Pre-Basque phonology (fwd)

[Larry Trask]

On Mon, 6 Sep 1999, Roz Frank wrote:

[on the reconstruction of Pre-Basque]

> As I recall in the exchanges that took place a couple of years ago
> in the issues of _Mother Tongue_ , there was a rather lively debate
> about several aspects of the phonological reconstruction. I refer
> specifically to the differences of opinion expressed by Bill
> Jacobsen, Jose Ignacio Hualde and yourself. Again I don't have the
> journal issues in front of me, but I do remember that there were
> several points of dispute which I don't think were over "phonetic
> details" but rather more substantive issues, e.g., the presence or
> absence of /m/ and/or /k/ in Pre-Basque as well as a couple of other
> items that escape me right now.

Hualde has since developed his position in an article.  In fact, he does
not challenge Michelena's reconstructed phoneme system at all.  Rather,
he proposes to assign different phonetic features to the proto-phonemes.
In particular, while he agrees with Michelena that Pre-Basque had no
voicing contrasts in word-initial plosives, he believes that the voicing
of initial plosives was facultative, rather than phonetically
consistent.  Whether you buy this or not (I don't), it has no
consequences for our reconstruction, but only for the subsequent
development of the system.

As for Jacobsen, he expressed skepticism about Michelena's conclusion
that no */m/ could be reconstructed for Pre-Basque, but he has never, so
far as I know, tried to follow this up.  In particular, mhe has never
tried to argue that */m/ must have been present in Pre-Basque after all.
In the absence of any such case, there remains no reason to query
Michelena's conclusion, which remains well substantiated.

> Also, I've always wondered about the p/b, k/g, t/d alternations in
> pre-Basque (and/or their *aspirated counterparts). Perhaps you could
> comment a bit on the distribution of these in modern Basque. It's an
> intriguing problem.

It certainly is, or rather they certainly are, since you've pointed to
two quite different issues.

Most of the voicing alternations are confined to word-formation, where
they are well understood, phonetically motivated, and unproblematic.

But there remains a residue of cases which are more refractory, a fine
example being <ebaki> ~ <epai> `cut'.  These cases are not understood,
but they are not numerous, and at present they constiture no more than
an intriguing but peripheral puzzle.  The instances are too few, too
scattered and too varied to allow us to draw any generalizations.
All we can say at present is that a few words variably exhibit /b/ or
/p/ (virtually all examples involve these two plosives) for reasons that
are not understood.

As for the aspiration, that is one issue on which we still do not have
full agreement.  The matter is much too complex to be discussed in
detail here.  Michelena's conclusion, which I endorse, was that *most*
instances of the aspiration are of suprasegmental origin, possibly
associated with the word-stress at an early stage.  Michelena left open
the possibility that *some* instances of /h/ (though not of the
aspirated plosives), at least in word-initial position, might have
resulted from the lenition of earlier consonants.  There is a tiny
amount of evidence to support this, but not enough to make a persuasive
case.

> [LT]

>> Potentially the most serious problem is the /h/, but I know of no one at
>> present who disputes M's conclusion that *most* instances of /h/ are of
>> suprasegmental origin.  However, it remains possible to disagree about
>> whether *some* /h/s are of segmental origin.  In practice, though, this
>> isn't much of an issue, and we can readily dispose of any difficulties
>> by reconstructing Pre-Basque -- contra M -- with a *phonetic* [h] in our
>> transcriptions, allowing users to draw their own conclusions.

> [RF]
> Examples?

Sure.  For <orri> ~ <horri> `leaf', Michelena reconstructs *<oRi>, with
his fortis rhotic.  If we prefer, we could write *<[h]oRi> instead.
It probably makes little difference, so long as we are consistent.

> [LT]

>> I'll give you one oddity for free: while word-initial */b/ is
>> *exceedingly* common in bisyllabic words, it is all but unknown in
>> monosyllables.  This is curious, and I have no explanation.  It may be
>> no more than an accident of survival.

> Does that mean that you would argue that /be/ "beneath" is really /pe/
> rather than seeing /pe/ as the allophonic representation of /be/?

The independent word is <behe> (two syllables) in the north.  As a
suffix, it is (quite regularly) reduced to <-be>, with a variant <-pe>
arising regularly after a voiceless consonant and then being extended to
other positions.

> Or do you count /behe/ as two syllables?

I certainly do, since it is pronounced as two syllables in the dialects
retaining the aspiration.

> What about /bal/,

I presume you mean <bal> `sheaf (of grain)', `pile (of grain) laid out
on the threshing-floor before threshing'.  There are two problems.
First, the word is about equally recorded as <bal> and as <bala>.
In all likelihood, then, the earlier form is *<bala>, since loss of
final /a/ from nouns is a very frequent phenmenon in Basque, probably
because suffixed <-a> is the omnipresent article, and speakers sometimes
conclude that a final /a/ in a noun is the suffixed article and remove
it.

Second, the word is almost certainly borrowed from Romance.  We have
Gascon <bale`> `threshing-floor', with a derivative <balade> `sheaves of
grain spread out for threshing', and Gascon, Bearnais <bale> `chaff,
husk (of grain)'.  Even Castilian has a derivative <ba'laga> `part of
the threshing-floor where the sheaves are laid out'.

> bart/

Western <bart> `last night' is clearly a reduced form of <barda>,
preserved today in the east.

> bat/

And <bat> `one' is pretty clearly derived from earlier *<bade>.

> /behi/

But <behi> `cow' is two syllables in the aspirating dialects.

> /bein/behin,

And <behin> `once' is also two syllables in the north.  The dialects
which have lost the aspiration have, as a result, acquired a number of
new monosyllables which are still bisyllabic in the north and which were
formerly bisyllabic in the south.  I don't count these as ancient
monosyllables, with good reason.

/beltz/,

Well, <beltz> `black' is a very interesting case.  Native words rarely
end in consonant clusters, and this is just about the only word I can
think of ending in the unusual cluster <-ltz>.  The word is surely built
upon the ancient element *<bel> `dark', not recorded as such but present
in numerous compounds and easily reconstructible.  In all likelihood,
the earlier form was *<beletz>, as proposed by Michelena.  An item
BELEX(-), occasionally BELEXS- or BELS-, is frequent in the Aquitanian
names and appears to represent the same word, with the last (and rarest)
variant seemingly already showing the contraction.  Michelena proposes
that the contraction took place because the adjective was regularly
postposed (as is normal in Basque), and because postposed items in
Basque frequently undergo otherwise irregular reductions.  (By the way,
note also that <beltz> is a moderately frequent element also in medieval
Basque personal names.)

/bihi/,

And <bihi> `grain' also has two syllables.  There is a tiny amount of
evidence pointing to original *<bini>, but not enough to be decisive.

> /bihur/,

And <bihur> `twisted' (and other senses) is also two syllables.

> /bost/,

The word <bortz> ~ <bost> `five' is also very interesting.  It is
certain that eastern <bortz> is more conservative, and that western
<bost> results from a familiar sound change.  But that still leaves
<bortz> among the fifty or so native monosyllables that I mentioned
earlier.  However, this too ends, unusually, in a consonant cluster, and
I myself suspect a lost vowel -- probably *<bortza>, conceivably
*<bor(r)etz>.  Michelena himself hinted at *<bortza>, but never endorsed
this openly.

> etc. It strikes me that the kind of study that Jon is undertaking
> would allow us to determine more accurately the distribution of such
> items.

But not for Pre-Basque, if he insists on taking all recorded Basque
words as his database.

> My impression (and it is merely that) is that in the case of
> relative frequency of monosyllables and bisyllables with
> word-initial /b/ we would find a higher number of bisyllabic roots,
> perhaps 3 to 1.

Much higher.  It's easy to list dozens of native bisyllables with
initial /b/, but almost impossible to find any native monosyllables with
initial /b/.

> However, certainly there are a significant
> number of monosyllabic parent stems also, particularly when they are
> compared to certain other monosyllabic parent stems with word-initial
> consonants.

This remains to be seen, at least for native and ancient words.

Whoops -- the workmen need my office.  Back later.

Larry Trask
COGS
University of Sussex
Brighton BN1 9QH
UK

larryt at cogs.susx.ac.uk