Three fish bone identification protocols used for determining taxa composition for Pacific island archaeofaunal assemblages are evaluated. The protocols include using the following: (1) the most commonly identified five paired cranial bones and ‘specials’ or unique elements; (2) an expanded number of cranial bones; and (3) the less common inclusion of all vertebrae. Explicit identification and quantification protocols are outlined for systematically incorporating all vertebrae which, predictably, increases the number of identified specimens for an assemblage, thus providing more bones useful for reconstructing live fish biomass (weight and length). Significantly, a range of unique archaeological vertebrae are useful for calculating minimum number of individuals. Using a well-preserved assemblage from Henderson Island, Pitcairn Group, southeast Polynesia, numbering 6480 fish bones (concentration index = 21 580 m3), we demonstrate differences in rank-order abundance from three taxon identification protocols. For example, when using all vertebrae grouper (Serranidae) and surgeonfishes (Acanthuridae) are more numerically equivalent than when relying mostly on cranial bones for identification for minimum number of individuals and number of identified specimens. This has important implications for making comparisons between sites or across regions where different identification protocols were used. This pilot study demonstrates that using all vertebrae for taxon identification and quantification, not just unique hypurals (terminal vertebrae) or those from sharks and rays (Elasmobranchii), should be standard practice for identifying a greater number of bones to taxon and thereby providing better reconstructions of prehistoric fishing and subsistence practices in the Pacific. Copyright © 2013 John Wiley & Sons, Ltd.