Long-branch Attraction

In inference of phylogenetic trees from molecular data, long branches — branches on which a large amount of evolution has occurred — can potentially be joined as sister taxa, even when a sister-taxon relationship is not the true relationship. The simplest case of long-branch attraction has four taxa, \(A\), \(B\), \(C\), and \(D\), with sister-taxon pairs \( (AB) \) and \( (CD) \) in the true tree, \(A\) and \(C\) representing long branches, and \(B\) and \(D\) representing short branches. Because little evolution has occurred on branches \(B\) and \(D\) from the time of the common ancestor, these branches can be joined by an inference method. Because much evolution has occurred on branches \(A\) and \(C\), recurrent mutations can make these taxa spuriously appear to more similar than they are (Felsenstein 1978). The result is an inferred tree \( ((AC),(BD)) \) with the long branches paired and the short branches paired — as in this Italian stone pine — rather than the correct \( ((AB),(CD)) \).