Ancient Basque Stops

[José Ignacio Hualde]

[ Moderator's note:
  The following post is in response to recent discussions between Larry Trask
  and Roz Frank.  Because it is strictly concerned with the interpretation of
  the Basque data, it is not strictly relevant to the Indo-European list, but
  I post it out of courtesy to Mr. Hualde.

  I have set the Reply-To: address to Mr. Hualde rather than to the list.  I
  will not post follow-ups to the Indo-European list; those wishing to pursue
  this particular question further should move the discussion to another venue.
  --rma ]

Good morning everyone. I hope I am allowed to post this message, even
though I am not a member of this list. Dale Hartkemeyer has forwarded me
two or three recent messages regarding the ancient Basque plosives and I
would like to be given the opportunity to clarify my position, since it has
been the subject of some exchanges.
The problem of the ancient Basque plosives, as stated by Martinet and
others before him, can be summarized as follows: " How come  Basque, which
has a robust opposition between voiceless and voiced oral stops in
intervocalic position, shows a much weaker contrast in word-initial
position?" From Martinet's structuralist standpoint this is a problem
because the word-initial position is supposed to be the one where the
greatest number of contrasts is found in any language. To solve this
problem, Martinet made up a story that has to do with an ancient contrast
between fortis and lenis stops which was later somehow replaced by the
modern voiced/voiceless contrast. Michelena adopts a version of this
hypothesis, which has become the standard account.
My view is different. Basque differs from most languages presenting
assimilation in voice across morpheme- and word-boundaries in that it is
the morpheme- or word-initial consonant that assimilates to the preceding
morpheme- or word-final one, instead of the other way round. So in Basque
/s+d/ becomes [st], etc., whereas in, say, Spanish, /s+d/ becomes [zd].
E.g. the initial /d/ of <dator> "s/he is coming" becomes /t/ in [estator]
"s/he is not coming", [menditi(k)tator] "s/he is coming from the mountain",
etc. Or, to give you another example, whereas <buru> "head"starts with a
/b/, the same morpheme starts with /p/ in, say, [ajspuru] "stone head".
Nowadays, there is little chance that Basque speakers will identify initial
[p] and [b] as allophonic variants,  bacause of (a) their familiarity with
Spanish or French and (b) because the assimilation rule tends to apply only
in restricted phrasal contexts. BUT assuming that this assimilation applied
more frequently in the past (as  Michelena also assumes) it stands to
reason that if <tator> and <dator>, <buru> and <puru>, and so on for lots
of plosive-initial words, are variants of the same word in different
phonological context, this would inevitably lead towards a merger of the
voiced and voiceless oral stops in morpheme- and word-initial position
(where the alternation is found) but not morpheme-internally. End of the
story. The more complicated Martinet-Michelena hypothesis (which in
addition requires an unexplained transformation from ancient to modern
Basque) is, in my view, simply not needed and has no serious evidence in
its favor. Thanks for allowing me to clarify my position.