Africa vs. Americas, Multilateral Comparison

ECOLING at aol.com ECOLING at aol.com
Wed Oct 6 17:34:56 UTC 1999


In comparing Joseph Greenberg's work on
African language classification vs. on
Amerindian language classification,
it is of course important to note differences, partly in order to
discover whatever limitations there may be on the use of the
Multilateral Comparison technique.
Does its undeniable usefulness in the history of linguistics
tail off to very little after some approximate time depth?
Does its undeniable usefulness become much less under
certain kinds of social or linguistic conditions?

Herbert Stahlke is among those who have pointed out that
one must include in the techniques available to linguists
also those techniques which permit one to discover new
hypotheses for later more detailed examination.
Multilateral comparison has always been such a technique,
for hundreds of years
(and I hasten to add, though it should not be necessary,
that it long predated Greenberg and is independent of
any particular application of it, by Greenberg or by anyone else).
One cannot demand at the beginning of new hypotheses
that they arrive fully-formed from the head of Zeus,
complete with proofs to the most exacting standards of
long-established disciplines.

(As Thomas Kuhn says, most
of the work of most scientists consists in "mopping-up" small
details, new ideas are usually attacked, anomalies are not even seen,
are rarely collected to help open doors to new syntheses.)

So empirical studies of Greenberg's application of this technique
in his work on African language classification and in his work on
Amerind language classification are very important.
In this, we can treat the African vs. American as a variable,
and consider the different outcomes (of course trying to factor
out any other variables which are not essential to the technique
of Multilateral Comparison itself, but which are "accidental"
in the philosophical sense, accidents of how Greenberg may
have applied the technique in one vs. in the other instance.

John E. McLaughlin has taken a sceptical approach towards this,
claiming that the two situations are "apples" vs. "oranges".
(I attach below only the crucial parts of McLaughlin's most
recent posting in reply to Stahlke.)
I believe such an approach is not in the interests of science as a whole,
because it essentially tells us not to study a body of data available
to us.  It is also impossible for anyone to be so omniscient as to
know in advance that nothing fruitful can be derived.
I believe McLaughlin's urging us not to even study this question
is similar to or influenced by the general rejection of Greenberg's
work, and wish to remind readers that there was a similar reaction
at first to Greenberg's African language classifications.
(Which does NOT imply anything at all concerning the quality or
success of Greenberg's work on Amerind neither that it is of
the same quality or success, nor that it is not.)

Rather, the point of a study of the differences between Multilateral
Comparison applied to the African and the American contexts
is precisely to discover the empirical facts, whether the technique
works similarly or differently in the two contexts, with as many
specific detailed answers to this as possible.

We can imagine that there is a different time depth,
and I for one think this is a major variable, and agree with
McLaughlin and with many others on this:

[I have deleted the word "comparative" from the following
quotation, because Greenberg's method is not the comparative method.
But McLaughlin's use of it I think reflects in some way the general failure
of historical linguists to appreciate the real necessity for
both discovery procedures and testing procedures --
one should not confuse Multilateral Comparison with
much more detailed Comparative-Historical techniques.]

[JM]
>You have also hit upon the main problem with G's work--
>he IS trying to extend the [...] method [...] back to between 12 and 40k

It is not at all certain that the time depths are so different,
if one takes Africa as a whole, and considers substrate languages
or families (at least this is what I understand from some of Stahlke's
commentaries, and he should be a good source for this).
Nevertheless, within certain SUBSETS of the African materials,
the time depths will of course be less, as also within certain subsets
of the American materials.  Perhaps the technique of Multilateral
Comparison will be useful within such subsets, but not between them?
It should be clear that the issue is much, much more complicated than
any global a priori judgment that the time depth in the Americas is
uniformly greater
(that is not exactly what McLaughlin says, but I think it underlies
the tone of his suggestion that the method is worthless in the Americas).

Stahlke wrote:

> McLaughlin writes:

> Hmmmm.
> That's the difference between the Americas and Africa.  Africa's had a stable
> indigenous population.  The Americas haven't.  Indeed, it's quite possible
> that northwestern North America has been the site of many groups of people
> from Asia, speaking different languages, landing on the shores of or walking
> across the "bridge" to a New World.

> >>>>>>>>>>

Stahlke responds:

> This apparent difference is deceiving.  The Khoi-San languages, with or
> without the Tanzanian pair, represent a clearly distinct group probably
> originating in southern Africa.  Although all but substratal information on
> pre-Bantu pygmy languages has disappeared, and the substratal information
> isn't any better than in most other parts of the world (worse, in fact), they
> must have represented at least one ancient language family that has
> disappeared.  Beyond those, the major migrations appear to have east to west
> (most of Niger-Congo) and north to south (Cushitic, Nilotic, and Bantu, in
> that order).  Nilo-Saharan, if Songhay belongs in it, may represent a central
> Sahara to Great Lakes migration.  Cushitic, as a branch of Afro-Asiatic,
> represents either a group that originated somewhere along the Red Sea or a
> migration from the Arabian Peninsula.  A-A probably is NE African in origin.
> All of these represent time depths of rather less than 40k:  N-C and A-A in
> 10k-15k range, and Nilo-Saharan and older.  At these time depths, it's hard
> to make a case that extra-continent vs. intra-continental orign makes much of
> a difference.

McLaughlin then responded as follows.
My comments follow about why McLaughlin's are not to the point.

[JM]
>You're missing the point.  You're describing a mixing and matching WITHIN
>the confines of native Africa.  But still within Africa.  You haven't
>described a single event here of an EXTERNAL group coming into Africa and
>affecting the mix of languages.  In the Americas, we're talking about just
>the opposite.  We're discussing groups from northeast Asia coming INTO
>America from outside a number of times overland during the life of the
>Bering bridge and by sea afterwards.  These immigrants would be bringing
>languages from OUTSIDE the Americas, not just moving around within the same
>group of languages as you're describing for Africa (and as also happened
>within the American continents).  This could have happened many, many times,
>although G's "Amerind" implies ONE migration by land at least 12k (but more
>likely much earlier) ago, one migration (possibly by sea) for Na-Dene, and
>one migration by sea for Eskimo-Aleut.

>Africa and the Americas are apples and oranges as far as linguistic history
>and population history is concerned.

***

Therefore it seems to be implied that we should not study the application
of Multilateral Comparison to the American context?

Why not?

McLaughlin's comments only imply that the Multilateral Comparison technique
may very possibly be unable to handle the relations of all of the languages
of the Americas, presumably on the grounds (I am attempting to fill out
what I think is McLaughlin's intent here, common sense with which I agree)
that there may be many migrations and the time depth between back to a common
origin in Asia, if there were such, would be so great as to be beyond the
reach of the technique, in general.

But that in no way implies that Multilateral Comparison technique
is not useful.  It may turn out (from future hindsight) to have been useful
only within certain subgroups of Amerind languages.
These might reflect the groups about which Greenberg felt more secure,
but not his highest-level classifications within Amerind.
After all, the results of Multilateral Comparison may equally be taken
to indicate the greatest probability of non-relation among those groups
linked only at the highest level of classification.  This statement,
while true, is often surprising to both supporters and detractors of
the technique.  They simply had not thought of it from that perspective.
Or there might be surprises, of other kinds.

[For a general explanation of Multilateral Comparison,
please see my earlier message, slightly revised, which
courtesy of Pat Ryan is now posted at his web site,
accessible from the bottom of his main page
<http://www.geocities.com/Athens/Forum/2803/index.html>
or more specifically from
<http://www.geocities.com/Athens/Forum/2803/comment-MultilateralComparison.htm
>
Please suggest any further revisions or improvements to
that explanation, I want it to be objective and balanced.
In agreeing with Pat Ryan to place it there, I insisted that it not
be taken as comment pro or con any other positions, and that
its scope not be exaggerated.  It is not about Greenberg,
it is about the technique of Multilateral Comparison.]

McLaughlin I think also misses the point of Stahlke's commentary.
That is that the time depths may be just as great in Africa,
when we consider the groupings which cannot be linked,
and the substrates, etc.  Why should it matter that for the Americas
we think in terms of migrations, whether across the Bering land
bridge or along older seacoasts, whereas for Africa the physical
distances and barriers are less.  Should that make ANY technique
applicable or not applicable?  Or should each technique merely give
slightly different results in different types of context?

The point of studying the application of the technique of Multilateral
Comparison to both Africa and the Americas is precisely to get a
better idea of what the differences might be.  The results should be
empirical findings, not a priori judgements in advance.
Saying the two are "apples and oranges" is to say
we should not and cannot consider them at the same time.
That is to block an entire domain of science, and to deny the use of
one of the most potent tools of science, the use of a variable or variables
(here Africa vs. the Americas, refined into more detailed variables)
to discover the effect of those variables on some outcomes
(here the results of using the technique of Multilateral Comparison).

I believe an underlying psychological reason for not treating the
African and the American applications of Multilateral Comparison
together is simply the great discomfort most comparative-historical
linguists have with techniques which are used to discover plausible
hypotheses for future work.  They are simply not comfortable at
the edges of new knowledge, because it cannot have the certainty
of the most exacting techniques applied in long-established fields.
There is obviously no criticism here of those more detailed techniques,
it is simply inappropriate to insist on them at the edges of new knowledge.
It manifests an inflexibility to be unable to adapt one's choice of
techniques
to the differences between long-established fields and the edges of
significantly new knowledge not limited to filling in details.
No one person may be comfortable working in both kinds of domains,
but that does not entitle either personality for criticising those
who prefer the other type of domain.
And, fortunately, there is a continuum of intermediate types of work,
at neither of the two extremes.

Each of us should work where we are most comfortable,
because we will do better work there.
Each of us should be supportive of those who work in
a domain requiring different sets of techniques
from those in our own preferred domain,
and should keep the channels open for
flow of significant information between domains of different kinds.

That means those who prefer to work in domains where the
more detailed comparative-historical method is applicable should make the
results of our studies available in forms which those working
in other domains can use without having to be specialists in our own
particular domain.  (And of course should not make claims to greater
certainty than we have a good basis for doing.)
[This is currently sometimes the case, often not.]

It also means those who prefer to work in domains where
Multilateral Comparison is applicable should make the
results of our studies available in forms which those working
in other domains can use without having to be specialists in our own
particular domain.    (And of course should not make claims to greater
certainty than we have a good basis for doing.)
[This is currently sometimes the case, often not.]

No one set of techniques has an exclusive right to all fields
of linguistics.

For a more succesful future for historical linguistics in general,
as for any other field,
it is essential that we use issues which generate public excitement
and channel that interest in productive ways.
Historical linguists shouldbe at the forefront of actively
exploring correlations between language, ethnicity, blood groups,
dentition types, DNA studies, and all other sources of information
which can give clues to the history of humanity.
That can give a better financial foundation to linguistics,
than can a narrow focus exclusively on filling in details within
accepted frameworks.

Paradoxically, public interest is also a very good way,
perhaps the best way, to get more respect and support precisely for
the most advanced and highly precise techniques of
Comparative-Historical techniques.
An attitude of generously providing the results of
the best Comparative-Historical techniques to all comers,
even to those exploring hypotheses which are not favored by
those more narrowly focused, is the best way of spreading
education in the value of those precise techniques.

The reaction to Greenberg's book on languages of the Americas
SHOULD HAVE BEEN from the very beginning
"I am a specialist in X, and can correct some of his data.
Let's see what differnence that would make in his conclusions."
Instead it was mostly "there are errors, so the work is worthless".

To those who have said Greenberg's judgements are purely personal and
not replicable, I can only reply, I had no difficulty selecting from
his data that which I considered more closely vs. less closely similar,
even if I might not have done it exactly the way he would have,
and exploring whether using only the more highly similar data had
any significant impact on his results.
I had no difficulty taking some of the lists of corrections (after they
were published much later by various specialists), and examining
some of Greenberg's results to estimate whether particular corrections
would make any significant difference.
(In neither case did I think the changed data would dictate
significantly different results, though now many years later,
with many more corrections available, I would like to have the time
to examine this again.)

I think the objection that Greenberg's work is not objectively replicable
is really less what the objectors think it is, and mostly simply their
discomfort with exploring a field where they cannot use their most
favored techniques, and where results have a much lower
level of security of being right than they are accustomed to.
But I have also learned from later exchanges that trained linguists
do have different judgments of what constitutes closer similarity,
and therefore now strongly advocate explicit studies of the technique
of judging similarity.

Personally, I am comfortable working both in filling in details
in given and accepted frameworks, and in exploring data sets which
may suggest changing frameworks, or in which we cannot have
an exact and detailed answer.  Not everyone is comfortable
across such divides.  That's OK.  But those of different preferences
should not be critizing each other, they should be cooperating
productively.

Sincerely,
Lloyd Anderson
Ecological Linguistics



More information about the Indo-european mailing list