Re: [tied] Bartholomae's Law and Grassman's Law

From: Piotr Gasiorowski
Message: 13301
Date: 2002-04-17

It's a handbook example of synchronic rule reformulation _after_ the historical operation of the sound change(s) that the rule in question corresponds to.
 
Bartholomae's Law turns the sequence [breathy voiced] + [voiceless] into assimilated [breathy voiced] + [breathy voiced], e.g. *-bH + t- > *-bHdH- (> Skt. -bdH-) or *-dH + s- > *-dHzH- (in Proto-Indo-Iranian sibilants could be breathy voiced too).
 
Grassmann's Law consists in the dissimilation ([+ aspirated] > [- aspirated]) of any occurrence of aspiration if followed by another occurrence of the same in the same root-size domain in a contiguous syllable, e.g. *bHudH-je-toi > budHyate.
 
Grassmann's Law is younger than Bartholomae's, but it interacts with some phonological processes in an apparently paradoxical manner. In pre-Vedic times *zH and *z^H (produced by Bartholomae's Law) were "deaspirated" into Indo-Aryan s and s., and any preceding obstruent became voiceless (as well as deaspirated) by assimilation: *-gHz^H- > -ks.-, *-dHzH- > -ts-, *-bHzH- > *-ps-. This change obscured locally the operation of Bartholomae's _and_ Grassmann's Laws, since it removed the features that had originally triggered them. One would expect to get historical derivations like the following:
 
*bHeudH-s-je-ti
*bHaudHzHyati (BL)
*baudHzHyati (GL)
 botsyati (deaspiration and devoicing)
 
or
 
*h1e dHugH-s-e-t
*adHugHz^Hat (BL)
*adugHz^Hat (GL)
 aduks.at (deaspiration and devoicing)
 
Such or similar forms are indeed found in the earliest and most archaising layer of Vedic grammar, but the already productive Vedic type (and the only possible one in later Sanskrit) was <bHotsyati, adHuks.at>. This "modernising" type owes its existence to the analogy of forms like the root noun *bHudH-s, where the loss of phonation contrasts in word-final clusters produced *bHuts (> Skt. bHut) prior to Bartholomae's Law. By the time of S'a:kalya (let alone Pa:n.ini) the new forms were so well entrenched in Sanskrit that in S'a:kalya's <padapa:t.Ha> to the Rigveda they are regularly substituted for the archaic ones even where the latter occur in the text.
 
The synchronic grammatical rule needed to account for the output of Grassmann's Law was reformulated as "aspirate throwback" (Hock's term) rather than the dissimilation of an underlying pair of aspirates; accordingly, the underlying form of the root was reanalysed as well: /budH-/ --> [bHut-]/[bHud-] instead of earlier (and diachronically correct) /bHudH-/ --> [budH-]/[bHut-]/[but-] in the right environments. In the restructured system there is no place for the allomorph [but-], and only [bHut-] can be derived with the new rules, more or less along these lines:
 
/budH+s/ /bu-budH+s+a-/ /baudH+s+ya-/
 bHuds    bubHudsa-      bHodsya-   (aspirate throwback)
 bHuts    bubHutsa-      bHotsya-   (voice assimilation)
 bHut     ---------      --------   (final simplification)
 
"Generalised aspirate throwback" applies also when historical breathy voiced stops are deaspirated in clusters such as *-gHdH- > *-gdH-, e.g. in Skt. dHug-bHiH reflecting *dHugH-bHis (like dHuk < *dHugH-s) vs. duh-aH < *dHugH-os, synchronically all derived from underlying /dugH-/. This means that Grassmann's law and a variety of deaspiration processes are compressed into a single synchronic rule. There is no throwback, hoewever, in forms like buddHa < *bHudH-to-, where Bartholomae's Law operates transparently (but it doesn't apply to stop+s clusters synchronically). 
 
/budH+ta-/  /dugH+bHis/
 buddHa-     --------- (BL [synchronic])
 -------     dHug-bHis (aspirate throwback)
 
If we retained /bHudH-/ as the underlying form, we'd get the illusion of Grassmann's Law operating before Bartholomae's: /bHudH-ta-/ --> budHta --> [buddHa-]. This, I suppose, is the paradox that Peter had in mind. The source of the paradox is the fact that the synchronic rules are subtly different from their historical models. The actual development was *bHudH-to- > *bHudHdHa- > *budHdHa- > buddHa- (phonologically regular).
 
I have omitted some details for the sake of clarity (please don't laugh), but I hope you can see what the moral is: rule reordering obliterates the fingerprints of the diachronically underlying changes. If we had no evidence of the older system and no comparative data from outside Indo-Aryan, and if we could only rely on the application of internal reconstruction to the "modernising" forms (the only grammatical ones in the classical language), Grassmann's Law and its interaction with other obstruent changes in pre-Vedic would be reconstructed incorrectly.
 
Piotr
 
 
----- Original Message -----
From: x99lynx@...
To: cybalist@yahoogroups.com
Sent: Tuesday, April 16, 2002 10:47 PM
Subject: [tied] Bartholomae's Law and Grassman's Law

"P&G" <petegray@......> writes:
<<For me, the major phonetic problem in Sanskrit is reconciling Bartholomae's
Law and Grassman's Law, both of which are needed, but each of which implies a
different order of events.>>

Peter- if you have the time - could you explain how you understand the two
laws "imply" two different orders of events?  I've heard this before but
don't understand how this conflict is detected.  I don't think you are saying
there are specific Sanskrit words where Bartholomae-type permutations show
erratic chronological occurence, i.e., before or after Grassman (aspirated
consonants lose their aspiration if followed by an aspirated consonant) in
the order in which they happened - I don't think.  I don't follow how the
implied chronology can be different for the two laws.