Regular Correspondences as Proofs of Relationship
We now consider the two most frequently used criteria, as mentioned in 2.2, which are supposed to prove relationship, namely regular sound correspondences and the reconstruction of proto-languages. In fact, these two criteria are ultimately related. Textbooks of comparative linguistics, in their methodological chapters, start with sound correspondences and then show how, by using them, one can reconstruct the ancestral language. Viewed in this light they are part of a larger whole, the initial and final stages respectively of the comparative method. Hence, one often encounters the statement that the validity of some particular stock has been demonstrated by the comparative method. Nevertheless, these two criteria deserve separate treatment.
Regarding sound correspondence, the detailed treatment in Greenberg (1987) can be summarized as follows. The statement that languages are related if they show regular sound correspondences is not so much denied as shown to be refutable in specific cases because there are so many alternative explanations of the sort universally accepted in historical linguistics, the most important probably being analogy. Here individual linguists may differ in regard to preferences for particular types of processes as explanation. There are those who believe that sound laws have no exceptions and who are, therefore, likely to favor regular sound changes to account for the difference in forms which appear similar enough to be likely etymologies. On the other hand, there are those who refuse to multiply sound correspondences and either reject certain etymologies accepted by the first group, or resort more frequently to other explanations including sporadic assimilation and dissimilation, metathesis and of course morphological analogy more frequently than the first group. These are, of course, tendencies rather than two clearly defined camps with fundamental methodological differences.
There is a further point which has been seen and accepted for some time now by many historical linguists, namely that where there have been extensive borrowings from one language into another, the two languages can show great regularity of sound correspondence, indeed, often greater than between cognate languages. A now famous case is that of Wolfenden who assigned Thai to Sino Tibetan on the basis of borrowings from Chinese into Thai. His thesis is now universally rejected but it is not always realized that in a broader classificational attempt it would be unacceptable, not only because Thai basic vocabulary is hardly affected but because every Thai resemblance to Sino-Tibetan points directly to Chinese as a source. If it were really Sino-Tibetan it would in its vocabulary frequently display resemblances to languages like Tibetan and Burmese but not found in Chinese.
