LINGUIST List 2.82

Tuesday, 19 Mar 1991

Disc: Language Families & Unstressed Vowels

Editor for this issue: <>

Directory

1. "Michael Kac", Language families
2. "Daniel L Everett", Re: Language Families
3. Osamu Fujimura, Re: Vowels
4. , Response: Stress
5. , Response: Stress and Chinese
6. , Response: Stress and Icelandic
7. John Coleman, RE: Reduced Vowels

Manaster-Ramer's comments about Nostratic prompt the following: Up until very
recently, about all I knew about the Nostratic hypothesis was that it had been
proposed by two (presumably crazy) Russians and that no one took it seriously.
>From somewhere I got the impression that the reason no one took it seriously
was that it wasn't done, as Manaster-Ramer says, 'by the book'. But this ap-
pears to be wrong (you can trust Alexis on points like this): it's all done
using classical philological methods. So why the near conspiracy of silence
about it? (People at Watkins' Presidential Address at the 1988 LSA will recall
a brief comment about 'the Nostratsophere', which about captures how American
linguists are conditioned to think about all this.)

A personal sidelight: A couple of years ago, after the Michigan conference on
Language and Prehistory, the Classics department here hosted a Russian linguist
named Vladimir Orel who presented evidence that Etruscan is a North Caucasian
language. When he announced that this was what he was going to argue for, my
immediate reaction was to expect some kind of wild-eyed strange stuff, but a-
gain it was strictly by the book: there's not a lot of Etruscan, but there
are etymologies for a good many words, and they're pretty detailed (at least
from the standpoint of someone who is admittedly not a historical linguist).

I will also confess to being taken with Alexis's twist on the Greenberg stuff
vs. Nostratic: if the latter is in fact wrong, doesn't that indeed have dis-
turbing implications?

It is very unfortunate indeed that Scientific American and Atlantic
Monthly have chosen to continue the unscientific drivel on language
families that was initiated in US News & World Report.

Why can't these reporters do any honest investigation before they 
print things like this (cf. US News' equally silly coverage of
the Gulf War)?? 

However, I realize that many nonspecialists think that some of this
stuff sounds plausible. So, perhaps this would be a good place to 
debate the issue.

I can only talk about Greenberg's work on South American languages
(which is part of the overall story that keeps getting revived). For
every South American language that I have done field work on (about 15),
Greenberg's data are either almost completely bad, or they are good but
what he says is nothing new or not at all conclusive. Cf. My article in
the forthcoming book from Stanford University Press on "Language and
Prehistory in the Americas". There is no conclusive evidence whatsoever
for a single "Amerind" grouping.

There is not, nor has there ever been, a mainstream historical linguist
who believes that the depth of reconstruction reported on in these three
magazines is at all plausible. 

However, since I am primarily a theoretical/descriptive linguist, I ask
that historical linguists reading this newsgroup join the discussion 
and put this nonissue to rest.

Dan Everett

About English reduced vowels, Marian Macchi has a very interesting finding
though preliminary and unreported. Using a pair of words like 'adopt'
vs. 'adept', she found that speakers of the same dialect, the same generation
split into two groups in terms of the direction of formant shift when asked
to repeat the words; one toward a unified schwa, one into two distinct
reduced vowels. Her e-mail address is mjmbellcore.com.
Osamu Fujimura, OSU, Speech & Hearing.

Another example of a language with a regular stress pattern
(i.e., a nonlexical one) which skips over reduced vowels (in
this case schwa) is Mandarin Chinese, I believe.

I find it difficult to believe that anybody would claim that]
English only has one unstressed vowel in ANY of its many varieties.
Even if you lack the Rosa's roses contrast, as many people no
doubt do, is there anybody who does have some contrast between
at least two of the last vowels of sofa, city, and yellow. 
Also, in initial position, there are likely to be many more
than just one vowel, e.g., about vs. iguana vs. obey, even
for speakers who lack additional contrasts such as request vs.
ricotta. Rich Rhodes (rrhodescogsci.berkeley.edu) has
a paper on unstressed vowels in English and, while I think
he is wrong in some cases about what is unstressed and what
is non-primary-stressed, it is quite enlightening and will
certainly help dispel the myth under discussion here.

One more language with a regular pattern of stress and several
(three, I think) reduced vowels is Icelandic. This in fact
the basis of Vennemann's wonderful argument against classical
generative phonology. Since stress is predictable it would
be assigned by a late rule, but the morpheme structure conditions
would have to duplicate the stress rule in order to capture the
restricted set of vowels found in the unstressed position in
polysyllabic morphemes. (N.B. I am not sure if the "reduced"
vowels are phonetically reduced in any real sense, but the
set of vowels in unstressed position is much more restricted
than in stressed. If anybody needs details, they can be
rounded up easily enough).

So dkbrentariucdavis.edu is "searching for any language where ... 
the stress will not occur on the predicted syllable unless the 
syllable contains a 'full' vowel. That is to say, not a reduced vowels".

`Reduced' is not a well-defined concept in phonetics or phonology. 
In English phonology, "reduced" means among other things, central,
short and unstressed. So are they searching for a language in 
which the stress cannot occur on unstressed syllables? This
is a circular definition, so surely any language will do.
For example, stress placement in English is regular, but stress
may not fall on reduced vowels (since they are exponents of
absence of stress). 

--- John Coleman