Re: [engelang] phonology - engelang messages

On Sun, Sep 16, 2012 at 9:06 PM, And Rosta <and.rosta@hidden.email> wrote:

Mike S., On 16/09/2012 22:47:
> On Sun, Sep 16, 2012 at 11:02 AM, And Rosta <and.rosta@hidden.email
> <mailto:and.rosta@hidden.email>> wrote: We have not discussed whether
> there will be tone, but -- assuming that if there were tone it would
> have low tone only on all word-final syllables -- having tone would
> affect the word boundary rules; word-final CVs could also be
> word-initial without ambiguity. So far we've been assuming that
> word-final CVs can't be word-initial.
>
> Well, sure. And if we had clicks, we wouldn't be as quickly filling
> up of operator space. That's not a serious proposal; I'm just
> pointing out that given a tone-based SSM, everything that we now
> have could easily change beyond recognition.

Exactly. Nobody's proposed clicks. (Livagian has them, but I think real-world users would be put off by them.) Whereas, you've described a simple tone-based word-segmentation system, and we haven't really discussed whether Xorban should use it. As I've written before, I see pros and cons for it.

Yes, true, I did mention that, and I did give it some brief thought that I didn't mention. My thoughts are these:

- If we are going with tones, it will obviously need to be simple, i.e. H & L. Even then, expect some people to be frightened.

- Tones need vowels, mainly. Xorban formally pretends that predicate stems don't have vowels, and lots of times they actually don't. So no matter how you slice and dice the "words", all the distinctions that tones could make right now would effectively be among closed class morphemes i.e. variables and operators. We don't really need to do that, precisely because they are closed class. In other words Xorban as it stands is a poor candidate for tonalization AFAICT.

- If we wanted to put vowels back in Xorban predicate stems, then we would have a good reason for tonalization. We could make the following rules:

simple-formula := stem variable (k variable)*
stem := C (H | C)* C
operators := (n | r | s | [etc...] ) variable
variable := L L*

Stems are forms starting and ending with a consonant (but not the same one) with zero or more H or C in between, and names could be, say, all the stems that start with "q". Presumably we'd keep the implicit buffer vowel crutch around for geminates and the like. Vowels outside of stems are L.

>> What is the phonological value of <h>, which I have been using
>> experimentally?
>
> Nothing but [x, h] makes any sense.
>
> Agreed, and probably only one of /x h/ should be allowed.

I don't know what you mean. You mean there shouldn't be a phonemic contrast between [x] and [h]? I agree; the contrast is too feeble. Or do you mean that the realization of /h/ shouldn't be allowed to range over [x, h]? I disagree; the restriction is pointless.

No phonemic contrast, but either phonetic realization for what would conventionally be spelled <h>.

>> Voiced velar fricative?
>
> I have been assuming that is the value of <x>.
>
> I cannot help noting that that would make Jorge's Xorban name,
> spelled as it is, sound like the name of an orc, but I understand the
> reasoning behind what you're proposing. "Jorge" would be "hrhe" and
> <x>=[G] would fill a gap in the putative phonology.

Yes.

> Then again, with [x] and whatever <'> is already on the table,
> adding [G] really starts to make things crowded.

<'> is [?], so there's no crowding.

When or if it becomes that.

>> I don't think there's a problem with <'> being [h] (which gives us
>> 25 extra variables) other than its stunning orthographic weirdness,
>> so long as <x> is used in predicates but left unassigned for
>> operators. Otherwise CCCi xi is going to clash with CCCi'i. This is
>> definitely an issue, download Audacity and try it for yourself.
>
> The orthographic weirdness is a problem. Furthermore, we have
> assigned <'> as an allograph of <q>, [?], and therefore bcda'afga'a
> is not "bcda'a fga'a" but rather "bcda 'afga'a".
>
> Isn't <x> as [G] also weird? I can't think of any language that does
> that. And while I'm board with <'> as [?] especially for the
> quotative function, <q> as [?] is pretty much limited to Maltese,
> isn't it? Yes I know you want to assign the most reasonable values
> to these extra letters; just saying.

<x, q> for /G, ?/ is less weird than <'> for /h/. Also less unwarranted.

IMVHO they're all weird.

> Really, the language as it is is speakable and has a clear SSM. If
> Jorge wants to keep things Lojban-looking then the only thing that
> IMHO needs actual fixing is the /h x/ distinction. I would recommend
> scrapping "x-" as an operator and making <'> and <x> allographs of
> the same phoneme. A slightly more complex proposal is to change <x>
> to <h>=[x], reintroduce <x>=[G], and forbid <h x> from being
> operators. Then <'> would be the neutralization of /x G/ between
> vowels and could be realized as any velar or glottal fricative [x G h
> h\] preferred by the speaker.

I can't digest that, because I'm not sure what counts as "keeping things Lojban-looking". But anyway, if keeping things Lojban-looking is for you one of the criteria for the phonology and/or orthography, then rather than argue over that, we can just do separate phonologies.

I switched to what I think is Jorge's point of view when I mentioned changing nothing except the phonemic <x>-<'> contrast. By Lojban-looking I just meant keeping a similar-looking orthography and (somewhat) similar-sounding phonology including <'>. I don't mind discussing as long as you like hearing ideas. If clicks and tones ultimately end up in the language, that's fine by me. I am not going to make a separate phonology. I am going with the flow on phonology related stuff.

> I agree with you on the basic seven vowels /a e i o u w y/=[a E i O u
> y 9] where /y/ could also be [@] and /a/ is any low vowel.

Given the presence of 9, acoustically more similar to [@] than [y] is, I'd have thought it wd be 9 that [@] wd be an allophone of.

Sorry for my sloppy notation; <y> is [9] or [@] and <w> is [y].

> I would say that /i u/ could be underlyingly semivowels or glides
> which I will symbolize as [I U] or consonants [j w], but can also be
> realized as a sequence of vowel and consonant. Yes they serve as
> separators. I would constrain vowel strings composed of /a e i o u/
> as follows: No geminates which means no /ii uu/; no sequences of two
> non-high vowels /ae ao/ etc. However everything else is allowed with
> the following sanctioned phonetic variation:

> 1) After a non-high vowel, /i u/ is pronounced [I U], but may also be
> pronounced [ji wu]. i.e. each of /ai au ei eu oi ou/ = [aI aU EI EU
> OI OU] or [aji awu Eji Ewu Oji Owu] respectively.
>
> 2) Before any vowel, /i u/ pronounced [I U] or [j w], but may also be
> pronounced [ij uw] i.e. each of /ia ua iu ui/ etc. = [Ia Ua Iu Ui]
> etc. or [ija uwa iju uwi] etc. respectively.
>
> 3) between two vowels, /i u/ are pronounced [I U] or more likely [j
> w], though even [jij wuw] is allowed.
>
> The second rule facilitates sequences like /Tia Pua/ where P is a
> labial obstruent and T is a coronal obstruent. So /tia pua/ can be
> [tija puwa].

We disagree on the details, but since your focus is elsewhere, I won't respond for the time being.

Okay.

> As far as /w y/ which slightly confusingly are [y 9] as in Loglan,
> although I know one is going to agree with this, I would use them in
> stems in the following way.
>
> stem := (y) root (y root)* root := C (w) C ( (w) C )*
>
> /y/ would not be elidable,

A stem would begin with unelidable /9/?

That <y> would be elided if the onset were pronounceable, or if preceded by another word, and it wouldn't have to be spelled either way.

> but /w/ would always be optional between any two consonants within a
> root without changing the meaning of that root. I would rebuild the
> lexicon from the ground up so that all "native" roots would be
> further constrained such that all <w> are easily (by SAE/Lojban
> standards) elidable.

Lojban standards don't seem very SAE to me, but I agree that the elidability of schwas should be proportional to the frequency of the root.

That would be the goal.

>> I am pretty sure I and And agree on the phonotactics of <y w>. If
>> we are going to use these for interjections, I don't like *yi or
>> *wu, but those are easily replaced with yay and waw.
>
> Jorge's yi would be [i], unambiguous in isolation, but it leads to an
> overly feeble contrast between /i/ and /iyi/.
>
> Are you sure? [ji] is not impossible to produce or perceive as long
> the first segment is maximally raised and the second lowered or
> diphthongized slightly, it just seems to be a minor hardship on some
> folks.

If [ji] had to contrast with [i], that would be unacceptable, and I think /i/ contrasting with /iyi/ is unacceptable, but my point is that in Jorge's scheme there is no syllable /i/, so word-initial [i] is /yi/.

We agree with this and I have already said that I am not keen on [ji] or [wu], by which I mean contrasting with [i], [iji], etc. I have a hard time getting my head around word-initial [i] being /yi/, even if it's entailed by a more general onset rule. This is not a proposal, but a question: if /yi/ is really [i], why not just say words can start with vowels and let everything fly. Geminated vowels at word boundaries could be long.

> Lojban attitudinals begin with q followed by a vowel.
>
> Yes though not all of them.

I'm not sure which UI count as attitudinals, but I don't dispute what you say.

I don't know either, so please ignore my claim.

> The way I look at it, the reason that most of them start with the
> glottal stop follows from a more general rule that all syllables
> start with a non-null onset.

Wouldn't it be more meaningful to say "the reason that all of them start with a consonant follow from a more general rule that all words start with a consonant"? Or would you rather not consider /./ a consonant?

I think the intent of Lojban .uV, .iV is that the onset be .u, .i and the nucleus be the V, so the /./ isn't necessarily filling an otherwise empty onset -- tho if V are always nuclei then your formulation works.

I think that Lojban's intent is made clear by things such as forms like <ia> being called "diphthongs", semivowels being spelled the same way as vowels, and semivowels not counting as consonants for the purpose of defining brivla boundaries. (Granted there are no rafsi like "kua" without the apostrophe to test this, but if there were, then my understanding is that they would be treated like CVV and not CCV; cf. "tai", not analyzed as CVC.)