Re: crows and the glottalic theory

From: Richard Wordingham
Message: 16612
Date: 2002-11-07

--- In cybalist@..., Miguel Carrasquer <mcv@...> wrote:
> Going back to basics and looking at the attested reflexes in the IE
> languages, we can distinguish 4 groups:

> 2) Armenian, Germanic (+ Celtic)
> *t = /th/, *dh = /d/, *d = /t'/
> Again typologically unproblematical, and even the *tegh/dhek-
> prohibition can be explained as a simple constraint on the
> co-occurrence of voiced [*dh] / unvoiced [*t *d] stops. The problem
> is explaining how the other systems may have originated from this
> (a "reverse Grimm's law" would be as bizarre as... Grimm's law
> itself), without e.g. *tH and *t merging, etc.

What's Celtic doing here? (Was Thracian meant?) I would have
thought that it belonged in Group 3 if any of these, though I'm not
aware of any Celtic reflex of the PIE *d ~ *dH contrast. Mind you, I
thought Germanic *t' was just a tidying up of the groups of observed
reflexes until I found one of Kortlandt's papers ( ) which postulate
Germanic *?t from PIE *d.

What then is the problem with a 'reverse Grimm's law'? Isn't the
second Germanic sound shift pretty much a repeat of Grimm's law? And
I recall references to Grimm's laws in other language families.

> 4) Latin, Greek
> *t = /t/, *dh = /th/, *d = /d/
> Unproblematic typologically, but the *b-gap and the deg-constraint
> historically inexplicable. In fact, far from being a possibility
> (pre-)PIE, this system is probably simply derived from the previous
> one (by merger of *th and *dh).

Except that pre-Latin didn't have a *th to merge with *dh.

> ... the following proposal to explain the *deg-constraint and *b-
> The original situation (likely to be ancestral to the whole
> language family, long before PIE) would have been a system with two
> kinds of stops (voiced and unvoiced or perhaps fortis and lenis: *p
> *b, *t *d, *k *g, etc.), and, suprasegmentally, two tones (marked:
> high, unmarked: low).
> The following unsurprising constraints applied: (1) one root (word)
> could only contain either all voiced or all unvoiced stops; (2) only
> one syllable in a word could have the high tone.
> For example:
> taka (= tàkà) táka (= tákà) taká (= tàká)
> [but no táká, taga, tága, tagá]
> daga (= dàgà) dága (= dágà) dagá (= dàgá)
> [but no dágá, daka, dáka, daká]

What phonemes do these voicing prosodies normally affect? Just

> Subsequently, the tones were lost, but high tone left a trace in
> (marked) glottalization of the consonant, while low tone gave
> (unmarked) aspiration, as follows:
> taka t?aka tak?a [but no t?ak?a, taga, t?aga, tag?a]
> daga d?aga dag?a [but no d?ag?a, daka, d?aka, dak?a]
> For vowel initial words, we perhaps had:
> haka ?aka hak?a
> haga ?aga hag?a

Are there examples of this in widely accepted reconstructions? Tai
tone splits conditioned by the phonation of plosives normally put the
boundary somewhere in the sequence voiceless aspirates, voiceless,
preglottalised, voiced, so I am surprised that high tone should
produce glottalisation.

> To explain the Indo-European state of affairs, all we have to do is
> postulate a merger of (ejective) **t? and (implosive) **d?:
> 1 2 3 4
> **t? , **d? > *t' ~ *t' ~ *'d ~ *d
> **d[h] > *th ~ *d ~ *dh ~ *th
> **t[h] > *tt ~ *th ~ *t ~ *t
> with the Germanic/Armenian solution based on the opposition **t? ~
> (> t(') ~ th), the others rather on **d? ~ **d (> (')d ~ dh). I see
> no way to merge the two solutions at the PIE level with only three
> (not four) series to work with, so either I have to posit *t? and
> for PIE istelf, or divide PIE into two dialects with rspect to the
> obstruent system.
> For the labials, we need the special rules:
> **p, **p? > *p (/p/)
> **b, **b? > *bh (/b/)
> This was undoubtedly triggered by the loss of the ejective feature
> *p?, an entirely predictable phenomenon. Implosive *b ( = /'b/) may
> be allowed to have had a marginal existence besides unmarked *bh ( =
> /b/), as for instance in modern Arabic (ba:b = door, 'ba:'b = dad)
> even English (bye-bye = /'baI'baI/), but for the most part it merged
> with *b in a simple opposition /p/ ~ /b/ for the labial series.

I'd like to propose a derived variant of this scheme.

Starting point:
1. Voiceless v. voiced (or fortis v. lenis) (as Miguel)
2. Voicing prosody (as Miguel)
3. Contrastive stress (as opposed to tone) We can probably get rid
of some stressed vowels, so that in some words only originally
unstressed vowels survive.

1. Plosives _following_ stressed vowels are preglottalised, as with
voiceless plosives in Cockney.

2. Overstrengthening causes ?p > ?, ?b > ? (c.f. [?t] > [?] in
Estuarine English). This is reminiscent of Peter's speculative
suggestion that
pre-PIE **b > h3! I recall arguments based on 'strength'
hierarchies, which came up with the odd observations that dentals
were stronger than labials in Germanic languages, but vice versa in
Romance languages, so the comparison with Estuarine is valid for ?p.

Variations are possible. I suppose we might even have ?p > h3, ?b >

3. Voicing contrast of preglottalised consonants lost (as Miguel). I
think the outcome of a subsequent loss of preglottalisation could go
either way; Tai-Kadai /?m/ > Siamese <hm> (once voiceless?) but /?b/
> /b/ (implosive for some speakers). Do we need to postulate a
stable dialect split on the basis of whether the preglottalised
plosive is [?t] or [?d]?

4. PIE branches go their own way.

As Proto-Tai > Siamese shows the change t ~ ?d ~ d > t ~ d ~ tH (even
though Proto-Tai already had /tH/), do we need to postulate that pre-
Greek ever had [dH] (Group 3 ancestry of Group 4)?

How much do these schemes help with Nostratic? We have five

PIE *dH PAA/PKartv *d
PIE *t PAA/PKartv *t
PIE *t PAA/PKartv t' (IS)
PIE *d PAA/PKartv t (IS)
PIE *d PAA/PKartv t' (Bomhard)

and only four phonemes (/t/, /d/, /?t/, /?d/).