Inconsistency of the Maximum-parsimony Method: the Case of Five Taxa With a Molecular Clock

The inconsistency of the maximum-parsimony method for the case of five taxa with a molecular clock was studied using an analytical approach and Monte Carlo simulation. The inconsistency occurs in the case of a symmetrical tree with short internal branches and long external branches but can be avoided by using slowly evolving sequences. The neighbor-joining method is consistent if the evolutionary distances between taxa are estimated accurately. However, for short sequences, estimates of distances are subject to large statistical errors, and the probability of obtaining a certain erroneous tree by this method can be larger than the probability of obtaining the true tree. The implications for phylogenetic reconstruction are discussed.