Culture prehistorians using Edward Sapir’s synthesis of North American language classification need to exercise one essential precaution: to bear in mind the approximate scale and implicit time depth of each connection. Sapir at times finds reason to infer common origin “in spite of great differences of structure.” The indications are that Sapir had actually examined material for every language or language family in his scheme, and made no groupings without having found specific evidence to support them. The new lexicostatistic method represents an externalization of one part of Sapir’s procedure, in that the data are always written out in the form of a standard vocabulary of noncultural words, constituting the diagnostic word list. J. W. Powell and most of his successors have evidently felt that in linguistics, unlike some other aspects of culture, it is possible to differentiate clearly between genetic and diffusional agreements. Language is that part of culture which is the least subject to diffusional changes.