Pre-Basque phonology (fwd)

[Larry Trask]

On Mon, 20 Sep 1999, Roslyn M. Frank wrote:

> [LT]

>> Fourth, we have a dictum in comparative reconstruction.  If variety A
>> has a contrast which is absent from related variety B, then, unless
>> there are very good reasons for doing something else, we reconstruct the
>> contrast for the common ancestor, and conclude that the contrast has
>> been lost in B.  Since the northern varieties of Basque have an /h/-zero
>> contrast, absent in the south, we therefore prefer to reconstruct the
>> contrast for the common ancestor, and to assume that the southern
>> varieties have lost it.

> The question is determining the rules for discovering what a
> "contrast" might be. That is one of the points I have been trying to
> make in my postings. For instance, as you well know, /h/ is
> considered a suprasegmental feature in many environments. You
> yourself have stated this.

No, I haven't.  I have pointed out that Michelena reached the conclusion
that the historical Basque aspiration was, at least in most cases, of
suprasegmental origin.  That doesn't mean that the aspiration is a
suprasegmental in the varieties retaining it today, and in fact it
isn't.

> Secondly, it is also a well known fact that in Euskera vowels are
> often separated by an intrusive consonant and not always the same
> one.

Correct.  Basque doesn't like hiatus, and it employs various strategies
to eliminate hiatus whenever this arises.  One of these strategies, in
the varieties retaining the aspiration, is to insert /h/ between the
vowels, but only when those vowels are in the first two syllables, since
the aspiration cannot occur later than the onset of the second syllable.
Other varieties sometimes insert any of /b d g r/.

> For example, we have the root-stem /ao/
> which also appears as /aho/, /abo/ apo/ and /ago/.

Not sure about ?<apo>, which I've never seen and can't find in any
dictionary, but the others are genuine.

> Indeed, the /abo/
> as in /abokada/ (/abo-ka-da/) "bocanada (Sp.)" or "mouthful", more
> literally, "an action done repeatedly with the mouth, mouthing over
> and over") is commonplace.

Yes, but <abo> itself is well recorded for `mouth', at least in
Bizkaian.

> Are we to assume an original /aho/ which
> was reduced to /ao/ in souther dialects?

On the basis of the evidence, this is clearly the best conclusion.

> Or was the original form
> /ao/ which then under certain conditions was pronounced as /aho/ so
> as to avoid the falling together of the two vowels:

As far back as we can reconstruct, the best choice is an original
*<aho>.  The origin of this is beyond our powers of recovery.

> so that root-stem wasn't in danger of undegoing permanent reduction
> (to */o/) rather than only a momentary one as has occurred in the
> case of some compounds (/aomen/ vs. /omen/). In other words there
> are many examples where there is an intervocalic /h/ whose status is
> unclear.

Agreed.  In particular, we cannot always tell whether a phonetic [h]
should be assigned to Pre-Basque or regarded as a later development --
though in very many cases we *can* make this decision.

> Next, in referring to an /h/-zero contrast, I assume you are
> referring back to your example of the minimal pair in /sei/ and
> /sehi/.

Yes, except that we need not appeal to minimal pairs, which are few.
The point is that the presence or absence of the aspiration is,
generally speaking, totally unpredictable, which is what we mean by
`contrastive'.

[on Basque <sei> `six']

> But I think in another venue you argued that <sei> was a loan word
> in Euskera.

No, I did not.  Quite the contrary: I have several times argued
*against* the proposal, put forward by several other people, that <sei>
is borrowed from Romance.  The problem is the phonology.  The Latin for
`six' was /seks/, and all the descendants of this in Romance varieties
in contact with Basque have a final sibilant, as far as I know, as in
Castilian /seis/ and French /sis/.  Now, when Basque borrows a Latin or
Romance word containing a final sibilant, it *always* renders that
sibilant with some sibilant of its own, without exception.  So, Latin
/seks/ should have produced a Basque *<zetz> or something similar, while
Castilian /seis/ should have produced a Basque *<seits> if borrowed
early or *<seis> if borrowed late.  But the only Basque form recorded
anywhere is <sei>, and hence I conclude that a borrowing from Latin or
Romance is impossible.

> [LT]

>> We don't really need minimal pairs.

> But we do, Larry. You offered one example which I challenged. Indeed, for
> the case to be made, one should be able to identify sets of minimal pairs
> in northern dialects that differ only in the presence/absence of the
> intervocalic /h/ and whose meanings are totally different. The following
> lists do not provide that sort of information.

[snip lists]

Not so, I'm afraid.  Minimal pairs are desirable but not essential.
If you want to argue that the absence of /h/ is original, and that
modern instances of /h/ result from insertion, then you *must* provide
an explicit statement of the circumstances in which the /h/ is inserted.
If you can't do that -- and I don't think you can -- then you cannot
maintain an analysis involving /h/-insertion.

[on the same list]

> But many of these forms are identical in northern and southern
> dialects??!! So what does this prove? What are you trying to argue?

The point is that the occurrence of /h/ in the north is *not
predictable* on the basis of the southern forms, which you propose to
take as conservative.

> In the case of <gei> I believe the northern variants include /gehi/.

Correct.  Basque <gai> ~ <gei> `material, quantity' does have a northern
variant <gehi>.  This raises another interesting issue which I don't
propose to go into here.

[still on that list]

> As I said, these are not examples of minimal pairs. Nor are these
> examples consistent in the sense that each of them has a northern
> and southern variants, e.g., one with /h/ and another without.

*Exactly* the point.  Northern varieties freely contrast the presence of
/h/ with the absence of /h/, in an unpredictable manner, while southern
varieties have only the absence of /h/.  Since we cannot predict where
the /h/s will occur in the north, we must reconstruct proto-forms
containing /h/ in the appropriate places, and recognize loss of /h/ in
the south.

> So that leaves us where we started: there are northern variants that
> have /h/ and southern ones that don't. And...

Indeed.  But the northern varieties are clearly more conservative here.
No other conclusion makes any sense.

[on the need to provide a conditioning factor for the introduction of
/h/, if that analysis is proposed]

> Could one not argue that the conditioning factor was the
> introduction of a consonant, a mechanism intended to keep the two
> vowels of the root-stem from falling together, especially in
> compounds where stress could have led to their reduction. The latter
> would have led to a loss of recognition of the root-stem and hence
> the meaning of the compound; and that could have affected the shape
> of the root-stem itself.  Mechanisms that allow for the maintenance
> of the phonological shape of root-stems would seem to me to be
> particularly important in a language such as Euskera, i.e.,
> typologically speaking.

No; this doesn't work.  It fails to account for the existence of
northern forms like <sai>, <goi>, <soin>, <goiz>, <lai>, <gau>, and so
on.  In your account, these should have acquired an inserted /h/, just
like <nahi>, <ohi>, <ahul>, <ahur>, and so on -- but they didn't.
So an analysis involving /h/-insertion to break up hiatus doesn't work.

> Again if were could come up with a list of minimal pairs in northern
> dialects that differ only in the presence/absence of /h/, we would have a
> very strong case. But we can't.

But we don't need minimal pairs.

How many minimal pairs can you cite in English for the consonant <sh>
(as in `ship' and `fish') and <zh> (as in `measure' and `vision')?
Not very many, right?  Does it follow that these two sounds constitute a
single phoneme?  No, it doesn't, because the distribution of the two
sounds is *unpredictable*.

Same question for the <th> sound of `think' and the <th> sound of
`this'.  And the same answer.

> Didn't they exist?  And if they didn't exist, why not?

Nobody knows how many earlier lexical items have been lost from Basque
without being recorded.  As for the paucity of minimal pairs, well, I
have my own ideas about that, but I don't have the evidence to make a
strong case, and so I'll keep silent.

> Indeed, what I would like to determine is whether there are any
> examples at all of minimal pairs other than that of /sei/ vs. /sehi/
> which we've discussed above?

I can't answer this question without doing a bit of digging.  But bear
in mind that <sei>/<sehi> *does* exist, and that's enough: the
occurrence of /h/ cannot be predicted.

Oh, there is this nice set: <ari> `busy, occupied', <hari> `thread,
fiber', <ahari> `ram'.  What do you make of this set?

Larry Trask
COGS
University of Sussex
Brighton BN1 9QH
UK

larryt at cogs.susx.ac.uk