I've been working on a PDF with a reconstructed grammar and phonology of pre-Indo-European. I'm trying to be as complete as possible, which, sadly, results in very slow progress. But to keep this blog somewhat alive, I'll at least post some of the things that I have already established, one of them is a phonology.
Indo-European phonology has been highly problematic from the start, voiced aspirates without voiceless aspirates, tons of gaps; a more common palato-velar than a normal velar. It's just not very realistic. Just to give you an idea how unrealistic it looks I made a table. I reconstructed the laryngeals on commonly reconstructed places, though many Indo-europeanists leave it at 'something laryngeal'.
What a monster! There's more gaps than there are phonemes, and there's already a lot of phonemes! And I've already merged some of the catagories, fior example the *i is techincally a Palatal, not a Palato-Velar.
The first step I took is, that I established that the Palato-velars can never be original palato-velars, since it's the most common type of velar. And in any language on this planet, such a velar is the plain velar.
If the palatovelars are velars, the velars obviously can't be velars, so the probably were uvulars.
Then there's the laryngeals. The fricative line is very weird in shape. If we were to assume that *h₁ was originally [x], which is perfectly imaginable, Southern Dutch underwent the shift Proto Germanic [x] > West Germanic [h] > Dutch [h] > Southern Dutch [ʔ]. Also Cockney English went through this same shift. This leaves the Pharyngeals to simply fill up the gaps of the Labio-Velar and uvular.
Then the voiced stops were probably creaky voiced, or maybe 'glottalised' whatever that may be. And the voiced aspirates must have been 'voiced'.
If we then reparse the little table into a new model, we get a lovely symmetrical phonology which looks quite realistic, somewhat like a real language. It's almost magic.
The rest of the gaps are common. no /f/ is not a big problem. No /ŋ/ is also normal, especially since, in front of velars there's an allophone of *n as [ŋ]. Labio-velar nasals don't exist. And Uvular nasals exist, but would never exist if the velar nasal doesn't exist either.
There's very few languages that that a contrast between a labial approximant and a labio-velar approximant, so this is also realistic. And a Uvular approximant is almost unheard of.
Also if Glen gordon is right that word final *t > *s at a certain point, then word final *k > *x makes a lot more sense than *k > *ʔ.
One slight problem still remains. I reconstructed labio-velars, but considering the similarity of the way *xʷ and *χ function it is likely that *xʷ was actually *χʷ, and thus the series would be labio-uvulars. In the latest version of Indo-European this was probably also the case, but in its earliest form probably to be avoided as a reconstruction.
As Glen Gordon is currently proposing, I would not be suprised if the Uvulars turned out to be allophones of the velars in an earlier stage. But I'm a bit reluctant to reconstruct it as such yet, since I have not reconstructed indo-european back far enough yet that this becomes apparent.
Note that *y and *w, in the earliest reconstructible face do not vocalise to *i or *u ever, so there's no use to transcribe them as such anymore.
Then there's the vowels. Vowels allow me to be very short. Late indo-european gives *e, *o, *ē and *ō. I think that the long vowels are completely allophonic, at least in an earlier stage. *é and *ó existed accented, while there was no contrast in unaccented syllables. When a language only has 2 distinct vowels, you do not expect [e] and [o], but a vertical vowel system of: *ə and *a, or *ɨ and *a. I reconstruct *ə in unaccented syllables and *ə́ and *á in accented syllables. I tend to write *ə́ as *é since the fonts I use at least, do not make the most attractive schwa with combining acute accent. It shouldn't cause any confusion.
So that's it! The new and improved phonological system of Indo-European. Feel free to make any comments.