AbstractDNA barcoding has been largely successful in differentiating animal species, but the most effective loci and evaluative methods for plants are still debated. Floras of young, oceanic islands are a challenging test of DNA barcodes, because of rapid speciation, high incidence of hybridization and polyploidy. We used character-based, tree-based and genetic distance-based methods to test DNA barcoding of 385 species of native Hawaiian plants constituting 20 lineages at the nuclear ITS(2) locus, nine lineages at each of the plastid loci trnH-psbA and rbcL, eight lineages at the plastid locus matK and four lineages with concatenated data. We also incorporated geographical range information and tested if varying sample sizes within a lineage influenced identification success. Average discrimination success was low (22% maximum) with all methods of analysis across all loci. The character-based method generally provided the highest identification success, there were limited benefits from incorporating geographical data and no relationship between number of species sampled in a lineage and identification success was found. Percentages of identification success are the lowest reported in a DNA barcoding study of comparable scale, and multi-species groups that radiated in the Hawaiian archipelago probably cannot be identified based on current DNA barcoding loci and methodologies.