Motivating the Root Restrictions of PIE

[David L. White]

[ Moderator edited.  Please do not send HTML mail to the list. ]

Sorry if you folks have heard this before, but I never got a response...

According to Lehmann, PIE shows three kinds of root restrictions.  Accepting
for now the traditional interpretation and using "D" to represent any voiced
plosive, "T" to represent any voiceless plosive, and "DH" to represent any
voiced aspirate plosve, these are:

                1) no /DeD/
                2) no /TeDH/
                3) no /DHeT/

Note that 2 and 3 are the inverse of each other, and are assimilatory, whereas
1 is dissimilatory.  The problem is that if the traditional interpretation is
correct, these restrictions (not to mention the /b/-gap) do not make sense.  So
here is one way (the only way I can see) that they could make sense.

        1)  The voiced plosives were orginally not voiced but pharyngealized.

Pharyngealization is a crude gesture, and if occuring on both sides of a vowel
would distort this in the direction of /o/.  If PIE was attempting to maintain
a distinction between /e/ and /o/, this would not be good.  Alternatively, the
restriction could be related to avoiding having pharyngealization noise mess up
a distinction of laryngealixed versus murmured vowels, see below.
Pharyngealization would also explain the /b/-gap, as pharyngealized labials
are, for good acoustic reasons, disfavored generally.  It could also explain
part of what JER has noted, that vowels in the vicinity of voiced sounds in PIE
tend to come out as /o/.  (The other part would have to be related to the vague
back-round resonance of sonorants generally, as seen in modern English "prison,
prism, bottle", where it is not entirely clear whether a [U] sound, as in
"foot", is to be regarded as present or not.)

        2) The voiceless plosives were orginally laryngealized (which is not
           the same as glottalized).

        3) The voiced aspirates were as traditionally posited, technically
           murmured.

More specifically, what I have in mind here is that the voiceless and voiced
aspirated series were originally associated with what Ladefoged and Maddieson
call "stiff voice" and "slack voice" respectively, which are in effect weak
laryngealization and weak murmur.  (Where what we would consider normal
phonation ("modal voice") does not occur, there is no point in going to the
trouble of producing the strong forms of laryngealization and murmur.)
Laryngealization and murmur are phonetic opposites. This would explain why they
could not co-occur, especially if the original distinction of phonation type
was on the vowels, and was only later reanalyzed as belonging to the
consonants.

Such a system would be typologically (or rather statistically) odd, even
unattested, but no more normal system motivates the root restrictions seen.  It
is not true, contrary to what is sometimes alledged, that languages never
contain unique features.  For example, the stress system of Estonian is (so I
have it on good authority) absolutely unique.

House-husbandry calls, more (on place) anon.  Congratiulations in advance to
anyone who spots what (I see) is wrong (though not uncorrectably) with all
this.

                                                       Dr. David L. White