I'll first explain what I'm thinking now. I'm considering revising my rule such that the loss of supershort schwa in late Mid IE *either* produces preceding gemination of consonants *or* produces lengthening of the immediately preceding vowel. The choice between the two would then be governed by a simpler rule:
If the supershort schwa is word-medial, it lengthens an accented vowel in an immediately-preceding open syllable, otherwise all supershort schwas geminate the immediately-preceding consonant instead. (Word-final supershort schwas never cause vowel lengthening in a preceding accented syllable.)So now for some examples that I hope will finally tie everything together into a simple, intuitive rule:
1) MIE *bérata "she carries"Now I think I understand exactly why this would be happening and I owe it to those blessed Slavic languages. Thank you, Slavs, for existing and rocking my linguistic world! Basically, vowel lengthening would only ever occur in accented syllables and it's therefore natural then why gemination, rather than vowel length, is triggered by word-final supershort schwas in words with antepenultimate accent (i.e. words where a word-final supershort schwa is preceded by another unaccented vowel). Gemination is in effect the compensatory lengthening that occurs in posttonic syllables while vowel lengthening only ever occurs on the stressed syllable. Finally, word-final supershort schwas in penultimate-accented words never trigger vowel lengthening even though the preceding vowel is accented probably because gemination is the default and favoured process. Afterall, these affected consonants are nearer to the eventually-omitted tautosyllabic schwa than an accented vowel in a preceding syllable.
→ *['be:rᵊt:ᵊ] (Lengthening & Gemination)
→ eLIE *bērd̰ (> PIE *bʰērt)
2) MIE *kahʷánasa "dog (nom.sg.)"
→ *[kᵊ'hʷa:nᵊs:ᵊ] (Lengthening & Gemination)
→ eLIE *kwānz (> PIE *ḱwōn)
3) MIE *kérad̰ "heart"
→ *['ke:rᵊd̰] (Lengthening)
→ eLIE *kērd̰ (> PIE *ḱēr)
4) MIE *kʷaita "what?"
→ *[kʷᵊit:ᵊ] (Gemination only; the preceding syllable is closed and unaccented)
→ eLIE *kʷid̰ (> PIE *kʷid)
5) MIE *hʷamaigása "urine"
→ *[hʷᵊmi'gas:ᵊ] (Gemination only; gemination is the favoured process)
→ eLIE *maigáz (> PIE *moiǵʰos)
6) MIE *nákʷtasa "night (nom.sg.)"
→ *[nakʷtᵊs:ᵊ] (Gemination only; closed syllables block length)
→ eLIE *nakʷts (> PIE *nokʷts)
It's all about compensatory lengthening in the end, whether involving long consonants or long vowels. Considering also that a long consonant is effectively a "fortis consonant" and considering that such consonants already existed as voiced laryngealized stops at this stage, long consonants would be quickly reinterpreted as their voiced counterparts via lenition. All of these processes now appear to me to be natural and attested in many other languages.
Here are some links to developments in real-world languages that one may find eerily similar to what I'm now proposing for late Mid IE. You may notice that supershort schwas in Mid IE are parallel to jers in Slavic in the way they behave and the processes that they apparently both trigger. Enjoy!
Darya Kavitskaya, Pitch-accent as a condition on Slavic vowel length (see pdf):
Slavic loss of jers and concomitant lengthening is eerily similar to my new ideas on Pre-IE...
"The development of compensatory lengthening (CL) through vowel loss is a common sound change in most Slavic languages. Slavic CL is a diachronic process whereby the second vowel (jer) in a CVCV sequence is lost with a subsequent lengthening of the preceding vowel (1)."
Frederik Kortlandt, On the relative chronology of Slavic accentual developments (2006), p.8 (see pdf):
"10.8. Stressed short vowels were lengthened and received a falling tone before a non-final lost jer in Slovene, e.g. bȋtka ‘battle’. This development was evidently posterior to the progressive accent shift (10.7)."
(June 13 2008) On second thought, MIE *kérad̰a "heart" should probably be only *kérad̰ since this would be more in line with the accentuation at the time. On the other hand, it's always possible that the word is a Kartvelian borrowing anyway (from Proto-Kartvelian *m̥k’erd- "breast") and therefore may not have been naturalized with a penultimate accent before Syncope took over. At any rate, with or without the final supershort schwa, the result would be the same but I suppose that we can do away with this word-final vowel for the sake of theoretical optimality.