Are Third Positions Really That Bad? A Test Using Vertebrate Cytochrome b

In the analysis of coding DNA sequences for phylogenetic purposes 3rd positions are commonly downweighted, or excluded, and 2nd positions are likewise given high weight. The basis of this is a belief that the larger numbers of 3rd position substitutions imply a larger amount of misleading information than the slowly evolving 2nd positions. This paper tests this assertion by splitting 10 vertebrate data sets into three parts consisting of 1st, 2nd, and 3rd positions only. The relative performance of these data sets was tested by a number of suggested methods for assessing the phylogenetic content in a data set. Contrary to the prediction, the relative performance was 3rd > 1st > 2nd. This implies that a downweighting of 3rd positions in favour of 2nd positions is not supported by the analyses of their phylogenetic content. There are several reasons to believe that the overall amount of change is a poor predictor of the level of homoplasy and these are discussed. The overall conclusion is that unless there is solid evidence that 3rd positions really are misleading, they should not be treated so by default.