58 COMPARATIVE LANGUAGE 



individual languages must be derived would seem unwarranted. It 

 is for this reason that I should be slow to assert that the Latin 

 subjunctive forms are semantically the product of a fusion of Indo- 

 European subjunctive (feras) and optative (faxis) forms. I cannot 

 see what obstacle should prevent our interpreting these forms as re- 

 flexes of a section of Indo-European speech in which the adaptation 

 of forms terminating in long a e o and those with the formative ie : I 

 to subjunctive and optative uses respectively had never taken 

 place; just as the Italic and Celtic r-forms of the passive have their 

 formal but not their semantic counterpart in Sanskrit. Thus, while 

 in dealing with the formal side of Indo-European speech the con- 

 struction of parent forms is a useful and convenient device and 

 cannot under any circumstances do harm, the case is different in 

 syntactical work with its emphasis on the semantic side. Here insist- 

 ence on a uniform parent language with well-defined semantic sys- 

 tems shared in by all sections of Indo-European folk seems fraught 

 with danger and must often tend to cloud the issue by injecting 

 foreign semantic elements, which in reality were, perhaps, never 

 present in the history of a mode or tense. It seems methodologically 

 wrong to assume that because certain formatives in a given num- 

 ber of languages can be formally united, their respective semantic 

 contents must also be unified under one denominator, which is to 

 be regarded as starting-point and fountain head from which the 

 meanings in the individual languages are to be historically derived. 

 Early formal identity of formatives may well go hand in hand with 

 primitive semantic differences due to separate and sectional develop- 

 ment. 



I turn from this general discussion of the value of inferred forms 

 and meanings to a number of problems connected with the different 

 departments of grammar, selecting a few which are of a more general 

 nature. 



In phonetics the problem of the causes upon which rests the 

 striking uniformity of sound-changes is not yet finally solved. The 

 investigation of the nature of phonetic changes has been, in the 

 main, confined to the causes which produce primary changes, that 

 is, those which originated in, and were created by, the individual, 

 who therefore plays an active part in their production. While these 

 changes have received detailed treatment, another phase of the 

 subject, namely, the cause which underlies the comparative uniform- 

 ity of these changes in a large number of individuals scattered over 

 a considerable area, has been touched only lightly and in a more or 

 less general way. There are two possible ways of accounting for such 

 uniformity. One theory (and it is important to note that the fore- 

 most authority on the psychology of language holds this view, cf. 

 Wundt's Volkerpsychologie, Die Sprache, vol. i, p. 391) explains it as 



