Several times this semester Joseph Greenberg's writings had been the subject of class discussions. Usually, the papers were dismissed as unconventional and unacceptable. The conclusions he arrived at seemed to be accepted as correct, but the methodology Greenberg used to reach them was not. After reading the assigned writings and completing a project which involved two fairly accepted linguistic practices (glottochronology and the comparative method) I have to disagree with convectional wisdom and side with Greenberg. There is a need for a method similar to Greenberg's. However, though it may be a sign of my naivete, there appear to be serious repercussions that result from the power of Greenberg's use of multilateral comparisons.
The project I undertook this semester was to apply glottochronology to Sephardic Hebrew, Moroccan Arabic, and Abore. All three languages are classified as Hamio-Semitic. Therefore, a proto-language from which the current forms of these languages can be derived should exist. However, I am not aware of any theory which states that every modern word in every modern language should be traceable back to a proto-form in a proto-language.
Glottochronology is actually built upon the principle that not all forms are traceable -- that change does occur (and occurs at a fixed rate). For this reason I was not surprised when cognates were few and far between for these three languages. But I was shocked and a little annoyed when only ten words (from the 100 on the Swadesh list) had cognates in both Arabic and Abore. This meant that the languages had diverged more than 7500 years ago. While that might very well actually be the case, the readings indicate that almost any two languages could end up with ten cognate words after seventy-five centuries.
I have to assume that far more went into determining the relationships between Arabic, Hebrew, Abore, and the Hamio-Semitic family than mere comparative method, because using that method on Abore and Arabic yields results that are no different from comparing English and Abore. History, archeology, genetics, and more thorough linguistic analysis might solidify the relationship of Abore to Hebrew and Arabic while castrating the concept of a proto-Englore, but the comparative method alone failed to convincingly support any kinship between Abore and Arabic.
From this frustration sprang a paradox. How can I determine to what languages the comparative method should be applied? An analysis of two randomly picked languages will probably yield some number of apparent cognates if all non-phonetic knowledge is ignored. Blindly assuming that apparent cognates are signs of a common proto-language, any two languages can be classified as related and summarily lumped together. Without a priori knowledge it seems very unlikely that anyone would be able to group together two distantly related languages -- after all, a cursory glance at Abore and Hebrew revealed about twelve possible cognates. Off the top of my head I can come up with nine possible cognates for English and Hebrew (shemesh, sun; katul, kat; hu, he; keren, horn; eretz, irth; har, hill; kar, kold; ayin, eye; sheva, seven [in a semi-phonetic transcription of both languages] ).
Early in the semester one of the topics presented was a theory of language change which involved circles. Languages could be grouped in a series of concentric rings and the further a ring was from the center the more change the language it represented had undergone. Different language families would form the basis for different sets of concentric circles. If the comparative method is used on two languages, the hope is that the derived proto-language will be on a circle closer to the center than either of the compared languages. But what happens if the two languages that are picked happen to look similar but are actually just overlapping outer rings of two different centers -- they are distant descendents of different proto-languages and belong to different families. Using the comparative method and no other tool, the resulting reconstruction would create an artificial mother tongue where none had previously existed. This is what happens if I attempt to create a proto-Heblish. How can I be sure that I am not positing an equally invalid language when I compare Abore and Arabic?
If several languages were available for comparison with each other, then it might be feasible to postulate a hierarchy of banality. That is, if several features are common to a group of languages, then those languages are most likely related. If a set of features (or a feature) in one group corresponds to a different feature in another group then those two groups stand a good chance of being descendents of an earlier parent language. This is my interpretation of Greenberg's method of multilateral comparisons.
Multilateral comparisons would reveal that although two descendent languages may appear to be closely related, each is more closely related to other languages and thus belong to different families. (Of course, there exists the possibility that all languages are descendents of a single proto-World, but that would constitute another essay altogether). Thus a need that is satisfied by no other method presented in class is fulfilled by Greenberg's method. It facilitates the determination of when it is appropriate for the comparative method to be applied.
Like Greenberg, I intend no slur on comparative linguistics. I am merely curious as to why it seems to be applied indiscriminately and its results accepted axiomatically. If there was a feeding order of linguistic procedures, then more utility could be derived from the comparative method. Multilaterally compare hundreds or thousands of languages (I get the impression Greenberg would like to say compare every language) and then apply the comparative method to those that are shown to be closely related. Perhaps iterations of this procedure could take the reconstruction back further than the 10,000 year limit proposed for the comparative method and glottochronology, but maybe not.
The opposition to multilateral comparison seems to be based on flaws in Greenberg's application of it and the lack of any robust proof of its validity. Greenberg is not omnipotent. As a mortal, it is not possible for him to be aware of the nuances or even the broad outlines or every language. If he attempted to pass himself off as being aware of such things he was overstepping his bounds. It was pointed out in class that several of the comparisons he used were not merely inaccurate, they were wrong. Phonemes with no lexical content (or with a content different from that with which they were compared) were used in demonstrations of his methodology.
There are at least two counters to this argument. The first is perhaps a little trite sounding, but it is important. At some point in a multilateral comparison of hundreds of languages, there will be languages (like Abore and Arabic) that are related but not in any obvious way. What Greenberg's method reveals (or what I think a useful variation of the method should reveal) is that the languages have some common thread, however tenuous, which binds them together. This thread may not be visible, important, or even present to a comparative linguist because it can not be quantified in precise rule, but comparative linguistics is intended to yield a relatively focused view of the short term while multilateral linguistics should provide a broad overview of the long term -- sewing various and sundry threads into a web of related languages. Thus the precise meaning of the sounds are not important once a certain level is reached. What is important (and a little amazing) is that a group of languages has something in common.
To counter my own rebuttal: There are only so many sounds that can be generated by the human mouth. Obviously some of them will be found in every language humans have vocalized. The strength of the comparative method is in identifying consistent patterns of similar vocalizations with similar meanings. Why should the consistent presence of merely one sound in various words which might have similar meanings provide any sign of a relationship? I have no answer for this and am forced to assume that there is enough valid evidence in Greenberg's notebooks to justify the conclusions he reached. I am more concerned with the theory of multilateral comparisons than with Greenberg's application of it.
Any assessment of the validity of multilateral comparison can not be done with the standards of comparative linguistics. As far back as the comparative method will go, it confirms Greenberg's results. However, the point of multilateral comparisons is to do what can not be done with the comparative method. Just as evidence from fields other than linguistics were needed to demonstrate the kinship of Abore and Hebrew, evidence from other fields will have to suffice as justification of Greenberg's method. However inefficient the studies may have been, the genetic work done by Cavalli and Sforza supported the results of multilateral comparison.
Without support from other areas of science, linguistics based solely on pronunciation and grammar has weaknesses. No method can recreate the hand motions or facial expression of speakers of early-German or proto-Mayan. And even with support from other the other sciences, neither the comparative method nor multilateral comparison can successfully recreate and classify proto-languages. The effectiveness of the comparative method can not be denied -- nor should its weaknesses. Reciprocally, the effectiveness of multilateral comparison should not be denied. It allows for successful employment of the comparative method and reveals familial groupings that would otherwise only be guessed at. Its weaknesses, including the lack of an appropriate benchmark for accuracy, should also be pointed out. But it should not be summarily dismissed as a fraudulent method foisted upon linguistics.