Basque 'sei'

[Larry Trask]

On Sat, 25 Sep 1999 s455152 at aix1.uottawa.ca wrote:

> While I have the highest respect for professor Trask, I must disagree with
> him on the impossibility of Basque SEI being a Romance loanword; see
> below.

[snip my doubts about a Romance origin for Basque <sei> `six']

> According to Gerhard Rohlfs (LE GASCON: ETUDES DE PHILOLOGIE
> PYRENEENNE, Second Edition (1970), p.145) final -s shifts to -j when
> followed by a voiced plosive, a liquid or a nasal: this is
> especially frequent in Eastern Gascon, and Rolfs quotes such
> examples as ERAY DUOY RODOS "the two wheels" (instead of ERAS DUOS
> RODOS): this -s to -j shift is also found in Bearnese, which is
> close enough to the Basque country. Now, according to the ALF, map
> 1235, the word for "six" in the area is found under various forms,
> /ses/ and /seis/ being the most common. One would therefore expect
> forms such as /sej/, /seij/ in front of voiced plosives, liquids and
> nasals, and if Basque had borrowed such a form (perhaps from an
> expression such as "six times", which I would expect to be something
> like /sej betses/ in Gascon), the attested phonological form (SEI)
> would be *EXACTLY* what one would expect.

This is very interesting, and I thank Stephane for the information.
Oddly, I've read a good deal of Rohlfs' work on Basque, but I can't
recall that he ever raises this possibility in his Vasconist work.

> There may be good reason not to believe SEI to be a Romance
> loanword, but on the basis of the above, it is plain that its
> phonological form is not one of them.

Well, no.  I'm afraid this conclusion is too strong.

First, the suggested scenario is rather convoluted, and it requires a
very particular sequence of events: the borrowing, not of the counting
numeral, as is most usual in borrowing numeral-names, but of a specific
expression which is nowhere attested in Basque, followed by an
extraction within Basque of a conditioned variant of the numeral-name
not found in isolation in the source language.  Not impossible, I guess,
but not exactly a straightforward analysis, either.

Second, there's the problem of the sibilant.  Basque has two contrasting
voiceless alveolar sibilants: a laminal, notated <z>, and an apical,
notated <s>.  Now, in early loans from Latin, Latin /s/ is almost always
rendered as the laminal <z>.  The same is true at all periods of loans
from Gallo-Romance: the laminal /s/ of Occitan and French is rather
consistently rendered by the Basque laminal <z>, not by the apical <s>.
In contrast, the apical /s/ of Ibero-Romance is equally consistently
rendered by the Basque apical <s>.

We therefore have a problem: a Gascon */sej-X/ should have come into
Basque as, at best, *<zei>, and not *<sei>.  But no such form as the
expected *<zei> is recorded anywhere: <sei> is the only form attested in
Basque.

Nor can we appeal to contamination from adjacent numeral-names: the
number-names for `5, 6, 7, 8, 9' in Basque are <bortz>, <sei>, <zazpi>,
<zortzi>, <bederatzi> -- all with laminals except <sei>.  (The other
number-names contain no sibilants.)

Hence the scenario described above further requires that, in this case,
instead of the normal treatment of Occitan /s/ as Basque <z>, /s/ was
exceptionally borrowed as <s>.

I don't suppose any of this is impossible, but, taken together, it
appears to me to constitute a very dubious case -- especially since all
the other Basque names of lower numerals are native.  I still think the
Basque-Romance resemblance here is best regarded as being on a par with
cases like English `much' and Spanish <mucho> `much', English `bad' and
Persian <bad> `bad', and English `him' and Laz (Kartvelian) <him> `him'.

Larry Trask
COGS
University of Sussex
Brighton BN1 9QH
UK

larryt at cogs.susx.ac.uk