On 04-10-13 05:52,
enlil@... wrote:
>>That's beatufully analogous to the generalisation of linking /r/
>>in non-rhotic English (the latter appears not after _any_
>>final vowel, but only after those that _might_ have resulted from
>>the loss of final /r/).
>
>
> It seems like any r from what I hear on the BBC. Listen to "Supernova
> in the Sky" by Oasis. I hear "Supernover in the Sky". What do you
> hear?
The same, of course. I suppose I didn't make my point cleaRRRR enough.
Let me try again. Most instances of final /&/ (as in <father>), /I&/
(<clear>), /U&/ (<pure>), /E&/ (<care>), /3:/ (<fur>), /O:/ (<score>)
and /A:/ (<far>) in RP-like accents reflect vowels once followed by /r/.
Though lost preconsonantally and before a pause, the /r/ continued to be
pronounced as a "linking R" when the next word began with a vowel (as in
<clear enough>).
This habit has been generalised (despite the spelling) also to those
instances of any of the vowels above that come from some other
historical source. When pronounced in isolation, <clear> rhymes with
<idea>, <score> with <saw>, <loafer> with <sofa>, <far> with <shah>, and
<fur> has the same vowel as <milieu> (in RP at least). Hence the
temptation to treat them in the same way with regard to sandhi rules
(even in root+inflection sandhi: <sawing> and <soaring> may be
pronounced the same).
So an older rule _deleting_ a final /r/ in sandhi (_unless_ a vowel
followed) has been replaced by a rule _inserting_ a final /r/ after the
vowels in question (_if_ a vowel follows), even if there is no
historical justification for the rhotic. It offends the ears of some
pedants, hence the term "intrusive R" to distinguish it from
"legitimate" linking R.
What may have happened in Greek was a similar generalisation. <eikosi>
'20', has a final <-si>, just like <perusi> or any other similar form
after which a "linking N" would have been historically justified (if
originally part of the locative ending). From the point of view of a
native speaker of Ancient Greek, the structure and etymology of both
<eikosi> and <perusi> was equally opaque. They didn't know they had an
old nom.du. in the one and an old loc.sg. in the other, so there was no
synchronic motivation for treating them differently.
Piotr