Re: NEW LANG: Telek

Hey.

This is getting longer and more involved; lot's of fun stuff
below about Telek!

On Tue, 25 Apr 2000, Marcus Smith wrote:

> At 4/25/00 10:50 AM -0600, you wrote:
>
> >> Pitch-accent:
> >>
> >> Words start at a low pitch, and gradually raise until reaching the peak
> >> syllable.  Pitch remains high on every remaining syllable (usually only
> >> one).  Determining the peak is done by the following rules:
> >>
> >> 1. The peak is either the ultimate or penultimate syllable, whichever is
> >> heavier, if tied, on the ultimate.
> >> 2. The peak must be heavy.
> >> 3. If both syllables are light, the vowel of the penult lengthens and
> >> becomes the peak.
> >> 4. The peak may not be on the initial syllable, disyllabic words peak on
> >> the final syllable.
> >
> >Doesn't rule 3 contradict rule 1?
>
> Stated this way, yes it probably does.  I should rearrange the order of the
> rules: 3 before 1.  Actually, I was thinking about this intuitively as: peak
> must be heavy, do not lengthen word final vowel, put peak as close to the end
> of the word as possible.  Rather OT-ish.
Yes! If we assume the following constraints

        WtS: Weight to Stress "stressed syllables are heavy"
        NonInit: initial syllables may not be stressed
        NoLongF: final syllables may not contain long vowels
        Align-R: the stress is as close to the right edge as
        possible

and rank them in that order, the right results should pop out,
as the following tableax show (oh no! he's including tableaux!)

           /.cv.cv.cv./ | WtS | NonInit | NoLongF | Align-R
           ------------------------------------------------
           .CV.cv.cv.   |  *! |    *    |         |   **
           .cv.CV.cv.   |  *! |         |         |    *
           .cv.cv.CV.   |  *! |         |         |
           .CVV.cv.cv.  |     |    *!   |         |   **
        => .cv.CVV.cv.  |     |         |         |    *
           .cv.cv.CVV.  |     |         |    *!   |

           /.cv.cv./ | WtS | NonInit | NoLongF | Align-R
           ---------------------------------------------
           .CV.cv.   |  *! |    *    |         |    *
           .cv.CV.   |  *! |         |         |
           .CVV.cv.  |     |    *!   |         |    *
        => .cv.CVV.  |     |         |    *    |

To get cases where both of the final two syllables are heavy, a
top-ranked constraint MaxMora will be needed to insure that
there is no deletion of underlying moras to satisfy NoLongF (and
get stress on the ultima).

> >> Phonotactics
> >>
> >> The syllable template is (C1)V(C2).  Within a root and derivational
> >> morphology, syllables divide as CV.CV.  The sequence CV1.V2C is impossible,
> >> and in such situations, the vowel closest to the stem deletes: CV1.C in the
> >> case of prefixes, CV2C in suffixes.  Inflectional morphemes divide into
> >> syllables at the boundary, so it is possible to have CV.VC or VC.VC.
> >
> >This looks like Lexical Phonology; if so, it seems odd that in
> >the first stratum (stem + derivational morphology) you avoid
> >hiatus by deleting the vowel closest to the stem (could this
> >also be a stem vowel?); I would have expected the other vowel to
> >delete.
>
> I'm not familiar with Lexical Phonology.
Suppose you have a rule which palatalizes a coronal preceding a
front vowel within a root.

        /ti/ -> [Ci]

This means that there will be no roots in the language with the
sequence *[ti], all of the /t/'s having been palatalized. Now
suppose that you have a suffix -i, marking, say, plural. Now you
have a root /sat/ 'rock', a plural suffix -i, yielding [sati].
The palatalization rule only applied within roots, so the
complex form will be untouched, and you have a surface sequence
[ti]. This can be generalized to the claim that phonological
rules can apply both before and after morphological operations;
this is the heart of Lexical Phonology.

In Telek, you have a system of deletion which handles hiatus
between stems and derivational affixes, but a different system
between stems and inflection. This would be a prime candidate
for a lexical phonological analysis.

> I probably should have phrased this better -- I'm not used to others
> reading my
> explanations for Telek :).  It is certainly the stem vowel that deletes.  The
> idea here (and perhaps this is unnatural, but I thought otherwise) is that
> since there is less material in the affix, deleting material from the stem
> makes it easier to recover the meaning of the word.  For example, if I were to
> delete from the affix, the hypothetical suffixes -ik and -ak would both reduce
> to -k; but by deleting from the stem, they remain distinct.
A functional principle which the OTers have adopted is that
preservation of underlying material in the root or stem takes
precedence over preservation of underlying material from
affixes. That's why I thought this was unusual. There's a
corollary which states that affixes will contain "unmarked"
segmental material and "marked" segments will be confined to
roots and stems.

> >> C1 may be any consonant word initially, but is mildly restricted after
> >> another consonant: the glottals ' and h may not occur after C.  When this
> >> happens, the glottal disappears, and the preceding C geminates.
> >
> >Are coda consonants moraic? If so, I would expect that this is
> >result of preserving syllable weight in the leftmost syllable.
> >If not, I would have expected the glottal to disappear
> >altogether. Interesting ...
>
> The glottal stop and h don't have a place of articulation in the mouth, so
> they
> disappear, but the time slot is still present, so the preceding consonant
> spreads to fill it.  (Autosegmental thinking.)
If timing is represented by moras in Telek, doesn't this amount
to the claim that onsets are moraic? This would truly be
unusual!

> >> C2 may be any consonant word finally, but is severely restricted before
> >> another consonant.  Before another C, C2 must be unaspirated or assimilated
> >> to a following homorganic stop, e.g., ts > ds, td > dd, dt > tt, tk > dk.
> >> Note that this is not a restriction on codas, because this does not hold
> >> word finally, or in codas followed by a vowel, a situation found at the
> >> inflection-stem boundary.
> >
> >It just seems wrong not to think of this as a coda-condition,
> >but I can't think of any really convincing arguments. I'd like
> >to see some evidence that there really is a syllable boundary
> >between a C and a following V; I believe such things happen, but
> >there is always compelling evidence for it.
>
> If it were a coda condition, I don't know how I would explain the lack of
> neutralization word finally.  The only evidence I have for the syllable
> boundary between C and V is that when asked to say _etanoom_ "our heads"
> slowly, native speakers say et-a-noom.
>
> The only way to be sure would be to have a monosyllabic, transitive verb.  If
> there were a transitive verb _a_, we could use _namina_ "I am X-ing you" to
> check the location of the peak.  It would be naMINA is the syllable boundary
> corresponds with the morpheme boundary, naMIINA otherwise.  Unfortunately,
> there are no monosyllabic transitive verbs, so this cannot be tested.  (In
> fact, the only monosyllabic verb is the copula _ee_ . but it only appears with
> suffixes or a single prefix).
That would clinch it. Thanks for some conlang phonology to think
about! (Mind you, I can afford the time only because the end of
the semester is here and I don't have to prepare lecture notes!)

Dirk

--
Dirk Elzinga
dirk.elzinga@m.cc.utah.edu