Abstract We reanalysed Yang & Patton's allozyme data, published in Aukin 1981, of Darwin's finches with a variety of distance and cladistic methods to estimate the phylogeny of the group. Different methods yielded different results, nevertheless there was widespread agreement among the distance methods on several groupings. First, the two species of Camarhynchusgrouped near one another, but not always as a monophyletic group. Second, Cactospiza pallidaand Platyspiza crassirostrisformed a monophyletic group. Finally, all the methods (including parsimony) supported the monophyly of the ground finches. The three distance methods also found close relationships generally between each of two populations of Geospiza scandens, G. difficilisand G. conirostris. There is evidence for inconstancy of evolutionary rates among species. Results from distance methods allowing for rate variation among lineages suggest three conclusions which differ from Yang and Patton's findings. First, the monophyletic ground finches arose from the paraphyletic tree finches. Yang and Patton found that the ground finches and tree finches were sister monophyletic taxa. Second, Geospiza scandensappears to be a recently derived species, and not the most basal ground finch. Third, G. fuliginosais not a recently derived species of ground finch, but was derived from an older split from the remaining ground finches. Most of these conclusions should be considered tentative both because the parsimony trees disagreed sharply with the distance trees and because no clades were strongly supported by the results of bootstrapping and statistical tests of alternative hypotheses. Absence of strong support for clades was probably due to insufficient data. Future phylogenetic studies, preferably using DNA sequence data from several unlinked loci, should sample several populations of each species, and should attempt to assess the importance of hybridization in species phylogeny.