LINGUIST List 3.327

Wed 08 Apr 1992

Disc: Greenberg and Mass Comparison

Editor for this issue: <>

Directory

1. Bill Poser, Manaster-Ramer on Greenberg
2. lyle campbell, classification methods, 3-316

Studies of the sorts that Alexis Manaster-Ramer mentions are to
be welcomed, and add to similar studies that have previously been
made. An early one is Callaghan & Miller (1962), which argues
that by Swadesh's criteria English is Macro-Mixtecan. Campbell
(1973) argues that using criteria proposed by Bender, Finnish,
Cakchiquel (a Mayan language), and Quechua are related. Campbell
(1988) argues that by Greenberg's criteria Finnish is Amerind.

Campbell (1988) has also argued, as Alexis does, that Greenberg
did not really follow his own method. However, I am reluctant to
accept Alexis' tentative conclusion that it is only Greenberg's
application of the method rather than the method itself that is
at fault. First, there are plenty of cases in which methods like
Greenberg's have been used and have led to error, and no reason
to believe that the method was improperly applied. For example,
Sir William Jones, whom Greenberg claims as a model, incorrectly
classified Pahlavi, "Afghan" (presumably Dari or Pashto), and
Malay as Semitic, incorrectly classified Tibetan and the rest of
Austronesian as Indo-European, failed to recognize that Malay is
Austronesian, and vehemently denied that Hindi was related to
Sanskrit. Franz Bopp, another of Greenberg's models, thought
that Indo-European, Malayo-Polynesian, and Georgian were related.
(These examples are discussed in Poser & Campbell 1992.) A nice
example of methods indistinguishable from Greenberg's is Chase
(1869), which argues that Yoruba is related to everything from
Egyptian to Chinese to German and thus that all or nearly all
languages are related. In this century Paul Rivet, using methods
just like Greenberg's, arrived at many claims of affiliation that
very few people accept, including claims that Greenberg himself
explicitly rejects (see _Language in the Americas_ p. 107). Now it
could be the case that all of these people too somehow misapplied
the method, but the case that they did so remains to be made, and
we have to wonder how it is that ALL practictioners of
superficial lexical comparison manage to misapply it and why
practictioners of the same method disagree with each other so much.

Second, there is good reason to expect superficial lexical
comparison to fail - we know that it is possible for borrowing to
lead to such similarities, and the probability of
chance resemblance is rather high when the criteria are as lax as
those used by Greenberg and most other practictioners of this
method.

Finally, it isn't clear that Greenberg's African
classification really provides a good precedent. First, there is
the question of how original Greenberg's African classification
was. Greenberg and his fans often talk as if he had started from
scratch, in Africa and in the Americas, but in fact in both cases
there was a signficant amount of previous work. Judging from the
discussion of the history of African classification on pp. 76-85
of Ruhlen (1987), it appears that his classification was heavily
dependent on previous work, especially that of Westermann.
It appears that, as far as higher-level groupings are
concerned, Greenberg's original contribution to African
classification consisted primarily of two things:

(a) the flattening of Afro-Asiatic, with the elimination of the
	grouping of Egyptian, Berber, and Cushitic into the
	Hamitic subgroup of Hamito-Semitic.

(b) the splitting off of Nilo-Saharan from Niger-Congo, which
	Westermann had grouped together.

He also removed various incorrect affiliations that had been
based on non-linguistic or purely typological evidence.

However correct and important these contributions may have been,
neither is an argument for distant genetic affiliation of the
sort that is the subject of the current controversy. Greenberg's
role was actually that of a ``splitter'', not a ``lumper''.
Greenberg's work on the classification of African languages therefore
does not provide any precedent for his claims about American Indian
languages.

Note: Greenberg is also credited with the addition of Chadic to
Afroasiatic. However, although Greenberg did adduce additional
evidence, according to Ruhlen the credit for proposing this
affiliation and adducing the main evidence for it goes to Johannes
Lukas.

Second, there is the question of whether the classification was actually
correct. The word is apparently still not in. Here is what one well
known Africanist has to say (Heine 1992):

	Although Greenberg's work represents considerable progress
	over that of previous writers, it leaves a number of questions
	open. His approach is largely inadequate for the PROOF
	of genetic relationship; it can do little more than offer
	initial hypotheses, to be substantiated by more reliable
	techniques like the comparative method. In a number of
	instances,languages or language groups have been placed
	in a given family solely on the basis of a handful of
	`look-alikes', i.e. morphemes of similar sound shape and
	meaning. The Nilo-Saharan family, in particular, must be
	regarded as a tentative grouping, the genetic unity of which
	remains to be established.

One also has to wonder why Indo-Pacific isn't taken into consideration.
Between his African classification and his American classification,
Greenberg (1971) looked at the Papuan (non-Austronesian) languages of
New Guinea and argued that they (plus Tasmanian) are all related.
This classification appears to be almost universally rejected.
Foley (1986), for example, doesn't even mention it. Why isn't
the failure of Indo-Pacific to gain acceptance as much of a
precedent as Africa?

REFERENCES

Callaghan, Catherine A. & Wick R. Miller (1962)
``Swadesh's Macro Mixtecan hypothesis and English,''
Southwestern Journal of Anthropology
18.278-285.

Campbell, Lyle (1973)
``Distant genetic relationship and the Maya-Chipaya hypothesis,''
Anthropological Linguistics
15.3.113-135.

Campbell, Lyle (1988)
Review of Language in the Americas.
Language
64.591-615.

Chase, Pliny Earle (1869)
``On the comparative etymology of the Yoruba language,''
Transactions of the American Philosophical Society
13.35-68.

Foley, William A. (1986)
The Papuan Languages of New Guinea.
Cambridge: Cambridge University Press.

Greenberg, Joseph H. (1971)
``The Indo-Pacific hypothesis,''
in Thomas F. Sebeok (ed.)
Current Trends in Linguistics, vol. 8.
The Hague: Mouton.

Heine, Bernd (1992)
``African Languages''
In William Bright (ed.)
International Encyclopedia of Linguistics.
Oxford: Oxford University Press. Vol. 1, pp.31-36.

Poser, William J. & Lyle Campbell (1992)
``Indo-European practice and historical methodology,''
to appear in the Proceedings of the 18th Annual Meeting of the
Berkeley Linguistics Society.

Ruhlen, Merrit (1987)
A Guide to the World's Languages: Volume I: Classification.
Stanford: Stanford University Press.

 It was encouraging to see Alexis Manaster Ramer's discovery that Greenberg
had not actually applied his method of mass (multilateral) comparison to ar-
rive at the classification of Tonkawa and Zuni. That Greenberg did not apply
his method to arrive at most of his classification has been pointed out
repeatedly in other contexts (references below). Thus it is by now well-known
that (1) Greenberg followed the lead of previous classifiers in both his
classification of African languages and especially in that of the American
Indian languages (relying extensively on Sapir for North America and Rivet for
South America to the extent of including classifications of theirs now known
to be erroneous); and, (2) Greenberg had already fixed most of the classifica-
tion before he began to assemble the data, without applying his much-praised
method.
 Thus for Africa, Greenberg (1955:106) said, in effect, that the classifica-
tion was arrived at before he had formulated the method of mass comparison:
"I was only dimly aware of the significance of mass comparison at the time of
writing." For the Americas, his classification was already codified before
(Greenberg 1953, 1956~[1960|) date of 'awareness', and an inspection of the
notebooks on deposit in the Stanford University Library upon which his 1987
book is supposed to be based shows that the languages and data were arranged
from the outset in a way which reflects this preconceived classification.
That the classification was already decided before the data were assembled
and without the benefit of any application of the method is illustrated in
the following quote from Greenberg (1953:283) himself:
 Even cursory investigation of the celebrated "disputed" cases,
 such as Athabaskan-Tlingit-Haida and Algonkin-Wiyot-York, indicate
 ~[sic| that these relationships are not very distant ones and, indeed,
 are evdent on inspection. Even the much larger Macro-Penutian
 grouping seems well within the bounds of what can be accepted
 without more elaborate investigation and marshaling of supporting
 evidence.
(For additional examples, evidence, and discussion, see Adelaar 1989,
Campbell 1988, and a host of reviews of Greenberg 1987.)
 Even Greenberg's view of mass comparision itself has changed over
time. In Greenberg 1949 and 1957 he advocated rather standard notions of the
comparative method, though with too much dedication to lexical evidence alone
for my taste. Thus he said (Greenberg 1949:79, same in 1955:1):
 Such lexical resemblances, whether in roots or inflectional
 morphemes are sine qua non for the establishment of all relations
 of a more obvious type ~[here speaking of Sir William Jones and the
 establishment of Indo-European|. Only when these have been
 established may one use more subtle methods such as those
 employed by Sapir in North America.
By 1987, however, lexical resemblances are argued to be better than the
standard comparative method and sufficient for the classification of American
Indian languages in general -- this aspect of Greenberg's method has been
almost universally criticized by specialists.
 In short, the references below and many others not listed here argue
persuasively that (1) Greenberg's methods don't work in any case, and (2) he
did not even apply them himself for the more telling parts of his classifi-
cation.
Adelaar, W.F.H. 1989. Review of Language in the Americas, by Joseph H.
 Greenberg. Lingua 78.249-55.
Campbell, Lyle. 1988. Review of Language in the Americas, by J. Greenberg.
 Language 64.591-615.
In press. The classification of American Indian languages & its implications
 for the earliest peopling of the Americas. The classification & prehistory
 of American Indian languages. A. Taylor, ed. Stanford U. Press.
Greenberg, Joseph H. 1949. Studies in African linguistic classification: 1.
 The Niger-Congo family. Southwestern Journal of Anthropology 5.79-100.
___. 1953. Historical linguistics and unwritten languages. Anthropology
 today, ed. by A.L. Kroeber, 265-86. U of Chicago Press.
___. 1957. Genetic relationship among languages. Essays in linguistics,
 chapter 3, 35-45. U of Chicago Press.
___. 1960~[1956|. The general classification of Central & South American
 languages. Men & cultures: selected papers of the 5th Internat. Cong. of
 Anthro. & Ethno. Sciences, 1956, ed. by Anthony Wallace, 791-4. U of
 Philadelphia Press.
___. 1987. Language in the Americas. Sanford U Press.
Poser, William. In press. The Salinan data in Greenberg 1987 ~[approximate
 title|. IJAL.

 In conclusion, Alexis, I second Poser's recent posting that not only were
the methods not applied (as we all seem to agree), they do not provide re-
liable outcomes even if one takes the trouble to try to apply them, as seen
in the many erroneous cases in the history of linguistics which have employed
methods comparable to Greenberg's (cited in Posner's posting).
 Lyle Campbell