28 BUEEAU OF AMERICAN ETHNOLOGY [bull. 40 



clear in the case of words like prepositions, conjunctions, or verbal 

 forms which belong to subordinate clauses. Thus it would be ex- 

 ceedingly difficult to imagine the use of words like and, for, to, were, 

 expressed in such a way that they would convey a clear idea, except 

 perhaps in forms like the Laconic If, in which all the rest of the 

 sentence is implied, and sufficiently indicated by the if. In the 

 same way, however, we who are grammatically trained may use a 

 simple ending to correct an idea previously expressed. Thus the 

 statement He sings beautifully might elicit a reply, sang; or a 

 laconically inclined person might even remark, in reply to the state- 

 ment He plays well, -ed, which by his friends might be well under- 

 stood. It is clear that in all these cases the single elements are 

 isolated by a secondary process from the complete unit of the 

 sentence. 



Less clear appears the artificiality of the word as a unit in those 

 cases in which the word seems to designate a concept that stands out 

 clearly from others. Such is the case, for instance, with nouns; and 

 it might seem that a word like stone is a natural unit. Nevertheless 

 it will be recognized that the word stone alone conveys at most an 

 objective picture, not a complete idea. 



Thus we are led to the important question of the relation of the 

 word to the sentence. Basing our considerations on languages differ- 

 ing fundamentally in form, it would seem that we may define the 

 word as a phonetic group wJiicli, owing to its permanence of form, 

 clearness of signifcance, and phonetic independence, is readily sepa- 

 rated from tJie whole sentence. This definition obviously contains a 

 considerable number of arbitrary elements, which may induce us, 

 according to the general point of view taken, sometimes to designate 

 a certain unit as a word, sometimes to deny its independent exist- 

 ence. We shall see later on, in the discussion of American languages, 

 that this practical difficulty confronts us man}^ times, and that it is 

 not possible to decide with objective certainty whether it is justifiable 

 to consider a certain phonetic group as an independent word or as a 

 subordinate part of a word. 



Nevertheless there are certain elements contained in our definition 

 which seem to be essential for the interpretation of a sound-complex 

 as an independent word. From the point of view of grammatical 

 form, the least important; from the point of view of phonetics, how- 



