Re: Phoneme analysis. (Motherhood statements?) (Long!)


"Ruud Harmsen" <realemailonsite@xxxxxxxxxxxxxxxxxx> wrote...

"John Atkinson" <johnacko@xxxxxxxxxxx>:

(3) Anything that doesn't have anything to do with the realisations of
current speakers (and therefore can't be shown to correspond to
something going on in their heads) has no place in a phonemisation of
their language. Thus, for example, for non-rhotic Brits with intrusive
R, the words NEAR, SQUARE, CURE, NURSE, START, FORCE and lettER have no
phoneme /r/, "underlying" or otherwise.

The fact they that they have intrusive r (and linking r) shows that
something DOES go on in their heads. So your statement is
self-contradictory.

No it isn't.

For a dialect with only _linking_ R, your statement would be true. There may indeed be such varieties, in areas where rhoticism is still common, or among people who are bi-dialectical, speaking both rhotic and nonrhotic varieties, or individuals who have been taught in school, or have taught themselves, to add /r/ only when there is an <r> in the spelling. (Brian belongs to that last group, I suspect.)

However, your average Brit non-rhotic has _intrusive_ R, which by definition means that /r/ is inserted after _all_ instances of various vowels (/i@/, /E@/, /u@/, /@/, /a:/, /3:/ , /O:/ in my dialect), before a word or morpheme boundary followed by any vowel. This occurs whether or not the corresponding words in rhotic accents finish with /r/. The rule is:

0 --> r / [@, a:, 3:, O:] _ # V (where # denotes word or morpheme boundary)

As you can see, the conditioning factor says nothing about the presence or absence of that putative underlying /r/ (or anything else) at the end of the word. Thus <pa> behaves identically to <far>, <comma> identically to <letter>, <sawing > and <soaring> are homophones.

(4) What happens in other dialects, and what happened in the past,
should never be taken into account in deciding what the phonemes of a
particular dialect are.

I think history IS important.

This doesn't mean that investigators shouldn't
consider the phonemisation of other varieties, including unrelated
languages, in making hypotheses about what might be happening;

So you do too.

No I don't (see the bit you deleted!). I'm saying that even unrelated languages (and a priori related varieties, including ancestral varieties) can give you ideas. The correctness of these ideas can _only_ be evaluated by looking at the here-and-now variety.

(6) Different dialects of a language should be phonemicised
independently.

How you identify which speakers speak a different enough dialect to
get its own phonemization?

Ideally, phonemise them both independently and _then_ look to see whether the two phonemisations are the same (or equivalent). In practice, of course, this may be too much like hard work!

(7) If two models predict equally well and one is "simpler" or "easier
to read" than the other, so much the better.

Define 'simpler' and 'easier to read'.

I can't, of course. That's why I put those words in scare quotes!

(8) Notation: It's nice if the symbols for a phoneme are close to the
IPA for the main allophones in the variety being phonicised. Thus, when
phonemising Brit, it's appropriate to write /I/ and /i:/ for the Brit
KIT and FLEECE vowels, since both length and openness are important in
distinguishing these vowels in these varieties; and, in phonemising my
own dialect, to write /E/, /E@/, and /ei/ for the DRESS, SQUARE, and
FACE vowels, since their main allophones are (close to) [E],

[e]. [E] is much to close to [&], and it varies a lot per dialect.

I specifically wrote: "in phonemising my own dialect". I did this intentionally because these three phonemes do indeed "vary a lot per dialect" within the group I called Brit.

[E@], and [ei].

Apropos of "vary a lot", the last is actually [a:i] in Broad Australian. Since I claimed to be phonemicising "my own dialect", perhaps I should have called the phoneme /V"i/, since my own colloquial pronunciation is [V"i], or perhaps [3i].

But being "close to the actual sound" is only one criterion of a nice notation, and /ei/ better satisfies the "easier to read" criterion, no? As well as being the pronunciation of this phoneme in other nearby accents (RP in particular), FWIW.

Thanks for the comments.

John.

.

Relevant Pages

  • Re: BBC does it again
    ... >> phoneme, and that long vowels are a sequence of two short vowels. ... Does this mean you don't agree that there are such languages? ... phonemic orthography is an ideal orthography. ...
    (sci.lang)
  • Re: BBC does it again
    ... phoneme, and that long vowels are a sequence of two short vowels. ... Does this mean you don't agree that there are such languages? ... orthographies are already used to writing English, ...
    (sci.lang)
  • Re: BBC does it again
    ... > phoneme, and that long vowels are a sequence of two short vowels. ... Does this mean you don't agree that there are such languages? ... > orthographies are already used to writing English, ...
    (sci.lang)
  • Re: BBC does it again
    ... convention in many Uralic languages, ... phoneme, and that long vowels are a sequence of two short vowels. ... orthographies are already used to writing English, ...
    (sci.lang)
  • Re: Drifting phonemes [was: Re: The AmE o sound]
    ... On Thu, 11 Nov 2004 01:57:47 GMT, Bob Cunningham ... if offered examples in close succession. ... accommodate both of our varieties, ... sense, if only on historical grounds, to call the phoneme ...
    (sci.lang)