NOTE 13E: We have assumed that all codons are equally likely, which is not at all the case. The fraction of synonymous mutations will vary with codon usage; see p. 543. Programs that estimate phylogenetic relationships from coding sequence must take such biases into account.