LINGUIST List 3.350

Sun 19 Apr 1992

Disc: Greenberg and Mass Comparison

Editor for this issue: <>


  1. Bill Poser, Greenberg and Mass Comparison
  2. , Re: 3.346 Greenberg and Mass Comparison; Endangered Languages

Message 1: Greenberg and Mass Comparison

Date: Sun, 19 Apr 92 00:25:24 -0Greenberg and Mass Comparison
From: Bill Poser <>
Subject: Greenberg and Mass Comparison

Alexis Manaster-Ramer writes:

 I do not really think that we should pay much attention to
 the rejection of Greenberg's classification of the
 languages of Papua and vicinity by people who do not
 themselves do comparative linguistics. And even people
 that do typically work on just one small family and so
 their judgement may not be relevant.

The point isn't that the near-universal rejection of
Greenberg's Indo-Pacific hypothesis in and of itself shows that
it is wrong. The point is this: Greenberg and his fans want to use
his past batting average as evidence that his methods and judgment are
reliable and hence should be accepted in the American case too.
They cite the African classification as having been successful.
(In any case, as I have pointed out, it wasn't strikingly original
and how successful it was is disputed.) But if they want to use
as the criterion for success the acceptance of Greenberg's past
attempts at classification, then Indo-Pacific must enter the
picture, and judging by the SAME criteria as the African case,
namely whether it is accepted by specialists, Indo-Pacific was a

In any case, it simply isn't true that the people who have
rejected Greenberg's Indo-Pacific hypothesis don't do
comparative linguistics. If Alexis will consult Foley's book, to
which I referred, he will discover considerable discussion of
the relationship of the non-Austronesian languages to each other
along with Foley's tentative evidence for a relationship between
the North-East Highlands family and Australian. Another critic
of Indo-Pacific is Donald Laycock, who has devoted considerable
effort not only to the description of Papuan languages but to
issues of classification. Here's what he (Laycock 1976;57) has
to say:

 To date, it can safely be said that there is no real
 evidence to link the NAN [Non-Austronesian - WJP]
 languages of New Guinea with any other linguistic group,
 though some slight indications of possibilities are
 discussed elsewhere in this volume (see 2.16.). In
 particular, Greenberg's `Indo-Pacific Hypothesis' (1971)
 which would interrelate `the bulk of non-Austronesian
 languages of Oceania from the Andaman Islands on the west
 in the Bay of Bengal to Tasmania in the Southeast' is not
 only far from proven, but also based on inadequate and
 insufficiently-analysed data (for example, comparisons are
 all too frequently made only of items within large groups
 of languages --- such as the Trans-New-Guinea Phylum ---
 that are already known to be related, so that there is
 little support for the wider relationships postulated.)

Both Foley and Laycock are familiar with diverse languages
of New Guinea, and both are also acquainted with Australian
languages. The idea that they are narrow specialists whose
comparative experience is limited to closely related
languages doesn't fit the facts. But I would also ask why
we should believe that people who haven't made a point of
claiming to demonstrate remote relationships should be
regarded as incompetant to evaluate such proposals? Surely
anyone trained in the comparative method should be able to
examine proposals of relationship. And surely it is those
most familiar with the languages who are best able to judge
whether the evidence adduced is correct. And doesn't M-R's
proposed criterion for evaluators contain a dangerous bias,
whereby only those convinced that they themselves have
demonstrated distant relationships are able to judge
others' proposals? We don't consider astrology valid
because astrologers say it is.

In the case of the Americas this argument, which Greenberg
too has made, is all the more peculiar because the great
Americanists are predominantly people with a wide
acquaintance with American languages and broad knowledge of
the comparative literature. Alexis may only have been
acquainted with Uto-Aztecan until his most recent work, but
leading Americanists, such as William Bright, Lyle
Campbell, Wallace Chafe, Ives Goddard, Victor Golla, Mary
Haas, Kenneth Hale, William Jacobsen, Terrence Kaufman,
Dale Kinkade, Michael Krauss, Marianne Mithun, Patricia
Shaw, and others are known for their broad knowledge.
Greenberg's critics include such distinguished Americanists
as Wallace Chafe, a specialist in Iroquoian and Caddoan
(distantly, if at all, related), Ives Goddard, the dean of
Algonquianists and a person of extremely broad knowledge of
the languages of the Americas, and Lyle Campbell, who has
not only done extensive comparative work but has personally
done field work on Mayan, Xincan, Lencan, Uto-Aztecan,
Mixe-Zoqean, Otomanguean, Jicaquean, Misumalpman, Chibchan,
and Wakashan languages. Eric Hamp, another critic, is
perhaps better known as an Indo-Europeanist, but has done
substantial comparative work on American languages as well
as field work on Quileute. People like these, I submit,
with their broad knowledge of the languages of the
Americas, their experience in doing solid historical
linguistics, and their detailed knowledge of particular
languages and families, are vastly better positioned than
someone like Greenberg who has done little or no
traditional comparative work and has no demonstrated
knowledge of ANY American language. As Leonard Bloomfield
is said to have remarked, "If you want to compare two
languages, it helps to know one of them."

Alexis writes:

 It does appear that Greenberg is right in claiming that
 historically much of the work on classifying the languages
 of the world proceeded in much the way that he has used.

Sorry, but this is precisely one of the points that Lyle Campbell
and I have disputed in the postings to which M-R is responding,
and our BLS and Pittsburgh papers. It simply isn't true, and M-R
doesn't advance the argument by repeating this unsubstantiated
assertion. Greenberg hasn't been able to produce valid examples
in support of this assertion that Indo-European was done using
methods like his. There isn't a hint of Greenbergian superficial
lexical comparison in most of the work. Where such methods were
used, as by Sir William Jones and Franz Bopp, they made horrible
errors, as did other practictioners of Greenbergian methods, such
as Pliny Earle Chase and Paul Rivet, as pointed out in my
previous posting. I could go into greater detail, but long-winded
though I may be, I think that posting our entire BLS paper (not
to mention its 150pp big brother) would be excessive. It is true
that some valid work was done in this way, e.g. the Powell
classification of North American languages, but the batting
average of the technique is poor except when, as in the case of
the Powell classification, the approach is used conservatively
and the languages grouped together are closely related.

I'd be interested in knowing what exactly Alexis is referring to
as Dolgopolsky's improved method. Is this the stuff cryptically
discussed in his 1964 _Voprosy Jazykoznanija_ paper?

It is true that if all languages are related then showing
that Greenberg's methods relate Finnish and Mayan doesn't
show that the method is unreliable. It does, however, show
that his methods do not discriminate between the remotest
relationships and much more plausible ones. I agree with
Alexis that his Tonkawa example provides a more direct test
of Greenberg's subgrouping since Greenberg claims that
Amerind and Na-Dene are separate branches of Proto-World. I
note that _Language in the Americas_ contains no evidence
whatever for Greenberg's proposed subgrouping of Amerind.
For his highest-level groupings, like such as Southern
Amerind, he gives no argument at all, but even for his 11
lower-level subgroups, for which he believes he has argued,
there is no actual evidence. He merely lists equations that
he asserts are restricted to these subgroups. This of
course does not establish anything - to present a prima
facie case for his subgrouping he would have to show that
the languages he groups together show greater similarity to
each other than to others, which he does not even attempt to
do. So at best Greenberg has established that the "Amerind"
languages are related to each other - his proposals about
the subgrouping of Amerind and for that matter, that
Amerind is a distinct branch of Proto-World, remain bald

It is too bad that some participants in this debate apparently
do not use email. Joe Greenberg apparently does use email but
not much. If Merritt Ruhlen wants to I would be happy to set up
an account for him.


Laycock, Donald A. (1976)
``History of Papuan Linguistic Research,''
in Steven A. Wurm (ed.)
New Guinea Area Languages and Language Study. Vol. 1: Papuan Languages
and the New Guinea Linguistic Scene.
Sydney: Department of Linguistics, Research School of Pacific Studies,
The Australian National University.
(= Pacific Linguistics Series C, no. 38).
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 2: Re: 3.346 Greenberg and Mass Comparison; Endangered Languages

Date: Sat, 18 Apr 1992 10:33 MSTRe: 3.346 Greenberg and Mass Comparison; Endangered Languages
Subject: Re: 3.346 Greenberg and Mass Comparison; Endangered Languages

Concerning Greenberg and Mass comparison.
I agree with Alexis Manaster-Ramer that if you work on just one small
family, your judgement on Greenberg's method might not be relevant, and
that comparing, say Finnish with Amerind does not prove the method does
whatever you would like it to do. However, some Amerindianists have
worked on a large number of families, and sure enough, like Greenberg,
we see all kinds of uncanny resemblances. I've been fortunate to be
able to do fieldwork on all three of Greenberg's families (Eskimo:
Siberian Yupik; Na-Dene: Navajo, Sarcee, Apache; Amerind: lakota and
Quechua) and so I might be in a better than average position as far
as feelings (I prefer to say feelings rather than views) concerning
mass comparison. There's obviously no way to disprove monogenesis, and
many of us feel that all Native American languages might ultimately be
relatable. The point is, has Greenberg proved that they are classifiable
only in his way? Is the classification replicable by someone else who is
completely indepnde? Do we accept the classification simply because no
one else is going to be working on the same data as hard as Greenberg did?
One finds a lot of similarities that cross-cut Greenberg's three-way
classification, for example a suffix -ni in Eskimo, a stem ni in Navajo
and Quechua, all meaning 'to say'. Is this Proto-World, Proto-Eskimo-
Nadene-Amerind, or does it not count? Some grammtical correspondences
 are more interesting: for example both Lakota and Navajo have only
one prefix thais of a VC shape, u(+nasal hook)k- in Lakota, and iid-
iNavajo. They are not too close phonetically but the intersting thing
is that they are the only prefixes with that shape and that they mean the
same thing, i.e. 1st person dualy question is then: how do we evaluate
this? How do we decide that this is/is not evidence for a Siouan/Atahbaskan
subgrouping? It seem to be less likely to be due to chance than the famous
1st and 2nd person n-/m- business.Of cour, this is not the only
similarity between Siouan and Athapaskan I've found, and I suppose the
only reason why I found them is that no other people (since Mary Haas) have
been looking in detail at Athapaskan and SiAgain, I'd appreciate
comments on this.
Willem J. de Reuse
Department of Anthropology
University of Arizona
Tucson, AZ 857
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue