A scheme of classification should depend on many organ systems, not on one. How should this be achieved, resolving conflict between different bodies of evidence? In the Treatise on invertebrate paleontology (vols N1-N3) special importance was apparently attached to one feature, the dentition. It is not clear how other criteria were integrated with this. The anatomical features listed in diagnoses of superfamilies and higher taxa in the Treatise were tabulated in a few alternative character-states. Several of these features were rejected as lacking classificatory significance. Nine anatomical features were selected as potentially useful, some drawn from diagnoses in the Treatise, and others from reviews in literature. A matrix of data on the character-states for these nine anatomical features, for all 40 Recent superfamilies of Bivalvia, has been analysed by computer. This compared each superfamily in turn with all others, and clustered those superfamilies showing the highest percentage similarity of phenetic characters. Each cluster was assigned those characters common to all members of the group, and the clustering process was repeated. Each stage in clustering was reported in terms of percentage similarity, and a dendrogram was drawn by data plotter. The investigation excluded variability below the superfamily, thus limiting confusion by convergence; unfortunately it also excluded information on phylogenetic relations between superfamilies (which could easily be added when available). The exclusively phenetic data are presented compactly in the dendrogram, which should be taken fully into account by anyone designing a phylogenetic classification. The six clusters in the dendrogram correspond generally with the classification in the Treatise, which offers confidence in the methods used, but some changes are suggested. The procedure is objective, and can be repeated with suitable amendments, e.g. for correction of errors found, for exclusion of unsuitable data, and inclusion of newly acquired useful data. It is suggested that the Bivalvia comprise only two subclasses, deposit-feeding protobranchs and suspension-feeding lamellibranchs, thus recognizing the major functional differences between these contrasted life forms. The lamellibranchs can best be interpreted as a matrix of about 12 x 12 families which are evolving partly in parallel, and among which the taxonomic differences are of a lower degree.