30 BUKEAU OF AMERICAN ETHNOLOGY [bull. 40 



I doubt very much whether an investigator who would record 

 French in the same way as we do the unwritten American languages 

 would be inclined to write the pronominal elements which enter into 

 the transitive verb as independent words, at least not when record- 

 ing the indicative forms of a positive verb. He might be induced 

 to do so on discovering their freedom of position which appears in 

 the negative and in some interrogative forms. 



The determining influence of the freedom of position of a phonetic- 

 ally fixed part of the sentence makes it necessary to include it in our 

 definition of the word. 



Whenever a certain phonetic group appears in a variety of posi- 

 tions in a sentence, and always in the same form, without any, or at 

 least without material, modifications, we readil}^ recognize its indi- 

 viduality, and in an analysis of the language we are inclined to con- 

 sider it as a separate word. These conditions are fully realized only 

 in cases in which the sound-complex in question shows no modifica- 

 tions at all. 



It may, however, happen that minor .modifications occur, par- 

 ticularly at the beginning and at the end, which we may be ready 

 to disregard on account of their slight significance as compared to 

 the permanence of the whole word. Such is the case, for instance, 

 in the Dakota language, in which the terminal sound of a permanent 

 word -complex which has a clearly defined significance will auto- 

 matically modify the first sound of the following word-complex which 

 has the same characteristics of permanence. The reverse may also 

 occur. Strictly speaking, the line of demarcation between what we 

 should comr^only call two words is lost in this case; but the mutual 

 influence of the two words in connection is, comparatively speak- 

 ing, so slight that the concept of the individuality of the word out- 

 weighs their organic connection. 



In other cases, where the organic connection becomes so firm 

 that either both or one of the component elements may never occur 

 without signs marking their close coupling, they will appear to us 

 as a single unit. As an example of this condition may be mentioned 

 the Eskimo. This language contains a great many elements 

 which are quite clear in their significance and strong in phonetic 

 character, but which in their position are so limited that they 

 always follow other definite parts of the sentence, that they can 

 never form the beginning of a complete phonetic group, and 



