isolates

[Sarah G. Thomason]

   Alexander Vovin's confidence that some of the isolates listed
by Larry Trask have been de-isolated by general consensus among
specialists is probably too sanguine.  A recent review in LANGUAGE,
for instance, expressed doubts about the evidence for connecting
Japanese with anything else (I don't remember the details of the
reviewer's arguments; the review appeared three or four years ago;
it's possible that the reviewer's focus was on Japanese + Austronesian
rather than on Japanese + Korean).
 
   And I have my own concerns about evidence linking Ainu with
anything else, to the extent that the evidence relies on the
reconstructions in Vovin's book on Ainu (which contains proposals
like Proto-Ainu *hd- for a large correspondence set in which most
dialects have w- and the others have segments which could easily be
reflexes of *w-; Vovin declines to reconstruct *w- here because
there are few words with this correspondence set, whereas both *h and
*d are reconstructible).
 
   Vovin's reasoning about when an isolate is not an isolate seems
odd: if a language has no relatives, it is an isolate, regardless
of whether it once had relatives.  More importantly, from a
methodological viewpoint, Vovin asserts that cases like Ket
and Yukaghir show us "how isolates come into being".  But what evidence
have we, for any language that we all agree is an isolate and that
has no attested former relatives, that it used to have some
relatives?  It's easy enough to imagine a situation in which no
split will occur, ever: just situate your hypothetical language in
a remote mountain valley (say), in a small area that only supports
a small cohesive population -- a single speech community in which (roughly)
everyone talks to everyone else -- and leave everyone there permanently.
No split.  A belief that there are no real-life cases of this general
sort (it doesn't have to be a mountain valley, etc.) is a  matter of
faith, not science.
 
   Vovin is right, of course, in saying that Japanese is not an isolate
even if it has no demonstrable relatives, if it is really a small
family of very close-related languages.  But then Proto-Japanese (still
on the hypothesis that it has no established relatives) would be/have
been an isolate, so it would still go into the total.
 
   -- Sally Thomason
      sally at isp.pitt.edu