Present vowel heightening
One of the greatest problems with Indo-European, is that the common reconstruction of the language's phonology is typologically unlikely to impossible. Lately I've been focussing on the vocalic system.
The general idea is that there were two or four vowels depends on how you treat *i and *u; For the sake of this article it will be easier to view it as four vowels. The four vowels are *e, *o, *i, *u. This system is unlikely; if anything we'd at least expect the vowel *a. There is also some reason to believe that there might be the vowel *a; but the actual phonetic distribution of this vowel is so incredibly small, that the system is still extremely unlikely.
I'd rather not deny this system never existed; But I'd like to find an older, more logical system behind all this. My 'gut' tells me that *o was *a at some point, especially considering Hittite having /a/ for *o. A system like *a, *e, *i, *u. Though a lot less unlikely still doesn't quite feel right. To account for seemingly random *e/o and *i/e alternations in some situation we might like to see a *ə. And I'd want to assume that then all later *e's are in fact secondary developments from *a or *ə.
So that's a lot of assumptions; now let us take a look at some of the things that might agree with me. First know that, as Hittite seems to indicate, Perfect and Present used to be two completely unrelated categories of verbs, rather than forms of one verb system. Therefore a Perfect and Present can both be in the present tense and past tense; as odd as that may sound.
Now let's look at the reduplicated forms of both a present and a perfect (I believe pre-Hittite-Indo-European did have reduplication). I only gave first, second and third singular, and third plural respectively; With the 1st and 2nd plural I'm not very confident on how archaic those forms are; They don't quite function like the rest.
Present Perfect
*dʰi-dʰéh₁-m+i *de-dóh₃-h₂e
*dʰi-dʰéh₁-s+i *de-dóh₃-th₂e
*dʰi-dʰéh₁-t+i *de-dóh₃-e
*dʰi-dʰh₁-ént+i *de-dh₃-ér
There's several remarkable things about these two paradigms. First the Present
has e-grade in the root, while the perfect has o-grade. But not only that, the present has a reduplication with *i, while the perfect has a reduplication with *e. Last but not least, the perfect has three endings ending in *e; while according to the rules of syncope, we wouldn't like to see final *e's or *o's at all. And in general we do not wish to see the vowels *e and *o outside of the accent at all. Nevertheless they're there, it'd be crazy to deny that.
Now let's look at both version of these forms as I would reconstruct them in their Pre-Syncope-Indo-European.
Present Perfect
*dʰi-dʰéh₁-me+i *de-dóh₃-h₂e
*dʰi-dʰéh₁-sa+i *de-dóh₃-th₂e
*dʰi-dʰéh₁-to+i *de-dóh₃-e
*dʰi-dʰeh₁-énta+i *de-doh₃-éra
Somehow though the *e's I reconstruct for the perfect did not syncopate like the one in say *-me did. This brought me to think that these vowels might be of a fundamental different nature than the normal *e. Let's assume that this vowel is *ə
For the sake of unification we'd like the reduplication vowels to originally be from the same source as well. And dare I say it, I'd personally feel for unifying the root vowels as well, into an *a from the *i,*u,*a,*ə-stage of Pre-Indo-European. One again quite a lot of assumptions but let's look at the paradigm first before I continue.
Present Perfect
*dʰə-dʰáh₁-me+i *də-dáh₃-h₂ə
*dʰə-dʰáh₁-sV+i *də-dáh₃-th₂ə
*dʰə-dʰáh₁-to+i *də-dáh₃-ə
*dʰə-dʰah₁-éntV+i *də-dah₃-érV
Note how I'm now using *V for unknown vowel because the previously used placeholder vowel <a> is now being used for an actual *a. Also note how in endings -me, -to and -érV I'm still writing the previous' stage's vowels. I do not pretend to know what these forms would be in this stage of Indo-European, but writing them all down as V would be silly, as we have significantly more data what they become in a later then the vowels I've denoted as V. And since *e nor *o overlap with the proposed system, they're not in the way.
Right so now onto the forms; what can we say? You'd almost like to assume that the *+i caused i-umlaut on all vowels; but this is sadly impossible for the secondary endings (without *i) also display e-grade, and it would be too far a stretch to assume that this was analogy. The other difference is that the all Present endings only have non-ə vowels for endings. Somehow this seems to have heightened the *ə and *a to *i and *e respectively.
Now sharp readers might want to point out the -érV is also a non-ə ending. They would be right, but realise that the vocalisme of the root becomes Ø in modern Indo-European, so there is a chance that the pre-syncope vowel there was actually *e and not *o, as we would expect from the present.
This heightening of the vowel because of non-high-vowels is very difficult to support phonetically though. I have tried to fit in some syncopated *i somewhere to account for the vowel-heightening, but this would be too far fetched. It almost looks like a sort of anti-vowel-harmony, where the vowels preceding the suffix disharmonise with the suffix vowel. I'm curious if anyone knows such a language, please tell me if you do.
I do not pretend to know how this will solve itself; nor am I sure if I'm thinking in the right direction, I'm going to need more proof (maybe in the nominals?) and more plausible soundshifts to be able to support this theory fully myself. But I just decided to put down my thoughts here, so that people can think along. So please any thoughts or questions are welcome.
Comments
Phoenix: "My 'gut' tells me that *o was *a at some point, especially considering Hittite having /a/ for *o."
Hot diggity! That's what I've been saying all along! (Actually, so has Allan Bomhard and John Kerns.)
Phoenix: "And I'd want to assume that then all later *e's are in fact secondary developments from *a or *ə."
Bomhard and Kerns covered that already back in 1994. I don't agree with everything they say and find a lot of the connections farfetched (because of the "lumper" bias to prop up the Nostratic hypothesis) but I think their observations on the Pre-PIE vowel system are astute at least. See Bomhard/Kerns, The Nostratic Macrofamily: A Study in Distant Linguistic Relationship (1994), p.76 (link). It was their position that PIE *e < Pre-PIE *ə and PIE *o < Pre-PIE *a. I've assimilated this view and added a few extra personal touches of my own.
Phoenix: "Last but not least, the perfect has three endings ending in *e; while according to the rules of syncope, we wouldn't like to see final *e's or *o's at all."
That's why years ago I added the Suffix Resistance exception to the Syncope rule. Monosyllabic open-syllable affixes (other than those formed from deictics like the nominative singular, the pronominal inanimate and the 3ps/3pp verb endings) retain their vowel. In this way, there is only *one* possible vowel in word-final position in the Pre-Syncope stage of IE, namely */ə/ (which I write as *a in my system). Note that these singular "perfect" endings are all monosyllabic. The loss or preservation of word-final schwa then is predictable and one vowel suffices.
Phoenix: "It almost looks like a sort of anti-vowel-harmony, where the vowels preceding the suffix disharmonise with the suffix vowel. I'm curious if anyone knows such a language, please tell me if you do."
Great idea but I've never heard of such a thing personally. I think you're just circumnavigating what I've already come to realize: Mid IE (Pre-Syncope IE) allows only one possible vowel in unaccented position (*a = /ə/) while allowing two (*e = /e/, *a = /a/) in accented position (or so the analysis tells me).
Keep in mind that what is analysed to be so is not necessarily the true nature. Sounds very zen, doesn't it? That is, if one should reconstruct a single vowel for unaccented positions in Mid IE, that doesn't mean that one should necessarily believe that the language was spoken this way since a "single vowel" could be a mirage for what is in effect two or three vowels that underwent a "silent merger" for all we know.
Great idea but I've never heard of such a thing personally. I think you're just circumnavigating what I've already come to realize.
I agree, the idea was too crazy. I was just sort of trying to get rid of ablaut. Which is of course, not necessary, but I thought I'd experiment with the thought. In the end though, I'd have to agree that trying to get rid of it is really hard to prove.
They idea of having an exception class for syncope, devoted to deictics seems rather crazy in my opinion. Why would they syncopate and not normal suffixes?
I'd sooner expect normal suffixes to syncopate, while deictics would stay resistant. Just a thought here that just popped into my head: Maybe we're not looking at one syncope. Maybe there was syncopation, and only later suffixation of deictics, which underwent their own de-accentuation and therefore loss of vowel.
Not sure how well this would work chronologically, seems to work in my head right now; but it's 3am. It would fix the exception status of the deictics though.
The loss or preservation of word-final schwa then is predictable and one vowel suffices.
So according to that all of the word final *a's would result in what we know in the conventional system as *e? Or did I misunderstand this? (Again, it's 3am, I'm going to bed ;))
"The idea of having an exception class for syncope, devoted to deictics seems rather crazy in my opinion. Why would they syncopate and not normal suffixes?"
Thanks for being brutally frank! I'm a crazy kind of guy :) Well, in Etruscan, for example, can is reduced to a suffix placed after the noun, -cn. This is to be pronounced with syllabic word-final /n/. In that language, an event of syncope occured around 500 BCE because it too acquired a heavy stress accent, fixed on the initial syllable. Nonetheless, the name Turan (a product of tur 'to give' and -an) never seems to be reduced to *Turn even in the latest stage of Etruscan circa 1st c. BCE. This then is a real-world example of a language that shows exceptions to its own syncope. Evidently, deictics here didn't "stay resistant" as you suggest.
"Maybe we're not looking at one syncope. Maybe there was syncopation, and only later suffixation of deictics, which underwent their own de-accentuation and therefore loss of vowel."
This seems unnecessarily complex compared to the scenario I suggest. You still need an added pre-IE vowel to explain 1ps perfective *-h2e. Furthermore, perhaps you should ponder on 1ps mediopassive *-h2or which I interpret to be a composite of the aforementioned 1ps perfective and a former postposition. The alternation of *e/*o here is significant and it seems to me like evidence of earlier schwa (via Schwa Diffusion and Merger which also explains the thematic vowel alternation). This then points me to earlier 1ps -hə in early Late IE. As I said, the word-final vowel survives Syncope due to the Suffix Resistance exception which preserves non-deictic suffixes of a CV structure.
As for why deictics were not immune to Syncope, there could be multiple reasons. For example, in the verbal paradigm, there would certainly be motivation towards analogical levelling of MIE's 1ps/2ps endings (*-am(ai), *-as(ai)) with that of the 3ps (*-ata(i)). Since Syncope aided in this levelling, it would be pointless to resist, hence resulting eLIE *-m(i), *-s(i) and *-t(i) (instead of 3ps **-ta(i)). In the declensional paradigm, yet again there would be motivation for analogical levelling which Syncope would aid. Hence when MIE genitive *-ása was reduced to eLIE *-ás, the nominative similarly reduced from *-sa to *-s.
So in other words, the deictics are simply following Syncope as expected while the other suffixes must have been motivated to buck this trend and do something abnormal. One reason is phonotactic constraint (i.e. the avoidance of excessive clustering) since most of these derivational suffixes were word-medial afterall and preceded the already clipped nominative *-s or 3ps secondary ending in *-t.
"So according to that all of the word final *a's would result in what we know in the conventional system as *e?"
Yes, exactly. Phonemically I write MIE *a even though it is for phonetic /ə/. (At this stage, stressed /a/ alternates with unstressed /ə/ under the same phoneme *a.) Normally, this unaccented *a is dropped by early Late IE (eLIE) unless exempt from Syncope due to an exception to the rule. When exempt, it simply remains schwa. However, be warned that my notation changes at this stage: I explicitly write *ə in eLIE in order to distinguish it from *e and *a which were at this point phonemically distinct in unaccented positions. Only schwa is later affected by the Schwa Diffusion and Schwa Merger rules which result in the fascinating *e/*o alternation of thematic vowels. So I suggest the following development of the 1ps "perfect": MIE *-ha > eLIE *-hə > PIE *-h2e.