Re: [tied] PIE dialects [was: Satem shift]

From: Piotr Gasiorowski
Message: 8040
Date: 2001-07-22

The main difference between a bona fide "branch" like Germanic and a fuzzy areal grouping like Graeco-Armeno-Aryan (henceforth "GAA") is that branches can be defined by pointing to shared inherited innovations unique to the group in question. For example, Grimm's Law, followed by Verner's Law and the Germanic stress shift affected all the Germanic languages and can be regarded as part of the definition of (Proto-)Germanic (together with several less celebrated but equally characteristic phonological changes). There are also a number of uniquely Germanic morphological features (like the weak/strong preterite system) and shared vocabulary, which confirm the correctness of the grouping. For Celtic, we have the loss of initial *p-, the change of syllabic liquids into *Ri, etc., again supported by shared grammatical and lexical inovations.
By contrast, the definition of GAA relies on a number of features that don't always occur in all the constituent branches, or if they do, they look easily borrowable (like the preterite augment) or are not exclusively GAA (e.g. the thematic genitive *-osjo occurs also in Old Faliscan, and the prohibitive particle *me: in Albanian and possibly in Messapic). Apart from the treatment of syllabic nasals (shared with Albanian, _not_ shared with Phrygian and parallelled by analogous processes elsewhere) there are no phonological traits to found a solid relationship hypothesis on. Grassmann's Law in Greek is independent of Grassmann's Law in Sanskrit (or in Tocharian, for that matter), and the "Graeco-Armeno-Iranian" lenition of *s (reenacted in Brittonic Celtic) is demonstrably a different process in each branch.
To return to Germanic: its main "constitutional" features are very young. Grimm's Law (or at least its final stages) probably operated not earlier than the 3rd c. BC, which means that the "pre-Proto-Germanic" period represents a vast time depth -- enough for any language to "speciate" if its closest relatives are pruned out. (Celtic expansion in Central Europe may have wiped out some of them). Late pre-PGmc was not merely a dialect of PIE. Even without Grimm's Law it surely had some "Germanoid" features like the merger vowel *a for PIE *o and *(h2)a, *uR, *uN for PIE syllabic liquids and nasals, and still older features shared with Italic or Italic and Celtic (e.g. *-ss- for PIE *-t-t- [-tst-]). The preterite system also appears old and the "strong" subsystem has Italic and Celtic connections. If we travelled sufficiently far back in time, we'd probably find a network of related dialects covering much of NW "Indoeuropia", without any clearcut splits. I wonder if this NW IE continuum could not after all be regarded as a distinct intermediate proto-language between the "Neo-IE" node (IE without the Anatolian languages) and Proto-Germanic, Proto-Italic and Proto-Celtic (not to mention minor members like Venetic or Lusitanian). As regards phonological innovations attributable to that distant stage, I'd propose the *-t[s]t- > *-ss- change mentioned above and perhaps the change of the voiced aspirated (breathy voiced) stops into voiced fricatives (assuming their early merger with voiced stops in Celtic) long before the remaining stages of Grimm's Law.
----- Original Message -----
From: markodegard@...
Sent: Sunday, July 22, 2001 12:13 AM
Subject: Re: [tied] Satem shift

This group probably needs to come to a clear understanding of what we
mean by 'sprachbund'. We also probably need to discuss when it is
proper to speak of a distinct stock ('proto-anything') vs. vague
labels such as 'late NW IE'.

In the case of Greek-Armenian-Indo-Iranian, we are speaking of a
cluster of related languages -- post PIE, but perhaps at a state where
'dialect' might still be used in its broadest sense.

One the other hand, when you set you mind to consider these languages
as they were way back then, they were still very closely related;
comparativists would probably group them *then* as a single stock, but
since each of these languages later fully differentiated into fully
independent stocks, we push the origin of each of the stocks to
practically the first moment something distinctive happened. English,
Swedish, Dutch and German are all very different but are currently
considered to be closely related; I think it's fair to say Greek,
Armenian and Indo-Iranian (or for that matter, what became Celtic or
Italic) were as closely related *then* as German and English are
*today* (and probably, closer).

It all depends on how you are looking at the question. Tree vs. bush,
wave vs. particle, phylum vs. clade, etc.