Quoting JS Bangs <[log in to unmask]>:

> There is only one language that I know of that challenges this theory:
> Standard Swedish. Or maybe Norwegian. I forget. Anyway, it has all of /i
> y
> } 1 u/, I believe, and so cannot be accomodated under this system. But
> there is also a significant difference in *lip* position between /y/
> and
> /}/, implying that the real featural difference might have to do with
> types of rounding. In any case, one kind-of-marginal counterexample
> isn't
> quite strong enough to dissuade most linguists.

You won't find [1] in standard Swedish. I don't know of any Scandinavian
dialect that has it, but that's not saying too much.

I've got the same lip position for /2 y }/, and another one for /u o/. BP
has /}/ in the other group ...

My 'lect's vocalic inventory would be something like this:

| i: y:        }:          u:
|    I Y                U
| e: 2 2:                  o:
|              8
| E E:                     O
|        a                 A:

(Add schwa as unstressed variant of /E/, the diphthong [aU] and some
allophones conditioned by neighbouring consonants.)

Say that the kind of rounding is the distinctive thing in the pairs [}:]~[u:]
and [8]~[U], and we're rid of the central~back distinction. Well, you've more
or less convinced me.