Re: phonology of Plan B
| From: | Jörg Rhiemeier <joerg_rhiemeier@...> |
|---|
| Date: | Wednesday, July 11, 2007, 21:35 |
|---|
Hallo!
On Wed, 11 Jul 2007 17:58:30 +0100, R A Brown wrote:
> Alex Fink wrote:
> > On Wed, 11 Jul 2007 09:24:03 +0100, R A Brown <ray@...>
wrote:
[...]
> >>>-- Then a morpheme /123/ can be [ged] or [iha]. It's because of these
> >>>systematic equivalences that I think the 5-phoneme analysis is correct.
> >
> > This is what I always assumed was going on with Plan B,
>
> Yes, I don't think either I or And or Jörg disagree that that is
> essentially what is going on. The difference is on the _interpretation_
> of what is happening.
Yes. The question is not whether there is something that can manifest
either as [d] or as [a] (or whatever pair of consonant and vowel);
the question is whether this something ought to be called a 'phoneme'.
In my opinion, not, because it is *not a contiguous domain in the human
phonetic space*. Actually, it is not something pronuncible or audible
at all - it is a *bit quartet* that is mapped onto *two distinct phonemes*,
depending on whether its position number in the utterance is even or odd.
> I think the morpheme really should be given as {123} - the use of the
> slashes denotes phonemes, and on a *purely phonological* level, I fail
> to see how, e.g. [d] and [a] can be considered allophones of a phoneme
> /3/. This looks to me like phonological nonsense.
Yep. This indeed looks like phonological nonsense to me, too.
It is like positing a phoneme in English with allophones [h] and [N] :)
> > though probably
> > through the influence of my own dabblings with binary conlangs of vaguely
> > similar flavour. These had words of completely variable length, not
> > necessarily even whole numbers of hex digits.
>
> Do you mean words or phonemes? The phonemes of Plan B can be of
> different length, but they must be a whole number of hex digits. Your
> system sounds rather more intricate :)
You surely mean *morphemes* (or words), not phonemes.
> > But when it came time to
> > assign ways of reading bitstrings, my first couple attempts took the lazy
> > approach of just giving each n-bit string a pronunciation for some
> > constant
> > n, in the first case n=7. So each word had seven allomorphs, completely
> > phonetically unrelated
>
> Seven, eh? That makes Plan B's two look very paltry ;)
>
> But that is the point, isn't it? The allomorphs are completely
> phonetically unrelated.
They are indeed. There is nothing that /Soubout/ and /eiZeZru:/
have in common, except that they are reflexes of the same bit pattern.
But bit patterns fall outside the scope of phonology entirely.
> [snip]
> > impossible to learn to do in real time in a human-spoken language. I'm
> > not
> > so sure about the alternations my and And's interpretation of Plan B
> > displays, but at the very least they wouldn't survive for long in the
> > wild.
>
> I'm sure you're right. While Plan B makes life very easy for a computer
> programmer, I do not think it would survive long "in the wild," as you
> put it.
I agree.
> > Anyhow, on a closer reading of Prothero's essay, it does seem to be
> > unspecified whether Plan B exhibits this behaviour, probably as a
> > consequence of the inconsequentality of how the phonetic realization is
> > done.
>
> Yes, indeed. It is fairly clear, I think, that he does not put any great
> importance on how the thing is pronounced. Cf.:
> {quote}
> * An alphabet. "bcdf ghjk lmnp stvz" is suggested, but the
> choice is not critical.
> * A pronunciation scheme which makes all sequences of
> letters equally pronouncable, thus decoupling the rest
> of the language design from the details of the human
> vocal tract.
> {/quote}
Yes. The whole pronunciation shebang is entirely secondary!
> The scheme I give does exactly that. I have given an alphabet of sixteen
> letters, namely: w y g k r l z s n ñ d t µ m b p
> I have given a pronunciation scheme which makes all sequences of letters
> equally pronounceable. See:
>
http://www.carolandray.plus.com/Loglang/PhonAndOrthog.html
>
> The difference is, of course, that my scheme will not give rise to
> arguments about what are and are not phonemes (with the possible
> exception of the status of [j] and [w]), and all morphemes have one shape.
> > What would you say, then, to the perspective that Pentaphon (and by
> > extension And's Plan B)
>
> It's And's Pentaphon & Jeff Prothero's Plan B ;)
>
> > has as its inventory five (sixteen) morphophones,
> > and that the phonemic level must be taken as secondary?
>
> I would say I need to think more about this.
>
> The concept "phoneme" is an abstraction and not all linguists accept the
> phonemic theory or phonemic analysis. But it generally works well enough
> for the familiar European languages and we use it as a matter of
> convenience on this list and most of the time we more or less understand
> what each other means. I think the phoneme concept is a useful tool for
> the phonological analysis of very many languages.
Yes. It works well with most natlangs, and also with naturalistic
conlangs that work like natlangs. But even with such languages, things
are not always simple, as the controversy about the vowel systems of
Northwest Caucasian languages illustrates. At any rate, phonology has
progressed beyond linear segmental phonology; not all theories proposed
are equally useful, but some phenomena (such as tones and vowel harmony)
can be very conveniently treated with non-linear models such as
autosegmental phonology.
> However, the concept 'morphophoneme' seems to me a further level of
> abstraction. While most on this list would, I think, happily go along
> with the phonemic transcriptions /naif/ and /naivz/, I'm less certain
> that everyone would be so happy with {naiF}{Z}; and there are problems
> with this sort of approach. What morphophoneme are involved in 'mouse ~
> mice'?
Synchronically, it is perhaps best described as suppletivism.
(Diachronically, it is quite simple, of course.)
> > The alternations
> > are sufficiently unnaturalistic that Crystal's definition of phoneme just
> > breaks down, IMO.
>
> This is the point, isn't it? Plan B is is too 'unnatural'. I used
> Crystal's definition for the sake of convenience. But I hold that by any
> classical definition of phoneme, pairing [a] and [d] as variants of a
> phoneme /3/ just doesn't make sense.
Yes. Plan B is simply unnatural, to the point that conventional
linguistic concepts simply break down.
> But on reading further on morphophonemics, as it's generally called in
> the US, or morphophonology, which seems to be term preferred this side
> of the Pond, I read that "systematic phonemics" is the preferred term in
> more recent generative theories. I also find that those who adopt the
> 'systematic phonemics' approach, distinguish between "systematic
> phonemes" and the 'autonomous' phonemes of traditional phonemic phonology.
>
> I know that And adopts (and I apologize if I am wrong about this) a
> generative approach to grammar; he also knows, I think, that I am
> skeptical about the concepts of 'deep structure' and 'generative
> grammar.' I suspect that And may well adopt the concepts of 'generative
> phonology'; I do not. I think that therein lies the difference between
> our ways of looking at Plan B.
I am not very fond of generative grammar, either. Some concepts make
sense, others do not. I have the feeling that this way of describing
language has little to do with how language actually works. Myself,
I follow a functionalist approach: language serves a purpose, and thus
everything in it has some more or less well-defined function, and form
follows function.
> I would have thought it was obvious that all along I have been speaking
> in terms of 'autonomous' phonemes of traditional phonemic phonology. I
> am fairly certain now that And is talking in terms of 'systematic
> phonemes.'
Apparently. No wonder you talked past each other.
> In the case of Plan B, there _is_ a deep structure (tho not very deep) -
> the bit-stream. In Plan B, the surface phonology is generated by the bit
> stream. The first four bits (quartet) generate a consonant, the next a
> vowel, then a consonant, then a vowel, etc. Whether we have a consonant
> or vowel depends upon the _position_ of the quartet in the bit stream.
Yes. That's the way Plan B works, and where it differs fundamentally
differently from natural languages and causes a naturalistic analysis
to break down.
> Because, for example, 1100 can generate either [raj] or [s], And is
> taking [raj] and [s] as allophones of /C/ (where C is the hex digit for
> twelve), and the allophones are determined by the position of each
> quartet in the string of bits. This may well be correct in terms of
> 'systematic phonemics', but it clearly does not accord with phonemes of
> traditional phonemic phonology.
I concur with this. I, like you, understand the term "phoneme" in the
traditional sense.
> I don't know whether And will agree with me or not, but it seems to me
> that our difference over what are and are not phonemes in Plan B are
> because we talking about different things: And has, I think, been
> talking in terms of 'systematic phonemes'; I certainly have not.
Yes.
> > An example in a similar spirit dropped on the ZBB a week or so back. How
> > would you phonemically analyse spoken Solresol, in which the segmental
> > content of all utterances matches the regular expression
> > ((do)|(re)|(mi)|(fa)|(sol)|(la)|(si))* ? Seven morphophones or ten
> > phonemes, or something else?
>
> That depends how it's uttered. I understood that Sudre intended Soresol
> to be whistled, hummed or 'sung without words'. Thus in its purest form
> it has seven phonemes.
Yes. No doubt; after all, the musical notes cannot be meaningfully
decomposed any further - they are the fundamental building blocks of
the utterance, as phonemic as they can be.
> If, however, one for whatever reason doesn't
> whistle, sing etc but merely _says_ the names of the names of the
> different notes, then the number of actual phonemes uttered will depend
> upon the L1 of the speaker.
>
> In a sense, the speaker has turned the Solresol utterance into a subset
> of his/her own L1. It's as though instead of saying "Bonjour" /bO~Zur/
> (five phonemes), I were to say: "bee oh en jay oh you ar" - a good deal
> more phonemes :)
>
> That's the great thing about Conlangs - you have interesting things like
> Solresol & Plan B to keep you thinking :)
Yes, these things are interesting. Even to me, who otherwise is more
interested in naturalistic conlangs, with sound changes and all that.
That, too, is interesting enough; but it is always refreshing to deal
with an unnatural engelang - it teaches you not to take *anything*
in a language for granted.
On Wed, 11 Jul 2007 20:46:14 +0100, And Rosta wrote:
> What an interesting thread!
Indeed!
> R A Brown, On 11/07/2007 09:24:
[...]
> > I asked 'what' because I did not understand what you meant. I see from
> > below you are referring to a peculiar feature of Plan B's _morphology_.
>
> I don't agree that it is *morphology*. But more below.
>
> > Part of the problem is, I think, exactly what we mean by 'phoneme'. I
> > was using it to mean "The minimal unit in the sound system of a
> > language, according to traditional phonological theories" [Crystal]. I
> > thought it was clear that I have consistently been talking strictly in
> > terms of phonology. This seems to be the case with Jörg also.
>
> OK, so far we agree...
>
> > I know that there are different definitions of 'phoneme' given by
> > differing schools of linguists. But the variant allophones of a phoneme
> > are IME normally considered to similar in some manner or other.
>
> Well, as you say, there are different definitions & overlapping
> traditions and so forth. I think (a) that the phonetic similarity
> of allophones is merely typical of natural languages, and not criterial
> in establishing phonemicity & allophony; and (b) that (a) is consistent
> with mainstream views in the discipline.
>
> (b) is, of course, not to say that you would be wrong to insist that
> I am wrong on (a). But, firstly, I think that you might now on reflection
> find the 16-phoneme analysis of Plan B less risible.
I understand what you mean. So you would accept phonemes that consist
of utterly different allophones, such as [raj] and [s], or, for that
matter, the famous [h] and [N]? I'd personally define a phoneme as
"a contiguous domain in the human phonetic space, disjoint from other
phonemes, containing sounds which are considered the same in the language".
Surely, there is a connection between [raj] and [s] in Plan B, but
I think it would stretch things a lot to say that they are "the same
phoneme". They are *the same bit quartet*, and bit quartets don't live
in the human phonetic space :) The whole problem is that while in human
natlangs (and naturalistic conlangs), phonemes are the fundamental
building blocks of utterances, in Plan B they are not.
> Secondly, phonological theories that do treat phonetic similarity as
> criterial for allophony would tend to be 'cognitivist' (rather than
> 'abstractionist') and very much focused on the phonology of natural
> language.
Yes. And they are preferable for the analysis of such languages.
When you allow theories which describe languages in ways that have
nothing to with how human beings produce and analyse linguistic
utterances, you probably can "prove" everything and nothing about
language.
> Such theories would be very unsuited to intrinsically
> unnatural engelangs like Plan B. Thus, Plan B, by its nature,
> needs to be analysed in terms of an abstractionist model without
> a bias to natlangs
I understand. So you say, "This is not like a human natural language".
OK, I fully concur with you. So you continue, "Hence, we should not
treat it as such". Hmm, yes. Plan B is indeed different. Some
linguists would even deny that it was a language. However, using
words such as "phoneme" with new meanings is usually misleading;
you should at least put them in quotes. Ray and I don't have the
least doubt that the sounds [ej] and [S] have a connection in Plan B,
but we don't consider it reasonable to call it a "phoneme" because
it is a bit quartet, and as I said above, bit quartets don't live
in human phonetic space.
[...]
> > While I imagine a language in which [b] and [u] might be variants of the
> > same phoneme, I find it difficult to see what [d] and [a] have in
> > common, still less what [ej] and [S] has in common on Plan B.
>
> Let's agree for the sake of argument that the (putative) allophones
> are phonetically dissimilar.
Yes. They are utterly different sounds.
> > Listing the morphemes as /1/, /2/, /3/ etc looks to me just a fudge.
> > Normally (always?) when morphemes are given between slashes the
> > character is a phonetic one that _broadly_ denotes the range of sounds
> > the phoneme has in a particular language.
>
> I've long held the view that phonemes should be symbolized by arbitrary
> symbols.
That's your opinion; it is not mine, not Ray's (I think), and not most
linguists'. Of course, you can represent a phoneme whose focal
representation is [f] by |gh|, one whose focal representation is [i]
by |o|, and one whose focal representation is [S] by |ti|, but such
baroque spelling conventions will mislead just about everybody who reads
your transcription. To anyone familiar with the Latin alphabet, the
letter sequence |ghoti| just doesn't connote [fiS]!
> And there were some years a few years back when I, when teaching
> Intro to Phonetics & Ponology, would use phoneme symbolizations based
> on spelling (such as tend to be found in american dictionaries) rather
> than phonetic symbols.
>
> Again I'd argue that tenability of using a broad phonetic representation
> as a phonological representation is an accident of natural language.
This is because in human languages, phonemes are *contiguous domains
in the human phonetic space*, usually with a central focal area where
the most typical manifestations of the phoneme are found, and this area
can be adequately identified by means of phonetic symbols.
> (--Because there is scant chance for phonetically dissimilar allophony
> to come into being and get learnt by children.)
I would say that the chance for a child to learn something like Plan B
as L1 is nil - because it does not conform to the way human language
works.
> >> -- Then a morpheme /123/ can be [ged] or [iha]. It's because of these
> >> systematic equivalences that I think the 5-phoneme analysis is correct.
> >
> > We are here, surely, dealing with a _morphophonemic_ level of analysis.
> > Morphophonemes are normally symbolized with upper case letters as, e.g.
> > English {najF} which some posit as the morphophoneme of English _knife ~
> > knives_.
> >
> > Yes, in Pentaphone one could consider that the morphophoneme {123} may
> > be be realized as [ged] or [iha].
>
> I confess myself a bit hazy about morphophonemics, since afaik it was
> largely abandoned 50 years ago. My understanding is that morphemes are
> composed of morphophonemes in order to account for phonologically-
> conditioned allomorphy. In that case, {najF} wouldn't be an example,
> since the f:v alternation is not phonologically conditioned.
> But Z (/s:z:@z/) would be an example, as would suffix {iC}, where C
> is k:s (ethnic, ethnicity).
>
> On the basis of that understanding, I really think Pentaphon is
> correctly analysed in terms of phonemes & allophones, and not
> morphophonemes. There is no warrant to see [ged]:[iha] as allomorphy,
> since this sort of alternation affects every single morpheme in the
> language, in an entirely predictable way.
Leave morphophonemics out of the game, yes. As you say, it has been
scrapped, because it turned out to be a cul-de-sac. But I don't really
see why [ged]:[iha], or [Soubout]:[eiZeZru:], is not allomorphy
just because *every* morpheme has two shapes. Who says that allomophy
has to be restricted to a small subset of morphemes, especially if the
distribution rules are as simple as they are in Pentaphon and Plan B?
> > As I say, it depends how one defines 'phoneme'. Jeff Prothero does not
> > use the term in his description of his language. It is also clear to me
> > that he was not particularly interested in how it was pronounced, but
> > simply a gave a ad_hoc scheme whereby a string of four-bit groups could
> > be given a human pronounceable sound, without bothering what this might
> > imply for phonological or morphophonemic analysis.
>
> I quite agree. And therein lies a prime example of the theoretical
> interest of engelangs.
Yes. They raise questions whose answers one would otherwise wrongly
take for granted.
> > But IMO treating
> > > /1/ [g, i]
> > > /2/ [h, e]
> > > /3/ [d, a]
> > > /4/ [f, o]
> > > /5/ [b, u]
> >
> > ... as five _phonemes_ merits the satire of Jacques Guy's "Plan C."
>
> OK, so to summarize:
>
> One can take either a concretist or an abstractist view of phonology,
> according to the degree to which phonology has or lacks phonetic
> substance (and, by extension, whether phonemes are phonetic
> generalizations over allophones).
>
> You are a concretist, and think abstractism deserves satire.
> (I of course do not feel slighted by this satire!)
>
> I am an abstractist at heart. Both abstractism and concretism
> have a lot going for them as ways of modelling natural language
> in human minds. But for an unnaturalistic engelang, only an
> abstractist analysis makes sense.
True. What you call concretism is appropriate for natural(istic)
languages but tends to break down in engelangs. What happened in
this thread was that Ray and I used certain terms in a "concretist"
framework while you used the same terms in an "abatractist" framework.
And this was the source of our misunderstandings.
... brought to you by the Weeping Elf