In Canis and Ursus the largest proportion of attachments of muscles of the shoulder and brachium on the scapula and humerus is direct; fewer attachments are aponeurotic or tendinous. In both genera most attachments can be associated with superficial osteological features (scars or delimitable surfaces); attachments that lack such features are direct. Most aponeurotic attachments are associated with rugose scarring whereas tendinous attachments are often associated with smooth surfaces. Although most attachments can be associated with osteological features the areal extent of attachment is often not inferrable from the bone. The inference of muscle size or functional significance from osteological features is problematic. The amount of myological information that can be deciphered from the osteology in Canis and Ursus is greater than that reported for particular members of other vertebrate groups which suggests that there may be differences in the degree to which muscles can be reconstructed from superficial osteology alone. Nonetheless, even in mammals such as the Carnivora, detailed muscular reconstructions in extinct taxa cannot be achieved without reference to the musculature of extant relatives. Such reconstructions rely on assumptions, that often have not been adequately tested, regarding the similarity of musculature in closely related taxa. This testing and well corroborated hypotheses of phylogenetic relationship are essential for the evaluation of the accuracy of reconstructions of the musculature in fossil vertebrates.