phonetic resemblances etc.

[Larry Trask]

----------------------------Original message----------------------------
Alexis Manaster Ramer writes:

[snip quote of me claiming that comparative linguistics is hard]

>  While I respect Larry and we usually agree about a lot of stuff, this is
>  just plain wrong, and I think it is a cop-out typical of the people who
>  like to criticize work on language classification but who rarely do it
>  themselves and who also seem to be reluctant to discussing specific
>  examples such as those I have repeatedly cited.

[snip]

>  What now follows is a crystal-clear example not only of an argument which
>  is not too large but also of one which I have repeatedly cited as showing
>  that Larry (and Rich and Sally) are quite wrong in their claims about how
>  language relatedness actually gets established, perhaps because, again, it
>  is not a topic they work on very much. .

>  GODDARD'S (1979) ENTIRE SET OF COMPARISONS AND ENTIRE ARGUMENT FOR THE
>  "COMECRUDAN" LANGUAGE FAMILY (COMPOSED OF MAMULIQUE, GARZA, COMECRUDO),
>  ACCEPTED BY CAMPBELL (1996 and passim).
>
>
>            Mamulique   Garza      Comecrudo
>
>  sun       atl        ai         al
>  moon      kan        an         eskan
>  water     aha(?)     axe        apanekla
>  road      --         aie        aaul
>  man       (kessem)   knarxe     na
>  woman     kem        kem        kem
>  sky       --         apiero     apel
>
>          Goddard's entire argument for this relationship follows:
>
>          The case for relating Garza and Comecrudo seems strong; the
>  agreement in the words for 'sun' and 'road' is particularly striking and
>  shows a consistent phonological pattern suggestive of an historical sound
>  law.  The sparse Mamulique data compare well with Garza and Comecrudo as
>  far as they go.
>
>  I cannot emphasize this too strongly.  This is the ENTIRE argument, given
>  by Goddard and fully endorsed by Campbell, two of the most vociferous
>  advocates of the position that linguistic relatedness demands
>  morphological and not just lexical comparisons and that correspondences
>  and not just phonetic similarities are necessary.

OK; this is interesting.  I'll put my cards on the table: no way on earth do
these data constitute proof of genetic relatedness among these three
languages, or even anything close to it.  At the *very best*, these data
suggest that there might be something worth pursuing -- assuming there exist
further data with which to do the pursuing.

Goddard and Campbell have a big advantage over me here: they are both
specialists in North American languages.  Consequently, they are able to draw
upon a vast amount of background knowledge which is not available to me:
where these languages were spoken, how reliable the sources are, what the
local patterns of borrowing are like, and so on.  When talking to one
another, they can take this background knowledge for granted, just as I can
do when I'm talking to my fellow specialists.  But an outsider, as I am here,
needs more.

Six words?  Just six words?  That's it?

To begin with, how can we even be sure that the cited words belong to the
languages they're assigned to?  Is that certain?  Is it established, for
example, that the Mamulique words were obtained from a native speaker, and
not from a Comecrudo speaker who was bilingual in Mamulique?

Second, these are presumably the best data available for supporting a
relationship.  Yet, oddly, while the other words are only vaguely similar, at
best, the word for 'woman' is identical in all three languages.  Very
strange.  How can we be sure that this is not merely a loan word, from
whatever source, which has passed into all three languages?  Are there
grounds for rejecting that suggestion?  If so, what are they?

Third, the words for 'moon' suggest a loss of initial /k/ in Garza only.  But
the words for 'man' suggest a loss of initial /k/ in Comecrudo only.  Yet the
words for 'woman' suggest a loss of initial /k/ in no language at all.  And
this is supposed to point to implicit sound laws.

I could go on, but that's enough.  If Goddard and Campbell are impressed by
these sparse data, then they must have access to background knowledge denied
to me.  I am not impressed at all, and I cannot regard these data as evidence
of anything at all, beyond a suggestion that a closer look at these languages
might well be justified -- assuming further data are available.

If somebody sent me comparable data in support of a claim that, say, Basque,
Zulu and Ainu were related, I probably wouldn't even bother to reply.

In case I have failed to make myself clear, let me reiterate: there is no
evidence here for any interesting conclusion at all.  The probability is high
that we are looking at nothing more than a few chance resemblances plus the
odd loan word.

Something more might be decidable if we had more data, but it seems we don't.
   Now, most often, when I see stuff in this vein, I am deeply suspicious of
the bona fides of the presenters, because I have plenty of experience showing
that zealous lumpers select and edit their data in unscrupulous ways.  In
this case, I have no such suspicions: I have the greatest respect for both
Goddard and Campbell, and in fact I know Campbell personally.  That
eliminates one big potential worry.  But the fact remains: on the basis of
these data, the probability that these three languages are genetically
related is only mildly greater than chance.

Finally, before anything gets personal, let me add that I also have great
respect for that prodigious polymath Alexis, but this time I think he is
wrong to present these sparse data as evidence for anything.


Larry Trask
COGS
University of Sussex
Brighton BN1 9QH
UK

larryt at cogs.susx.ac.uk