Re: THEORY: phonemics (was: RE: [CONLANG] Optimum number of symbols

Mike S:
> On Tue, 21 May 2002 01:27:25 +0100, And Rosta <a-rosta@...> wrote:
> >
> >1. I reject the classical notion of a phoneme inventory, because there are
> >different sets of contrasting segments for different phonotactic positions
> >(e.g. syllable-initial and syllable-final). Criteria for identifying
> >members of different sets with one another rest on a set of rather
> >woolly criteria involving some notion of phonetic similarity or details
> >of morphological alternations. If the phoneme inventory is to be
> established
> >by rigid application of the principle of contrast, then the total
> >phoneme inventory should consist of the separate inventories for
> >the different phonotactic positions, with phonotactic position treated
> >as a phonological feature of each phoneme.
> >
> >2. The more usual objection to phonemics (as I gather most of us know;
> >please don't take my explanation to be patronizing) is that it overly
> >privileges the segment. A full analysis of a phonological structure
> >requires recognizing elements both below the level of the segment (i.e.
> >structure internal to the segment) and above the level of the segment
> >(e.g. at rime, syllable, foot level). So, to answer Julien, the
> >validity of the segment is not being denied; what is being denied
> >is its privileged, 'emic', status. To my mind, Julien wrongly
> >takes 'phoneme' to mean 'phonological segment'; 'phoneme' carries
> >extra theoretical baggage with it.
> >
> >I do fully accept that phonemics is a very useful tool for creating
> >writing systems -- indeed, that was the driving force behind the
> >early development of phonemic analysis. But as I said in another
> >message, the writing systems that result are practical but kludgey,
> >especially at the conceptual level.
Do note that final paragraph of mine, so as to make sure we don't
inadvertently end up arguing about something we in fact agree on!

>
> I am intrigued by your rejection of the notion of the phoneme, and
> what it suggests in the way of producing a phonological decription
> of a given language, but I am still a little unclear in a few areas.
> The following are the questions foremost in my mind:
>
> (1) In a given language, each phonotactic position will specify
> a set of contrastive phonological segments.  I am gathering that
> each contrastive phonological segment in a given set/phonotactic
> position will be invariant to about the same degree that a phone
> is invariant in a classical phonemic description of a language.
> Is this correct?
I'm afraid I can't work out what 'phone' and 'invariant' mean here,
but if we confine our attention to a single phonotactic environment
then the criteria for establishing the phoneme inventory would be
the same as for classical phonemics.

> (2) If my assumption in (1) is correct, would you object to
> referring to the 'contrastive phonological segment' simply as the
> 'phone', as long as we carefully stipulate that nothing analogous
> to the allophone of phonemic theory exists here?  After all,
> it seems to me that your sets are indeed composed of phones.
If 'phone' is supposed to mean a particular type of phonetic
segment, as it standardly does, then this would not be satisfactory.
A phoneme can have multiple allophones which may be discontinuous
in phonetic space (e.g. [t] and tap [D] as allophones of English
/t/ in foot-internal intervocalic position).

> (3) If such an approach is to be useful in producing a phonological
> description of a given language, it seems to me that you would have
> to exhaustively identify and name every unique phonotactic position
> in the language.
That's right.

> I am rather unclear as to how we would economically
> go about identifying and naming every such position.
Essentially, you would identify the minimum number of positions such
that phonemes themselves need not be mentioned in phonotactic rules.
The only phonotactic rules would be those that define the positions.

> (4) Assuming that (3) is resolved in some manner, do you have any
> feeling as to the relative efficiency of describing the phonology
> of a given language using your approach in comparison to using
> the approach of phonemic theory, generally speaking?  In other words,
> which approach in general can be expected to produce the smaller
> number of production rules?
I think my approach is much more principled and less arbitrary and
probably of about equal descriptive efficiency. But it is less
efficient as a basis for an alphabet.

For example, consider fricative + plosive clusters in English onsets.
The fricative can be 's' (or, very marginally, 'S', as in _shtook_).
What can go in the plosive position? Well, there's a three-way
place contrast but no voice contrast, so the 'positional model'
would identify 3 positional-phonemes here. These would be identifiable
with neither /p, t, k/ nor /b, d, g/, since (among other reasons)
these sets are defined by a voice contrast. The trad analysis must
identify them with one or the other set, for no very principled
reason. But when it comes to writing it down in a script, it
is more efficient -- in that it requires fewer overall symbols --
to recycle characters in different positions, even if that implies
some sort of spurious systemic identity between their values in
different positions.

--And.