The Classification of American Languages
International Journal of American Linguistics
Anthrop., N. S., 23 : 367 et sq., is a discussion of the theoretical point of view one should adopt in classifying American languages. On the whole I am very much in sympathy with his remarks : see my paper on American languages in the Journal of the Washington Academy of Sciences, VII : 222 et sq. (I917). But there is one point which I think Dr. Boas overlooks when discussing the borrowing of morphological features, admitting that he has made it very plausible that a number of borrowings occur
... of borrowings occur where they had not been previously suspected. And this is, that if the morphological resemblances between two supposedly distinct but contiguous stocks were entirely due to borrowings, by the doctrine of chances we should expect to find similar borrowings in another supposedly distinct but contiguous stock. And this is not the case in at least certain instances. Thus Athapascan, so far as is known, has been in iust as intimate contact for a very long period with Salishan and Esquimauan as with Tlingit, but there is but slight resemblance structurally between Athapascan,Salishan and Esquimauan. On the other hand admittedly there is a very decided structural resemblance between Athapascan and Tlingit, even if the amount of vocabulary held in common is small. Or again, Algonquian, so far as we know, has been in just as intimate contact with Iroquoian, Siouan, and Muskhogean for at least several hundred years as it has with Esquimauan. Yet structurally Esqui-mauan and Algonquian resemble each other, and similarly Siouan and Muskhogean: but observe that the first pair does not resemble the second. pair nor does either member of the first group resemble either one of the second. The contrast between Algonquian and Esquimauan on the one hand and Athapascan on the other is another case in point. Similar cases occur in the southwest and also northwest. Now if the above were entirely due to borrowing we should expect to find resemblances equally distributed where supposedly distinct stocks are contiguous. If the resemblances are confined to one or two features, they may be safely ascribed to acculturation; but when there are far-reaching structural resemblances between two or more supposeldy distinct (and especially contiguous) stocks we may legitimately infer an ancient genetic connection which perhaps can no longer be proved owing to very early differentiation.The actual application of the above principle on a large scale is quite another thing. We are pro-.bably not yet in a position to make final announcement of such ancient genetic connections; though tentative results might properly be made public. The recent efforts to prove genetic connections on a large scale have been deplorable from a methodological point of view. Enthusiasts have cast all prudence to the winds; still their work has not been entirely in vain, for they have at least called attention to problems which must be faced sooner or later.