Quoting John Cowan <cowan@...>:
> Ray Brown scripsit:
>
> > The tendency, if there is a shifting, seems to be towards one of the
> > apexes of the 'vocalic triangle' as, I think, John's email implies.
>
> At first I thought so too. But the u-umlaut process is rounding, not
> backing: it mutates /i/ > /y/ and /e/ > /2/. Presumably it could even
> transform /A/ > /Q/.
Doesn't Icelandic make something patently weird along the lines /a/>/2/? Seem
to recall something about _altar_ being _altörum_ or _ölturum_ in some
inflected form.
Andreas (sleep deprived)