Nuclear ribosomal SSU and chloroplast rbcL data for heterokont algae and potential outgroups were analyzed separately and together under parsimony. A series of taxon sampling experiments were performed. Traditional heterokont classes (e.g., diatoms, phaeophytes, etc.) were monophyletic in most analyses of either data set and in both together. Relationships among classes and of heterokonts to outgroup taxa are sensitive to taxonomic sampling. Stability of heterokont algae relationships was analyzed by deletion of outgroup classes, starting with those presumably most distant from the heterokonts to those closest. In only three analyses using SSU data were the heterokonts monophyletic: with all available outgroup taxa, with the outgroup being hyphochytrids alone or with the outgroup being hyphochytrids plus Developayella. For rbcL data, heterokonts were monophyletic when all available rhodophytes were used, when any combination of rhodophytes tested were used, and when rhodophytes plus bacteria were used. Any addition of cryptophyte and/or haptophyte sequence resulted in a non-monophyletic heterokont algae. In both cases, distribution of heterotrophic stramenopiles among the heterokont algae changed as different outgroups were used. Combined data was similarly equivocal about monophyly of the heterokont algae. The literature suggests that this result is obtained regardless of whether parsimony or likelihood methods are used, and that is because taxon sampling affects all inference procedures. We suggest not only that more molecular data be obtained, but that morphological evidence for monophyly of the heterokonts be re-examined. Finally, a community effort needs to be made to synchronize character sampling across different data sets to increase the number of taxa sampled and to create parallel data sets for all data types.

Key words: bootstrap support, combined data, heterokont algae, outgroups, rbcL, SSU rDNA, taxon sampling