Marine fishes from the northwest Atlantic Ocean were analysed to determine whether barcoding was effective at identifying species. Our data included 177 species, 136 genera, 81 families and 28 orders. Overall, 88% of nominal species formed monophyletic clusters based on >500 bp of the CO1 region, and the average bootstrap value for these species was 98%. Although clearly effective, the percentage of species that were distinguishable with barcoding based on the criterion of reciprocal monophyletic clusters was slightly lower than has been documented in other studies of marine fishes. Eelpouts, sculpins and rocklings proved to be among the most challenging groups for barcoding, although we suspect that difficult identifications based on traditional (morphology based) taxonomy played a role. Within several taxa, speciation may have occurred too recently for barcoding to be effective (e.g. within Sebastes, Thunnus and Ammodytes) or the designation of distinct species may have been erroneous (e.g. within Antimora and Macrourus). Results were consistent with previous work recognizing particularly high levels of divergence within certain taxa, some of which have been recognized as distinct species (e.g. Osmerus mordax and Osmerus dentex; and Liparis gibbus and Liparis bathyarcticus), and some of which have not (e.g. within Halargyreus johnsonii and within Mallotus villosus). The results from this study suggest that morphology-based identification and taxonomy can be challenging in marine fishes, even within a region as well characterized as Atlantic Canada. Barcoding proved to be a very useful tool for species identification that will likely find a wide range of applications, including the fisheries trade, studies of range expansion, ecological analyses and population assessments.