So let me talk about QAR (the Quasi-Penultimate Accent Rule) of Mid IE. I deduced based on the wandering accent seen in many of IE's nominal and verbal paradigms that Mid IE, the stage of PIE that I position before the event of Syncope (the loss of most unstressed vowels), must have had a regular accent placed mostly on the penultimate syllable (i.e. second-from-last syllable) of words. In a few instances however, the accent was antepenultimate (i.e. third-from-last syllable) but interestingly these exceptions seem to always show up like clockwork when a postclitic has been attached. These postclitics are things like *-sa, the nominative case ending derived from *sa "the" (> PIE *so); and *-ta, attached to the 3ps ending and derived from *ta "that" (> PIE *to-). I deduce that they've been attached within the Mid IE period (6000 to 5000 BCE), which means that before these additions Old IE had a perfectly regular accent on the penultimate syllable.
Proterodynamic and hysterodynamic are terms used to describe how the accent behaves in nouns when they're declined with case suffixes[1]. In proterodynamic nouns, the accent wanders between the root and the suffix of the noun stem. In hysterodynamic nouns, the accent wanders between the suffix and the case ending. More basically however, we can just combine the patterns together and state simply that their accent shifts between one syllable and the immediately following syllable. I realized a long time ago now that these two paradigms must have once been one and the same paradigm and that they were the result of a loss of unaccented final vowels (Syncope) and the underlying penultimate accent before Syncope which obscured their relationship and the nature of the wandering accent.
But maybe none of this is sinking in as long as I talk all this linguistics babble so I'll illustrate what I'm talking about with a few paradigms of PIE nouns which exhibit typical alternations in the stem between fullgrade and zerograde because of an alternating accent, thereby showing you my thoughts on their earlier case forms:
| *ḱwon- "dog" | Proto-IE | Early Late IE | Late Mid IE |
| nominative "the dog" (subject) | *ḱwōn /kʰwo:n/, *ḱuwṓn /kʰuwó:n/ | *kwán-s | *kahʷána=sa /kʰəˈhʷanəsə/ |
| accusative "the dog" (object) | *ḱwón-m̥ /kʰwónm̩/ | *kwán-m̥ | *kahʷána-m /kʰəˈhʷanəm/ |
| genitive "of the dog" | *ḱun-ós /kʰunós/ | *kun-ás | *kahʷaná-sa /kʰəhʷəˈnasə/ |
| locative "at, by the dog" | *ḱun-í /kʰuní/ | *kwán-i | *kahʷána-i /kʰəˈhʷanəj/ |
| *doru- "tree" | Proto-IE | Early Late IE | Late Mid IE |
| nominative-accusative "the tree" (subject) | *dóru /d̰óɾu/ | *d̰áru | *d̰árau /ˈd̰aɾəw/ |
| genitive "of the tree" | *dreu-s /d̰ɾéws/, *dorw-ós /d̰oɾwós/ | *d̰reu-s | *d̰aréu-sa /d̰əˈɾewsə/ |
| locative "at, by the tree" | *dréw-i /d̰ɾéwi/ | *d̰ráw-i | *d̰aréu-ai /d̰əˈɾewəj/ |
| *nepot- "grandson" | Proto-IE | Early Late IE | Late Mid IE |
| nominative "the grandson" (subject) | *népōts /népʰo:ts/ | *népat-s | *népata=sa /ˈnepʰətʰəsə/ |
| accusative "the grandson" (object) | *népot-m̥ /népʰotʰm̩/ | *nepát-m̥ | *napáta-m /nəˈpʰatʰəm/ |
| genitive "of the grandson" | *népt-os /néptʰos/ | *nept-ás | *napatá-sa /nəpʰəˈtʰasə/ |
| locative "at, by the grandson" | *népt-i /néptʰi/ | *nepát-i | *napáta-i /nəˈpʰatʰəj/ |
| *wodr- "water" | Proto-IE | Early Late IE | Late Mid IE |
| nominative-accusative "the water" (subject) | *wódr̥ /wód̰əɾ/ | *wád̰r̥ | *wád̰ar /ˈwad̰əɾ/ |
| genitive "of the water" | *wedén-(o)s/*udén-(o)s /we-, ud̰én(o)s/, *wedn-ós/*udn-ós /wed̰-, ud̰nós/ | *wed̰n-ás | *wad̰an-ása /wəd̰əˈnasə/ |
| locative "at, by, in the water" | *wedén(-i)/*udén(-i) | *wed̰én-i | *wad̰én-ai /wəˈd̰enəj/ |
It's almost inevitable that there will be some changes that may seem irregular here. They're caused either by analogical leveling (i.e. the spread of a particular form or feature across a paradigm) or by sound changes that I've added to my QAR or Syncope rules. For example, some may be curious why the nominative ending *-s disappears in the paradigm for *kwon- "dog" but this omission happens after many PIE nouns whose stems end in resonants (*l, *r, *m, *n)[2]. A characteristic lengthening of the preceding vowel is the sign of a missing case ending that once was there (e.g. *ph₂tḗr "father" < *ph₂térs, *dʰǵʰōm "earth, ground" < *dʰǵʰoms). In my theory, I place this loss of the nominative ending of these nouns at the end of the Late IE period (around 4500-4000 BCE).
NOTES
[1] Fortson, Indo-European Language and Culture: An Introduction (2004), pp.107-108 (see link). Proterodynamic and hysterodynamic are also known as proterokinetic and hysterokinetic.
[2] Fortson, Indo-European Language and Culture: An Introduction (2004), p.104 (see link).
UPDATES
(Jan 25 2008) I forgot the palatal diacritic on the *k of *ḱwon- "dog" and so I changed that and also added the Lindeman variant with intervening *-u- in the nominative. While I don't personally believe that Indo-European had palatal consonants at all and that they should be reinterpreted as plain stops, I am respecting traditional convention here to minimize confusion of readers who will be more familiar with the standard notation.
I'm curious why you have earlier /kʰ/ for PIE k, and /hʷ/ for w.
ReplyDeleteThis has to do with one of a few rules I notice concerning the loss of some laryngeals during Syncope. I call this "Laryngeal Vocalization" where laryngeals vocalize and become vowels in some cases. If *k- in late MIE were aspirated and labialized *hʷ (> PIE *h₃) were pronounced with full lip-rounding (unlike at the PIE stage where the rounding would have been "mild" to distinguish from biphonemic *-h₁w- in mediofinal positions), it's then natural that the syllable would become phonetically indistinguishable from true *kw- /kʰw/ in early Late IE.
ReplyDeleteI suspect forms like *wedéns and *dréus reveal an earlier distinction between the animate genitive, in *-ós, and the inanimate genitive, in *-s. Forms such as *udnós and *drwós would then represent a dialectal innovation of regularizing the animate genitive across both genders. What do you think?
ReplyDelete- Rob
Follow-up:
ReplyDeleteI wonder if the Hittite form witenas represents *w@déns (where '@' represents schwa) as opposed to *wedéns. This may have an important implication: the dialect that gave rise to Hittite may have separated from the rest of IE before the advent of syllabic resonants.
- Rob
Rob: "I suspect forms like *wedéns and *dréus reveal an earlier distinction between the animate genitive, in *-ós, and the inanimate genitive, in *-s."
ReplyDeleteA relationship between gender and ablaut is pretty obscure considering animate genitive *dem-s "of the house". Whereas, a connection between ablaut and accent (assuming an earlier stress accent)is quite readily apparent and intuitive. It would seem that the retraction of accent to preceding syllables after Syncope is the optimal explanation for the ablauting genitive ending.
Rob: "This may have an important implication: the dialect that gave rise to Hittite may have separated from the rest of IE before the advent of syllabic resonants."
It certainly isn't an important implication considering many simple examples such as nu < PIE *nu "now" and kuis < PIE *kʷis "who?". All Indo-European languages derive from a language with syllabic resonants.
First of all, great article! Very interesting, realistic stuff.
ReplyDeleteI like the VRs > V:R theory. Something we see in Greek, and therefore feels very natural.
I suspect forms like *wedéns and *dréus reveal an earlier distinction between the animate genitive, in *-ós, and the inanimate genitive, in *-s. Forms such as *udnós and *drwós would then represent a dialectal innovation of regularizing the animate genitive across both genders. What do you think?
Beekes (1995) mentions this suspicion as well. Though I must say above presented theory more or less claims otherwise, and is quite convincing.
I wonder if the Hittite form witenas represents *w@déns (where '@' represents schwa) as opposed to *wedéns.
I can't help but smile at this one. If you've read the blog post before this one, and especially my comments, I tried to argue something similar myself.
The conclusion that Pre-Anatolian split IE has no vocalic consonants must be wrong though. Unless you'd like to suggest that Pre-Anatolian split IE had both a /w/ phoneme and an /u/ phoneme (Does not seem very pleasant to assume).
As for the 'schwa' issue, keep an eye out on my blog at http://phoenixblog.vox.com/ I'll be writing an article on Hittite phonology soon.
Glen: "A relationship between gender and ablaut is pretty obscure considering animate genitive *dem-s "of the house". Whereas, a connection between ablaut and accent (assuming an earlier stress accent)is quite readily apparent and intuitive. It would seem that the retraction of accent to preceding syllables after Syncope is the optimal explanation for the ablauting genitive ending."
ReplyDeleteTo be honest, forms like *dems always bothered me. In almost every IE noun (in "traditional" IE), there are no inflections that cause ablaut in the root. So I find it hard to believe that forms like *dems, or worse yet, *peds (gen. sg. for "foot"), existed. What do you think were the phonological conditions by which such ablaut could have arisen? There's also the question as to why, if *doms became *do:m, *dems did not become *de:m.
As to the animate/inanimate distinction in the genitive, my current hypothesis is that the animate genitive had a prevocalic element attached (i.e. *-Vs), while the inanimate genitive did not (i.e. *-s). Thus there is no actual ablaut going on (aside from qualitative ablaut in the animate genitive, but that's another story).
Glen: "It certainly isn't an important implication considering many simple examples such as nu < PIE *nu "now" and kuis < PIE *kʷis "who?". All Indo-European languages derive from a language with syllabic resonants.
Has anyone defined the reflexes of IE syllabic resonants in Hittite?
Phoenix: "I can't help but smile at this one. If you've read the blog post before this one, and especially my comments, I tried to argue something similar myself.
The conclusion that Pre-Anatolian split IE has no vocalic consonants must be wrong though. Unless you'd like to suggest that Pre-Anatolian split IE had both a /w/ phoneme and an /u/ phoneme (Does not seem very pleasant to assume)."
It would seem to me that some pre-stage of IE (or perhaps the earliest stage of IE itself) must have distinguished between /w/ and /u/ as phonemes. Certain facts need to be taken into account:
1. Evidence in languages such as Hittite where one seem to find sequences of */w/ + */@/. These could not conceivably come from earlier */u/.
2. The entire labiovelar series. :P
- Rob
Rob: "What do you think were the phonological conditions by which such ablaut could have arisen?"
ReplyDeleteTo me, *dems and similar forms seem to be caused by a retraction of stress from an originally accented case ending *-ós towards the stem. A former alternation of accent would then explain the alternation of *o and *e in the stem. I can't see any other solution that makes sense.
Rob: "There's also the question as to why, if *doms became *do:m, *dems did not become *de:m."
... Which might hint that it was *demós at the time of this sound change. However, note also *n̥s "us" and the accusative plural *-ns which also evidently survived.
Rob: [...] my current hypothesis is that the animate genitive had a prevocalic element attached (i.e. *-Vs), while the inanimate genitive did not (i.e. *-s)."
This is ad hoc overanalysis. You forget that you need to first prove the morphemic status of such a vowel and you have not. It's only a random assumption among many.
Rob: "Has anyone defined the reflexes of IE syllabic resonants in Hittite?
Yes, I just did. It's futile to question something so well-founded.
Rob: "2. The entire labiovelar series. :P"
I hope you understand that the labiovelar series is monophonemic in nature and was kept very distinct from the sequences of velar stop and semivowel *w. In *ḱwōn "dog (nom.)", one should pronounce it with heavy lip-rounding even to the point of saying *ḱuwōn (c.f. Lindeman variant). On the other hand, the lip-rounding of labialized velar *kʷ was comparative weak which explains its merger with *k as in Indo-Iranian.
Glen: "To me, *dems and similar forms seem to be caused by a retraction of stress from an originally accented case ending *-ós towards the stem. A former alternation of accent would then explain the alternation of *o and *e in the stem. I can't see any other solution that makes sense."
ReplyDeleteThe question is, what would cause the retraction of stress? Is it a conditioned or an unconditioned sound change? If the former, what were the conditions? Do you know of any other evidence for such a change?
I find this scenario unlikely because it seems to posit two different cases of syncope. Otherwise, it would seem that one round of syncope is sufficient to explain the alternations in IE noun inflections.
Finally, what evidence do you have for positing *dems at all, besides Greek de:spóte:s and Sanskrit dampati?
Glen: "... Which might hint that it was *demós at the time of this sound change. However, note also *n̥s "us" and the accusative plural *-ns which also evidently survived."
Given that both of those forms coexisted, there are the following hypotheses:
1. A different process (or set of processes) operated on each. Ceteris paribus, this would mean that the originating forms were different somehow. We would then need to figure out the difference(s).
2. The same process (or set of processes) operated on both forms, rendering them effectively identical to start out with. Something else happened later to one or the other to render them different from each other.
Which of the above do you think is more likely and why?
Glen: "This is ad hoc overanalysis. You forget that you need to first prove the morphemic status of such a vowel and you have not. It's only a random assumption among many."
How else do you explain the difference between these "aberrant" heteroclitic and other* forms and the "regular" athematic forms?
* I see a connection between the "aberrant" heteroclitic forms and the s-stem adjectives (e.g. Greek eumené:s 'of a good mind' < *h1su-menéss) and the neuter i/u-stems (e.g. *dóru/déru vs. *dréus). Notice that these are all inanimate.
Glen: "Yes, I just [defined the Hittite reflexes to IE syllabic resonants]. It's futile to question something so well-founded."
Sorry, where did you do that again? (I mean that as an honest question -- no sarcasm there.)
Glen: "I hope you understand that the labiovelar series is monophonemic in nature and was kept very distinct from the sequences of velar stop and semivowel *w. In *ḱwōn "dog (nom.)", one should pronounce it with heavy lip-rounding even to the point of saying *ḱuwōn (c.f. Lindeman variant). On the other hand, the lip-rounding of labialized velar *kʷ was comparative weak which explains its merger with *k as in Indo-Iranian."
Oh, of course I understand that. :P
My point was about the labiovelar series only. The sequence *kw could be explained as coming from **kVw, where V is a vowel other than /u/.
- Rob
This might be, by far, the longest comment ever created in the history of Blogger, hahaha. Bon apétit, mes camerades.
ReplyDeleteRob: "The question is, what would cause the retraction of stress? Is it a conditioned or an unconditioned sound change? If the former, what were the conditions?"
If there is a similar retraction of stress seen in Narten presents: *bhērti "he carries" vs. *bhérn̥ti "they carry", then *dems can derive from an earlier form *demós in the same way that *bhérn̥ti was surely once *bherénti. A formerly alternating accent has been regularized onto the root. Accent regularization is motivated by greater grammatical simplicity for its speakers and this is why acrostatic nouns are so common in PIE, but they don't represent the accent of Pre-IE before Syncope. These forms are restricted to the Late IE Period.
Rob: "I find this scenario unlikely because it seems to posit two different cases of syncope."
I find it necessary because otherwise accent and/or morphology is not regular in the Pre-IE stages into which I've delved so intensely.
For me, the Indo-European case suffixes originate from postpositions that have been agglutinated to the noun stem in early Proto-Indo-Aegean (circa 8500 BCE). So the genitive started out originally as a dative postposition *si, attached to the noun stem just before Vowel Centralization was complete. Vowel-final stems in this stage (including in *-i & *-u) received *-si while consonant-final stems received *-a-si with an intervening vowel between stem and case suffix. After Penultimate Accent Shift, Old IE inherited these case morphemes as *-sə for vowel-final stems and *-ásə for consonant-final stems. This became Mid IE *-sa and *-ása due to neutralization of height in final vowels. This predictably produces *-s and *-ás after Syncope in early Late IE which then becomes *-s and *-ós after Vowel Shift in the latter half of the Late IE Period. Everything is perfectly devised so that the reason for *h2weis "of the bird" is lucid (< Mid IE *hawéi-sa < early IAeg *háwi-si). :)
As far as I know, there is no kind of alternation like *h2awi-/*h2wei- in stems ending in anything other than semivowels. So *dems must be a different phenomenon somehow and my theory strictly predicts earlier *demós which is hardly a strange result.
Rob: "Finally, what evidence do you have for positing *dems at all, besides Greek de:spóte:s and Sanskrit dampati?"
Hey, don't blame me :) It seems to be standard (see here with more evidence of dems).
Rob: "Which of the above do you think is more likely and why?"
Phew! So many questions, haha. Well, I know that accusative plural is ancient, from Mid IE *-am-as, a simple agglutination of accusative singular *-am and the plural *-as. This regularily produces *-m̥s after Syncope and assimilation to *-n̥s is sure to follow. (The simple nominative plural *-as didn't reduce to *-s but rather was strengthened to *-es during Syncope in order to maintain number contrast, by the way. But I digress)
I also believe that 1pp *n̥s dates directly to Syncope, from Mid IE enclitic *nas.
On that topic, I still have some things to work out... or rework out, so bear with me.
Rob: "How else do you explain the difference between these "aberrant" heteroclitic and other* forms and the "regular" athematic forms?
The connection with gender just isn't intuitive, sorry. I believe that accent position has a better chance of explaining most of this phenomena and, as I suggested, any accent stemming from the Mid IE stage should be predictable according to QAR unless obscured by subsequent innovations in Late IE.
Rob: "Sorry, where did you do that again? (I mean that as an honest question -- no sarcasm there.)"
No prob. Just so that we're on the same page, a "syllabic resonant" in Proto-Indo-European is any of the following: *u, *i, *l̥, *m̥, *n̥ or *r̥. So it suffices to show that since Hittite kuis preserves a syllabic resonant (namely *i in *kʷis), your original claim that "the dialect that gave rise to Hittite may have separated from the rest of IE before the advent of syllabic resonants" is effectively dismantled.
Rob: "My point was about the labiovelar series only. The sequence *kw could be explained as coming from **kVw, where V is a vowel other than /u/."
But this has been my view for a long while. The labiovelar series comes from *k neighbouring *u before Vowel Centralization in early Indo-Aegean, whereas *kw is only dated to Late IE after Syncope. In my theory, your "V" can only be one thing: unaccented *a.