>
> Something more compositional (e.g. _H _LH _L _HL) is simpler, and far far
> more transparent, if admittedly longer.  I hereby give you a friendly push
> in that direction.


duly noted.  i kept forgetting what the pinyin numbers referred to anyway.
they were also hella inconvenient for my scheme since i had 1, 2, and 4 but
no 3.


> On reading this, this was the first variation / analysis that came to my
> mind, one where the length distinction wasn't lost.  Short vowels are one
> mora, long are two.  Each mora is unmarked (= low) or high; every long
> vowel
> has to have at least one high mora, and both may be high at once; short
> vowels cannot have high morae.  Which I think is a fine system, and more
> likely to explain why things would come about this sort of way.  (Of
> course,
> you could then take my system and lose length but retain the pitches.)
> The flavour of this reminds me of the Slavic accentual developments.
> Someone else here can probably recommend a better place to read about them
> -- I was somehow cowed by most treatments I bumped into until finding Geoff
> Eddy's take on them for a Latin-to-Slavic bogolang:
>  http://sleepingsages.com/kalinin/GeoffEddy/bogo.html
> but I guess the first place I'd point (besides WP) is
>  http://www.kortlandt.nl/publications/art066e.pdf
>

WOW.  thanks for the articles.  also, i approve of the mono/di-moraic
distinction.  i was sort of planning on doing that anyway.  BUT i have a
question:


> To generalise a bit what Roger's been saying, the basic principle is that
> the prominent element receives the prominent tone (usually high, but it can
> be low), and the non-prominent tones on deleted elements accrete onto it
> and
> make contours.
>
> Your basic problem, at this level, is that both changes (2) and (3) have
> the
> first of two tone-bearing units survive (*V*CV and *V*G resp.), so they
> should both turn out prominent-tone-first.
>
> If you don't want them to merge, my best advice is to turn one of them to
> HL
> (perhaps (3)), then do some shifts in the tone system, then turn the other
> one ((2)) to HL now that the old HL is out of the way.  Perhaps like this;
> this switches the outcomes of (1) and (2) but my gut feeling is it's more
> plausible that way.  [View in fixed pitch.]
>
>  V:      VCV     VG      protolang
>  V_HH    VCV     V_HL    first tonogenesis
>  V_H^    VCV     V_HH    tone shifts stage 1: raise second elements
>                          (^ means extra-high)
>  V_LH    VCV     V_HH    tone shifts stage 2: lower the H^ unit as a whole
>  V_LH    V_HL C  V_HH    second tonogenesis
>

i like your analysis and thank you profusely for the suggestion.  i'm
wondering whether i can extend the idea a bit using this layout of
diphthongs from Classical Sanskrit:

a           ā          ai          āi
/a/         /a:/       /e:/        /a:i/
forget about /e:/.  my question: can i interpret the diphthong as a long
vowel + diphthong = 3 morae?

IF this can work ... i can set up my initial conditions thus:

V:             VCV             V:G
V_HH        VCV             V_HHL

this doesn't seem to me to be an outrageous stretch.  but tell me what you
think about the following chain shift:
V_HH        VCV            V_HHL
V_HL        VCV             V_HHL
V_HL        VCV             V_HH
wherein first the last mora of HH "weakens" (lowers) and second the last
mora of HHL drops out.  if THIS is acceptable, i can then do your extra-high
trick thus:
V_HH        VCV            V_H^
V_HH        VCV            V_LH
V_HH        V_HL           V_LH

thoughts?

matt


On Fri, Nov 12, 2010 at 5:48 PM, Alex Fink <[log in to unmask]> wrote:

> On Wed, 10 Nov 2010 14:50:07 -0500, Matthew Boutilier <[log in to unmask]>
> wrote:
>
> >i have some ideas for tonogenesis.  they aren't the standard voiceless
> cons.
> >> high tone, etc., but i think they're plausible so i thought i would
> >shamelessly post them and hope for comments.
> >
> >in chronological order (if it matters; also, i'm using the pinyin numbers
> >for simplicity):
>
> Something more compositional (e.g. _H _LH _L _HL) is simpler, and far far
> more transparent, if admittedly longer.  I hereby give you a friendly push
> in that direction.
>
> >1) V: > V_1
> >e.g. lo:kam > lo1kam
> >basically, vowel length ceases to be phonemic (chronemic) and long vowels
> >are realized as ones with *high *pitch
> >
> >2) VCVCV > V_4CCV
> >e.g. todumag > to4dmag ( > to4zma )
> >the syncope rule: wherever syncope occurs, the preceding vowel
> "compensates"
> >(to use a sufficiently vague term) by developing *falling* tone
> >
> >3) VS > V_2
> >(S = semivowel)
> >e.g. pelwoi > pelwo2
> >basically, falling diphthongs become pure vowels with *rising* tone.
>  yeah,
> >i know diphthongs =/= vowel+semivowel but here the difference is trivial
> >enough.
> >
> >[4) none of the above - which will apply for most of the syllables in a
> big
> >word, and most monosyllabic words - implies no tone, or neutral tone, or
> >whatever.]
>
> On reading this, this was the first variation / analysis that came to my
> mind, one where the length distinction wasn't lost.  Short vowels are one
> mora, long are two.  Each mora is unmarked (= low) or high; every long
> vowel
> has to have at least one high mora, and both may be high at once; short
> vowels cannot have high morae.  Which I think is a fine system, and more
> likely to explain why things would come about this sort of way.  (Of
> course,
> you could then take my system and lose length but retain the pitches.)
>
> The flavour of this reminds me of the Slavic accentual developments.
> Someone else here can probably recommend a better place to read about them
> -- I was somehow cowed by most treatments I bumped into until finding Geoff
> Eddy's take on them for a Latin-to-Slavic bogolang:
>  http://sleepingsages.com/kalinin/GeoffEddy/bogo.html
> but I guess the first place I'd point (besides WP) is
>  http://www.kortlandt.nl/publications/art066e.pdf
>
> >thoughts?  does this make any sense whatsoever?  should 2) and 3) be
> >reversed?  wiki, fount of all knowledge, says that falling diphthongs
> often
> >have higher pitch on the more prominent first element ... but how
> universal
> >does this have to be?
>
> To generalise a bit what Roger's been saying, the basic principle is that
> the prominent element receives the prominent tone (usually high, but it can
> be low), and the non-prominent tones on deleted elements accrete onto it
> and
> make contours.
>
> Your basic problem, at this level, is that both changes (2) and (3) have
> the
> first of two tone-bearing units survive (*V*CV and *V*G resp.), so they
> should both turn out prominent-tone-first.
>
> If you don't want them to merge, my best advice is to turn one of them to
> HL
> (perhaps (3)), then do some shifts in the tone system, then turn the other
> one ((2)) to HL now that the old HL is out of the way.  Perhaps like this;
> this switches the outcomes of (1) and (2) but my gut feeling is it's more
> plausible that way.  [View in fixed pitch.]
>
>  V:      VCV     VG      protolang
>  V_HH    VCV     V_HL    first tonogenesis
>  V_H^    VCV     V_HH    tone shifts stage 1: raise second elements
>                          (^ means extra-high)
>  V_LH    VCV     V_HH    tone shifts stage 2: lower the H^ unit as a whole
>  V_LH    V_HL C  V_HH    second tonogenesis
>
> Alex
>