14 Global Etymologies
Yüklə 356,86 Kb. Pdf görüntüsü
|
|
14. Global Etymologies
285
excluded from his Indo-European family. 2 For Indo-European, and for the Indo-Europeanists who came to dominate historical linguistics, the problem of classification was essentially resolved by Jones, and the later additions of a few more obvious branches, such as Tocharian and Anatolian, did not alter this state of affairs. The problems that Greenberg confronted, however, when he set out to clas- sify the languages of Africa, were quite different from those facing a historical linguist investigating an already-defined family. Greenberg was confronted by over 1,000 languages, only some of which fit into well-defined families (e.g. Semitic, Bantu), and among which there was little understanding of the rela- tionships. Under these circumstances, where does one start? Obviously the only way to begin is by the comparison of basic lexical items and grammatical formatives in all the languages, which inevitably leads to a classification of the languages into a certain number of groups defined by recurring similari- ties. This is exactly what Jones had done when he identified Indo-European, stressing, as he did, “a stronger affinity, both in the roots of verbs and in the forms of grammar, than could possibly have been produced by accident.” He said nothing of sound correspondences or reconstruction, for in fact these concepts came to prominence (despite the earlier work of Rask, Grimm, and Bopp) only in the second half of the nineteenth century. We believe, in short, that there is really no conflict between Greenberg’s method of classifying languages and what is often referred to rather inexplic- itly as “the standard methodology.” The standard methodology is used to investigate family-internal problems; it does not—at least as it is explained in the basic textbooks referred to above—tell one how to identify language fam- ilies. Accordingly, it does not tell one how to classify the world’s languages. This, rather, is what Greenberg’s work does, and it is, furthermore, how Greenberg views what he does. It has recently been alleged that he himself subscribes to the view that his methods differ from the standard methodol- ogy: “Greenberg (1987) makes clear that he believes such groupings [as Al- taic, Hokan, and Amerind] cannot be reached by the standard comparative method; a wholly different method, mass comparison, is required” (Nichols 1990: 477). That this is, in fact, exactly the opposite of Greenberg’s views is shown in the following: Statements from certain American Indianists that I have rejected comparative lin- guistics and have invented a new unorthodox method called mass or multilateral comparison are repeated again and again in the press. However, as I clearly stated in Greenberg (1987: 3), once we have a well-established stock I go about comparing and reconstructing just like anyone else, as can be seen in my various contributions to historical linguistics. However, as I pointed out long ago in regard to my gen- erally accepted African classification, the first step has to be to look very broadly, 2 The term Indo-European was not introduced until the nineteenth century. 286 14. Global Etymologies on at least a continent-wide scale, to see what the obvious groupings are. How can one start to apply the comparative method until one knows what to compare? (Greenberg 1990: 8) RECONSTRUCTION It is remarkable how frequently reconstruction is confounded with tax- onomy. For a moment’s reflection should make it clear that one can only begin reconstructing a proto-language after one has decided which languages belong to the putative family. Until one has delineated a set of seemingly re- lated languages, collectively distinct from all others, by the methods outlined at the outset of this chapter, there is simply nothing to reconstruct. (After the fact, of course, reconstruction and (re)classification may enjoy a fruitful feedback.) And as for the supposed validating effect of reconstruction, would anybody claim that a bad reconstruction invalidates a well-defined family such as Indo-European? Or that a brilliant reconstruction could show that Slavic, Ob-Ugric, and Basque form a valid family? As a process, reconstruction is entirely different from taxonomy, and the two should not be confused. It is for this reason that Bynon’s claim that Greenberg uses multilateral comparison as a “substitute” for reconstruction really makes no sense, and it is certainly not anything that Greenberg has ever written or said or even suggested. SOUND CORRESPONDENCES Perhaps the greatest source of confusion in recent taxonomic debates has been the role that sound correspondences, for example Grimm’s Law, play in classification. It is clear that many historical linguists see regular sound correspondences as playing some crucial role in identifying valid linguistic taxa. In reality, sound correspondences are discovered only after a linguistic family has been identified, for the simple reason that sound correspondences are properties of particular linguistic families. They are not—and could not be—a technique for discovering families. When the Indo-European sound correspondences were worked out in the nineteenth century, not for a minute did any of the Indo-Europeanists imagine that they were “proving” Indo- European, the validity of which had not been in doubt for decades. There are several reasons why sound correspondences have become en- meshed with taxonomic questions. First, it is sometimes alleged that it is only by means of regular sound correspondences that borrowings can be dis- criminated from true cognates. It has long been recognized, however, that loanwords often obey regular sound correspondences as strictly as do true cognates, a point emphasized on several occasions by Greenberg (1957, 1987). Campbell (1986: 224) makes the same point: “It ought to be noted that such
14. Global Etymologies 287
agreements among sounds frequently recur in a number of borrowed forms, mimicking recurrent sound correspondences of true cognates.” Another alleged use of sound correspondences is to discriminate superficial look-alikes from true cognates (see the quote by Bynon above), and cognates, it is claimed, do not look alike and can only be recognized by means of sound correspondences. Thus, the commonly accepted Indo-European sound cor- respondences show that Armenian erku ‘2’ and Latin duo ‘2’ are cognate, despite their different form, whereas English bad and Farsi bad are not cog- nate, despite their identity of form. Campbell has aptly criticized such views: Identical or very similar sound matchings do not necessarily imply loans or weak evidence of genetic connection. . . . With a time depth approaching that of the Indo-European languages of Europe, the Mayan correspondences are on the whole identical or are the result of single natural and recurrent changes. Proto-Mayan *p,
the over thirty Mayan languages. All other correspondences are very similar. Even English, after its many changes, reflects Proto-Indo-European *r, *l, *m, *n, *s, *w, and *y unchanged, on the whole. A quick survey of once-disputed but now established remote genetic relationships reveals that identical (or very similar) sound correspondences are not that unusual . . . . Therefore, identical correspondences should not be shunned nor too speedily at- tributed to borrowing. While longer separation may provide greater opportunity for unusual and exotic correspondences to develop in cases of distant genetic relation- ship, it is in no way necessary for such developments to have taken place nor for correspondences to be non-identical” (1986: 221–23). Indeed, when one looks at the reconstructions that have been proposed for almost any family, one is able to find modern languages that preserve the pro- posed ancestral forms virtually unchanged. To cite just a few examples, Proto- Indo-European *n¯epot- ‘nephew, son-in-law’ is strikingly similar to modern Rumanian nepot, and Proto-Indo-European *m¯
out change in Latin, Old English, and Sanskrit. Proto-Austronesian *sepat ‘2’ is almost identical with Rukai sepate, and Proto-Autronesian *mat
is identical with Rukai mat s a. Proto-Uralic *tule ‘fire’ is preserved in Finnish tule-, and Proto-Uralic *mo´ska ‘to wash’ differs little from Estonian m˜ oske-. At an even greater time depth, we find that Proto-Nostratic *nato ‘female rela- tion by marriage’ has survived, in Uralic, as Finnish nato ‘husband’s or wife’s sister’ and, in Dravidian, as Malayalam n¯ att¯ un ‘husband’s sister, brother’s wife,’ while Proto-Nostratic *p h alV ‘tooth’ survives in Dravidian as Telugu palu and in Altaic as Ulch palu. At a time depth perhaps even greater than that of Nostratic, we find Proto-Australian *buÑku ‘knee’ preserved in Dyirbal buÑku. In the etymologies we present below, connecting all of the world’s language families, the situation is not all that different from that within the families
288 14. Global Etymologies just discussed. There are, in fact, many examples of sound correspondences of the transparent variety discussed by Campbell. This initial stage of the analysis is necessarily characterized by the identification of easily recognizable similarities, just as was the discovery of Indo-European or any other family. The refinement represented by exotic sound correspondences of the erku– duo variety inevitably awaits a later stage in the analysis—the second stage, which we have called “historical linguistics.” And it is important to recognize that the work of this stage leads almost invariably to a refinement of the
Among the world’s language families, there are no doubt exotic sound correspondences as well that we have not detected. It should be noted, nev- ertheless, that as early as 1986 one of us (Bengtson) proposed some global sound correspondences, and the Russian scholar Sergei Starostin (1991) has re- cently published the most explicit statement of interphyletic sound correspon- dences to date. His brief table of Nostratic–Dene-Caucasian correspondences, though not quite global in scope, accounts for a vast expanse of the linguistic world. Nostratic, for Starostin, includes ten of our 32 taxa (Kartvelian, Indo- European, Uralic, Dravidian, Turkic, Mongolian, Tungus, Korean, Japanese- Ryukyuan, and Eskimo-Aleut), and Dene-Caucasian, for Starostin, includes Caucasian, Sino-Tibetan, Yeniseian, and Na-Dene—to which one may confi- dently add both Basque and Burushaski (Bengtson 1991a,b). Thus, Starostin’s equations account for roughly half of our 32 taxa, as well as the vast majority of the Eurasian land mass. We find nothing in Starostin’s correspondences that is inconsistent with the etymologies proposed below. ON THE LIMITS OF THE COMPARATIVE METHOD It has recently been widely asserted that the comparative method in linguis- tics produces reliable results only for the past 5,000–10,000 years. According to Kaufman (1990: 23), “A temporal ceiling of 7,000 to 8,000 years is inherent in the methods of comparative linguistic reconstruction. We can recover ge- netic relationships that are that old, but probably no earlier than that. The methods possibly will be expanded, but for the moment we have to operate within that limit in drawing inferences.” Similar statements from a host of other scholars are given in Chapter 11, where such beliefs are identified as the central myth of historical linguistics (Chapter 13 further analyzes such myths). The origin of this myth, we believe, is an attempt by Indo-Europeanists to “explain” why Indo-European has no known genetic connections—in our view yet another myth. The fact that Indo-European is intimately connected with numerous other families has been demonstrated beyond a reasonable doubt by the Russian Nostraticists (Illich-Svitych 1971–84), a demonstration that is complemented and extended by Greenberg (to appear). 14. Global Etymologies 289
We have shown that in numerous cases sounds (particularly stable ones like nasal consonants and liquids)—and even entire words—have persisted over time spans greater than 8,000 years virtually unchanged. This raises the question why these evidently quite stable sounds must suddenly change beyond recognition, or disappear entirely, beyond the supposedly insuperable threshold of 10,000 years. If we can use modern languages to reconstruct proto-languages that existed at least 6,000–8,000 years ago (e.g. Proto-Indo- European, Proto-Uralic, Proto-Dravidian, Proto-Austronesian), why cannot such earlier languages themselves be compared (as in fact we will do) in order to discern still earlier groupings? Would it not be one of the more remarkable coincidences in the history of science if Indo-European, the family in terms of which comparative linguistics was discovered, turned out to define the tempo- ral limit of comparative linguistics as well? That there is no such coincidence is amply demonstrated in the etymologies we give below. We feel it is time for linguists to stop selling the comparative method short and to apply it consistently to the world’s linguistic taxa, without preconception. The present chapter represents a step in this direction, an initial step that shows that all of the world’s populations are linguistically connected. The culmination of these efforts will be a comprehensive subgrouping of this single linguistic family. BAD SEMANTICS Another criticism of global etymologies in particular, and of long-range comparison in general, is that such liberties are taken with semantic change that literally anything can be connected with anything else, and it is certainly true that many global etymologies proposed over the years have been semanti- cally unconvincing. But for just that reason we have constrained the semantic variation of each etymology very tightly, and few of the semantic connections we propose would raise an eyebrow if encountered in any of the standard ety- mological dictionaries. They are in fact semantically more conservative than many proposed connections in Pokorny (1959), the standard Indo-European etymological dictionary. Whatever damage this often alleged defect may have done to earlier programs of long-range comparison, we believe that it does not affect the etymologies presented below. ERRORS IN THE DATA Another often-cited criticism of long-range comparison is the presence of errors in the data, errors that invalidate the overall hypothesis. This is a specious argument, for it ignores both common sense and the standard mea- sures of statistical significance. Genetic classification is not analogous to a mathematical proof, wherein one false step undermines the complete demon- 290 14. Global Etymologies stration. Rather, the cumulative weight of all the evidence completely swamps the effects of whatever random errors may be scattered through the work. As Greenberg has often stressed—and has in fact shown in his work—multilateral comparison yields valid genetic classifications even from decidedly degenerate data. An example was Greenberg’s classification of Australian languages in 1953, using little more than the vocabularies published by E. M. Curr in 1886–87. The notion that data must be pristine and copious flies in the face of commonly accepted historical method. It is all well and good for Kaufman (1990: 18) to demand at least 500 items of basic vocabulary and 100 points of grammar before “serious comparative work” can be carried out, but the fact remains that Indo-Europeanists have classified Lydian as Indo-European, without dissent, on the basis of a handful of words, as noted by Greenberg (1990: 10). Similarly, David Payne (1991: 362) reports that “all that remains of the [Shebayo] language is a vocabulary list of fifteen words collected at the end of the 17th century. . . . Despite the paucity of data from this language, it is quite clear that it is Arawakan.” Historians and historical linguists—not to mention paleontologists working from handfuls of bashed fossils—use what- ever material is available; they do not demand that the evidence be complete or immaculate. DISTRIBUTIONAL DIFFERENCES It is often alleged that one can find anything in linguistic data if one looks for it hard enough. Thus the global etymologies we present below are a tribute more to our industry and enterprise than to real genetic connections. Such a view is widespread among linguists who have never actually compared large numbers of languages (or language families), but those of us who have done this kind of work know the reverse to be true. “Wanting” to find something is of very little help if it is not there. Greenberg (1987) points out that the Amerind family has two general words for females, tuna ‘girl’ and kuna ‘woman.’ Both roots are abundantly attested throughout North and South America, and both are found in all eleven branches of the Amerind family. What is interesting about their distribution, however, is that whereas kuna is widely attested in the Old World, as we show in etymology 11 below, we have found no trace of tuna in the Old World. If it were really so easy to find anything one looks for, why did we fail to find tuna in the roughly 4,500 Old World languages, when it is so readily observed in the approximately 500 New World languages? The evolutionary analysis provides a simple and natural explanation: when the Amerind forebears first entered the New World they brought with them the word kuna ‘woman,’ and only later did they invent the word tuna ‘girl.’ That there is no trace of tuna ‘girl’ in the Old World is because it never existed there.
14. Global Etymologies 291
GLOBAL ETYMOLOGIES For each etymology, in what follows, we present a phonetic and semantic gloss, 3
have not attempted to present a unified phonetic transcription for all sources, we have adjusted certain transcriptions from time to time to avoid potential ambiguity. In the first etymology (but not elsewhere) yod has been normalized to j in all citations. Ejectives have been normalized to p’, t’, k’, etc.; V represents a vowel of indeterminate timbre; ˘ı is used for the Old Church Slavic soft sign and ˘ u for the hard sign; and ∼ separates alternative forms. In the two interrogative etymologies (10, 17), interrogative and relative uses are not distinguished (‘who?’ as in “Who is that man?” vs. ‘who’ in “The man who came to dinner.”). The intimate connection between the two is well known and uncontroversial. Most of the cited forms are, however, true interrogatives. The source of the information for each family represented in a given entry is indicated by an abbreviation in brackets at the end of the entry. The number following the abbreviation is either the etymology number in the original source (if there is one) or the page number there. Since the existence of these roots as characteristic features of the language families cited has already been established by other scholars, and is not for the most part in question, we do not give the complete documentation for each family, limiting ourselves in most instances to an indication of the range of semantic and phonological variation within the family. The reader who wishes to see every relevant form for a given family should consult the sources cited. For Amerind, however, we give extensive citations, in order to counterbalance the fallacious criticism that has been directed at Greenberg’s work. Parts of etymologies that are problematic, by dint of either phonetic or semantic divergence, or by restricted distribution, are preceded by a question mark. The lack of a semantic gloss following a form means that that form has the same meaning as the preceding form. We make no claim to being the first to discover any of the etymologies listed below. The pioneering work of Trombetti, Swadesh, Greenberg, Illich- Svitych, Dolgopolsky, and Starostin has identified numerous widespread roots. What we have tried to do is to make each etymology more complete and more soundly documented in this incarnation than it may have been in previous ones. With this goal in mind we have weeded out certain families from pre- 3 We do not deal here with reconstruction, and these glosses are intended merely to characterize the most general meaning and phonological shape of each root. Future work on reconstruction will no doubt discover cases where the most widespread meaning or shape was not original.
292 14. Global Etymologies vious proposals, where the root was phonologically or semantically too di- vergent, or too weakly attested, to be convincing. But we believe we have also uncovered some additional etymological connections that had previously gone unnoticed. To a very great extent the recognition of these similarities has been made possible by the lower-level classificatory work of Greenberg Yüklə 356,86 Kb. Dostları ilə paylaş:
|