocated
to one partition, and onewhere the trnL-F and the non-coding part
ofmatKwas treated as a different partition fromthe coding part of
matK. For each partition, the optimal model of sequence evolu-
tion was calculated on a randomly chosen MPT from the set of
obtained trees from parsimony analysis of the entire data set,
using the Akaike Information Criterion (AIC) criterion as im-
plemented in Modeltest 3.7 (Posada and Crandall, 1998). Four
chainswere run for 5million generations and sampled every 1000
generations. This was repeated twice, and these independent runs
were compared to make sure that similar estimates of substitution
model parameters, topology and branch lengths were obtained.
The number of “burn-in” generations needed before the various
log-likelihood values reached stationarity was determined by a
graphical plot, and the first 1000 sampled generations were dis-
carded. Swapping among chains and acceptance of proposed
changes to model parameters were monitored to ensure that ef-
ficient mixing had occurred. Results fromthe run with the highest
harmonic mean of the −ln likelihood are reported here.