If word and morpheme boundaries are ignored then there's
no complementary distribution of /h/ and /N/, is there?
Doesn't /N/ occur in the onset of the syllable in 'singer'
[sI.NIN] and 'thing-a-ma-jig' [0I.N&.m&.d^z^Ig]?

David

--- In phoNet@yahoogroups.com, "Richard Wordingham"
<richard.wordingham@n...> wrote:

> --- At http://groups.yahoo.com/group/cybalist/message/22698 , Piotr
> Gasiorowski <piotr.gasiorowski@i...> wrote:
> > To use a real-life example, English /h/ and /N/
> > ("ng") are in complementary distribution, but no-one has ever
> dreamt of
> > simplifying the system by assigning them to the same phoneme.
>
> Actually, that idea can be scotched for those English idiolects

which

> distinguish <lock>, <loch> and <long>. In these cases, the final
> consonant of <loch> is best analysed as /h/ even if it is [x].
> Syllable final /h/ may get re-inforcement from Moslem names such
> as 'Rahman'.
>
> Richard.