Typology and the phonetics of laryngeals

[Miguel Carrasquer Vidal]

Jens wrote:

>On Wed, 4 Oct 2000, Herb Stahlke wrote:

>> [...] but the resulting fricative series /s, x', x,
>> x^w/ is, as a fricative series, bizarre.  I looked through
>> Hockett's Manual of Phonology and couldn't find anything even
>> close.  Why no labial?

>If you call /H1/ a simple /h/ which it must have been, at least some (or
>most? or all?) of the time, and take the voicing of /H3/ as reason to call
>it /gh/ (gamma, voiced velar spirant), but keep /s/ and /x/, you get VERY
>close to Dutch.

But not *that* close.  In the north, we have /s/, /x/ and /h/, but no
/G/.  In the south, they have /z/, /s/, /G/, /x/ and /h/.
Additionally, there's either /f/ or /v/ and /f/.  [I'm not counting
written <w> as a fricative: it's a bilabial, labiodental or labialized
labiodental approximant or stop, depending on dialect].

Interesting that you say "at least some (or most? or all?) of the
time".  I've been thinking along those lines myself.  We distinguish
the laryngeals by their vowel colouring and vocalic reflexes in Greek,
mainly, but that doesn't mean that every laryngeal that gave /e/ in
Greek must have come from the same unitary PIE laryngeal phoneme.  In
the case of *h1, I agree that some, or most of the time, we're dealing
with /h/: (some) *h1('s) aspirate(s) a following or preceding stop,
some *h1's give /h-/ in Armenian and Albanian.  On the other hand, *h1
must sometimes have been a simple glottal stop /?/: I believe a root
like *h1es- "to be" is more likely to have been /?es-/ than /hes-/ (I
mean, maybe it was */hes-/, but I don't think it's likely that *all*
roots beginning with *h1V- had /h/).

As to *h3, I don't think /G/ is very likely.  At least in late PIE, I
think there were no voiced fricatives.  Earlier voiced *z, as in the
nom.sg. which lengthens the thematic vowel, later merged with *s, so
it's very unlikely that *G, if it ever existed, did not merge with *x
(*G is usually the first voiced fricative to go, cf. Dutch).  Also, a
voiced velar fricative does not explain the o-colouring, so we should
at least have /Gw/, and that would be strange indeed, to have /Gw/
without /G/.

My own proposal would be to split up *h3 into /xw/ (labialized *h2,
aspirating, Hittite h-), /?w/ and /hw/ (labialized *h1, partially
aspirating, Hittite 0-) [allowing Jens to withdraw his "vote for
chaos"].  "Voicing" *h3 can then be from *Gw > *xw, if the traditional
reconstruction of the stops holds, or from *?w if the glottalic theory
is correct.  A possible confirmation of either thesis might come if we
could discover cases of voicing *h2 (< *G) or of voicing *h1 (< *?).

Now if we have labialized laryngeals (*h3), we probably have
palatalized ones too (*?y, *hy, *xy).  The first two surely give *h1,
the last one too (as phonetic [g]), judging by the *h3 ~ *h1
alternations in the dual endings, which I trace back to auslautend
nominative *-(a)ku > *-xw > *-h3 and oblique *-(a)ki > *-xy > *-h1.

=======================
Miguel Carrasquer Vidal
mcv at wxs.nl