michelson: American Indian languages 223 



cally (as far as is known) to any American linguistic stocks, but 

 nevertheless the totality of their features compels us to classify 

 them with American Indian languages. To account for these 

 facts we must assume either that these tribes are the descendants 

 of the forefathers of American Indians who did not follow their 

 brethren in the migration from Asia to America (for such a 

 migration is firmly established by the facts of physical anthro- 

 pology), or that there has been in comparatively recent times a 

 migration backward from this continent to Asia. In as much 

 as the bulk of American Indian languages are spoken on this 

 continent, from a purely linguistic point of view the latter 

 hypothesis is the most probable. But the essential fact in any 

 case will remain unchallenged, namely, that we have American- 

 oid languages spoken in northeastern Asia. 



Let us now turn to the almost unparalleled number of lin- 

 guistic stocks on this continent. We have an apparent anomaly 

 as compared with most parts of the world. The point at issue is 

 whether such a multiplicity of stocks is original or not. In the 

 first place, in the study of American Indian languages we are at 

 a decided disadvantage as compared with the study, say, of 

 Indo-European languages. Suppose that modern English and 

 modern Russian were the sole survivors of the entire stock. It 

 would be impossible to prove absolutely that they were both 

 genetically descended from a common ancestor, no matter what 

 we might surmise. It is only because we have continuous 

 written records of both covering several centuries, and have the 

 aid of other related languages which have even earlier records, 

 that we can absolutely prove this. For this reason it is clear 

 that there always will be American Indian languages whose 

 genetic connection we may suspect, but which we can not prove. 

 It may be urged that we can actually see what has taken place 

 in the development and differentiation of languages which 

 have been historically transmitted, such as Indo-European 

 languages, and that we should apply the principles derived from 

 such a study to American Indian languages in determining the 

 stocks. The methodical error in such a procedure lies in this, 

 namely, that there are less than a half a dozen, different stocks in 



