Rob wrote:
> But if "true long vowels" are shortened when atonic, shouldn't they
> then be considered the same as other short atonic vowels?
Old _short_ vowels were simply dropped pretonically.
> What would cause the supposed "accent attraction"?
The tendency of full vowels to be accented plus the frequent preference
for initial stress (which may be overridden by other tendencies but has
to do with the universal demarcative function of stress: it is
"attracted" by the initial word boundary).
> Is there a verb _sva:pjati_ in Vedic?
No, the suffix apophony of the two subtypes of causatives was levelled
out there (and Brugmann's Law levelled out the quantitative contrast in
the root), which is why we have Vedic sva:p-áya-ti alongside
ma:n-áya-ti, but cf. Latin so:pio: and Germanic *swo:f-jan- 'lull to
sleep'. The originally Narten character of the root is also visible in
such derivatives as iterative *swép-sk^e/o- (Av. xVafsa-) and *swép-no-,
where we would normally expect nil-grade vocalism.
Piotr