Genetically Relevant Criteria
In all the preceding discussion there has been frequent reference to similarities and degrees of similarity among languages e.g. that English is more similar to German in genetically relevant ways than it is to French. However, there has been no explicit treatment of the question as to what kinds of similarities are involved. After all, languages can be like or unlike each other in innumerable ways. However, there has been an implicit answer to this question in the immediately preceding section, similarity in specific word forms, or more exactly morphemes, involving sound and meaning simultaneously. Resemblances of other kinds, e.g. in sound only as in the existence of glottalic ejectives in both languages, or meaning only as in agreement in a tense system including a future vs. non-future opposition, or in word or morpheme order are irrelevant and would normally be considered typological. Of course, languages can have typological agreement with other languages of the same stock which is part of their common inheritance from their common linguistic ancestor. However, such characteristics do not figure in our carrying out the classification. They are, as it were, a bonus deriving from it and useful in considering the relative stability of typological traits and their historical development.
However, as self-evident as it might seem that it is from similarities in sound and meaning simultaneously that one finds the materials for genetic classification, there is a strand of thinking in keeping with the strong tendency towards formalism in linguistics to dismiss phonetic similarity. All that would count would be regularity of correspondence in abstraction from phonetic content. Of course, semantic similarity is not suppressed. For some, it is narrowed to semantic identity which in practice can only mean translation equivalence. The reason for this is that they wish to make it as difficult as possible to prove that anything is related to anything else.
I earlier thought that the importance of phonetic similarity was so obvious that it required no defense. We have seen in the previous section how differential similarities as in Welsh tri, Italian tre vs. Finnish kolme ‘three’ play a key conceptual role at the very initiation of comparison leading to classification. Those who oppose this obvious and sensible approach emphasize the frequency of accidental resemblances and the existence of phonetically unusual correspondences. Let us consider examples of each in turn.
Meillet (1958:92) gives the following example. French feu has nothing in common with German Feuer ‘fire’, while the German form goes back to an original represented by words which are quite different, Greek pur and Armeni an hur. What Meillet fails to consider is how we know that French feu is only accidentally similar to German Feuer.
We know it because if we compare French feu with the Spanish form fuego and Italian fuoco, we see that it goes back to a form in which the second consonant is a velar. Compare also French peu ‘few’ with Italian poco. Why do we compare French with Spanish and Italian in this and other instances? A glance at Table 7 in Greenberg (1987:24) will show why. From the very beginning, in word after word French aligns with Spanish, Italian and other Romance languages, while English lines up with German, Dutch etc. Besides this, in Meillet’s example, German Feuer, English fire, etc. are not all that different from pur and hur since p>f>h is a very commonly encountered form of change and they agree in their final r.
A favorite in recent discussion has been the correspondences between Greek dw and Armenian rk as seen in the word for ‘two’ and a few other words. All this shows is that phonetically improbable resemblences may occur in correspondences. However, they are rare. No similar instance to dw~erk has ever been found but t~d or t~t are commonplace.
More importantly, it has nothing to do with classifying Armenian as Indo European. Scores of phonetically obvious resemblances between Armenian and other Indo-European languages exist and were noted earlier. Indeed, Hübsch mann, the founder of modern Armenian studies who first established the affiliation of Armenian on a firm basis does not have this correspondence in his seminal paper on the position of Armenian (1875), what Meillet was later to call the ‘bel article’ on this subject and he never accepted this correspondence.
