Retroflex sounds

Miguel Carrasquer Vidal mcv at WXS.NL
Mon Jun 22 11:18:33 UTC 1998

Michael Witzel <witzel at FAS.HARVARD.EDU> wrote:

>(2) if  (1) does not work, and waht I also had in mind:  has anyone an
>opinion on:
> Nostratic **t  > PDrav. *.t,  etc.?

Illich-Svitych reconstructs three series of Proto-Nostratic stops,
along the lines of Afro-Asiatic and Georgian *d, *t and *t' (voiced,
unvoiced, "emphatic" or glottalic).  For PN > PDrav., the following
sound laws are postulated:

PN         PDr

*t-        t-
*-t-       -t-, -tt-

*d-        t-
*-d-       -t.-, -t.t.-

*t'-       t-
*-t'-      -t-, -tt-

[Other Proto-Dravidian retroflex sounds are the result of PN *-L-
(voiced lateral fricative), and PN *-l'- and *-n'- (palatal
lateral/nasal, respectively)]

That is to say, the 3 series merged in initial position.  In
intervocalic position, the voiced series gave the retroflex sounds,
the other two merged.

The problem of the three series of stops in Proto-Nostratic is far
from settled.  For instance, Alan Bomhard's reconstructions exchange
the reflexes of the unvoiced and glottalic series as compared to
Illich-Svitych.  Looking only at Dravidian, the system of stops seems
to be governed by two parameters: (1) fortis (TT) vs. lenis (T), and
(2) plain (t) vs. retroflex (t.), giving a total of 4 possible
intervocalic stops (t, tt, t., t.t.).  In the initial position, the
oppositions are neutralized (all merge to t-).  This is indeed
similar to the parameters found in Afro-Asiatic, Georgian and
Indo-European [glottalic version], which are (1) voiced vs. unvoiced,
(2) plain vs. glottalic, and is also comparable to the fortis-lenis
distinctions in Uralic (-t- vs. -tt-).

> From a strictly phonetic point of view, a development ejective
(glottalized) -> retroflex is rather plausible.

Miguel Carrasquer Vidal
mcv at

More information about the INDOLOGY mailing list