09 March 2013

Tense Wars, and the Survival of the Perfect


Before I address the fate of *bʰuH- in Germanic, a brief interlude is in order to provide my guests with some necessary background knowledge. The last common ancestor of the Germanic languages was spoken not much more than 2,000 years ago; the time-depth of the Indo-European family is likely to be about 6,000 years ago (with a wide margin of uncertainty). This means that perhaps as many as 4,000 years elapsed between the breakup of PIE and the stage we can reach by carrying out comparative reconstruction within Germanic. The ancestor of the Anatolian languages was the first to split from the rest of the family, followed by the ancestor of Tocharian. The remaining branch (let’s call it Neo-Indo-European) was ancestral to all the modern IE groups. It underwent further splits, diversifying into a number of distinct lineages, Proto-Germanic among them. After diverging from its closest relatives, Proto-Germanic continued evolving on its own, and developed a number of unique innovations inherited by its descendants but not shared with the rest of Indo-European.

De bello temporum
By the time Proto-Germanic began to split up, its grammar had been affected by thorough upheavals. In the verb system, aspect had lost its importance as a grammatical category. Another basic opposition – that of tense, present versus past – remained. The inherited forms of the stative aspect (the “IE perfect”) had acquired a new interpretation. The original perfect had referred to a current state brought about by a past action; now the focus had shifted to the action itself, and the perfect began to encroach on the territory of the old past tenses (the aorist and the imperfect), threatening to make them redundant. Only in a few cases did the perfect retain its stative meaning. For example, the perfect of the verb ‘to see’ (PIE *woid-/*wid- > Germanic *wait-/*wit-) survived as a present-tense verb meaning ‘to know’ (= ‘to have seen’) – a phenomenon found also in other IE languages (Sanskrit véda, Gk. oîde ‘(s)he knows’). The form of the perfect had been modified too. For example, most perfect stems had lost one of the typical traits of that category – the partial doubling (reduplication) of the root syllable. But the perfect kept some of its special inflectional endings, as well as the characteristic vowel alternations that distinguished it from the related present stems. They are still visible in sing vs. sang or drive vs. drove.

The battle of the tenses ended with the crushing victory of the perfect. The aorist and the imperfect died out almost completely. One solitary survivor was the imperfective past tense of the verb ‘to do’ (from PIE *dʰeh₁-). Its reduplicated imperfective (PIE 3sg. *dʰi-dʰéh₁-t, 3pl. *dʰé-dʰh₁-n̥t) still survives as the past tense did. It was also employed as an auxiliary verb that formed a periphrastic past tense in combination with a past participle. In this way, a host od secondary (derived) verbs without an inherited perfect of their own could acquire a semantically equivalent past tense. It was a useful function and it helped to keep the last imperfect alive. However, the auxiliary suffered the common fate of grammatical words – phonetic attrition. No longer a free-standing word, it degenerated into a clitic and then a suffix. Its last visible trace is the past-tense ending of “regular” verbs (English -ed), and even that actually reflects the suffix of the participle fused with the ex-auxiliary (that’s why loved is today both a participle and a past tense).

The aorist fared no better. It did not survive at all as a past tense in Germanic. To be sure, some derivatives of old aorists lived on as present-tense forms, but most of those had long been divorced from their historical source. For example, Proto-Germanic  had no past tense corresponding to the root aorist *gʷem-/*gʷm- ‘come, go’, but it did have a present stem (*kʷim-i-/*kʷem-a- > Goth. qiman, OE cuman ‘come’) continuing the pre-Germanic “simple thematic present” *gʷém-e/o- (3 sg. *gʷém-e-ti, 3pl. *gʷém-o-nti). It probably started out as the subjunctive mood of the root aorist. Such subjunctives began to evolve functionally into present-tense forms soon after the separation of Anatolian (cf. Vedic jám-a-ti ‘he comes’), and this new type of present became very productive in the Neo-IE languages. But there was also another type of present related to root aorists, with the zero-grade of the root and an accented suffix. For example, the root aorist *gʷerh₃-/*gʷr̥h₃‘devour’ produced the present *gʷr̥h₃-é/ó-, found in some branches of Neo-IE. Such presents retained much of the perfective force of the aorist: they referred to telic (goal-oriented) actions. They can also be found in Germanic, e.g. *wig-i/a- < *wik-é/ó- ‘conquer, fight’ (attested also in Celtic).

The discussion of ‘to be’ can be continued now, but it will have to wait till the next post.

40 comments:

  1. I think your idea of a Proto-Germanic language descending from "PIE" after a number of successive splitings is very simplistic, not to say incorrect. And in particular, the 3-way verbal system (present/imperfect/aorist) was an innovation of the Indo-Greek group not shared by other IE languages.

    In pure evolutive terms, the distinction stative/active is older than the aspect.

    ReplyDelete
  2. If you mean the family tree model is simplistic, I partly agree. If you mean that PIE is a fictitious entity and that PGmc. emerged from a variety of substrates as a mixed language, I disagree.

    I'm not sure what you understand by a three-way verbal system. Im my book, PIE (or at least Proto-Core-IE, after the separation of Hittite, had three aspects and two tenses. The present and the imperfect belonged to the same aspect; their forms were based on the same (imperfective) stems, so the only difference between them was the presence or absence of overt tense markers. If not the full system, then at least its fragments are well attested also outside Greek and Indo-Iranian.

    ReplyDelete
  3. ... I mean Anatolian, not just Hittite, of course. And since traces of the imperfective-perfective ("present-aorist") distinction can be seen even in Anatolian, loss in that branch is more likely than the rise of the contrast in the rest of the family ("Core IE").

    ReplyDelete
  4. This is more or less the Neogrammarians' vision in which "PIE" was a "full-enriched" entity where some of its features were "lost" in its descendants in a regressive way. Of course, I disagree with such approach.

    The Spanish Indo-Europeanist Rodríguez Adradosw, who has studied in deep the IE verbal morphology, is among the scholars who see Anatolian as a far more archaic group than the rest of IE languages. In particular, he considers the perfect as a post-Anatolian innovation.

    He also proposes a more refined tree model than the traditional one, with several intermediate stages.

    ReplyDelete
  5. You are using a straw man argument now. I have never said PIE was grammatically richer than any of its descendants. It's clear that lots of innovations arose after the breakup of PIE. The feminine gender, for example, and the whole conjugational class of "simple thematic presents" -- to cite only the most familiar examples. Anatolian is basal but not so archaic (in the sense of being typologically close to PIE). Quite the opposite, Anatolian was higly innovative on the one hand, and it had lost many protolanguage features on the other. It is not possible to decide at present if the perfect as such (with its stative semantics) goes back to PIE, but at least there was some kind of "protoperfect" with reduplication and *h2a-conjugation endings. We can see its traces even in Anatolian.

    ReplyDelete
  6. The active is a voice, and the stative is an aspect, so they are not in opposition to each other. Perhaps you mean active/(medio)passive. The first paragraph is a load of utter nonsense. PIE is by definition the language in which the shared inherited features of all the known IE languages (including Anatolian) coalesce as you move back in time. Whatever the lateral influence from other sources, the inherited core is large enough to guarantee the existence of such a language, and our reconstruction methods allow us to recover large parts of its structure. It doesn't matter that the reconstruction is approximate and open to modification, and that the coalescence of different reconstructed features can be traced back to different chronological layers. So Anatolian didn't "lose" anything is simply a non sequitur.

    ReplyDelete
  7. Hey, Octavià! It's difficult to discuss like that. I reply to your comment and then by the time I post my answer your comment is gone.

    ReplyDelete
  8. In Adrados' model, Anatolian split from a more archaic proto-language which he calls IE II aka "monothematic". After that, IE II evolved into IE III, which in turn split into IE III A aka "polythematic" (the Indo-Greek group) and IE III B aka "bithematic" (the rest). He also detects traces of an even more archaic stage which he calls IE I aka "pre-flexional".

    ReplyDelete
  9. "More archaic" than what? If Anatolian split off from "IE II", then "IE II" = Proto-Indo-European by definition (as the most recent common ancestor of the whole family). It doesn't matter that it isn't like the Neogrammarian idea of PIE from the time before Anatolian and Tocharian were discovered.

    ReplyDelete
  10. Sorry, Piotr! I have to rehearse my own writings in order to produce a quality product.

    The thing is that Adrados' "IE II" doesn't have the same features that your "PIE", no matter what the standard definition might be. In fact, for Adrados the actual "PIE" would be the much older "IE I", which his former disciple Villar dates to the Gravettian period.

    My own view is closer to Villar's than Adrados', as it involves language replacement. This way, some of the features shared by IE languages come from the language (or rather a group of closely related dialects) spoken by Kurgan people. But others, shared by particular groups of IE languages but not others, must come from the replaced languages, some of which might be related to Kurganic.

    ReplyDelete
  11. To put it simply, Kurganic (roughly equal to Adrados' "IE III A") would be the direct ancestor of the Indo-Greek group, but only a superstrate to the other IE languages.

    Interestingly, some IE-ists (Georgiev, Woudhuizen, Adrados) consider Etruscan to be an Anatolian language or a close relative of that group. However, most linguists (including myself) would agree that Etruscan doesn't belong to the IE family because it lacks core IE lexicon and viceversa. While some of the isoglosses shared by Anatolian and Etruscan might be due to areal contact (possibly Trojans were Etruscan-speaking ruled by a Luwian aristocracy), others might reflect part of the European linguistic landscape prior to Kurganization.

    For example, the IE numeral *Hok´te-H3(u) '8' can be considered as a fossilized dual of a numeral '4' reflected in Etruscan huθ. Although purely anecdotical, this illustrates my point.

    ReplyDelete
  12. The actual PIE is not some hazy mirage of a pre-inflectional stage (spoken by people apparently unable to utter more than one syllable at a time) but the language system accessible to us via the application of the comparative method. You can then try to infer something about its remoter pre-stages using internal reconstruction within PIE, but we don't call those vaguely discernible pre-stages "actual PIE", and I don't know any serious mainstream linguists who would date them all the way back to the Upper Paleolithic.

    Anatolian is "monothematic" in the sense that all the forms of a verb are based on the same stem. But this state of affairs is unlikely to be primitive. Most of the Anatolian verb stems contain suffixes (as well as other morphological devices, such as reduplication) which in other languages of the family derive imperfective (present) stems from root aorists. What are they doing in Anatolian? Were they added to verb roots just for the sake of embellishment, like decorative spandrels? Adrados' views are well known to people working in the field, and regarded as part of the eccentric fringe. He likes to accuse others (including authors of handbooks) of sticking to the outdated Neogrammarian model and of ignoring the Anatolian evidence, which is totally unfair and possibly reflects his own lack of familiarity with recent developments in IE and Anatolian studies.

    ReplyDelete
  13. I'm afraid that being part of the mainstream or the "eccentric fringe" doesn't make a theory better or worse, so the "majority view" argument doesn't appeal to me. I also remind you that the Upper Paleolithic chronology isn't Adrados' but Villar's, and I personally won't go that far.

    Although I'm not familiar with the particulars of the IE verbal system theory, I think the "archaicness" of Anatolian is patent in other areas such as noun morphology (e.g. lack of masculine/femenine), phonetics (conservation of "laryngeals") and lexicon, so Adrados' criticism isn't totally unjustified.

    Note: When I said "stative/active" I meant two different verb classes, not the idiosyncratic use in IE studies.

    ReplyDelete
  14. I'm not saying that the "majority view" must be correct, but it is shared by so many experts for pretty good reasons. Namely, it is supported by careful analyses of the available data, so any alternative should respect the totality of the evidence and explain the same facts at least equally well to merit serious attention. We do hot hesitate to abandon even a well-established model if is not compatible with facts. One example I gave earlier -- that of the simple thematic presents -- is instructive. The type is totally absent from Anatolian, rare in Tocharian, and we have a convincing explanation of its origin and of the pattern of its gradual emergence. The realisation of those facts was enough to convince most IEists that presents like Vedic bhárati were not really PIE depite the great productivity of their formation in languages such as Greek, Indo-Iranian, Germanic, Celtic, Italic, Balto-Slavic, etc.

    The fact that Anatolian is archaic in some respects does not prove anything with regard to other features. To think otherwise is a logical fallacy -- the same that made 19th-c. linguists think that the simple vowel system of Sanskrit must be primitive just because so many other aspects of Sanskrit looked archaic.

    ReplyDelete
  15. More than a logical fallacy, I think it's a question of probability. Also as regarding phonology, a 3-vowel system is more archaic than a 5-vowel one, although it doesn't mean it undergone the same evolution in all the IE language. In fact, Villar proposes a 4-vowel for earlier stages of the IE family. But since you mentioned Sanskrit, the fact contemporary IE-ists still reconstruct the so-called "voiced aspirated" for PIE is consistent with the outdated Neogrammarians' view (yes, I used those taboo words!).

    Notice that I don't use the term "PIE" because the notion of a single protolanguage ancestor to all IE languages is itself inconsistent with linguistic data.

    ReplyDelete
  16. No, a system with 3 vowels is only smaller than one with five vowels. In real life, both splits and mergers take place, so the size of an inventory doesn't tell you whether it's "primitive" or "evolved". As for "voiced aspirated" stops, the common understanding is that they were breathy-voiced or had some similar kind of phonation, not necessarily the same as in Sanskrit or the Modern Indo-Aryan languages. We know that they sometimes merged with "plain voiced" stops, sometimes changed into aspirated ones, and sometimes into fricatives (voiced, as in Germanic, or unspecified for voice, as in Italic). If you prefer some non-standard reconstruction, justify it and explain how they have produced the attested reflexes.

    ReplyDelete
  17. Like Newton's gravitational theory, the traditional PIE model can still be useful for some practical purposes, but it's far from being an ultimate theory like Einstein's.

    ReplyDelete
  18. From external comparison and internal data, I gather that traditional "voiced aspirated" (series III) were actually plain voiced. By contrast, traditional voiced (series II) weren't actually voiced but rather ejective or pre-glottalized in an earlier phase. Also traditional voiceless (series I) were phonologically aspirated, although this feature was only relevant in Germanic and Armenian (and partially also in Celtic). This is more or less what the so-called "glottalic" theory states, although they're other proposed variants.

    In this model, Grimm's Law in Proto-Germanic becames greatly simplified, reduced only to the spirantion of the voiceless aspirated series. The hardest thing to explain is the shift of series III in Italic and Greek, but I think this an areal feature due to language contact which also involved Etruscan (or rather its ancestor).

    ReplyDelete
  19. For example, the IE numeral *Hok´te-H3(u) '8' can be considered as a fossilized dual of a numeral '4' reflected in Etruscan huθ. Although purely anecdotical, this illustrates my point.

    Except that it's now quite clear that huθ means '6'. See Artioli, Nociti and Angelini (2011).

    ReplyDelete
  20. In this model, Grimm's Law in Proto-Germanic becames greatly simplified, reduced only to the spirantion of the voiceless aspirated series.

    This ignores the fact that PGmc. *b, *d, *g, *gʷ were actually voiced fricatives rather than stops. It also fails to account for the different development of Celtic loans in Germanic before and after Grimm's Law, and does not allow one to connect Kluge's Law with the Grimm/Verner complex of changes in a phonetically plausible way.

    ReplyDelete
    Replies
    1. Some of the oddities of the Proto-Germanic consonantal system could be explained if it was based upon a tenseness rather than a voice contrast. So it's time these old "laws" were reformulated in a entirely new framework.

      However, I'd retain the term "Grimm's Law" for the spiration of pre-Proto-Germanic tense/voiceless stops, which were phologically aspirated (as well as for similar changes in other languages), and call his original proposal "Great Consonant Shift", after the "Great Vowel Shift" in Modern English.

      Delete
  21. Except that it's now quite clear that huθ means '6'. See Artioli, Nociti and Angelini (2011).
    Paraphrasimg Einstein, comparative linguists don't play dice.

    ReplyDelete
  22. There's direct evidence that Etruscan huθ is '4'. The Pre-Greek toponym Hyttenia is translated as Greek Tetrapolis '4 cities'.

    ReplyDelete
    Replies
    1. Besides that, there's another evidence: In the Tlomb of Th Charons, the 4th figure of the demon Charon is labelled Χarun huθ 'the fourth Charon'. See Giulano & Larissa Bonfante: The Etruscan Language: An Introduction, revised edition.

      Delete
  23. What's that got to do with Etruscan? Can you demonstrate that Pelasgian was related to Etruscan? Can you prove that the Pelasgian name also meant 'the Four Cities'?

    The dice argument is waterproof in comparison, since in either of the two (and only two) attested numerical combinations 3 and 4 are on the opposite faces of the die.

    ReplyDelete
    Replies
    1. What's that got to do with Etruscan? Can you demonstrate that Pelasgian was related to Etruscan?
      Although it would be naïve to assume there was only one Pre-Greek language, as Beekes do, some Pre-Greek material has Etruscan counterpats. For example, Greek ksanthós 'yellow' ~ Etruscan zam(a)θi 'gold'.

      I suppose Pelasgian *hutt reflects an intermediate form with *-kt- > -tt-.

      Delete
  24. On the contrary, in Antiquity, the sum property of dice wasn't exactly invariable, due to imprefect manufacturing or other reasons. So this "evidence" cam hardly be reliable.

    ReplyDelete
  25. Let me explain it for you. The authors of the study examined 93 Etruscan dice, 91 with pips and 2 with words. Of the 15 possible combinations only two are attested, and even those two do not occur at random. One of them (1-2, 3-4, 5-6) is used on all dice before the 5th c. BC, and the other (1-6, 2-5, 3-4) on all dice after the 3rd c. BC. In both 3 and 4 are paired. That proves that śa = 4 whatever the arrangement of the dice with words. Therefore, huθ = 6. Of course you may deny reality, stop your ears and repeat "this is not evidence". But don't ask me to take such contrarianism seriously.

    ReplyDelete
  26. I'll read the article and gave you a more educated opinion.

    ReplyDelete
    Replies
    1. Unfortunately, this is a pay-per-view article.

      Delete
    2. It's behind a paywall, but I have emailed you a copy.

      Delete
    3. Thank you very much.

      Thne problem is the authors aren't linguist themselves (notice how they mispelled śa as **sa), so they rely on the view of specialists and they limit themselves to decide about the distribution of 4 and 6. But this is a logical fallacy, as the dice argument applies to the relative complementary (on dice sides) distribution of numerals 1-6, 2-5, 3-4 and NOT to their actual value.

      So if we know from independent evidence (see above) that huθ = 4, then it follows θu = 3. So it turns out that the wrong assigned numerals where 1/3 instead of 4/6. So the Etruscan numerals 1-6 would be:

      ci = 1
      zal = 2
      θu = 3
      huθ = 4
      maχ = 5
      śa = 6

      Delete
    4. The omission of a diacritic is irrelevant, and no great linguistic expertise is needed in such a simple matter. The identification of θu, zal, ci as '1, 2, 3' does require linguistic arguments and has been securely established mostly on the basis of their use in the Liber Linteus. The only arguments in favour of huθ = 4 are those you have already presented, both of them hopelessly weak. The interpretation of Hyttenia is based on untestable assumption about Pelasgians and Pelasgian toponyms, and Χarun huθs (sic! not **Χarun huθ) does not meant 'the fourth Charun': huθs is a genitive, not an ordinal numeral.

      Delete
    5. The omission of a diacritic is irrelevant, and no great linguistic expertise is needed in such a simple matter.
      I won't call the correct identification of the Etruscan numerals a "simple matter". I also disagree with your belief about Pre-Greek, as there's other evidence which contradicts it (see above).

      I also don't quite understand your argument about the genitive of a numeral can't be used as an ordinal.

      Delete
    6. Let's forget Etruscan huθ for the moment and get back to IE '8'. Kartvelian *o(ś)tx(w) '4' (Georgian nad Mengrel otx-, Svan woštx(w), Laz o(n)txo-) is apparently a loanword from Paleo-IE (a better term than "PIE") which includes the dual suffix. This would imply the numeral was originally a dual of '2', and would make it cognate to Uralic *kakta and Omotic *gutto '2'.

      This root has been studied by Nostraticists such as Dolgopolsky (who didn't include the IE numeral, though) and its original meaning can be reconstructed as 'to join, to couple', reflected in Baltic *kek(e)t-ā 'troop' and Slavic *četá: 'couple', in turn related to Latin catēna 'chain', caterva 'crown, troop', presumably borrowed from Etruscan.

      What we don't know is the reason of the shift from '4' to '8' at an early date, which left a vacant niche occupied by different words in Anatolian and the rest of the historically attested IE languages.

      Delete
    7. Frankly, I'd rather not do it here. We have strayed far enough from anything related to the topic of my post.

      Delete
    8. One day or another, I'll post about these things on my own blog, so feel free to make any comments there.

      Delete
  27. Here are the authors' conclusions:

    The results of the mathematical analysis indicate the following:

    • Only two of the 15 theoretically possible non-enantiomorphic combinations of six numbers on the cube face are observed in the ancient Etruscan dice.

    • The two observed combinations—that is, (1–2, 3–4, 5–6) [i.e., numerical difference between opposite faces = 1] and (1–6, 2–5, 3–4) [i.e., numerical sum of opposite faces = 7]—have a marked temporal distribution, with the first one being invariably in use before the fifth century bc, and the second one being invariably in use after the third century bc. In the fourth century bc,
    both combinations were in use in central–southern Etruria.

    • The assessed use of only two numerical permutations in ancient Etruscan dice, and the consistent use of 3 and 4 as opposite numbers, allows unambiguous assignment of the Etruscan numerals 4 and 6 to their graphic representation
    sa and huth, respectively. Combinatorial analysis of the Etruscan dice finally allows straightforward solution of a longstanding linguistic problem.

    Incidentally, the two dice with words turned out to be of the "modern" type (with numbers on the opposite faces adding up to 7)

    ReplyDelete