permissible IE roots?

[petegray]

Permissable roots:  Yes there are exceptions.   The "root constraints" refer
to a striking imbalance in proportions of combinations, so that we can
suggest some kind of constraint, which was overridden in "Lall-woerter" or
borrowings or in expressive noises.   The constraints are far from an
invariable rule.

(1)  The IE /b/

Bird's summary of basic IE roots from Pokorny has 28 roots with initial b, 4
with medial b, and 27 with final b.   Note that this includes 4 overlaps:
the roots *bab = to swell and *bamb (listed twice) = to swell and = bump;
and the root *baba = babbling.   That's a total of 55 basic roots.

About a third of these are clearly onomatapoeic.   Compare the 28 with
initial b against the 20 with initial g' and the 22 with initial kw.
Initially, there is nothing particularly odd about /b/!   The peculiar lack
is in medial and final position.

(2) Root constraints

(a) Yes, for the purposes of root constraints, palatalised k' is treated
differently from plain k, but kw appears to be treated the same as plain k.
The evidence is:

k' co-occurs with k.  10 cases with initial k',  1 with initial sk', 1
medial, 3 final.
g' occurs three times with g, only in the order gVg';     g'h with gh twice.

kw co-occurs with k' twice, once each way round, but not with k.

(b) Yes, the presence of initial s- devoices the following consonant, and
appears to neutralise voicing rules:  so roots like *stegh are not uncommon.

(3) Another constraint

G & I suggest that only roots with two voiceless stops (or initial s which
devoices the initial stop) show zero grade!

Peter