Re: [tied] Abstractness (Was Re: [j] v. [i])

> > Why not? The two phones are in complementary distribution in Avestan and
> do in fact there represent rule-governed realizations of /h/, IIr. /s/.
>
> Richard has already made a similar point. We discussed the phenomenon of
> "rhinoglottophilia" some time ago on the phoNet list. What I mean in the
> present context is not that [h] and [N] are necessarily different in
> phonemic terms, but that they are not combined into a single phoneme in
> English (despite their complementary distribution) BECAUSE there are no
> independently motivated rules relating them to each other (unlike the case
> of [t] : [?] in British English), so their hypothetically allophonic status
> can't be supported. One needs really compelling evidence to unite two phones
> that are so different phonetically. The minimal pair test is not a
> watertight criterion: English [h] and [Z] ("zh") don't contrast either
> because of their defective distribution, while [N] and [Z] may be found in
> similar contexts (e.g. <hanger> : <azure>) but accidentally fail to occur in
> a minimal pair because of their rarity.

It's not watertight only because you require the two words in the pair
to be existant in the language. How many English people wouldn't say
that hang and zhang are two different words even though zhang doesn't
mean anything in English. With this criterion we wouldn't have the
mismatch between the "intuitive" phonematic status of [h] vs [Z] and
the minimal pair test. Not so?

best wishes

Harald