Kristian Jensen <[log in to unmask]> writes:

> Jörg Rhiemeier wrote:
>
> > [formula vs. naturalistic approach to protolanguages]
>
> I went for the formula approach.
>
> > [internal reconstruction]
>
> That's the basically the way I pictured how linguists reconstructed
> Proto-B. The formula-like regularity of the reconstructed inventory
> is a direct result of the complete lack of material.

Assuming that an irregular pattern evolved from a regular one through
a series of sound changes that messed up the original paradigm is in
corcordance with Occam's Razor; assuming the inverse development
without solid evidence for a previous irregular system is not.  Hence,
a methodologically sound internal reconstruction will, in most cases
at least, yield a system more regular than the earliest attested one.

> > > [nasal harmony in B]
> >
> > I.e. /bu/ -> [bu] but /bu~/ -> [mu] (or something like that), as I
> > already guessed above.  Do voiceless stops yield voiceless nasals?
> > And what about continuants?
>
> All voiceless sounds are transparent to nasality. That is, they don't
> host nasality by turning into nasals themselves, but they don't block
> the spread of nasality either. So voiceless stops and fricatives remain
> stops and fricatives, even when the surrounding sounds are nasals.
>
> Approximants, being voiced, are hosts to nasality -- i.e., they become
> nasalized.

I.e. /B l y Y R/ > /m n n' N N'/ (n'=palatal, N'=uvular nasal), the
same way as /b d j g G/ > /m n n' N N'/, or is there a difference?

> > [outline of a vowel harmony system]
>
> You may be interested to know that there are actually languages
> with vowel height harmony. I seem to recall them being in Africa and
> Central America. The harmonic rules are something to the effect of
> having only high and low vowels in one set, and only mid vowels in
> another set.

Tungusic and Chukotko-Kamchatkan languages have a high-low harmony
with a high set /i e u/ and a low set /e a o/, if I am not mistaken.
Note that /e/ is in both sets without being neutral (high /e/
corresponds to low /a/, while low /e/ corresponds to high /i/);
probably, the high set was /i @ u/ or something like that before /@/
merged with /e/.

But back to Boreanesian now.

> -----<snip>-----
> > > Okey dokey! Here are some made-up examples just to keep things simple.
> > >
> > >    /su + diw/             > sudiw
> > >    /su + kay/             > sukway
> >
> > I.e., /u/ labializes following velar.
>
> Yes!
>
> > >    /pkway/                > pukway
> >
> > /p/ is labial, hence the epenthetic vowel is [u].
>
> Nope. First, since /k/ is peripheral, it can host the rounding of /w/.
> In effect, /kw/ is a permissable consonant cluster. Second, since /k/
> is peripheral and does not block the spread of rounding, the epenthetic
> vowel between /p/ and /k/ is [u].
>
> Actually, upon reflection, you could be right! If the underlying
> structure is /pwkay/. It would surface the same way [pukway], since
> /pw/ is a legal cluster (viz. /p/ is peripheral and compatible with
> rounding). So you could argue, /pw/ is labial_ized_, hence the
> epenthetic vowel is [u]. /w/ would then assimilate with this epenthetic
> [u]. And with /k/ another peripheral consonant, the rounding proceeds
> rightward. Like I mentioned before, it is extremely difficult to tell
> whether the rounding comes from consonants or from vowels. For all
> practical purposes, the underlying form could also just as well be
> /pukay/.

Either way or the other, the epenthetic vowel is [u].

> A better representation of [pukway] is through the autosegmental approach:
>
>    R LF
>    /\||
>    pkay
>
> (where R is the autosegment for rounding, L for low, and F for fronting.
>
> Here I don't have to worry about whether rounding comes from a vowel, as
> in /pukay/, or from a consonant, as in /pwkay/ or /pkway/. All three
> possibilities would surface as [pukway], making it impossible to state
> where rounding is coming from.
>
> I'll show what's going on autosegmentally in the next examples.

Yes, that will help.

> >
> > >    /si + pkway/           > sipukway
> > >    /T@ + pkway/           > T@pukway
> >
> > What is /T/?  A voiceless dental fricative, or something else?
> > A laminal stop (or fricative)?
>
> Sorry. Laminal denti-alveolar stop. It's just me being uncreative in
> the transcription.

I guessed that.

> Autosegmentally this is:
>
>    F   R LF        F  R LF
>    |   /\||        | /|\||
>    s + pkay     > sipukway
>
>    H   R LF        H  R LF
>    |   /\||        | /|\||
>    T + pkay     > T@pukway

One moment.  /s/ and /T/ are both laminal, aren't they?  So why
do they yield different epenthetic vowels?  Is there still
another rule I have missed?  Or are the left sides mis-typed and the
examples ought to look like below?

    F   R LF        F  R LF
    |   /\||        | /|\||
   si + pkay     > sipukway

    H   R LF        H  R LF
    |   /\||        | /|\||
   T@ + pkay     > T@pukway

If it was like this, things would look much clearer.

> > >    /d + pkway/            > dupukway
> >
> > /d/ is apical and thus neutral, hence the epenthetic vowel
> > duplicates the vowel of the following syllable.
>
> Something like that. More precisely, the rounding of /w/ spreads
> across two peripheral consonants: /p/ and /k/.
>
> Autosegmentally this is:
>
>        R LF          R  LF
>        /\||        ///\\||
>    d + pkay     > dupukway

As clear as it could be.

> > >    /d + Tway/             > d@Tuway
> > >    /d + pkway + Tway/     > dupukwa:cuway
> > >    /bkwi + T + pkway/     > bukwicupukway
> >
> > I can't see what is going on here, mainly because I don't know what
> > /T/ is.
>
> Yeah... sorry'bout that.
>
> Anyways, in the first one, /T/ is laminal and thus is incompatible
> with rounding. *[Tw] is therefore an illegal cluster and an epenthetic
> [u] is inserted. /T/, being laminal, also blocks the spread of rounding.
> Rounding cannot therefore spread backwards in the first one. The
> epenthetic vowel is by default [@].
>
> The second one shows that rounding spreads to the same prefixed light
> syllable because the intervening consonants are peripheral. It also shows
> /T/ palatalizes to [c] since laminal consonants are hosts to the fronting
> of /y/ (I should have written /j/). Loss of /y/ lengthens /a/.
>
> The third one should be clear by now. I hope.
>
> Autosegmentally this is:
>
>         RLF                R LF
>         |||                /\||
>    d + Tway           > d@Tuway

Perfectly clear.  /T/ blocks the spread of roundness, thus the
epenthetic vowel is [@].

>        R LF    RLF        R  L FR LF
>        /\||    |||      ///\\/\|/\||
>    d + pkay + Tway   > dupukwa:cuway

As you said, /T/ palatalizes due to preceding /y/, the resulting
cluster [yc] is simplified to [c] and the preceding vowel is
compensatorily lengthened.  Clear.

>    R F       R LF       R F   R  LF
>    /\|       /\||      /|\/\///\\||
>    bki + T + pkay   > bukwicupukway

Yes, /i/ palatalizes /T/ again.

> > > Hope that helps.
> >
> > A bit at least.  A lot of things have become clearer to me.
>
> I'm glad that someone shows great interest.

It has almost developed into a private thread between us two
(everybody who reads this is invited to join); well, it is pretty
advanced theory-heavy matter that probably baffles quite a number of
list members.  But it is highly interesting stuff which inspires me to
experiment with phonology as well.

Jörg.