PIE *gn- > know/ken

Miguel Carrasquer Vidal mcv at wxs.nl
Sat Feb 27 23:58:22 UTC 1999

philps at univ-tlse2.fr (Dennis Philps) wrote:

>Could anyone tell me at what stage in history the initial voiced consonant in
>PIE roots such as *gno- "know/ken/can" or *gen- "knife", etc. became devoiced,
>and why this devoicing concerns some roots (e.g. *gen- "knife", etc.) but not
>others (e.g. *ghen- > "gnat").

That's because *g and *gh are completely different PIE phonemes.
An Ancient Greek would have asked why one set, *g, remained
voiced (gi-gno:-sko: < *g(e)n-), while the other became devoiced
(*gheimn > kheima).

In fact, if we look at all the IE languages, we can distinguish
between those that keep *d and *dh voiced and distinct (Indic *d
~ *dh), voiced and the same (Iranian, Baltic, Slavic, Celtic,
Albanian *d), one voiced the other voiceless (Greek, Italic *d ~
*th), one voiceless the other voiced (Armenian, Germanic *t ~ *d)
and both voiceless and the same (Hittite, Tocharian *t).

This, coupled to the fact that voiced aspirates like /bh/, /dh/
and /gh/ are very rare in the world's languages and always
presuppose the existence of voiceless aspirates /ph/, /th/, /kh/,
which are absent from reconstructed PIE, has led to some
alternative hypotheses about the actual pronunciation of what we
write as PIE *d and *dh, etc.  The best known is Gamqrelidze's
"glottalic theory": *d was glottalized /t'/, while *dh was simple
voiced /d/.  And *t had allophonic variants [t] ~ [th]. As stated
by Gamqrelidze, the theory makes no sense.  There is no evidence
for aspiration of *t outside of Germanic and Armenian (and
possibly Celtic, if we count *p[h] > h > 0) and strong evidence
from Greek and Italic that it wasn't.  There is no reason,
outside of Germanic and Armenian, to think that *d was unvoiced
while *dh was not.  And I won't even go into Gamqrelidze's
attempts to generalize Grassmann's Law (the dissimilation of two
consecutive aspirates in Sanskrit (*dhidh- > didh-) and Greek
(*thith- > tith-)).

The most acceptable solution from my point of view is that PIE
did not have any voiced stops at all.  Instead it made a
distinction between fortis and lenis stops (as in Finnish, Danish
or Hittite), where the fortis (tense) stops (*t etc.) were always
voiceless and pronounced longer/with more energy ([t:] or [tt]).
The lenis (lax) stops (*d and *dh, etc.) were less energetic/
shorter, and had voiced allophones.  They came in two kinds, one
aspirated (*dh = [th]), the other not (*d = [t]).  Or,
equivalently, one glottalized (*d = [t']), the other not (*dh =

Miguel Carrasquer Vidal
mcv at wxs.nl

More information about the Indo-european mailing list