RELATIONS OF COMPARATIVE GRAMMAR 35 



coming from it. If we say, for example, that the Indo-European 

 form for "is" was *6sti, there is every reason to believe that we are 

 coming as near the exact truth as when we say that the Greek form 

 was IO-TI, the Latin est, the Gothic ist, etc. To be sure, we are ignorant 

 of the precise physiological character of each sound in the Indo- 

 European * esti, we do not pretend to know the exact quality of the 

 e, or whether the t was a pure dental or an alveolar, like our English t. 

 But the finer nuances of pronunciation are unknown also in the case 

 of the Greek or Latin form. All that one claims for the assumed 

 * esti is that it represents the parent form as nearly as our ordinary 

 written symbols ever represent the spoken form. However, the 

 assumed Indo-European forms differ widely not only in the degree 

 of certainty which attends their reconstruction, but also in the degree 

 of accuracy intended by them, 1 and, while every such reconstructed 

 form implies a belief in its existence on the part of the one who 

 employs it, they are in general best regarded as convenient formulse, 

 furnishing the means of expressing briefly the combined evidence and 

 its interpretation, but subject to change with the progress of the 

 science. Such formulse are indispensable to such a highly organized 

 science as the Indo-European Comparative Grammar of the present 

 day, and from the fact that but little use is made of them in Semitic 

 Comparative Grammar the Indo-Europeanist is prone to infer, sub- 

 ject to correction^ that it is still on a stage of development parallel 

 to Indo-European Comparative Grammar of the time of Bopp. 



I have said that the comparison of related forms was not an end in 

 itself, but a means of reconstructing the parent form. But I do not 

 wish to imply that these parent forms are of great intrinsic interest 

 or that the reconstruction of the parent speech is the ultimate aim. 

 No one is ambitious to speak this hypothetical language, nor does it, 

 as Bopp fondly hoped, furnish the key to the problems of primitive 

 linguistic development. Indeed, this language which we arrive at by 

 reconstruction is itself a highly developed form of speech, which has 

 behind it thousands of years of history which is forever inaccessible 

 to us. 



Its value lies rather in the light which is thereby reflected on the 

 history of each individual language belonging to the group. Each 

 language contributes its share of evidence for the reconstruction 

 of the parent speech, and each in turn is illuminated by it. The real 



1 For example, in the reconstructed *p3fe(r), "father" (Skt. pita, Grk. irar^p, 

 Lat. pater, etc.), no such degree of accuracy is claimed for the first vowel as for 

 that of * esti. Indeed, the a is merely a convenient symbol for a certain vowel 

 which appears, in a whole series of words, in Sanskrit as i, in the European lan- 

 guages as a, but which must have differed originally both from i (which is i in Eu- 

 ropean as well as in Sanskrit) and from a (which is a in Sanskrit as well as in 

 European), and which moreover appears as the reduced grade of a long vowel. 

 The usefulness of the symbol is not impaired by the fact that the original pro- 

 nunciation of the vowel cannot be determined. 



