Re: Order of Some Indo-Iranian Sound Changes

--- In cybalist@yahoogroups.com, "david_russell_watson" <liberty@...>
wrote:

>
> --- In cybalist@yahoogroups.com, "tgpedersen" <tgpedersen@> wrote:
> >
> > --- In cybalist@yahoogroups.com, "david_russell_watson" <liberty@>
> > wrote:
> > >
> > > Is it known, or is there any way of knowing, which of
> > > the changes came first?
> >
> > I know there is a problem with the ordering of these two
> > rules in IIr., but I can't remember what it was. However,
> > I do remember I think I solved it:
> > http://tech.groups.yahoo.com/group/cybalist/message/47124
>
> I see. You wrote at that time:
>
> > Three traditional rules for IE:
> >
> > 1 RUKI (Balto-Slavic, Iranian)
> > 2a *-dhT-, *-dT- -> *-zd-, *-tT- -> *-st- (Balto-Slavic, Iranian)
> > b *-dhT- -> *-ddh-, *-dT- -> *-dd-, *-tT- -> *-tt- (Indic)
> > c *-dhT-, *-dT-, *-tt- -> *-ss- (Celtic, Germanic, Italic)
> > where T is any stop
> > 3 stop + consonant -> corresponding fricative + consonant
> > (Iranian)
> > 4 stop -> fricative etc (Germanic, Armenian)
>
> The changes in 2 involved the dissimiliation of pairs
> of dental stops only, which aren't part of the Satem
> change of affricates to fricatives before dental stop,
> though. None of the affricates were dental then.
>

They are part of the rule sequence I propose to reshuffle and redo.

> In Indo-Iranian, at least, the changes you give in 2a
> are patently later than both of the changes I'm asking
> about,

They are later in the traditional rule set. That's not 'patently'.

> and the changes in 2b didn't happen at all, as far as I know.

Bartholomae. Barrow p. 88-91.

> > Sequence 1 < 2, 2 < 3, 2 < 4
> >
> > 1 must have come before 2, as Beekes (A Grammar of Gatha-Avestan)
> > points out, since the sibilants in outcome of the assibilation
> > rule otherwise would have been affected by the RUKI rule which
> > they aren't (Avestan cit + ti -> cisti, not **cis^ti).
>
> That change isn't shared with Sanskrit, though, which
> has citti-, and so is post Proto-Indo-Iranian, whereas
> the change I'm asking about is common Satem.

And?

>
> > It seems strange that the assibilation rule which seems to have
> > applied over a large area in some form, should have come after
> > the RUKI rule which is limited to Balto-Slavic and Iranian (and
> > Armenian, I think).
> > Also, the Iranian rule 3 seems to be similar to and in competition
> > to the assibilation rule.
>
> I don't understand how it's in competition.


> > Also, it is strange that almost all IE languages agree that the
> > otherwise non-affricated dentals of PIE suddenly should be
> > assibilated when they meet.
>
> T. Burrow, I think it was, explained this as due to the
> purely phonetic insertion of [s] between two voiceless
> dental stops in P.I.E., or [z] between two voiced ones,
> later eliminated in some branches, but phonemicized in
> others.

That makes no sense phonologically.

> > Therefore I propose the following sequence of rules to replace
> > those above:
> >
> > 1 stop + stop -> corresponding fricative + stop (PIE)
>
> This would make *x^t of *k^t then, which *x^ would be
> equivalent to, or later merge with, the *s´ arising due
> to RUKI? I haven't thought it completely through, but
> that might work, by itself.



> However in this step you've produced a slew of clusters
> beginning with bilabial fricatives, dental fricatives,
> and velar fricatives, both voiceless and voiced, all of
> which have to be converted back into stops before we
> arrive at Sanskrit, and you haven't even mentioned Greek.
>

But not in Sabellic, Germanic and Iranian. And the conversion id
driven by paradigm regularization, which is a natural tendency, not an
arbutrary rule. That's got to count for something, although it's hard
to quantify.

> > 2 RUKI (Balto-Slavic, Indo-Iranian)
> > 3a *-Dd- -> *-zd-, *-Tt- -> *-st- (Balto-Slavic, Iranian)
> > 3b *-Ddh- -> *-ddh-, *-Dd- -> *-dd-, *-Tt- -> *-tt- (Indic)
>
> Sanskrit normally has 'dd' and 'ddh' out of P.I.E. *dd
> and *ddh, though.

What I should have done here is make rule 3a apply in (Balto-Slavic,
IIr) and changed 3b to
*-zdh- -> *-ddh-, *-zd- -> *-dd-, *-st- -> *-tt-
since Sanskrit already has a(n almost defunct) rule *-zdh- > -ddh-,
which makes one suspect it once also had *-zd- > -dd-, *-st- > -tt-,
later removed by generalization.

> - edit -
>
> > The trick of this new set is that so to speak the common elements
> > of rule 2abc have been factored out and placed in PIE,
>
> Is it really worth the cost, though?
>

The benefit is the romoval of a rule in each of Germanic, Sabellic and
Iranian.

> > and only that part of the rule within the RUKI area that could get
> > phonemes "in harm's way" by changing them into sibilants is placed
> > after the RUKI rule.
>
> T. Burrow's suggestion does the same thing while invoking
> fewer sound changes.

No, not when you look at the whole picture.

> > At the same time we get rid of the whole Germanic sound shift by
> > replacing it with generalizations of allophones that were already
> > present in PIE.
>
> But at the same time you create a whole new shound shift
> for Indo-Aryan.
>

You have to account for the difference between Iranian and Sanskrit
one way or the other anyway. The savings are elsewhere.


Torsten