Campbell's def. of "language"

[Herb Stahlke]

[ moderator re-formatted ]

I too found Bartholomew's review engaging and fair.  One can, however, be a bit
more critical about some of Campbell's statements on the probability of distant
genetic groupings proving valid.  Such estimates are guesses that depend on
one's confidence in the guesser.  While I regard Campbell's scholarship in
Meso-American and related languages highly, I find he has a way of putting
forth his opinion on untestable matters with more authority than is
appropriate.  In his Historical Linguistics (MIT 1999), p. 163, for example, he
says in a table that Africa has "c. 20+ families.  I've been in African
linguistics since the mid-60s, and I've never confronted this number, but he
provides no justification for it.  On pp. 164-5, his table 6.2 lists "Some of
the better-known language families" of the world and rates the present state of
comparative studies in each.  Of the 29 families listed, he rates present
knowledge as good or better for 13, of which 8 are Western Hemisphere groups.
No African group, including Bantu, is rated better than "moderate," and he
leaves out groups like Jukunoid, Akan, Chari, Gbe, Central Khoisan, etc., where
the state of work is quite high.  He also rates Sino-Tibetan as "much needed"
without noting the extensive work going on both on the larger family and on
some of its subgroups.  A significant number of those he rates as good or
better are groups he himself has worked on.  I suspect the table reflects his
own knowledge and experience as much as it reflects scholarly reality.

By the way, unlike Larry, I think it was, I've found Campbell's intro very well
done, in spite of, perhaps also because of, matters where I differ strongly
with him.  I'm planning on using it next semester in an intro course for
graduate students who have had a phonetics course and a first phonology course
as prerequisites.  I expect them to be able to handle Campbell's challenges
well.

Herb Stahlke

Lloyd Anderson writes:
>>> <ECOLING at aol.com> 10/07 12:13 PM >>>

Nevertheless, Campbell is an extremely competent scholar.
For several proposed distant genetic groupings,
he provides probability estimates that the groupings
will ultimately prove valid, and separately from that,
a confidence level (based on how adequate he feels the data is
on which he bases his estimate of ultimate relationship).
For example, for Maya - Huave - Mixe/Zoquean,
he estimates the probability of ultimate relationship at 30%,
and gives a 25% confidence level for that estimate.