On 2008-02-12 09:25, Pavel A. da Mek wrote:
>> (neither was long *e:, for that matter).
>
> This is less understandable;
> the /e:/ from contraction could maybe be seen as /ee/ coloured to /ea/ and
> then conracted to /e:/, but why was not coloured /e:/ from compensatory
> lenghtening?
I'm not quite sure about *e: from compensatory lengthening. Do you mean
"Szemerényi" length in the nom.sg.? It is not compensatory, stictly
speaking, since the vowel is lengthened also when no segments are lost,
as in *dje:us. Secure examples with *-e:h2(s) or *-h2e:C-s may be
difficult to find. What was definitely not coloured was "Narten" length,
as in Hitt. hekur 'summit' < *h2e:k^-wr., as well as "sigmatic" length
(*g^ne:h3-s-), which only shows that length must be pretty old in those
cases. It seems that only the relatively "weak" vowel *e was prone to
laryngeal colouring, while "strong" vocalism (*o, *e:) was generally
immune to it.
Piotr