Recent discoveries of new fossil hominid species have been accompanied by several phylogenetic hypotheses. All of these hypotheses are based on a consideration of hominid craniodental morphology. However, Collard and Wood (2000) suggested that cladograms derived from craniodental data are inconsistent with the prevailing hypothesis of ape phylogeny based on molecular data. The implication of their study is that craniodental characters are unreliable indicators of phylogeny in hominoids and fossil hominids but, notably, their analysis did not include extinct species. We report here on a cladistic analysis designed to test whether the inclusion of fossil taxa affects the ability of morphological characters to recover the molecular ape phylogeny. In the process of doing so, the study tests both Collard and Wood's (2000) hypothesis of character reliability, and the several recently proposed hypotheses of early hominid phylogeny. One hundred and ninety-eight craniodental characters were examined, including 109 traits that traditionally have been of interest in prior studies of hominoid and early hominid phylogeny, and 89 craniometric traits that represent size-corrected linear dimensions measured between standard cranial landmarks. The characters were partitioned into two data sets. One set contained all of the characters, and the other omitted the craniometric characters. Six parsimony analyses were performed; each data set was analyzed three times, once using an ingroup that consisted only of extant hominoids, a second time using an ingroup of extant hominoids and extinct early hominids, and a third time excluding Kenyanthropus platyops. Results suggest that the inclusion of fossil taxa can play a significant role in phylogenetic analysis. Analyses that examined only extant taxa produced most parsimonious cladograms that were inconsistent with the ape molecular tree. In contrast, analyses that included fossil hominids were consistent with that tree. This consistency refutes the basis for the hypothesis that craniodental characters are unreliable for reconstructing phylogenetic relationships. Regarding early hominids, the relationships of Sahelanthropus tchadensis and Ardipithecus ramidus were relatively unstable. However, there is tentative support for the hypotheses that S. tchadensis is the sister taxon of all other hominids. There is support for the hypothesis that A. anamensis is the sister taxon of all hominids except S. tchadensis and Ar. ramidus. There is no compelling support for the hypothesis that Kenyanthropus platyops shares especially close affinities with Homo rudolfensis. Rather, K. platyops is nested within the Homo + Paranthropus + Australopithecus africanus clade. If K. platyops is a valid species, these relationships suggest that Homo and Paranthropus are likely to have diverged from other hominids much earlier than previously supposed. There is no support for the hypothesis that A. garhi is either the sister taxon or direct ancestor of the genus Homo. Phylogenetic relationships indicate that Australopithecus is paraphyletic. Thus, A. anamensis and A. garhi should be allocated to new genera.