> We always hear that PIE initial *b is rare but I don't quite see the
> basis for it.
Partly it's to do with the nature of the words that show it (many are
"babble-words"), and partly to do with the nature of the attestation. There
are very few - perhaps only two - which can be shown to certainly PIE *bel
and *bak). There are strong reasons for resisting the rest as good PIE
words.
We have (initial, medial final):
bab bab badyos baita bak bal bamb band bata bau bdel be: bed bed bel (x2)
bend bu (x3) bistle blat ble: blou bol brangh brukw bronk
abel baba cerbero abo(n)
ab bab bamb dumb dheb ereb jab gheub ghreib g'hemb k'elb k'eub kob lab leb
plab remb reub (s)kamb skeub skwerb (s)nerb treb ub wab
/b/ occurs 28 times as an initial letter in Bird's list of Pokorny's basic
roots. Compare g' (20 times initially), gwh (12 times initially), kw (22
times initially).
Each root with /b/ is found on average in 3.3 languages. (compare gwh, found
in 4.8 languages for each root, on average)
Peter