IE-Semitic connections

Larry Trask larryt at
Wed Feb 3 09:21:03 UTC 1999

On Fri, 29 Jan 1999, Glen Gordon wrote:

> In retrospect, I forgot about Basque "zazpi". Hmm, but I wonder what the
> mainstream theory is on that word. Larry Trask mentioned that Latin "s"
> becomes "z" in Basque borrowings. Latin, at least Vulgar Latin, as far
> as I am aware, reduced /m/ before vowel to a nasal vowel. I can't see
> how "zazpi" can be related to a t-less Semitic form because of the
> second -z- which has to be a soft representation of a previous /t/ (what
> else could it possibly be??) Hence zazpi might actually come from a late
> version of Latin septem which perhaps was pronounced /sept(s)@~/ at the
> time (@~ = nasal schwa).

> To be honest it probably can't be explained as a Classical Latin
> borrowing (cf. Lat apta- > hauta- "choose") but couldn't this be a late
> Latin borrowing? It's definitively not from a t-less Semitic form. Maybe
> an _m-less_ one at most and then we have to explain why Semitic *b
> becomes Basque /p/!

Basque <z> represents laminal [s], and Latin /s/ was indeed regularly
borrowed as Basque <z>.

A Latin or Romance origin for <zazpi> seems extremely unlikely, since
all the lower Basque numerals appear to be native, and there is no known
western Romance form of `seven' that could serve as a phonologically
plausible source for <zazpi>.

Apart from the name for `nine', which is partly analyzable, the Basque
number names up to ten are opaque and unanalyzable.  There is no
standard etymology for <zazpi>, but there is a proposal on the table,
due to Luis Michelena.

The Basque for `five' is <bortz>, and for `two' <biga>.  The
instrumental suffix is <-z>.  Michelena proposed that <zazpi> derives
from *<bortz> + <-z> + <biga> `two with five' (Basque is head-final).

To make this work, we need several things.

First, <bortz> must derive from earlier *<bortza>, to account for the
<a>, since, in word-formation, the vowel that is invariably inserted to
break up impermissible consonant clusters is <e>, not <a>.  But this is
plausible, since word-final consonant clusters in Basque are rare and
anomalous, and are known in a few cases to derive from a lost final
vowel, especially <a>, as in western <bart> `last night', from <barda>,
preserved in the east.

Second, <-z> must have been once a comitative suffix, rather than an
instrumental suffix.  This conclusion is actually supported by a modest
amount of evidence.  Moreover, it is well known that comitative markers
frequently acquire instrumental senses, and the modern Basque comitative
endings are of transparent and unquestionably late origin.

Third, <biga> must have been reduced to <bi> in this formation.  But
this is totally unproblematic: we *know* that <biga> is always reduced
to <bi> when in final position in a compound or phrase.

Fourth, the initial <bor-> must have been lost.  This is not regular,
and it constitutes the weakest link in the argument.

Fifth, if <bor-> were indeed lost, then the affricate <tz>, finding
itself in word-initial position, would absolutely have to change to
fricative <z>.  This is an ironclad rule of Basque phonology.

Finally, the original <b> absolutely must be devoiced to <p> when
preceded by a voiceless sibilant like <z>: this is another ironclad rule
of Basque phonology.

So, on balance, this proposal doesn't look too bad.  We have

	*<bortza-z-biga> --> *<bortzazpiga> --> *<bortzazpi> -->

Only the loss of the first syllable is problematic.

Larry Trask
University of Sussex
Brighton BN1 9QH

larryt at

More information about the Indo-european mailing list