Re: Allophones or Separate Phonemes?

Kristian:
> I'll first summarize my problem here for people who have just tuned
> in to the thread: Are sounds that are in complementary distribution
> but quite _dis_similar to each other allophones or separate
> phonemes?
>
> And Rosta wrote:
> >From your post I haven't been able to work out the rationale for
> >that choice of symbol. Unless things have changed in some recent
> >revision of the IPA, the symbol represents a velar approximant
> >glide. Why not use the symbols for pharyngeal fricatives (hook-top
> ><h> and barred <h>)?
>
> Because there aren't any specific IPA symbols representing
> pharyngeal approximants. So I could have chosen other symbols, but I
> didn't. Besides, in the modern language they are now realized as
> voiced and voiceless velar approximants in syllable-final position -
> at the phonemic level, that is. So the turned m with a long right
> leg seems appropriate.
I see: that makes sense. Though, BTW, the IPA does provide away for
diacriticizing symbols for fricatives to make them signify the
corresponding approximant: for apical/laminal/dorsal fricatives
the lowering diacritic is used (currently a small T beneath, I think;
I've no up-to-date chart here). I don't know what is used for
bilabial or radical fricatives, though.

>
> -----<snip>-----
>
> >If you're going to do a traditional phonemic analysis, then this
> >seems the obvious analysis: two phonemes, /?/ and /h/ that can be
> >syllable-final or syllable initial, and with the phonetic
> >realizations that you describe.
>
> Note however that on a phonetic level, both [?] and [h] in
> syllable-final position are a result of the two contrastive
> phonations of syllable codas in Boreanesian. Heavy syllables ending
> in a stiff voice always end in a glottal stop, while heavy syllables
> ending in slack voice always ends in voicelessness or [h]. So I
> guess syllable-final [?] and [h] are prosodic units (rather than
> segmental) that apply to heavy syllables as a whole. Would that
> change things?
I would see "?" and "h" as two contrastive elements that
diachronically have shifted their locus from the syllable coda node
to the rhyme node (with a corresponding change in phonetic
realization). But in the message I posted I was assuming a more
traditional descriptive framework, wherein all contrastive units are
segments.

>
> -----<snip>-----
>
> >These are pseudoproblems that arise only when your minimal unit of
> >phonological contrast is the segment. Well-known problems arise: In
> >English are /t-/ and /-t/ the same phoneme? (standard answer is
>
> >yes, even though their allophones are markedly different) Are /h/
> >and /N/ the same phoneme? (standard answer is no, because of their
> >dissimilarity even though they're in complementary distribution).
> >
> >Modern phonological theory separates out "structural" or "prosodic"
> >properties like "syllable-intial/final" from "segmental" or
>
> >"melodic" properties like "nasal" or "noisy" or "occluded".
>
> I think the problem with Boreanesian phonology is that both [?] and
> [h] are phonemic both segmentally (syllable-initial) and
> prosodically (syllable-final).
Or, in other words, melodic elements "?" and "h" are located on the
onset node and on the rhyme node.

> >For an expository description -- as opposed to a full analysis --
>
> >of Boreanesian, I see no problem in using phonemic analysis, so
> >long as you accept that the analytical system is itself imperfect.
> >
> >In my view, there is another problem that can face either kind of
> >phonological analysis. The problem lies in trying to cram all of
> >phonology into a single level of phonological representation. This
> >(in my view, again), is not how phonology works: there are two
> >levels. One is the level that interfaces with the lexicon, which
> >represents only phonological *contrasts*. The other is the level
>
> >that interfaces with phonetics, specifying what the phonetic
> >realization should be. It is entirely usual for the repertoire of
>
> >representational elements to be much larger for the phonetic
> >interface level than for the lexical interface level.
>
> I'm well aware of the two levels of phonological representations. In
> fact, in the grammar I'm writing, the phonemic representation of
> Boreanesian is meant to be a kind of practical orthography. Examples
> are given in this orthography, but since it does not specify
> phonetic realizations, some examples are given with a more narrow
> transcription on the phonetic level to give readers a feel for the
> phonological processes involved.
What does the Boreanesian writing system itself do? (The Livagian
ones represent only the lexical level of phonology.)

> -----<snip>-----
>
> >In this instance I think the solution is as follows. At the lexical
> >interface, you can have just one "melodic" unit - call it /S/, say.
> >This then entails that [s-], [-:h] and [-jh] are noncontrastive,
> >without you having add to the phonology a stipulation restricting
> >/s-/ to syllable initial position and /-jh/ to syllable final.
> >But at the phonetic interface level the distinction must be
> >represented, because of course it is *systematically* reflected
> >in the pronunciation of Boreanesian.
>
> I had a feeling that this _might_ be the case. But how would I
> represent such a unit in IPA; with an under-ringed lower-case-j,
> or with a c-cedilla? The problem here is that [s-] and [-jh] are
> rather dissimilar making it difficult to decide which symbol to use.
A common strategy in such cases is to use something that is clearly
*not* an IPA symbol, such as a large capital letter (e.g. <J>).

Using IPA symbols for phonological representations is always bound to
be rather confusing and potentially misleading.

> Furthermore, the [h] in [-jh] is more a mark of the slack
> phonation - a prosodic unit applying to heavy syllables as a whole
> rather than a segmental unit.
Melodic element "S" attached to an onset is realized as [s], while
"S" attached to a rhyme is realized as 'slack phonation'.

> So then again, this might _not_ be the case afterall. Both [s-] and
> [-jh] are noncontrastive - they are in complementary distribution
> being that one only occurs in syllable-initial position and the
> other in syllable-final position. But despite all this, I would
> still consider [s-] and [-jh] quite dissimilar. You mentioned a
> similar situation with English /h/ and /N/ where they too are in
> complementary distribution but are quite dissimilar. In fact, as you
> know, they are represented by different symbols altogether.
That is because noone (as far as I can recall) thinks it worthwhile
treating h and N as varieties of the same phoneme. It buys only the
slightest of representational simplifications, and probably obscures
certain generalizations one might want to make about, say {f, h}
or {m, n, N}.

> Another
> dissimilarity is the fact that [-jh] bears a prosodic feature that
> applies to heavy syllables as a whole (ie., [h]). The sound [s-]
> bears no such feature.
But see my analysis above.

> Boreanesian has evolved in such a way that
> the feature of phonation is segmental in syllable-initial position
> and has become prosodic in syllable-final position. So might [s-]
> and [-jh] be different phonemes altogether due to their
> dissimilarity despite their complementary distribution?
As I think I've shown, there are straightforward ways of treating
these as instances of the same melodic unit attached to different
prosodic nodes. However, whether you actually adopt that analysis
depends on whether you think it is better than the alternatives.

> -----<snip>-----
>
> >There doesn't seem to be a contrastive unit ":h". You have either
> >/-@H/ or /-IH/. [Where @ could be taken as notation for a
> >melodically unspecified structural segment].
>
> That is essentially the analysis I had in mind. In fact, I'm
> representing this "@" as a velar approximant in my grammar whose
> allophones lengthen central vowels, similar to how an palatal
> approximant would lengthen the front vowel and how a labio-velar
> approximant would lengthen a rounded back vowel. I chose to write
> "@" in this and the last post because of the limitations provided by
> emailing.
>
> This velar approximant only occurs in syllable-final position. So if
> I were to follow an analysis similar to [s-] and [-jh], then one
> could say that the unit that is realized as [-@h] in syllable-final
> position is the same segmental unit that is realized as [h-] in
> syllable-initial position. Similarly, how would such a unit be
> represented in IPA?
It couldn't (and shouldn't...).

>
> However, [h-] and [-@h] could also seen as dissimilar due not only
> to the lack of an approximant in one of them, but also to the
> difference in the types of phonological units involved. The [h] in
> the former is segmental while the [h] in the latter is segmental.
"the latter is nonsegmental" I presume you meant. See above for a
solution.

> So
> once again, might [h-] and [-@h] be different phonemes altogether
> due to their dissimilarity despite their complementary distribution?
>
>
> -----<snip>-----
>
> >> Comments, please!
> >
> >My comment is that I very much hope that you *will* study
>
> >linguistics at university.
>
> We'll see about that. I have 'til after summer to make a final
> decision. 8-)
>
> Thanks!
> -kristian- 8-)
Have you checked whether it is possible to do Anthro + Lx together?

--And.