Re: Double-segmentation (Was: brz, or Plan B revisited)

On Sun, 25 Sep 2005, Jörg Rhiemeier wrote, in reply to Patrick Littell:

> > On 9/22/05, Jrg Rhiemeier <joerg_rhiemeier@...> wrote:
> >
> > >
> > > There are many ways to achieve self-segregation.  Jeff's solution
> > > is elegant and original, but far from the only one.  A simple
> > > self-segregation system I once came up with has morphemes of the
> > > following structures:
> > >
> > > C
> > > CVC
> > > CVCVC
> > > CVCVCVC
> > >
> > > etc., i.e. alternating consonants and vowels beginning and ending
> > > with a consonant.  In this system, all morpheme boundaries are
> > > marked by consonant clusters, and every consonant cluster marks
> > > a morpheme boundary.  For example, _blaraktalmin_ can only be
> > > segmented as b-larak-tal-min.  If every word has to begin with
> > > two consonants in a row (i.e., with a C morpheme), word-level
> > > self-segregation is also achieved.
> >
> > Quite clever!  Self-segmentation is pretty easy, but a good method of
> > double-segmentation (morpheme and word) like this takes some thought.
>
> Yes.
>
> > This would also require, of course, that C morphemes could only occur
> > word-initially and word-finally.  Internal C morphemes would lead to
> > word-level ambiguity.
>
> C morphemes would have to be restricted to *one* end of the word,
> *either* initial *or* final.  Otherwise, a single C between two
> other morphemes would be ambiguous as for to which word it belongs.
> Unless, of course, each word is required to have one C morpheme
> on *both* ends.
>
> But I'd now say that C morphemes should not exist, because they
> lead to awkward consonant clusters.  I'd propose something else:
>
> 1. The language is exclusively suffixing.  Nothing (except another
>    root followed by /a/, see 5.) may precede the root.
>
> 2. Roots are alternating consonant-vowel sequences, beginning end
>    ending with a consonant, with at minimum one vowel, i.e. CVC,
>    CVCVC, CVCVCVC, etc.
>
> 3. Suffixes are alternating consonant-vowel sequences, beginning
>    with a vowel and ending with a consonant, with at minimum one
>    vowel, i.e. VC, VCVC, etc.
>
> 4. All vowels in a morpheme are the same, and only /e/, /i/, /o/
>    and /u/ occur.  Each suffix has two allomorphs, one with the
>    vowel /i/ and one with the vowel /u/.  The /i/ allomorph is
>    used after a morpheme with a back vowel; the /u/ allomorph is
>    used after a morpheme with a front vowel.
>
> 5. In compounds, a vowel is inserted between the two roots that
>    occurs neither in roots nor in suffixes: /a/.
>
> If I haven't made a mistake, this should be self-segregating at
> both morpheme level and word level.  A change of vowel indicates
> a morpheme boundary before it (with the exception of /a/, which
> marks a boundary between two roots); two consonants without a
> vowel in between mark a word boundary.
>
> > I would picture some grammatical category like person or number that
> > could occur on nouns, verbs, etc., marked obligatorily with a C
> > morpheme prefix.  (Or as a suffix, of course.)  Say, person is marked
> > on nouns for either the inherent person features of a noun or those of
> > its possessor, and on verbs for the subject.  (Object could be marked,
> > too, 'cuz there's space for one more C morpheme word-finally.)
> > Nothing particularly unnatural about this.
>
> That's also what I was thinking of when I proposed C morphemes
> at word boundaries.
>
> > Also possible, although weird, is a requirement that the C morpheme
> > must be the 2nd morpheme in the word, or second-to-last, or etc.
> >
> > --------------
> >
> > A couple other methods of naturalistically self-segmenting these on
> > the word level:
> >
> > 1) The final consonant of a morpheme must be a stop, and word-internal
> > sandhi rules cause them to fricativize:
> >
> >    kotuk-qap-t-mit => kotuhqafsmit
>
> An interesting idea.  But the consonant clusters /hq/ and /fsm/
> in your example are hideous.
>
> > 2) The final consonant of a morpheme must not be a stop, and all words
> > undergo a word-final stop mutation.
> >
> >   kotun-qam-s-min => kotunqamsmit
> >
> > And, of course, the many variations of these using different consonant
> > series, word- and morpheme-initial mutations rather than final ones,
> > etc.
>
> Or encode morpheme boundaries in vowels, as I did above.
>
> > Any more ideas?
>
> See above.
>
> Greetings,
> Jrg.
Hi Jörg,

I'd say your scheme was both logical and ingenious.

BTW, I disagree that the consonant clusters arising from
the other scheme (Patrick's, if I remember correctly),
are hideous - they do seem to be quite pronounceable, and
to respect the kind of sonority hierarchy shown at
http://en.wikipedia.org/wiki/Sonority_hierarchy.

I rather like the sound of "kotuhqafsmit" - sure beats
trying to pronounce Nuxálk (Bella Coola)!

But back on-topic - what kinds of schemes occur in
natlangs, and with what kinds of frequency, for both
morpheme- and word-level segregation?  And if such data
is available, are there any plausible theories to account
for them?

Regards,
Yahya


--
No virus found in this outgoing message.
Checked by AVG Anti-Virus.
Version: 7.0.344 / Virus Database: 267.11.6/111 - Release Date: 23/9/05

Reply