Re: [tied] Re: Satem shift

From: Piotr Gasiorowski
Message: 8619
Date: 2001-08-19

----- Original Message -----
From: tgpedersen@...
Sent: Wednesday, August 15, 2001 12:39 PM
Subject: [tied] Re: Satem shift

--- In cybalist@......, "Piotr Gasiorowski" <gpiotr@......> wrote:

> Your "regularisation of paradigms" has nothing to do with either
> change. They are phonological, not morphophonological.

[Torsten:] Phonology and morphophonology may have different numbers in the Dewey
decimal system, but rules of either type have one thing in common:
they affect words.
[Piotr:] Actually, phonological change affects segmental and suprasegmental PHONOLOGICAL FEATURES in a given PHONOLOGICAL environment. It may be sensitive to word boundaries and word-internal morphological boundaries but is not conditioned lexically or semantically (i.e., it doesn't affect names of animals as opposed to names of plants, or adjectives as opposed to adverbs, or nominatives as opposed to genitives). Morphophonological change such is analogy-driven and usually consists in the levelling-out of irregularities created by earlier phonological changes. Irregularities arise precisely because phonological change is blind to paradigmatic patterns and tend to play havoc with them.
> > But the whole raison d' for the *k-series, distinct from the *k'
> and
> > *kW, apart from the inconclusive Albanian evidence, was that they
> > went *s in some satem-languages and *k in others? Which is
> > what you would expect with a sloppy pre-literate generalisation?
> I don't think you understand the Satem developments. The *k series
> did not change into sibilants anywhere.

[Torsten:] I don't think you understood what I'm saying. The *k-series does not
change into sibilants anywhere because it was posited to account for
the cases where centum -k- corresponds to satem -k-. Arguing for the
reason d' of something assuming the existence of that something is
called begging the question, and that's what you're doing here.
[Piotr:] Your original wording (cited above) was hardly helpful. What does "they" refer to? The procedure that leads to reconstructing the *K series is the standard application of the comparative method. The method is not foolproof and MAY lead to an erroneous reconstruction of phantom three-way contrasts (cautionary examples are given in most handbooks of historical linguistics) in cases where there happens to be a mismatch between two-way contrasts in the daughter languages and no daughter language shows clear traces of three different series. Because of this danger, many IEists are sceptical about reconstructing three series rather than two. So was I, until I reconsidered the evidence. (See below.)

> The reason why we reconstruct
> *k as distinct from *k^ and *kW is that some instances of Satem *k
> correspond to *k, not *kW, outside the Satem group. There are some
> more recently discovered "triple reflexes" of the three series
> in "centum" languages; perhaps the most convincing case is Latin
> (Schrijver 1991, discussed here ca. 3500 messages ago).

[Tortsten:] I checked the discussion, and you didn't seem much convinced then,
but you are now?
[Piotr:] Yes, more than I was at that stage of the dispute. I'm more optimistic now about the reality of the three-way contrast. If you read the rest of the thread in question, you'd see how I gradually abandoned much of my former scepticism. Peter and Miguel offered some good arguments for the independent existence of the *K series, and I accepted them. While it is clear that many Baltic and Slavic *K's are due to messy development, interdialectal borrowing and special phonetic conditioning, there also seems to be an irreducible residue of "original" *K's justifying the multiplication of dorsal phonemes in PIE.

[Torsten:] Since you didn't understand it the first time, I'll be nice and
explain again my idea of how VASTLY, ENORMOUSLY banal, trivial
palatalisation, and after that paradigm regularisation might account
for the centum/satem phenomenon.

In the followig, WLG (without loss of generality) let k stand for k,
g, gh (pre-glottalic theory).

Assume we have two k-series: *k and *kW
Most word are inflected in PIE. As to vowel, they can have e-grade, o-
grade and zero grade (I'll leave out the extended grades; the
important thing here is the front/back feature). So we have in a


or, in the case the first is a velar stop:


which becomes, by palatalisation


Suddenly, social upheaval. The word-smiths and druids are killed (or
they fall out over some grammatical or other silly question; other,
more serious reasons are of course more likely). Two groups arise.
One generalizes the /k/. They are centum people.


Another generalizes the /c^/. They are satem people:


And they take the c^ futher down the road towards s^ and s.

But, as Piotr pointed out, in a pre-literate society these processes
cannot be carried out to perfection. So in some cases the satem-
people generalize the wrong way, faultily:


After this, the satem people have very few k's left. So they are free
to go *kW > *k.

Enter the linguist. What does he find?
The "correct" correspondence

    centum      satem
    k-          s-

the "faulty" correspondence

    centum      satem
    k-          k-

and the kW correspondence

    centum      satem
    kW-         k-

What does he do? He posits three *k-series: *k', palatal, k^ plain
and kW labiovelar, while fretting over why it is that the k- k-
correspondence varies in extent in Baltic and Slavic.

But there needs only have been two k-series in the beginning, as I
have demonstrated.

[Piotr:] There are several difficulties with an argument like this. First, what made the two groups of people regularise just about every paradigm in the same direction? (Variation, such as we find in Baltic and Slavic, is a marginal phenomenon, way below the expected level of messiness resulting from morphophonological generalisation.) Such regularity is typical of sound change but not of paradigmatic levelling, which is not a sweeping mechanical process but a tendency that affects word after word independently, sensitive to extragrammatical factors like frequency of use of a given lexeme and its forms, its social significance, etc. In English, for example, the past participle and the past tense of strong verbs have been levelled out one way or the other quite at random.
There is another, more devastating counterargument: the Satem shift takes place also in environments where front vowels are absent, e.g. preconsonantally: *k^lewos, *ok^to:, *tek^s-, *g^noh3-, and also in words like *(d)k^mtom (why "satem" rather than "katem"?). The *o-grade *k^omt- is attested, but **k^emt- is not. Without addressing such issues and analysing concrete linguistic material your scenario is an exercise in armchair linguistics.
Last, but not least, I don't believe in a primary split between "Centum" and "Satem" (such an idea has long been abandoned), so I also object to the sentence "Two groups arise". There was a Satem innovation, and what we call "Centum" is simply the primitive state of affairs within IE. In evolutionary terms, the "Centum" languages do not form a valid genetic counterpart of "Satem.