17 Jul 2007

Thoughts on the early Indo-European subjunctive 1ps ending

I had an intriguing thought a while back. No doubt it's published but you never know.[1] Maybe I'm original for once. So just in case, I want to pass it on to you all. It was a flash I had after I wrote something previously about the subjunctive and its development into the present-future indicative of non-Anatolian/non-Tocharian languages of the Indo-European family of languages.

What always nagged my curiosity was why on earth Proto-Indo-European (PIE) had a first person singular in *-oh₂ for thematic present-future forms (e.g. *bʰér-oh₂ "I carry") yet *-mi for athematic ones (e.g. *káp-mi "I take"). In the past I came up with a whole bunch of completely lame excuses, but I assure you all I was desperate in the absence of guidance from wiser academics who didn't seem to touch the subject in a comprehensive and comprehensable way. However, I think, my dear Watsons, I've finally got it!

It seems to me that the first person of the subjunctive implies something that is different from all other persons. In most languages, the subjunctive is the mood we use when we want to indicate a hunch, emotional feelings, a hypothetical situation or any other idea that lies outside the strict statements of empirical fact that the indicative is meant to cover. Naturally, however when we convey our own feelings or our own potential acts in the first person, we are surely speaking with the utmost authority. Afterall, who else knows best what I will do but I myself. (Speaking of which, I will no doubt go to a local café again to ponder on early Indo-European grammar some more, hehe.) The other persons of the subjunctive are different because we can never be as sure what the potentiality of an action or the true feelings of another person really are, were or will be. Thus we can say that the first person subjunctive is a more certain statement of irrealis when compared to all other persons.

With that revelation, let's agree then that original PIE (the one that preceded the oldest divergeance, that of the Anatolian and Tocharian languages) likely contained the following conjugated forms:
  • indicative present-future 3ps *bʰēr-ti "he carries"
    (later *bʰér-e-ti via the subjunctive)
  • indicative preterite 3ps *bʰēr-t "he carried"
  • subjunctive 3ps *bʰér-e-t "he would carry"
  • mediopassive 3ps *bʰer-ór "he carries (something for himself)"
    (later *bʰer-e-tór or *bʰer-e-tói)
Notice the mediopassive form which was used to express passives (i.e. "It was given") or reflexives (i.e. "I shave (myself)" or "I sit (for myself)"). In its earliest form, it's known to have used the same endings as in the perfective but with the addition of a mediopassive suffix *-r and a shift of the tonal accent to the suffix. This form is witnessed in Celtic, Anatolian and Tocharian branches. The use of perfective endings was a natural consequence of the fact that these endings marked largely intransitive actions to begin with. Reflexives and passives can be seen as a subclass of intransitive verbs. So naturally then, a specialized mediopassive ending was formed out of the originally general perfective ending. However, if we have to accept this chain of events, it implies that before the mediopassive was created, the perfective endings themselves had been used to mark reflexives in an earlier stage of (pre-)PIE.

Now we come back to the 1ps of the subjunctive which low and behold seems to contain a perfective ending, something that I contend marked the reflexivity inherent in the subjunctive of that person. When stating, for example, "I would go", one is in effect saying "I would go (speaking for myself personally)". It's no wonder then, considering it this way that we should find the subjunctive 1ps terminating in *-oh₂ (that is, that ubiquitous thematic vowel *-e/o- plus the old perfective ending *-h₂(e) that marked the ancient reflexive as it existed before the adoption of a specialized mediopassive conjugation).

Now I can make sense of the suppletive pattern we see in the subjunctive-turned-future-indicative endings:
  • 1ps *bʰér-o-h₂ "I would carry (speaking for myself)"[2]
    (later *bʰér-o-h₂ "I will carry")
  • 2ps *bʰér-e-s "you would carry (as far as I know)"
    (later *bʰér-e-si "you will carry")
  • 3ps *bʰér-e-t "he would carry (as far as I know)"
    (later *bʰér-e-ti "he will carry")
If the above hasn't already been published, then please publish the damn thing so that we can all move on with our busy linguistic-obsessed lives and adequately reconstruct far older protolanguages sometime in this century. Thanks a bunch.

[1] The above thought, by the way, is built on ideas already published (see Hittite and the Indo-European Verb, by Jay H Jasanoff) but I'm just not sure whether anybody explained the subjunctive in this way before. Hmmm...
[2] Some people, I wager, will be still confused about why there is no *-i on the later thematic 1ps indicative *bʰér-o-h₂ but in order to explain this we need to appeal to Mandarin Chinese and its particle zài as a real-world parallel of both grammar and etymology with that of the marker *-i in Proto-Indo-European. Just as Mandarin zài, a particle literally meaning "being there" used to mark continuous actions, is barred from perfective forms with particle le, so too would *-i be likewise disallowed in PIE's perfective aspect, despite many IEists who like to entertain the possibility of the two endings coexisting together. If what I'm saying is correct (that the 1ps subjunctive was in effect an old reflexive perfective), then we have a grammatical reason to justify the attested avoidance of *-i in being used in *-o-h₂.

(Aug 3 2007) I don't know what possessed me to put the accent on the thematic vowel of the subjunctive paradigm (i.e. I wrote **bherét instead of proper *bhéret) but I took my medication and I feel better now. The accent should be on the root and so I've changed all instances of it here. Good thing I review my own crap. In fact, I can't explain the alternating *e/*o pattern of the thematic vowel unless it is unaccented so this suits me just fine. I plan to talk about thematic vowels further in an upcoming blog entry.

(Mar 19 2008) I corrected the following "In its earliest form, it's known to have used the same endings as in the perfective but with the addition of a mediopassive suffix *-r and a shift of the tonal accent to the penultimate syllable." The part in italic bold was changed to "suffix". I assume I was mixing up different stages of Pre-IE when I wrote this because this statement is only valid for Mid IE.


  1. I'm recruiting hosts for the Four Stone Hearth. Please e-mail me!

  2. Hi, I've been following your blog ever since I ran into your last entry on the PIE subjunctive. Once again your theory seems to be correct.

    The bit on -h2+i also seems to be correct, although Anatolian languages clearly reflect -h2i as an ending (as -hhi), which make it seems a bit suspicious in my eyes, but on the other hand, it seems plausible that this is an Anatolian innovation, especially due to the lack of proof of the *-i in other languages.

    I wouldn't dare saying that this is unpublished, but being a student of Comparative Indo-European Linguistics at the university of Leiden, I would expect my teachers to tell me about this, or at least touch upon the subject. But none of them have. Next year I'll be following a course on the reconstruction of Hittite though, for which we'll undoubtedly get a fair amount of Jasonoff's work to read. We might discuss it by then, I'll keep you updated ;)

    If it really hasn't been published though, I would actually make work of getting it out if I were you.

  3. Hey, welcome back, Phoenix!

    It's always tempting to assume that everything in an older branch such as Anatolian is an archaicism but naturally that can't logically be the case. Even Anatolian has innovated, surely. I don't know of a language that doesn't innovate. The question is what has it changed in the original system and what was the original system.

    Ever notice that in the "rest of IE" there are some verbs that are "durative" (or "present") and some that are "aorist" by default? Ever notice that Anatolian verbs are either "mi-class" or "hi-class"? Ever suspect that they could both be united into a higher system, a kind of IEist version of the "Theory of Everything"?

    Let's say, oh I dunno, the original system had verbs which were either inherently "transitivish" or "intransitivish" by default and that they were conjugated with the -mi and -h2e endings respectively.

    Now, let's also say for kicks that verbs could originally switch back and forth between the two conjugations (similar to how you can convert a durative to an aorist or vice versa; but not quite like in Anatolian, afaik). Could there also be, synonymous with verb transitivity, an automatic difference in aspect between the two forms, perhaps? Could there be tonnes of linguistic parallels in modern languages to support such a thing?

    So let's give the "intransitivish" verbs an inherent perfective nuance and "transitivish" verbs a non-perfective one. Over time, Anatolian loses track of what the original meaning of the mi- and hi-class verbs was and, in confusion, decides to tack an -i onto the hi-class to even things out for symmetry's sake. Meanwhile, most other branches end up using the h2e-conjugation strictly for the perfective because... it already was being used as a perfective from the beginning anyway.

    But I'm crazy and I didn't sleep a wink last night so this idea is likely swiss cheese. Crazy, sleepy people shouldn't publish things ;)

    (By the way, if I recall, Nostraticist Allen Bomhard already suggested in 1996 a plausible idea that the mi/h2e-conjugation derived from some earlier pre-IE objective/subjective contrast but sadly Nostraticists tend not to be very detail-oriented.)

  4. Continuing on the idea of transitivish vs intransitivish, there's actually some proof for this in Anatolian. Though sparse.

    It can't be said that hhi is typically intransitive, but it does have the tendency. What's also interesting is that there's two types of derivations of verbs.

    First there's the causatives (To make ....), second the fientives (to become .... )

    for the causatives there's the suffixes -nu and -ahh, first for mi verbs, second for hhi verbs.

    The fientives though only have one suffix -êss, and this only conjugates according to the mi conjugation. Not completely strange ofcourse, because if you turn into something, you can think of that as a form of transitivity.

    It's also interesting to see that if these suffixes are used on adjectives, there's sometimes free variation between the -nu- and -ahh- suffix.

    Like "marsa-" `profane, unholy' > marsahh-/ marsanu `to make unholy'.

    This does seems to give the idea that Indo-Anatolian indeed had the transitive vs intransitive split with mi vs h2e.

    On quite a different note, but I'm curious what you think of it; the presentive suffixes.

    As you probably know, the aorist vs present root is often contrasted by suffixes/infixes. In greek you often see -nu-, nC-an and -ske/o suffixes. but also ie/o (phainô vs ephanon from an earlier phanyô vs ephanon).

    some of these suffixes have been very important for the spread of thematic verbs in most indo-european languages. But they don't often seem to have a semantic meaning any more. Like in Greek they're simply there to indicate the present/durative.

    But we'd like to think that Indo-Anatolian did not haven an Aorist; or rather that the aorist was simply the standard form, and these thematic suffixes were there to modify the meaning.

    But even in Hittite there's already verbs that simply take a suffix; without a non suffixed form, and without a clear difference in meaning with forms where we do say both the suffixed and unsuffixed form. Would you have any thoughts on the meaning/function of these suffixes?

  5. Oh dear goddess. I should just make another blog ;) Have fun with my data dump. You asked for it.

    phoenix: "It can't be said that hhi is typically intransitive, but it does have the tendency."

    Yes, I'm aware which is why I didn't commit to the literal idea of an intransitive-transitive system in PIE and used the word "transitivish". A subjective-objective system (as in Uralic and according to Bomhard) is very "transitivish" without being overtly so, yet still exhibiting some of the aspectual differences (cf. Nenets grammar) that I need to nail Pre-IE conjugation for good. If only I can somehow tie active-stative contrasts in with animacy, I can finally get to work on Indo-Aegean. Ick. Grammar hell! Get out the Tylenol!

    phoenix: "Not completely strange of course, because if you turn into something, you can think of that as a form of transitivity."

    You mean *-eh1-s-? Thinking less "transitive" and more aspectual, the use of *-mi would be because of its inchoative nature. It's certainly non-punctual.

    phoenix: "[...] -nu-, nC-an and -ske/o suffixes. but also ie/o [...]"

    Oh boy. You're delving into Pre-IE. Perilous. Okay, come to the dark side with me, my friend :) I like to understand *-sk^e- as *-s- (originally describing a state of an action; later "aorist") plus *-g^e- (emphatic particle *g^e, agglutinated in "early Late IE"). So it originally conveyed an iterative or intensive, I presume. You may say poppycock to an agglutinated *-g^e-... but then how else can one explain 1ps pronoun *eg^o: so elegantly: a verb, parallel in formation to Inuktitut uvanga thanks to that very emphatic particle *g^e, attached to locative particle *e to form the ancient thematic stem *eg^e- "to be here"?

    The suffix *-ye/o- is found in presentives, in denominal verbs and I'd dare even say fossilized to the causative in *-eye/o-. All from a common origin: the early Late IE enclitic relative pronoun *yə (accented *ya), agglutinated to the verb centuries before PIE proper. The schwa became *e before unvoiced phonemes and *a before voiced ones during Schwa merger. This initial split of schwa to *e/*a was at first caused by length differences triggered by voicing (shown to exist in English itself, in fact). Then by way of Vowel Shift *a was pushed back and rounded to *o in most positions. Schwa Split and Merger combined with Vowel Shift were some of the latemost phonetic changes which would produce that ubiquitous "thematic vowel" alternation in PIE.

    Or maybe I'm crazy. If all else fails, my theories make for a great conlang.

  6. First of all, thanks for being so kind for reading my blog.

    Speaking of the -s- and -sk- suffixes. Tocharian B uses them a lot (I bet you can tell which class I followed last year by now :P).

    The s-e suffix performs no special function in Tocharian; it's simply a present suffix.

    The sk-e suffix is different though. First it's use as a present marker. It's also used in compound suffixes like na-ske (<*nh2-sk-e) amd nä-ske (<*nu-sk-e), which is odd, to say the least.

    But there is one even stranger use for the ske suffix. It is used to make causatives.

    Words that have the so called a-preterite can have causatives in ske suffix. But in the past tense of these causatives this ske suffix is lost, and the differentiation is made by a palatised initial and an a grade vowel in the root, which takes the accent.

    so for example the roots kärs- `to know'

    has the preterite śarsa `he knew'(a is an accented ä here).

    The causative is then formed from the preterite rather than the present. Giving: śarsäske/ṣṣä `to let know',

    But the causative preterite would then lose it's suffix again but has a different stem formation than the non-causative preterite.

    The preterite is śārsa `he let know'.

    It's pretty interesting to see that even when a verb's function is expressed by one of the present suffixes, it simply refuses to be put into a preterite.

  7. Hello, my name is Josef Jarosch and if you are interested in my thinking and especially in my publications you should try www.josef-j-jarosch.de.
    I read your thoughts about the thematic first person singular ending with interest. For me, -h2 is not an ending, but the hic-et-nunc-marker. There are three of them, and in thematic stems, the distibrution is 1sg h2, 2sg i, 3sg i, 1pl s, 2pl h2 (again!) and 3pl i. So probably the whole cluster is -o-m-h2, usually turning into -o-h2, but in one-syllable stems into -o-m (lat. sum I am and not so^). If you drop the hic-et-nunc-marker (in the imperfect and injunctive), the old and original ending -m comes back. This also easily explains "the attested avoidance of i in 1sg": i here = h2.
    Your observation that in IE - as in any language - some verbs are "durative/present" (duration of time: potentially infinite) by default and others are "punctive/aorist" (duration of time: zero) by default is completely correct, as well as an original assignment of stative endings : durative and factive endings : punctive. But some verbs, for no recognizable reason, behave "the other way round"; I call them escapists. "stative endings : punctive" is responsible for transitive deponents and Hitt. hi-verbs; "factive endings : durative" is responsible for the commonest IE verb Engl. is (Lat. est; other fine examples are Lat. fit becomes (infinitive fieri still regular "stative ending : durative") or Lat. it he goes). So in this sense it is correct when you say verbs could originally switch back and forth between the "two conjugations", usually as a rather rare exception, but in Hitt. the difference is practically abandoned. Do you want to know more about, say, aspect or the origin of thematic verbs?
    For terminology:
    stative ending : durative = STATIVE
    (Engl. "he may")
    stative ending : punctive = FACTOSTATIVE (no Engl. example)
    factive ending : punctive = FACTIVE
    (all Engl. verbs with 3sg -s)
    factive ending : durative = STATOFACTIVE (Engl. "he is")

  8. J. Jarosch,

    A solution that explains the obscure with more obscurity is a waste of time. In this case, you
    imply a source from three "hic-et-nunc particles" (presumably *h₂o-, *so- and *i-) by pure whim and treat it like the panacea we're all looking for without bothering to explain in acceptable linguistic terms how the very distribution of these endings would have evolved into PIE in the first place! Your proposed system is not grammatically and semantically coherent.

    Since the verbal 1pp thematic *-mes is mirrored by the pronominal form *ns 'us', there is little doubt what the underlying 1pp pronoun stem of the two morphemes are (ie. *mes). There is also no doubt what *-s here is: the syncopated form of the more common animate nominative plural marker *-es. "We" afterall is formed out of a pluralization of "I" in many languages and serves as a much more natural solution than what you suggest for *-mes.

    The existence of syncopated plural *-s is supported further by the same morpheme fossilized in the animate accusative plural -ns (< *-m-s).

    Your theory can never be correct when ignoring such basic facts and less obscure solutions as these. Occam's Razor must be respected. The particle *so, undeclined for case and gender in earliest PIE and a late interloper of the distal *to-paradigm, hardly explains the 1pp let alone anything else here. As well h₂o-, *so- and *i- don't constitute a grammatical structure in PIE so it seems pointless to ponder why it is that you've assembled them arbitrarily here for this task.

  9. To add to my previous comment...

    This blog entry was made in 2007, but I've made more recent contributions which may clarify the diachronic developments I envision for Pre-IE back a few thousand years.

    Read Paleoglot: The trouble with the PIE 1st & 2nd person plural endings (3).

  10. you imply a source from three "hic-et-nunc particles" (presumably *h₂o-, *so- and *i-) by pure whim and treat it like the panacea we're all looking for without bothering to explain in acceptable linguistic terms how the very distribution of these endings would have evolved into PIE in the first place! Your proposed system is not grammatically and semantically coherent.

    I think, thought I agree it's a bit difficult to distill from the text, Jarosch reconstructs a *h2 particle as a hic-et-nunc particle. Implying that some of the *-i's found go back to a vocalised Laryngeal? Maybe the 3rd sg *-i he's talking about is the one of the passive aorists in sanskrit like avāci. It's all a bit vague, I'd like for him to elaborate.

    But at least 1 hic-et-nunc particle is a whole lot less crazy than 3 of them ;-)

  11. Jarosch clearly stated above: "For me, -h2 is not an ending, but the hic-et-nunc-marker. There are three of them, and in thematic stems, the distibrution is 1sg h2, 2sg i, 3sg i, 1pl s, 2pl h2 (again!) and 3pl i."

    If he is stating that three different particles were employed to indicate a single mood, his solution is already a failure.

    For starters, these particles share NOTHING in common other than that they're demonstrative. As I said *so is ultimately undeclinable and thus very different from declinable *i-. Whereas *h₂o 'next to, by' was technically used as an adverb (and thus undeclinable), again different from the usages of undeclinable *so and declinable *i- which are both completely non-adverbial. Already his account is wildly incoherent. Ergo dismissable. I prefer to be impatient with incoherence since it saves time.

    I fear that Jarosch may not even understand the fundamental semantic purpose of these 'hic-et-nunc endings'. Since the primary marker *-i so clearly marks mood and is semantically *independent* of person involved, there's simply no sense in employing three different hic-et-nunc particles based on person! That is, there's nothing meaningfully different between an indicative/declarative mood of the 1st person singular as opposed to the same mood in the 3rd person plural. A mood is a mood.

    So the mood must be one and the same despite person. Thus in ALL persons a SINGLE marker *-i should have once been applied to mark presentive/affirmative/realis actions, not three. The fact that some persons omit this ending suggests to me that once the individual usage of the hic-et-nunc morpheme came to be obscured and confused with the personal endings to which they were attached, analogy played a part in reshaping and obscuring the formerly regular pattern. Hence the reason for my later, à-propos article.

  12. Just realized a problem with another of Jarosch's statements: "So probably the whole cluster is -o-m-h2, usually turning into -o-h2, but in one-syllable stems into -o-m (lat. sum I am and not so^)." (bold emphasis my own)

    If it were true that 1ps thematic **-omh2 in Pre-IE were simplified to *-oh2, Sihler's mention of a widely known PIE rule on page 98 of New comparative grammar of Greek and Latin (1995) is directly contradictory: "After a resonant or *s, a word-final laryngeal dropped, with compensatory lengthening of the preceding vowel."

    There is quite simply no such lengthening observed in the thematic vowel of the 1ps thematic ending, nor is there any reason that the resonant rather than the word-final laryngeal should be dropped in direct violation of this widespread observation. Oopsy daisy! Lol.

  13. In PIE, there is a set of endings called secondary endings. These secondary endings are the original ones. They are used in the aorist, injunctive and imperfect. They are 1sg -m, 2sg -s, 3sg -t, 1pl me, 2pl. te, 3pl ent. There is no mood or demonstrativity in them. There is no trace of a "plural s" nor is there any hint of a connection with "ns" 'we'. Similarities with pronouns seem to be sheer coincidence. (When I read Szemerenyi who said 2sg must have been "t" because of "tu" 'you' I gave up taking things seriously. An s is a sound, it is a good sound, it is a sound sound.)Later "presents" are in principle, even though very common, also secondary (so why actually talk about them if we have the primary/original forms anyway?)They have an additive called hic-et-nunc-marker (in my system: "infinity enabler"). You can find out the form of the additive if you compare the original secondary endings and the processed/altered primary endings. Doubtlessly, the additive is 2sg i, 3sg i, 1pl s, 2pl h2, 3pl i. 1sg is a problem - that's why we are talking here. In cases concerning endings, it is not only "sound rules" (Sihler), but sometimes "continuants", and omh2 -> oh2 (Lat. sum would obey the sound rule) might be such a case. The Sanskrit passive aorist is not only secondary, but tertiary: secondary "strong" alternant, secondary (and unique) -o-ablaut in the middle, only 3sg, no ending at all (i is from verbal i-stems). We should try to think more "originally".

  14. "They are used in the aorist, injunctive and imperfect."

    Not exactly. In PIE proper, there was as yet no formalized "aorist" aspect distinct from an "imperfect/durative" one. They both employed the same endings, save that actions with aorist nuances coincidentally avoided *-i. Read Jasanoff, Hittite and the Indo-European Verb (2005). I consider PIE a "tenseless" language like Mandarin.

    "They are 1sg -m, 2sg -s, 3sg -t, 1pl me, 2pl. te, 3pl ent. [...] There is no trace of a 'plural s' [...]"

    Surely you mean *-me-, not just *-me. The second hyphen properly allows for dialectal variance vis-à-vis terminating plural ending (ie. *-s or *-n).

    Via the wave model of language change, (Pre-)PIE dialects were open to either retain older plural personal endings in *-n, or to innovate with a new plural in *-(e)s taken from nominal morphology.

    That being said, there is no proof that *-mes didn't form long before the dissolution of PIE since there was naturally no such thing as a "PIE singularity", so to speak. That's a taxonomic myth of yore.

    "Similarities with pronouns seem to be sheer coincidence."

    Proposterous. Personal endings in any language can hardly be formed from much else than agglutinated pronouns in an earlier stage. So Szemerenyi's instinct that the 2ps ending is somehow derived from its pronominal counterpart is straight-forward common sense, regardless of personal mockery.

    Furthermore, see my Pre-IE Sibilantization rule (*-t# > *-s) which relates both the sigmatic plural and sigmatic 2ps in PIE to earlier homophonous forms in *-it whose word-final *t are still carried on in Uralic and Eskimo-Aleut. I wonder too if the lenition seen in the Altaic plural *-r₂ is a very early isogloss shared with Indo-Aegean (Pre-IE).

    Sibilantization is dependent on later changes, QAR and Syncope, since the proper identification of such sibilantized word-final stops is also predicted by accent position. An oxytone accent predicts a lost word-final vowel before the event of Syncope and coincidently the plural and 2ps endings are not accented this way.

    "Doubtlessly, the additive is 2sg i, 3sg i, 1pl s, 2pl h2, 3pl i. 1sg is a problem [...]"

    Indeed, but this isn't my burden to bear.

  15. I should expand further on this particular assertion expressed above, at least for the benefit of other readers who appreciate facts:

    Jarosch: "There is no trace of a 'plural s' nor is there any hint of a connection with "ns" 'we'."

    Yet, beyond the fact that Sanskrit -mas and Latin -mus (both from none other than 1pp active *-mes) so obviously retain a plural ending in *-(e)s, there's the fact that Winfred Lehmann too analyses the accusative plural as accusative singular *-m plus plural *-s, according afterall to the simplest of common sense, on page 150 of Theoretical Based of Indo-European Linguistics. I merely add detail to this by explaining that MIE acc.pl. *-am-as regularly shortens to *-ms via Syncope while the retained vowel in nom.pl *-es is irregularly preserved by a quite understandable avoidance of homophony with singular nominative nouns already terminating in *-s (from agglutinated MIE *-sa, cf. PIE deictic *so).

    Considering that even Hittite and Luwian anzas secure reconstructed zerograde 1pp *n̥s at the PIE level (alongside reanalysed fullgrade *nos in other branches), together with coexisting 1pp verbal -mes, the evidence for a vestigial shortened plural marker *-s, and the support of other IEists in these connections show, I dare say that there can be no cogent opposition. Surely if acc.pl. *-mas (Syncope) > *-ms > *-ns, ergo surely too MIE unstressed 1pp pronoun *mas > *m̥s (Syncope) > *n̥s 'us'. Consistency is important.

    I could go on and on, because there are so many interwoven facts we must consider, but these above facts in themselves suffice to show why your analysis is not just ad hoc and unexplanatory but false. There is only one so-called "hic-et-nunc marker" in the conjugational system, *-i, whose uncanny grammatical and etymological parallel I've already found in Mandarin progressive zài (在) (< 'to be here/there').

  16. Your comment on nominal endings is OK. No objections.
    Together with the dual and the middle, we have 18 verbal endings. The majority of them is either unknown, or uncertain, or hypothetical, or incomplete. In the middle, the original secondary endings are 1sg h2e, 2sg th2e, 3sg e,3pl r. The additives that make them processed primary endings are i, i, i, and s (again!). Some of these endings might resemble personal pronouns. But the meaning of the middle is intransitive, reflexive, dynamic, reciprocal, passive, gerundivic (and in some cases - see the infectostatives -zero) etc. So how can one possibly conceive the idea of an intransitive pronoun?
    Our starting point was the question why thematic 1sgs do not have the ending m. The question cannot be answered, but it can be minimalized. My suggestion (a continuant of -o-m-h2 with h2 here being the additive hic-et-nunc marker) answers three questions: 1. Why does - mistakenly - the thematic 1sg not have a hic-et-nunc marker? 2. Why does Lat. sum 'I am' have an m? 3. Why do imperfect and injunctive have an m? To me, the question seems to be asked the wrong way. The question needs to be: Why is the distribution of the additive athematic i - i - i - s - h2 -i, but thematic h2 - i - i - s - h2 - i? This question cannot be answered, as little as questions why certain roots have exactly this ("their") form, consonants, vowels etc.
    In the course of my studies, I found out that PIE can be divided in two stages, PIE and pre-PIE (PPIE). PPIE did not have reduplicative variants (see my website), it did not have a perfect (which I nowadays call, with a double name, because of its endings Nactostative-Nactofactive), and it did not have thematic stems. The hypothesis makes it a lot easier to deal with linguistic problems, because a good many of them simply do not come into existence. In linguistics, PIE now has the position of e.g. French: you do not need it, because you have Latin. Hittite has the reduplicative variants, but it has no perfect and no thematic stems, so it left the community before the perfect was formed. This was formed in inner-PIE (or PIE proper)by using the alternant inventory of the infective again and (only partly) equipping it with stative endings. From the alternant inventory of the infective the perfect carried on three reduplicative variants of the infective, e, resonant and zero. The original verbal system in PPIE had the verbal categories confective ("aorist"), infective (presents from aorist roots) and stative (e-grade throughout the paradigm, with the zero-grades having moved into confective and infective to constitute the middle there). I have told you about the escapists, i.e. verbs that take, for no detectable reasons, the opposite ending. I call them confectostatives, infectostatives and statofactives. So now you know why my theory for the Hitt. hi-verbs (#8) is called infectostative theory and why I talk of a "six-ashlar-theory", with the ashlars being confective, confectostative, infective, infectostative, stative and statofactive. I could tell you a lot more. Maybe you send me your publications and I send you mine?

  17. J.Jarosch: "Your comment on nominal endings is OK. No objections."


    "Together with the dual and the middle, we have 18 verbal endings."

    Dual number cannot and must not be reconstructed for PIE verbs:

    A) It's not attested in Anatolian at all.
    B) Unlike the verbal dual, the mi/hi-dichotomy is most definitely of PIE origin.
    C) Given B, the mixture of 2ps hi-ending *-th₂e with 3ps mi-ending *-t(i) in these dual endings conclusively betrays their late origin.

    Without active duals, there can naturally be no mediopassive duals, and thus your version of PIE grammar cannot have ever existed.

    "Our starting point was the question why thematic 1sgs do not have the ending m. [...] My suggestion (a continuant of -o-m-h2 with h2 here being the additive hic-et-nunc marker) answers three questions: 1. Why does - mistakenly - the thematic 1sg not have a hic-et-nunc marker? 2. Why does Lat. sum 'I am' have an m? 3. Why do imperfect and injunctive have an m?"

    What frustrates me the most about your question is that it ignores how PIE is currently being reconstructed. In PIE proper, there simply is no formal distinction between "injunctive", "aorist" and "preterite (secondary)" - they are one and the same tenseless form.

    So you should instead be asking something more informed: Why is *-m absent in the *Post-IE* thematic primary and present in the thematic secondary? Unless one can frame the question correctly, I doubt one can find a correct solution.

    Since Jay Jasanoff has already shown that thematics postdate the dissolution of PIE and largely derive from the subjunctive, the absence of *-m is in fact related to the original *subjunctive* 1ps, not the genuine primary 1ps ending *-mi once exemplified by both *h₁és-mi 'I am' and Narten present *bhēr-mi 'I carry' (displaced in Post-IE by former subjunctive *bhér-oh₂).

    Do read Jasanoff. Until you comprehend that thematic roots just don't exist in PIE in the way they do later on, I'm afraid your views are too antiquated for me to accept.

    "To me, the question seems to be asked the wrong way. The question needs to be: Why is the distribution of the additive athematic i - i - i - s - h2 -i, but thematic h2 - i - i - s - h2 - i?"

    It would be wonderful if you could inform yourself of the fact that the POST-IE thematic comes from the PIE subjunctive rather than asking irrelevant questions.

  18. Just a quick question about the Dual. Surely it would be a stretch to try and reconstruct it for Pre-Hittite split PIE, but past that stage it becomes quite reconstructible doesn't it?

    Skt -vas, Av. -vahi OCS -vě Lith. -va for the 1du do point to *-ue(s)
    Just like sanskrit tom and gr -ton give 2du secondary *-tom

    Reconstructing the forms remains difficult, but I would say there's a bit too much corresponding data to consider it an even later innovation than post-Hittite PIE.

    I'm having some trouble to figure out how the dual was formed though. You'd expect it was taken from the pronouns, or a combination of verbal endings + dual noun endings.

    I'm curious what you make of it, and whether you agree with Beekes' reconstructions.

    In case you don't have his Introduction around the endings he reconstructs are:
    1du *-ues
    2du *-tHe/os
    3du *-tes
    1du *-ue
    2du *-tom
    3du *-teH2m

  19. "Surely it would be a stretch to try and reconstruct it for Pre-Hittite split PIE, but past that stage it becomes quite reconstructible doesn't it?

    Yes, we're in agreement on that but apparently not in agreement of what "Indo-European" itself means. To me, any grammatical model for Indo-European that completely ignores Anatolian and ignores "undesirably enigmatic" elements of Tocharian grammar does NOT represent Indo-European - pure and simple. The model we're looking for must explain both the origin of the familiar durative-aorist-perfective system *AND* the origin of Anatolian mi- and hi- classes at the same time. The traditional model, as explained in Encylopedia Britannica in great detail for example, fails to do that.

    So ironically, that "IE verb system" is not "IE" at all.

    "I'm curious what you make of it, and whether you agree with Beekes' reconstructions."

    These reconstructions are like one would find elsewhere and there's nothing a priori I would object to here.

    However, as I previously said, the use of singular "durative" 3ps *-t(i) as the basis for 3pd is a clear hint that the system can only have arose once the mi/hi-dichotomy started eroding. The 3ps ending *-t(i) is ancient, related to the deictic stem *to-, and was applied originally only to the *mi-conjugation, perhaps to supply a form for the 3ps *mi-conjugation that was finally distinct from the endingless imperative. The *h₂e-conjugation had *-e in the 3ps, yet nonetheless -t- is present even in the 3pd perfect, a strong indication that the entire dual was a late innovation, occurring only after Anatolian (which is, mind you, without any of these dual forms at all) left the scene.

    Also, it must be remembered the following grammatical hierarchy:

    singular > plural > dual

    This means that a dual can only arise from a system of singular and plural, and not surprisingly this new Post-IE dual incorporates various elements from the pre-existing singular and plural paradigms shared by all IE dialects in a rather hodgepodge fashion. Furthermore, no single dual set of endings has been preserved because all branches have pursued their own innovations independently. This also suggests that the dual system was only developing once the PIE community had already fractured into several distinct dialects.

  20. Thanks for the clarification.

    So ironically, that "IE verb system" is not "IE" at all.

    This is what I keep on saying too. It's almost frustrating to see how hard people attempt to reconstruct stuff in non-Graeco-Aryan languages by using a clearly graeco-aryan model of the Indo-European verbal system.

    But let me just make clear that I correctly understood the implication. You feel that any Indo-European after the Anatolian split was already to divergent in dialects to be able to speak of a non-anatolian Indo-european as a single entity?

    It's funny, I've been very convinced of strongly divergent dialectal regions in this period but it never occurred to me that they may have been there, or developed immediately after the Anatolian shift. But it's an attractive view.

  21. Phoenix: "You feel that any Indo-European after the Anatolian split was already [too] divergent in dialects to be able to speak of a non-anatolian Indo-european as a single entity?"

    Not exactly, but your question can only be answered with subjective opinion. In my opinion, it's next to impossible for "IE 2" (the non-Anatolian IE dialect area) to not have been fractured already into its own dialects to some unquantifiable extent. Attempting to logically define 'single entity' is futile.

    People often forget that a language area, even after fracturing into mutually incomprehensible (or almost incomprehensible) dialects, can still disseminate new features developed *after* the dissolution of the language. Such features then would appear to a thoughtless linguist as though they occurred before the dialects had developed (until one started looking more closely at the details and focusing on the proper chronology of each change that appears to have taken place).

    In this case, 1) if the oldest conjugational system in PIE uses a suppletive system of *mi- and *h₂e-sets which only after PIE had sired the triaspectual system of durative-aorist-perfective, 2) if the characteristic 3p ending *-t is strictly a *mi-set ending within the singular-plural subset, and 3) if the singular-plural system can only logically precede a dual add-on, then the awkward presence of *t in the perfect (< *h₂e-conjugation where *-e is instead to be expected) denies the presence of a dual conjugation in PIE proper. The second person dual is likewise shown to be a post-IE innovation in the same fashion. (And if this reasoning is sound, what a pity that the dual 1p in *-we- then can no longer be used to justify separate inclusive & exclusive pronouns in the 1pp, tsk tsk. Lol.)

  22. Hi you guys, thank you for discussion. I appreciate it.
    Please proceed from the fact that I know Jasanoff by heart. I really admire him for his gigantic knowledge of languages. But whenever he draws conclusions in direction of PIE, he gets it terribly wrong. The reader's or observer's task is to put his conclusions in an order of "falseness": which is the falsest, which are the least false? The falsest is certainly to call t^ezzi an aorist. Aorists never have an i. It sure can be called a back-formation, but from an imperfect (reduplicative variant zero) and not from an aorist. His second falsest is the derivation of thematic stems from subjunctives. In Germanic, as well as in a lot of other languages, an -´e-grade proterodynamic-thematic subjunctive was never formed; this is communis opinio seen correctly by generations of serious linguists. It is not possible that a form does not exist and never has existed but its successor exists. A family who do not have children cannot have grandchildren. And again: I asked you whether you can think of an intransivite pronoun that could have produced an intransitive verbal ending. (Endings, by the way, are not grammar. They are 1% of grammar, if that much.) Now there are thematic nouns. Could you ever think of a nominal subjunctive? From p. 221 - 223, Jasanoff-2003 talks about problems he should have solved before writing his book. How incomplete - or wrong? - he is you can see from his sentence p. 223 "But the stems of tud´ati-presents and thematic aorists are formally indistinguishable, and a theory that accounts for the one ought to account for the other." To be correct, he should have added: ... and at the same time also for nominal stems. I read Jasanoff, but he does not read Jarosch; so automatically and normally, I know a lot more.
    You say I ignore how PIE is currently being reconstructed. Currently, it is not reconstructed state-of-the-art (as other sciences steadily do in their fields), but state-of-1850. (Saussure, who lived and worked around 1900, had it state-of-1900, but too much of him has completely been forgotten again). This is simply due to the fact the the newest, most recent, and most developed literature is consequently ignored and suppressed. You say that inj., aor. and pret. are one and the same tenseless form. But they are definitely not, which is crucially important here, one and the same aspectless form. Aorist and imperfect are totally antagonistic in meaning, as antagonistic as can ever be. Aorist denotes single action (duration of time: zero) and imperfect denotes repeated action (duration of time: potentially infinite). The aorist OI ´adar means 'he pierced once', the imperfect OI ´adar means 'he pierced repeatedly'. Synchronically, of course, they have the same form (with the injunctive here given a tense). Diachronically, however, their forms have the most antagonistic and different form you can think of. The aorist ´adar has, diachronically, the five sounds ´e d´er -t. But the imperfect ´adar has, diachronically, a reduplicative variant zero. Reduplicatives (and of course, their variants as well) denote the repetition of an action. Please read my website and what I say about Meid's question. The question is how singularity develops into repetition. There is only one answer, and no alternative, this one or none: Singularity develops into repetition by saying something singular repeatedly. It is one of the very rare cases in linguistics that a form is identical with a meaning.

  23. Thematic stems - aorist, present, noun - develop as follows: say there is a "false-friend"-paradigm, root bher 'to carry', reduplicative variant zero, of the form bh´er_bh r-´. bh´er is the strong alternant or "mother stem" x=0/y=1, bh r-´ is the weak alternant or "ovarial stem" x=0/y=0(the center of the system). Taking over the -e- from endings verbal -´ent nominal -´es, the "ovarial stem" forms out an "embryonal stem" bh r-´e- x=0/y=-1. As such, this "embryonal stem" is not viable, but sometimes there is an incubator that helps it to survive: the -io-suffix that forms the causative, -h2- that forms nominal collectives or feminines, doubling the -´e- in Lat. subjunctives a-conjugation, or thematic vowels -e- under the protection of the entire thematic system (aha! here they are, the -e-colored thematic vowels - they are embryonal stems!). Under non-accent, -e- is colored to -o- and generalized; this is the "filial stem" bh r-´o-, x=0/y=-2, fully developed and viable. And on it goes: x=1/y=0 would be bh =n= r-´ (n-infigation here not realized); x=2/y=0 would be bh´er-, x=2/y=1 would be "strong alternant" bh´er, x=2/y=2 would be "strong alternant" bh^e´r (the example you often mention), x=3/y=0 would be bhor-´, x=3/y=2 would be bh^o´r (= Lat. fu^r 'thief'). x=4 - x=7 would be the same with -i-suffix, x=8 - x=11 would be the same with u-suffix, and so on (the system is fully worked out). From this x-axis, thematic languages would always tend to go into y=-2; Hittite always tends to go into y=2, even with very recent suffixes (iyannai).
    What ails me is that you say "The model we are looking for must explain both the origin of the familiar durative-aorist-perfect system and the origin of Anatolian mi- and hi-classes at the same time." I told you about the six-ashlar-theory; this is exactly what the the six-ashlar-theory explains. So again: The six ashlars are:
    I. confective (traditionally aorist)
    II. confectostative ("middle d´eclass´e"; transitive aorist deponents, Gk. ap´edoto 'he sold')
    III. infective (presents from aorist roots)
    IV. infectostative ("middle d´eclass´e"; transitive present deponents, Hitt. hi-verbs, originally infectostativized only in the "strong alternant")
    V. stative (non-ablauting for reasons of the development of "proto-middle")
    VI. statofactive ("active d´eclass´e"; Lat. est it fit etc., Gk. apedoth^e 'was sold')
    So I hope to hear from you soon again!

  24. Jarosch,

    "I read Jasanoff, but he does not read Jarosch; so automatically and normally, I know a lot more."

    Your conceit is limitless. These silly, unreferenced statements of yours show either that you're insistent on your ignorance or that you're a warped comic with an imaginary audience. Terms like "ovarial stem", "infectostative" and "ashlar" are non-descriptive gobbleygook terms of your own lonely making that no credible linguist uses if one wishes to be understood by anyone other than oneself.

    Either way, it's very isolationist and not tolerated for long on my blog. Let's go through all the facts that you've completely mangled so that readers can know exactly why my gracious patience with you is wearing thin.

    "[Jasanoff's] second falsest [claim] is the derivation of thematic stems from subjunctives."

    Be serious. Read Fortson, Indo-European Language and Culture (2009), p.106, which mentions the attested usage of subjunctives for future nuance in both Greek and Indo-Iranian.

    "The falsest is certainly to call t^ezzi an aorist."

    You've misread. Read Mallory/Adams, Encyclopedia of Indo-European Culture (1997), p.472. The Hittite **present**, tēzzi 'he/she says', reflects an innovative **post-IE** form *dʰeh₁-ti, from former aorist *dʰeh₁-t 'he/she puts' attested as such in Armenian & Indo-Iranian. Note the concordant change in meaning of the Hittite verb. If your thought here is to make the PIE verb a root "durative" instead, that may possibly be argued, but please don't misquote or misrepresent authors.

    "Now there are thematic nouns. Could you ever think of a nominal subjunctive?"

    Your two violations in logic here are:

    Non sequitur: Thematic vowels in nouns need not a priori come from the same source as that in thematic verbs.

    Strawman: No one has argued for a "subjunctive noun" in the first place, nor does anyone need to.

    Shape up or ship out. The semantic evolution of a subjunctive to a future tense has been explained to you and published a thousand times. It's up to you to learn these facts or take up another hobby for which you're more suited.

    "The six ashlars are:I. confective [...] II. confectostative [...] III. infective [...] IV. infectostative [...] V. stative [...] VI. statofactive"

    Sigh. No sensible person can be interested in your ashlar cult when your grasp of the most basic facts is so evidently meager. You seem to have a greater desire here to oppose everyone on everything than to reach outside yourself and learn something new.