22 May 2008

Rethinking the reduplicated perfect in Indo-European

In How old is the reduplicated perfect in Indo-European? I was attempting to get a handle on a reasonable dating of the reduplicated perfect in Proto-Indo-European (PIE) within the context of my grandiose theories on Pre-IE. As a recap, the reduplicated perfect form is demonstrated by the form *bʰe-bʰor-e meaning 'he/she has carried' where we can see that the first sound of the root *bʰer- is repeated and the root is given the vocalism *o. which is then followed by the personal ending *-e for the third person singular. Additionally, the perfect forms have a completely different set of personal endings from the present forms. So, we need to figure out how this might have come about. I previously suggested that i-reduplicated presents (e.g. *di-deh₃-ti 'he/she gives') were older than the reduplicated perfects but after meditating on this the whole day yesterday, I believe I've been missing the opportunity for more efficient solution.
So instead, I would like to explore the possibility that it is to the contrary the reduplicated perfect that dates to the Mid IE (MIE) stage before the Syncope rule while i-reduplication seen in the presentive forms is a much later feature of Indo-European conjugation. The reason for my change will become clear as I walk you through the proposed evolution. I already have common examples like the genitive-declined form *pedós 'of the foot' which contrasts with nominative *pōds showing *o/*e ablaut alternation that can be attributed to the change of unsyncopated unaccented MIE *a in the root to early Late IE (eLIE) *e in order to avoid making declension too obscure with overly erratic root vowel deletions in the paradigm of a noun (i.e. the Paradigmatic Resistance exception of Syncope).

Given that, it just makes perfect sense to exploit this preestablished exception (i.e. MIE unaccented *a > *e rather than deletion) to explain the appearance of *e in the reduplication of the perfect. We may then simply reconstruct an MIE pattern *Ca-CáC- as the antecedent to the later so-called reduplicated perfect of the form *Ce-CoC-. in much the same way as MIE *pad̰ása becomes eLIE *ped̰ás and later *pedós. On the other hand, i-reduplication can then be blamed on the schwa found in the middle of the Late IE period which would be a natural vowel to initially use when reduplicating a present form to convey the iteration of an action. Thus, eLIE *Cə-CéC- may serve as a more straight-forward antecedent to later *Ci-CeC- via my already-stated rule "preaccented eLIE > PIE *i".

However, there's still the unresolved controversy concerning what the real function of the reduplicated perfect in PIE itself and its preceding stages was exactly if it were ancient. For this, I feel the need to appeal to the tenseless language I'm most familiar with, Mandarin. I will suggest briefly that perhaps there was originally a simple two-fold distinction between 'non-completed' actions in e-grade and 'completed' actions in a-grade (later o-grade due to Vowel Shift at the end of the Late IE period). This grammatical structure would find a parallel then in Mandarin where completed actions are marked with the particle 了 le placed after the verb as in 我看了 wo kan le 'I looked'. There is also a further marker in Mandarin for the progressive, expressed when 在 zai or 正在 zhengzai is placed before the verb as in the example 我在看 wo zai kan 'I'm watching/reading'. It must be stressed that the completive is not the same thing as past tense since there is the example of (现在)下雨了 (xianzai) xia yu le meaning 'it is raining (now)' (i.e. 'it has finished starting to rain (now)', so to speak) whereby le appears in effect to be marking an inchoative action[1]. Confused? Excellent!

Back to PIE then, perhaps likewise in MIE there was once a "completive" aspect using the form *CaC- together with the *h₂e-set of personal endings distinct from the *mi-set. Then a reduplicated form developed out of this in MIE (*Ca-CáC-) to express an action that was continuative at some point but was thereafter completed. So perhaps a simple form in MIE like *bár-a came to mean "he/she has carried (once)" while *ba-bár-a meant "he/she was carrying (but now he/she is done)" which would oppose the durative past *bér-ata "he/she was carrying (and may still be carrying)." The two meanings between the simple a-grade and the reduplicated a-grade would only be minutely different but the latter form would be barred for inherently stative verbs like *waid̰- 'to know' which could only be used in the form *wáid̰-a meaning "he/she knows" (or possibly with inchoative meaning "he/she came to know") but never with continuative aspect as in **wa-wáid̰-a "he/she was knowing (but now he/she is done)." If you think about it, the reduplicated form could easily lend a resultative nuance if analysed in this way since the reduplication would have originally stressed the non-stative quality of the verb (either "repetitive" in nature as for punctual actions, or "continuative" as for non-momentaneous ones) while the *h₂e-set of personal endings would ensure a completive aspect in contrast to the non-completive *mi-set. In other words, the reduplicated a-grade would both stress an action rather than state as well as its eventual endpoint. Thus a natural rift would grow over time between the inflection of resultative verbs and of stative verbs within the originally united category of completive aspect. Additionally, the reduplicated a-grade might still have avoided crystallizing into a formal perfective as long as non-reduplicated forms of active verbs still had a function within this system.

So could this mean that *bʰor-e continued to mean "he/she has carried (once)" in PIE while *bʰe-bʰor-e meant "he/she was carrying (but now he/she is done)" until the Anatolian branch seperated from the Indo-European core? Does this mean that the simple CoC- construct remained the preferred resultative-stative form until after PIE? Is the mi-class/hi-class division found in Hittite to be understood as a derivative of this aforementioned two-fold system whereby the inherited hi-class verbs are largely either punctual or stative in nature?

[1] Soh/Gao, Perfective Aspect and Transition in Mandarin (2006), University of Minnesota, p.111 (see pdf).


  1. I don't feel that this *e to avoid making declensions too obscure is some sort of paradigmatic levelling, but simply a phonotactic matter.

    no matter in which part of the world you live, a word like pdása is never easy. I guess that a schwa was just inserted in such clusters which you'd write *a Which would make it look like an 'unsyncopated form' while actually it's a insertion of a schwa to break up the difficult cluster.

    Then the exact same thing would obviously happen in the reduplicate d perfect.

    Then, when this shift had completely gone through and pretonic *a > *e; Reduplication developed again, but this time the result of the again inserted schwa was different resulting in *i.

    This would actually take care of many of that weird *e/o ablaut patterning.

    I get the idea that you assume that in MIE you only had the vowel *a which was [a] when accented and [ə] when unaccented.

    But I'm guessing there's also an *e {or /ə/ phonemically), that is only distinguished from *a when accented. But I'm not really sure if I've ever seen it pass in your reconstructions. But am I right in this assumption?

    A system of *i *u *a/á and *é? That's what seems right to me at the moment. Later due to the schwa differentiation both unaccented *e and *o developed.

  2. Phoenix: "I don't feel that this *e to avoid making declensions too obscure is some sort of paradigmatic levelling, but simply a phonotactic matter."

    Really? Then I trust you know of a monosyllabic root in PIE that shows a genitive-declined form with an asyllabic root, i.e. of the form *CC-os. I know of no such paradigm and so this is why I think this is less a matter of phonotactics and more a matter of paradigmatic levelling.

    As for reduplication, I figure that a schwa as a reduplicated vowel is more likely than any other vocalism used. In this new version of my account, we have schwa initially arising in both instances of reduplication (n.b. my unaccented *a in MIE is phonetically a schwa as well). It sounds like we're in perfect agreement at this point though.

    Phoenix: "But I'm guessing there's also an *e {or /ə/ phonemically), that is only distinguished from *a when accented."

    Yes, that's correct except that I suspect that *e was already fronted to /e/ in Mid IE.

  3. though technically not CC-ós but CC-R-ós we have:
    *uód-r > *udrós. If it would really be a form of paradigmatic levelling, I would guess that the genitive would also be *uedrós here.

    It's difficult to find good examples of CC-ós, since those are only root nouns, of which aren't that many of course.

    But yeah you'd expect *déms > *dmós while I believe we get *demós.

    It's just difficult to say whether it's a phonotactic or levelling constraint. But for now, it's not that important; I guess it'll become apparent as I go along reconstructing some more Pre-IE ;-) Either way know where the *e is going to show up is fairly predictable now.

  4. Phoenix: "though technically not CC-ós but CC-R-ós we have: *uód-r > *udrós."

    Not valid, I'm afraid. First, the genitive should be *udnós because it's a heteroclitic inanimate. It's easy to forget though and I still mess up on that.

    Next, note that in your example, the root portion of the word is not (and never was) asyllabic: *udn-. Therefore it doesn't qualify as a valid test of my rule since *u carries syllabicity even though it's technically a consonant *w on a certain abstract level.

    My original point remains that **CC-os is avoided because it was the asyllabicity of roots and stems when inflected that was avoided.

    My rule however doesn't affect derivations and so we may find *pd- (rather than *pod- 'foot') when accent-stealing affixes are attached to form a new word with its own independent declensional paradigm. Note yet again, that even in derivations, the stem is never truly asyllabic, whether in strong or weak case forms.

    Phoenix: "If it would really be a form of paradigmatic levelling, I would guess that the genitive would also be *uedrós here."

    Negatory, my ever-curious comrade, negatory. Instead, MIE genitive *wad̰an-ása 'of the water' quite simply became *ud̰nás in early Late IE via Syncope. It was *not* affected by Paradigmatic Resistance because it was never ever at any point asyllabic. While the vowel *u is masking an underlying consonant in PIE, it remains nonetheless a vowel on the phonetic level, and we need to think about this rule that I propose on a phonetic level because there is where the motivation for it lies most of all. Alternative forms like *wednós I gather would have been more recent. It would easily be coined within the Late IE period based on a pre-existing *a/*e alternation (later becoming *o/*e) which was present in other inherited paradigms like that of *pod-.

    Phoenix: "It's difficult to find good examples of CC-ós, since those are only root nouns, of which aren't that many of course."

    Nyah, nyah. Can't catch me! ;) I defy you to find any purely asyllabic stem in PIE.

    Phoenix: "But yeah you'd expect *déms > *dmós while I believe we get *demós."

    Or *dems, although I believe this latter form is a post-Syncope development coming from *demós.

    Phoenix: "I guess it'll become apparent as I go along reconstructing some more Pre-IE ;-)"

    Yes! It's fun for the whole family! :)