Darwin-L Message Log 5:45 (January 1994)

Academic Discussion on the History and Theory of the Historical Sciences

This is one message from the Archives of Darwin-L (1993–1997), a professional discussion group on the history and theory of the historical sciences.

Note: Additional publications on evolution and the historical sciences by the Darwin-L list owner are available on SSRN.

<5:45>From WILLS@macc.wisc.edu  Sun Jan  9 16:53:37 1994

Date: Sun, 09 Jan 94 16:54 CDT
From: Jeffrey Wills <WILLS@macc.wisc.edu>
Subject: Re: Greenberg and other controversial beliefs
To: darwin-l@ukanaix.cc.ukans.edu

	I want to thank Bob O'Hara for suggesting Sibley as a parallel for
Greenberg. At the heart of the matter, I think, is the same methodological
problem with these "Buckshot Methods": not that the method is controversial
or eccentric, but that there is no method. No method, by which I mean no one
besides these inspired few could reproduce the same results. And yet why are
his "results" of such interest to certain anthropologists and others? In the
case of Greenberg (I cannot speak for Sibley), I think the method works this
	1) you start with the fruits of the discipline's work. You don't
challenge demonstrated connections, but try to move on to new terrain. What G.
observes from past work is that--golly gee gosh--most language families are
	2) Taking this insight of contiguity (never expressed nor perhaps
realized consciously) G. hypothesizes super-families.  How does this
hypthothesis arise? Because of substantial work making lots of little
connections between families? No--that hardly exists. Rather Greenberg's
search for superfamilies is theory-driven rather than data-driven (as Wegener
might have been). Of course, Greenberg is not starting entirely from scratch:
he has Sapir's six superstocks for North America and the undemonstrated
hypothesis of other big groupers. In fact, so far G. has done nothing new--
unless he would actually try to argue/prove his supergroups. This leads to:
	3) A naked theory without data is a public scandal. Lest he
be arrested for indecent exposure (and because he genuinely believes in these
superfamilies) G. looks for cognates in these hypothesized kin. It is as this
stage that the points of Victor Golla and others about probability in sets of
large random data take over. G's method is really one of randomness; with
unlabeled data sets I think he would/could create entirely different families.
In fact Lyle Campbell (Language 64.606-9) elegantly plays G's game ("method"
would be to misunderstand the matter) and finds striking matches of Penutian
(one of G's West Coast assemblages) and Finnish.
	It is at this stage that Greenberg makes his "contribution". But alas,
Campbell documents frequent significant problems with G's data in practically
every category you can name: sound-and-meaning isomorphism, borrowing,
semantics, unmatched segments, onomatopoeia, lack of grammatical similarities,
distributional oddities, repeated cognates, false cognates, erroneous
reconstructions, erroneous morphological analysis and even spurious forms. It's
obvious that G. enjoys matching up words in dictionaries (a harmless sport) but
to pretend he is engaged in linguistics without good knowledge of his data set
is not so harmless.

	To return to my question: why do the results of this non-method method
appeal to others in anthropology or genetics? Because G's "results" are really
nothing more than his assumptions: that people living near each other once
spoke the same language (let's call them Group A) and that  group B next to
Group A must have originally spoken the same language (supergroup AB) etc.
Remember that Greenberg's main thesis in Lang.in Am.(p. 38) is that "all the
languages of the Americas, except those of the Na-Dene and Eskimo-Aleut groups,
fall into a single vast assemblage."  G's "breakthrough" is to demonstrate the
linguistic unity of a large hemispheric group who naturally share overlaps in
culture and genetics. How can you go wrong? In fact, I should have preceded
this message by a sample poll of the non-linguists on Darwin-L: How many of you
think all the languages in the Americas are ultimately related to each other?
Thinking that "Amerindian" is ultimately united is hardly counter-intutitive
because in fact it is the linguistic expression of a non-linguistic assumption
(neighbors are related). In one sense the real achievement of linguistics has
been to identify surprising, non-adjacent language groups (Finno-Ugric, or the
connection of Yurok and Wiyot on the West Coast to the rest of the Algonquin
	In reference to the comments on the profit of controversial theories,
remember that Greenberg's classification in essence has been known since 1956
and has neither won acceptance nor stimulated any significant research.

	Alternative Theories for REALLY BIG LANGUAGE FAMILIES.
	It has been reasonably suggested several times on this list that the
best way to drive out Folly is to ignore it and promote Truth instead. But here
is my dilemma (although other linguists on this list may wish to give me reason
for hope): I don't think we will ever have the data we need to answer these
questions about languages more than a few thousand years before their
documentation. Language mutates at a sufficiently fast rate that there is very
little embedded in it which gives evidence of its origins after a few thousand
	Unlike anthropologists who can hope to discover cultural artifacts or
palaeologists who can hope to discover fossils and use genetic analysis, we
have no reason to think there was any recorded language (and certainly not a
significant amount of it) much earlier than we have found it (dates varying by
region).  I think the SuperGrouper language game is like asking: "What was the
origin and the subsequent historical distribution of marriage rites?" A
fascinating question, which we will never stop asking, for which we have
contemporary cross-cultural evidence galore. We can even think it likely that
all hominids had mating rites or that there was monogenesis or polygenesis of
such rites and we can even hope to discover a few artifacts which we might plug
into these theories, but to graph a full "tree" of marriage rites or language
or other evanescent facets of human culture would need whole new categories of
evidence. Anyone out there with the Great Galactic Sonograph should please
speak up.
	Perhaps I have a bad attitude, but given this data deficit, how would
you suggest we try to advance alternative theories for matters on which we have
no evidence or even reasonable expectation of evidence?
			Melancholy in Madison.

Jeffrey Wills

Your Amazon purchases help support this website. Thank you!

© RJO 1995–2022