Phylogenetic relationships among Labroid families, based on whole mitochondrial DNA sequences. Shown is the 50% majority rule consensus tree of the 4,700 pooled trees from two independent Bayesian analyses for dataset #2 (3rd codons RY-coded). The dataset comprises unambiguously aligned nucleotide sequences of 13,393 bp from 76 percomorphs and two outgroups; we set five partitions (1st, 2nd and 3rd codon positions from 12 protein-coding genes plus tRNA and rRNA genes). Partitioned Bayesian analyses were conducted using MRBAYES 3.1.2 , with the best-fit model of sequence evolution [ref. 54; GTR + I + Γ] being set each partition and all model parameters variable and unlinked across partitions. Numerals beside internal branches indicate Bayesian posterior probabilities (PPs) (shown as percentages) for dataset #1/#2/#3. Single numerals are given when analyses for all the datasets have shown the same values, and clades denoted by broad lines indicate those supported by 100% PPs in the all datasets. An unresolved trichotomy is indicated by an arrow, and topological incongruities among the datasets are denoted by open arrowheads (dataset #1 vs. #2) and filled arrowheads (dataset #1 vs. #3). Note that the species of the Labridae and those of the remaining three labroid families (Cichlidae, Pomacentridae, and Embiotocidae) form different monophyletic groups, respectively.