LINGUIST List 6.505

Wed 05 Apr 1995

Disc: Comparative method

Editor for this issue: <>

Directory

1. Jacques Guy, Labial-to-dental shift in a child's language (was: comparative method!)
2. Jacques Guy, Phonetic changes and the comparative method

In a recent ... can you call that article?... article, then, on the
comparative method, I mentioned the bilabial-to-dental shift
independently innovated in Greek, and by number of languages of Vanuatu.

 My wife says she had developed, when very young, a personal language,
of which she could only remember two words: [Zita] "bucket", and
[akagunga] "hat" or perhaps "my hat". Her mother remembers a little
more, and was a willing informant the other day, shortly after she
identified Fijian "vuaka" with our cat's [wako], which inspired my
send-up of Proto-World. Here is the entire corpus I elicited. Not much
to work from, but still interesting.

Key: IPA, with:
[S] = long s (esh) i.e. English sh, French ch
[Z] = the voiced equivalent, z with a curlicue. i.e. French j
[~] = nasalization (IPE tilde over preceding letter)
[E] = (epsilon), i.e. open e
[O] = open o

 French English
[nene] me'me' nanny, grandma
[nana~] maman mummy
[nana~ janin] maman Jeannine
 (her mother Jeannine)
[tata kanE~] papa ca^lin (?) cuddly(?) daddy
 (this is the term she used to refer to her father, Rene')
[tete tesa] pe'pe' ??? grandpa ???
 (her grandfather Gaston)
[Sita] seau bucket
[akakwa] ga^teau cake
[akagunga] chapeau hat
[kaza] casserole saucepan
[SetakOnSa] c'est pas it isn't so!
 comme c,a!

Her mother says that she had a word for rake (ra^teau), but she could
not remember it, only that like [Sita], [akakwa] and [kaza] it had
nothing with the French word. [kaza] referred specifically to the
saucepan used for washing her hair.

You can clearly see that the rest is derived from French, with a
general labial-to-dental shift. There is one instance [l] --) [n],
quite common in Vanuatu (Shark Bay /tEn/ ( *tolu "three"). [s] --)
[S] (in [SetakOnSa] "c'est pas comme c,a!") we find in Portuguese
and Limousin.

My wife has often asked me if I knew of a language like "hers", that is,
the part which is not clearly derived from French. She sorts of believes
in metempsychosis and all that. She was sorely disappointed when I
read out some Mayan, Nahuatl and Quechua to her. Because she looks very
much like the famous Mayan jade "Maize God" she thought she was an
American Indian in a former life. "They like sound nothing like *my*
language!". I pointed out to her that there were hundreds of languages
in North America spoken by people with a distinct resemblance with the
"Maize God", from Geronimo to Sitting Bull, if photographic memory
serves, so all hope is not lost yet.

Nota. Ross Clark rightly riles me in a personal e-mail for confusing
Fijian "levu" with "balavu". He is right, "long" is "balavu", not
"levu" (big). Come to think of it.... Swahili -refu "long, tall"
might have confused my grey cells. Swahili -refu, Fijian levu and
balavu, there is another brick towards the Proto-World ziggurat,
a.k.a. the pre-tower of Babel!

I wrote this over the week-end, before I read Lloyd Anderson's long
communication (this morning). So, it does not answer the questions
therein satisfactorily. However, it tells how and what made me come
to my jaundiced view of the comparative method. I take this view:
case not proven. The onus of demonstrating that something is, or
works, at any rate, has always been on the proponents of the theory.
I will properly address the (interesting and well-put) questions
of Lloyd Anderson another time, leisure permitting (I am no longer
into linguistics, you know). Here we go...

I was browsing desultorily through Ruhlen's "On the Origin of
Languages", noticing at every page how his method or, rather, lack of
it, would ensure that one would never realize that Sakao and Shark Bay,
spoken a mile apart, were closely related, when came back to my mind
the question of natural, unidirectional phonetic changes. Consider these
cognates:

 four stone his/her/its eye(s) my eye(s) rat
 Sakao ijED jED mDan mDEG A
Shark Bay TaR TaR nati:n nataN i:f

Key: E = IPA (epsilon) T = IPA <theta> R = trilled r
 D = IPA (delta) G = IPA <gamma> N = nasal velar
 B = IPA (beta)
  = schwa, lips spread
 A = back rounded A (very close to Hungarian "a")

It is only when you have records of additional, phonologically
conservative languages that you stand a chance of working out what
happened:

 four stone his/her/its eye(s) my eye(s)
 Tolomako Bati Batu natana nataku
 Tsureviu Bati Batu matana mataku

and eventually that Sakao /A/ and Shark Bay /i:f/ must have descended
from *aGw(aeo) otherwise unattested. As for the four words above, very
obviously, we should reconstruct: *vati "four", *vatu "stone", *mata-
"eye", *-na "his/her/its", *-ku "my". Elementary, my dear Watson.

But isn't the _dissimilation_ a --) i observed in Shark Bay rather, er,
odd, my dear Holmes? (*a goes to /i/ in Shark Bay only before *C(aeo)
where C = any consonant) And, as we gather more related languages, and
as we encounter Vao, with interlabials, and Lolovuevue, with
labio-velars, we see these regular correspondences (in which B =
bilabial, D = dental, I = interlabial, W = labio-velar, co-articulated):

 Tolomako Tsureviu Vao Lolovuevue
*B(aei) D(aei) B(aei) I(aei) B(aei)
*B(ou) B(ou) B(aei) B(ou) B(ou)
*D D D D D
*? B(aei) B(aei) B(aei) W(aei)

Dutifully, we reconstruct *? as a labio-velar series. All is nice,
clean, and clear: Tolomako and Vao (along with Sakao and Shark Bay) are
the odd men out, the "dental" family as it were. Lolovuevue is the most
conservative. Labiovelars went to bilabials everywhere, except in
Lolovuevue where they stayed so. Bilabials went to interlabials or
dentals in the "dental" family before front vowels and stayed bilabials
everywhere else. It's all very natural, and we contentedly note that it
is all quite similar to Indo-European, with Greek, P-Celtic and
Q-Celtic, so there you are: another lovely universal, what were you
worrying about?

How come, then, that Tolomako and Tsureviu share some 95% cognates?
This is off the top of my head, I haven't checked a cognate count, and I
don't need to. Once upon a time, when I was fluent in Tolomako, and
merrily chatting away with people of Tolomako village (on the eastern
shore of Big Bay) I suddenly realized, after perhaps ten minutes, that
they were often using bilabials where the Tolomako I spoke had dentals.
I soon found the explanation: some sixty years before, the Catholic
priest in charge of the Tolomako village area (I forget his name),
finding the region too insalubrious, had his whole flock leave and
settle in Port-Olry, another Catholic mission, on the north-eastern
shore of Espiritu Santo, in Sakao-speaking country. "La nature ayant
horreur du vide", Tsureviu speakers moved into the now vacant Tolomako
country. Written records I have seen show that the Tolomako language of
60 years ago was identical with today's. So, the "dentalization" of
Tolomako was not borrowed from its new neighbours, Sakao and Shark-Bay.
At any rate, Tolomako, Sakao and Shark Bay are so phonologically
different that they are totally mutually unintelligible. Nor are
they lexically close, Tolomako having about 40% cognates with Sakao
and Shark Bay, Sakao perhaps 44% with Shark Bay (*). How could Tolomako speakers
have turned Tsureviu /pei/ "water" into /tei/, but kept the /p/ of
Tsureviu /pei/ "good" (probably cognate with Maori /pai/), turned
Tsureviu /mata/ "eye" into /nata/ but left /mata/ "snake" unchanged
(Lolovuevue /Nmata/, Bau Fijian /Nata/). The corresponding Sakao words
would have provided no basis for the correct decision: Sakao has /ro/
for "water", /BOGBOG/ for "good", /mDa ~ mDE ~ nAD/ for "eye", /jo/ for
"snake"!

At any rate, when I first attempted to reconstruct Proto-Vanuatu on the
evidence I had, I reconstructed *interlabials. Sure, I was a bit worried
that my *interlabials never occurred before *back vowels. I was very
happy when I came across the languages of Oba, also known as Lepers'
Island. Then, lo and behold! I could reconstruct labio-velars instead, a
much more respectable set of consonants: we have them in PIE, don't we?
But... er, they never occurred before *back vowels either, except in
onomatopoeic words.

Yes, I know the "correct" answer: Proto-Austronesian, reconstructed
from a much greater number of languages, has neither labio-velars nor
interlabials. To me, that is begging the question.

And that is why I don't buy this business of classifying languages on
shared phonological innovations any more. One thing leading to another,
I started thinking. And that is why I no longer buy this business of
natural, unidirectional sound changes any more either.

Either that, or I'll be lumping Greek with the dental family of Vanuatu
because the shift p -) t is quite exceptional, and the probability that
it was innovated independently rather than inherited is negligible. And
if you won't follow me on these grounds, you have no choice but to admit:

1. either that the labial to dental shift is common,

2. or that the observed frequencies of phonetic changes are not a sound
 ground for classification and reconstruction.

Footnote.

(*) Why do I write "Sakao perhaps 44% with Shark Bay"? Don't I know?
I don't. I just looked it up Tryon's "New Hebrides Languages". He's got
two Shark Bays there, which he calls Shark Bay I and Shark Bay II.
I took the first list, he took the second. Looking at them now, I am sure
that they are from the same language, same dialect (the phonology *is*
weird, the morphology is not transparent, hence the probable discrepancies).
I did most of the cognate recognition for Espiritu Santo, perhaps all
(I don't remember, it's 25 years ago), but that was long before I had
worked out their diachronic phonologies. Result: Sakao has 44.2%
with Shark Bay I (231 items compared) and 41.4% with Shark Bay II
(237 items compared).