Phonemes, was: A sample of my newborn conlang

Sorry if this gets posted twice...

On Mon, 28 Jan 2002 13:47:55 +0000, Stephen Mulraney
<ataltanie@...> wrote:
[...]
>However, consider the mutation (the
>i-affection mentioned above) that sends, say (working with phonemes
>for the moment) [a] to [ae] to [ai]: With regard to phonemes, that's
>the end of the line, and affecting [ai] just gives [ai] again. But,
>phonetically we get the following: /A/ > /Ae/ > /ai/ > /(ai):/. That
>is, there's a further 'long' allophone of [ai] used in i-affection,
>and these allphones are I think rising. The extra allophone is notated
>in the orthography (as an accent over the middle of the diphthong, in
>ascii I double the second vowel) so the plural of a hypothetical noun
>*<cutai> would be *<cutaiir> (compare with *<cutae> > *<cutair>). As
>far as I know that plural-final -r (and other features in places where
>i-affection occurs) always occurs; so their are no minimal pairs making
><ai:> a phoneme.
If I understand you correctly, your <aii> is still a phoneme, I think.
Since it's conditioned morphologically, not phone*ically. Unless you
forbid all the final consonants that can invoke it (<r> etc.) to be
anything but (part of) specific endings; *then* it would be an allophone
occurring before such consonants. But this doesn't seem to be what you
meant, does it?

>A question: to what degree is the definition of a phoneme via minimal
>pairs meaningful?
It seems increasingly common to deny minimal pairs as the minimum
requirement.

That is, if you find that some words are stably pronounced with certain
sound, *and* can't formulate a rule explaining such forms as phone*ically
conditioned (i. e., there's no obvious rules for complimentary
distribution, in terms of phone*ics, with some other similar sound), then
the sound in question must be considered a phoneme (or a "potential
phoneme", at least). At any rate, it cannot be counted as an "allophone".

>e.g. in the extreme case, suppose only one minimal
>pair establishing a phoneme <X> as distinct from <x> occurs; is always
>then legitimate (from a speaker's and not a linguist's point of view) to
>make that distinction everywhere else in the languge, e.g. coining a new
>form <Xyz> alongside <xyz>?
In fact, these are two independent questions. If you have minimal pairs
(proving which is often another problem, though), then you certainly deal
with phonemes. But some phonemes (or phonemic distinctions) can have
a (phonetically) restricted distribution. For example, in Mandarine one
has phonemically distinct <s>, <sh> and <h> before back vowels; but before
front vowels (including /y/) one has only <x>, and it's not easy to say
whose allophone it is, synchronically (historically, it has same sources
as <s> and <h>).

>Moreover, if the minimal pair only occurs in
>a very constrained (non-phonetical) environment - say in a particular
>*grammatical* situation, (e.g. the imperative <flbXy> as opposed to the
>3rd person pres.ind.act <flbxy>) does this still establish the phoneme
>as firmly (in the sense mentioned in the parenthesis above) as have
>two series of minimal pair e.g. of basic vocabulary.
Again, if one can *prove* that the context factors involved are of purely
phone*ic nature, then you deal with phonemes. The problem is that many
morphemic boundaries can be addictionally marked with various phone*ic
phenomena, often difficult to detect (e. g., prosodics, sound tenseness,
etc.). Sometimes it seems more correct to classify certain phones as
(additional) boundary markers (of allophonic nature, associated with
intonation, syllable tenseness pattern, etc.), rather than independent
phonemes.

>Finally, what if the environment is not grammatical but 'contextual',
>so that one cannot tell (by meaning) <rdrx> from <rdrX> except by
>referring to the context; and that it is an error to use <rdrx> in the
>context appropriate to <rdrX> (& vice versa)? Still?
This, again, depends on the nature of the context which sometimes is
difficult to determine. For example, if relative clauses require certain
intonational patterns associated with them, one has to do a lot of
research to tell if certain phones restricted to clause-initial position
are conditioned syntactically or phonetically.
>
>(An example of the last might be "What about this flrb[xX]y?"
>Where following "I'm hungry", the word is <flrbxy> and means
>"tasty pre-dunked muffin" while after "What'll we do tonight?"
>the word is <flrbXy> meaning "movie starring Steve McQueen")
I am not sure what you call "context" here (or what is the relation
between the the two words: why do you analyze them as *same word* in
*different contexts*, rather than the reverse?).
>
>(A provisional response might be that 'rare' phonemes will tend to be
>(s)quashed by analogy, thus collapsing a minimal pair (but languages can
>deal with that), while more common phoneme might tend to spread, also
>by analogy, becoming more established?)
I am not sure if such a tendency exists. Do you refer to any specific
examples?

[...]

>Stephen
Basilius