The river otters of the Western Hemisphere were recognized as a distinct group of related species within the genus Lutra by Pohle (1920) in his revision of the Lutrinae. Since then a need has remained for further clarification of the relationships within the group, as well as that of the group as a whole to other forms in the subfamily. A large number of specimens was compared in an attempt to determine these relationships. Materials consisted largely of skulls, but also used were skins and postcranial material, available in smaller numbers. A number of bivariate relationships of the skull and dentition were analyzed allometrically and the overall similarities of nearly all species of Lutrinae were estimated using taxonomic distance. Other qualitative and descriptive comparisons were made to complement these methods. The amount of morphological variation within known wide-ranging species and between valid sympatric species were criteria used to infer the specific status of the taxa that were compared. Results of these comparisons led me to conclude that there are probably four Recent species of river otter in the Western Hemisphere, corresponding to the North American Lutra canadensis, the Neotropical L. annectens-enudris-platensis group, the Chilean L. provocax and the southern Pacific coastal form L. felina. The latter is the most distinct species and probably represents the earliest invader of the relatively isolated Chilean faunal province. Another Chilean species, L. provocax, is more similar to the other large Neotropical forms. Evidence further suggests that the relationship of the American river otters with L. lutra and other Old World species of that genus is not as close as was formerly assumed. Aside from differences in skull and dentition, the marked differences in the known bacula may be regarded as significant evidence of their distinctness. The analysis of overall similarity suggests that Pohle's platensis group merits generic rank. Infraspecific variation was analyzed, but not in Neotropical species because of insufficient geographical representation in samples. I propose that the platensis group of Pohle be elevated to generic rank, for which the name Lontra Gray 1 843 is available, and that L. annectens, L. enudris and L. platensis represent one species, which should bear the name Lontra longicaudis Offers 1818.