The Proto-Comparative Method
By the term ‘proto-comparative method’, I do not mean the total method of constructing a protolanguage, but rather an initial stage of the comparative method which logically and in actual practice but without explicit recognition, precedes what is usually taken to be the initial stage in the comparative method, namely, the discovery of sound correspondences.
The pre-sound-correspondence stage was illustrated briefly in Greenberg 1987 (especially 24–26). There in Table 7 equivalents of nine common words are given for 25 languages in Europe. By the time one has gotten to the second word a three-fold division into Indo-European, Finno-Ugric, and Basque and even the major subgroupings of the first two (the third being an isolated language) become apparent. In comparing for example the third item, equivalents for the English word ‘three’ we are making many judgements of differential phonetic similarities, the meaning here being kept constant. For example, we judge that Welsh tri is more similar to Italian tre than either is to Finnish kolme and in that, in turn, Finnish kolme is more similar to Estonian kolm than either is to the Welsh and Italian forms. Even these four equivalents for ‘three’ give us a grouping of Welsh and Italian (Indo-European) as against Finnish and Estonian (Finno-Ugric). But how do we arrive at these judgements? In doing so, we have as it were applied the comparative method in embryonic form. We compare the t of Welsh tri with the t of Italian tre and not with e in the latter word. So also the two r’s and the vowels are being compared. This set of equivalents for ‘three’ and the groupings it gives are already important evidence leading to the correct classification and, of course, there is no recurrence of sounds within the words so there is no correspondence of t to t or r to r in the usual sense. When, however, we have found more words such correspondences will be found and the t and r in the words for ‘three’ will be examples of them. We might, therefore, call them proto-correspondences.
There is further in the same section in Greenberg (1987) a discussion of the equivalents for the PIE word for ‘tooth’ as leading back quite naturally to a single form that might be characterized as approximately *dant or *dent. This once again is based on exiguous data, without yet assuming anything like a complete original sound system or a regular set of sound changes. Such relatively amorphous hypotheses are the actual and the logical precursors of more fully elaborated but never conclusive formulations resulting from the application of the comparative method. They are the true first steps in the method, never recognized in textbooks of comparative linguistics but worthy of fuller study and consideration.
