Re: PIE Soundchanges - Grassman & Bartholomae

Paul Bennett wrote:
> I know there's some controversy about how exactly Grassman's and
> Barthlomae's laws took place, but I'm working on Thagojian 2.0, and I
> *think* I've come up with a way to succinctly express a wild stab at
> something approaching them. How does this look...?
>
> Ch > C / ChV? _ s
> Ch > C / _ VCh
> (C)h(C) > ($1)($2)h
>
> Actually, Henrik: are rules like that third one allowed in SCH files? More
> to the point, given the syllable-based approach of SCH, is a rule like
> that even easily to implement across syllable borders?
>
Can't answer that question; But is #3 an IE rule, or just a Sanskrit or
Indo-Aryan rule??

IIRC rules of that type are non-kosher but  _tolerated_ in generative
phonology (structurally they're like grammatical transformation-rules); but
otherwise it's very cumbersome, using only rewrite rules, to deal with two
or more segments at a time.

It could be: 1. C > Ch /Ch__(additional env. X probably needed)
   2. Ch > C/_ChX

So the classic ex. /budh+ta/ > **budhtha > **budtha (plus a cluster voicing
rule) > /buddha/; whether that was the actual sequence of events is
debatable.

This happens to be a preoccupation of mine at the moment, since I'm trying
to reduce the 50-60+ narrative pages of Gwr sound changes to a series of
rules, hopefully using distinctive features. (And how rusty I've become.
Grr) Surprisingly, there aren't that many rules!! Equally surprisingly (or
not :-(( ), I'm finding all sorts of inconsistencies.

A comparable problem:  what would be the best way to accomplish--

*pátiC ult. > **pajt (the -C is relevant, since *páti simply > **pat)

one way: 1. V(-str) > homo.glide or "reduced" / _C
2. C > 0 / [reduced V]__#
3 (metathesis transform., not a rewrite)
    [V+str]C[V 'reduced']# ==> [V+str][V reduced]C #
     1     2 3           4      1      3         2 4

Another way, using assimilation (cumbersome IMO and requires a no-no [IIRC]
zero-rewrite rule)
1. 0 > [V reduced] / [V+str]__C[V reduced]C#

2. a later clean-up rule deletes the final reduced VC.

Interestingly, the reverse happens in case of an unstressed 1st syl. V (e.g.
*bidák) in certain environments; the same 1234 ==> 1324; but I'm pretty sure
the two rules can't be conflated to one. (*bidák > bdyák > dyak ult. >
**dzak

I've also been experimenting with ternary valued features +1/0/-1 or simply
+/0/-. It look promising (though such features are controversial)--
+stress -- 0 stress (= -str in standard binary notation) -- -str =
reduced/glide status.