Cryptic and hybridizing species may lack diagnostic taxonomic characters leaving researchers with semi-diagnostic ones. Identification based on such characters is probabilistic, the probability of correct identification depending on the species composition in a mixed population. Here we test the possibilities of applying a semi-diagnostic conchological character for distinguishing two cryptic species of blue mussels, Mytilus edulis and M. trossulus. These ecologically, stratigraphically and economically important molluscs co-occur and hybridize in many areas of the North Atlantic and the neighboring Arctic. Any cues for distinguishing them in sympatry without genotyping would save much research effort. Recently these species have been shown to statistically differ in the White Sea, where a simple character of the shell was used to distinguish two mussel morphotypes. In this paper, we analyzed the associations between morphotypes and species-specific genotypes based on an abundant material from the waters of the Kola Peninsula (White Sea, Barents Sea) and a more limited material from Norway, the Baltic Sea, Scotland and the Gulf of Maine. The performance of the "morphotype test" for species identification was formally evaluated using approaches from evidence-based medicine. Interspecific differences in the morphotype frequencies were ubiquitous and unidirectional, but their scale varied geographically (from 75% in the White Sea to 15% in the Baltic Sea). In addition, salinity-related variation of this character within M. edulis was revealed in the Arctic Barents Sea. For every studied region, we established relationships between the proportions of the morphotypes in the populations as well as between the proportions of the morphotypes in samples and the probabilities of mussels of different morphotypes being M. trossulus and M. edulis. We provide recommendations for the application of the morphotype test to mussels from unstudied contact zones and note that they may apply equally well to other taxa identified by semi-diagnostic traits.