The molecular phylogeny of parabasalids has mainly been inferred from small subunit (SSU) rRNA sequences and has conflicted substantially with systematics based on morphological and ultrastructural characters. This raises the important question, how congruent are protein and SSU rRNA trees? New sequences from seven diverse parabasalids (six trichomonads and one hypermastigid) were added to data sets of glyceraldehyde-3-phosphate dehydrogenase (GAPDH), enolase, alpha-tubulin and beta-tubulin and used to construct phylogenetic trees. The GAPDH tree was well resolved and identical in topology to the SSU rRNA tree. This both validates the rRNA tree and suggests that GAPDH should be a valuable tool in further phylogenetic studies of parabasalids. In particular, the GAPDH tree confirmed the polyphyly of Monocercomonadidae and Trichomonadidae and the basal position of Trichonympha agilis among parabasalids. Moreover, GAPDH strengthened the hypothesis of secondary loss of cytoskeletal structures in Monocercomonadidae such as Monocercomonas and Hypotrichomonas. In contrast to GAPDH, the enolase and both tubulin trees are poorly resolved and rather uninformative about parabasalian phylogeny, although two of these trees also identify T. agilis as representing the basal-most lineage of parabasalids. Although all four protein genes show multiple gene duplications (for 3-6 of the seven taxa examined), most duplications appear to be relatively recent (i.e., species-specific) and not a problem for phylogeny reconstruction. Only for enolase are there more ancient duplications that may confound phylogenetic interpretation.