Schematic presentation of the protocol used to simulate heterotachous alignments. Sequences were generated similarly as in ref.  under two different sets of branch lengths of equal weight (w = 0.5). In ref. , the branch lengths were altered by swapping the values of p and q (a). In our case (b), a single parameter (τ) allows to adjust the level of heterotachy from fully homotachous (τ = 0) to extreme heterotachous (τ = 1) conditions, while keeping the averaged branch length constant. Our branch lengths are (1 + τ) p and (1 - τ) q for the first partition and (1 - τ) p and (1 + τ) q for the second partition. 100 replicates of 5,000 nucleotide positions were simulated for each partition assuming a uniform JC69 model  using SeqGen  and were concatenated before phylogenetic inference using PAUP* .