the pie verbal endings (revised) 2015

Upload: uvolenvrach

Post on 09-Mar-2016

9 views

Category:

Documents


0 download

DESCRIPTION

Carrasquer Vidal

TRANSCRIPT

  • The Proto-Indo-European verbal endings (revised) Miguel Carrasquer Vidal, October 2015 Proto-Indo-European (PIE) inherited two conjugational systems, usually referred to as the mi-conjugation and the hi-conjugation, after the endings of the respective first person present forms in Hittite.

    The mi-conjugation endings

    First person singular The endings can be reconstructed straightforwardly as:

    *-m *-mi The present tense ending is regularly derived from the past tense ending by adding the particle *-i, which, as we will see, is the most common strategy for deriving the present tense forms in the mi-conjugation1. It should be obvious that the ending *-m originates in the agglutination of a first person singular pronominal element. The nominative of this personal pronoun has the suppletive form *h1eg-in PIE, but all the oblique forms are based on a root *m-. Based on the nominative of the second person pronoun (*tu), we can reconstruct the lost nominative form as pre-PIE **mu, and since the ending *-m(i) denotes the subject of the verb, **mu is the logical candidate for the agglutinated pronoun underlying the first person singular ending. Reconstructing the ending as *-mw offers the benefit that it can help explain a few irregular forms, most notably the present thematic first person singular ending, which is to be reconstructed as *-, instead of expected -o-mi (the past tense does have the expected form *-o-m). There is also the Luwian ending -wi (< *-mwi) instead of *-mi. After a consonant (verbal roots in PIE were required to end in a consonant) and at word end, the labialization was regularly lost, but between vowels (as in the 1sg thematic present) the labialization should have been preserved. Presumably, the development was **--mw-i > **-wi, and then, as in the locative singular of the u-stems, *-owi > *-owu > *-u. This ending is attested in Tocharian B -eu, -au. Elsewhere, the ending was simplified to *-. Alternative explanations of the ending *- assume the presence of a laryngeal in the ending, presumably as a cross-over from the hi-conjugation endings. However, the thematic vowel should have developed to *e before the voiceless laryngeals *h1 and *h2, yielding - or -, which leaves *-oh3 > *- as the only viable laryngeal explanation. However, an ending *-h3, otherwise utterly unattested, raises more questions than it solves.

    Second person singular The endings are usually reconstructed as:

    *-s *-si

    1 The traditional terms for the two sets of endings are primary (present tense) and secondary (past tense), which has the advantage of de-emphasizing the connection with tense, which is not always present, but has the disadvantage that the so-called primary endings are in fact mostly derived from the secondary ones.

  • Following up on the model of the first person singular, we can hypothesize that this ending is to be derived from an agglutinated pronominal element *-tu. The regular development in the Auslaut would have yielded *-sw, which was subsequently delabialized everywhere except between vowels, as in the thematic present ending *-e-sw-i. This is confirmed by Slavic 2sg thematic *-e, with -- instead of -s-. Tocharian A and B both show an ending -t in the 2sg. If we assume for a moment, for reasons that will become clear when we discuss the plural forms, that pre-PIE had different sets of endings for transitive and intransitive verbs, the Tocharian 2sg endings might be derived from the agglutination of absolutive *ti instead of ergative *tu. However, the Tocharian 3sg forms show that an ending *-t should have been lost in Tocharian, as would have *-s(w), for that matter. Therefore, the Tocharian ending -t is best explained as an inner-Tocharian development, where the regularly expected 2sg ending *- was extended with the pronoun *tu (> Toch. *t).

    Third person singular Judging by 3pl *-n-t (which must be derived from plural **-n + *-t), the ending was originally *-, to which subsequently an element *t was added, presumably the demonstrative pronoun stem *ta, functioning as a third person marker. It is possible that this marker was originally confined to the intransitive paradigms. The PIE endings are:

    *-t *-ti As we mentioned above, the ending *-t was lost in Tocharian A (which preserves the PIE secondary endings). The ToA ending *-m is most plausibly explained as 3sg *-, extended in the present tense with the particle *nu (> *n) now. Tocharian B -s regularly continues the primary ending *-ti.

    First person plural The Hittite endings are: -wn -wni These can be derived from an agglutinated pronominal element: **-mu-n The presence of the nasal labial *m(w) explains why final *-n did not develop into *-r in this form. Outside of Anatolian, East-Greek *-mn must go back to the same form (likely influenced by the forms we will be discussing in a minute), which means that we are surely not dealing with an inner-Anatolian innovation. In my theory of the pre-PIE personal pronouns, the form *mun is an absolutive, corresponding to ergative *mutu. Agglutination of the latter form to the verbal root would have yielded: **-mutu > *-mwsw > *-msw, which is faithfully reflected by Armenian -mk` (thematic -uk` < *-omesw), with regular preservation of final *-sw as -kh in Armenian. The labialized final may also have been a factor in the formation of the by-form *-mos besides *-mes in e.g. Latin -mus, Slavic -m.

  • Regular addition of hic et nunc *-i would have yielded the following primary and secondary endings: *-ms(w) *-ms(w)i, and a present tense form *-msi is indeed attested in e.g. Vedic -masi, Toch. A -ms. In the system leading to Greek, however, a distinction between primary and secondary forms failed to develop at all in the 1st (and 2nd) person plural: we have -men in East Greek and -mes in West Greek throughout. Another strategy for differentiating the primary and secondary forms arose when the 1pl ending of the hi-conjugation *-me was adopted in the past tense of the mi-conjugation. The ending *-ms was then reinterpreted as a present tense form, yielding: *-m *-ms This is the system attested in Vedic (-ma ~ -mas), and indirectly in Tocharian (B -m(o) ~ A -ms).

    Second person plural As in the first person plural, we have an opposition between Anatolian on the one hand, and most of the other languages on the other. Hittite has: -tn -tni We can again set up an original paradigm with transitive and intransitive forms: *-tun *-tutu, which should have led to: *-tr *-tni *-t *-t Notice that the formation of the primary form with *-i must have predated the sound law *-n# > *-r#, and that the development *-twtw > *-t is idiosyncratic, but typical of pre-PIE sequences *tw..*tw (cf. the 2pl personal pronoun *tutu > *(j)sw, with loss of initial *tw, or the nom. sg. of the demonstrative pronoun *tutu > *s, without the expected final sibilant). To be sure, a form *-ts(w) is attested in Latin -tis and Armenian -yk`, but it must certainly be secondary, in the light of Greek -t (without primary ~ secondary distinction), Vedic -th ~ -t (which uses a completely different method to distinguish primary and secondary endings), and Tocharian A -c < primary *-t. If the distinction *-t ~ *-ts(w) had been original, there is no reason for it to have been lost in Greek or Vedic. It therefore must have arisen analogically, on the model of the first person plural *-m ~ *-ms(w), in areas where that system had become established (such as in the dialects ancestral to Latin and Armenian). Evidence for the ending *-tr outside of Anatolian can be found in Tocharian B 2pl -cer, which can be derived from a variant *-tr, with analogical influence from the hi-conjugation 3pl ending *-r. In Hittite iself, expected *-ter was replaced by analogical -ten for obvious reasons (primary -tni, 1pl -wn).

  • In areas where the ending *-t displaced *-tr (or analogical *-tn), different strategies developed for distinguishing the primary and secondary endings. Addition of *-i was apparently not an option after the final vowel of *-t, or the *-i had been lost during the developments that led to *-twetw(i) developing into *-t in the first place. In Greek, as was the case with the 1pl, the forms remained indistinct. An opposition *-t ~ *-ts(w) developed, as we saw above, in e.g. Latin and Armenian. In Indo-Iranian, the primary ending *-th2 arose on the analogy of 2du primary *-th2s, as will be discussed below.

    Third person plural The third person plural offers less problems than the other plural forms. The reconstruction is

    straightforward:

    *-nt *-nti

    We can see that the addition of the 3rd person marker *-t predated the sound law *-n > *-r, as is

    only obvious after having established that the addition of primary *-i also predated that sound law.

    Note the total lack of reflexes of the expected ergative ending *-tu > *-sw, which became

    established as the nominal plural marker, and was completely eliminated from the verbal realm.

    First person dual The expected intransitive and transitive endings are: **-mu **-muku, which would have regularly yielded: *-wn *-wh2 The first form merges with the plural, the second form underlies Balto-Slavic *-v (Lith. -va), Gothic -u, -wa and Vedic -va . On the model of 1pl *-m ~ *-ms, an opposition was established between secondary *-wah2 and primary *-wh2s (Ved. -vas, Goth. -(o)s).

    Second person dual We can again set up: **-tu **-tuku, which would have regularly yielded: *-tr *-th2 The first form merges with the plural, the second form underlies Balto-Slavic *-t (Lith. -ta, OCS -ta).

  • On the model of 2pl *-t ~ *-ts, an opposition was established between secondary *-tah2 and primary *-th2s (Ved. -thas). This form was the model for the Indo-Iranian 2pl primary ending *-th2. The long vowel in the pre-Lithuanian first and second plural endings *-m and *-t can be explained on the analogy of the dual forms. In Proto-Balto-Slavic we must have had: primary secondary 1pl *-ms (*-ms) *-m (*-m) 2pl *-ts *-t 1du *-ws *-w 2du *-ts *-t In East-Baltic, this was reshaped to: primary secondary pl -ms -m -ts -t du -ws -w -ts -t, and subsequently, the primary endings were lost.

    In Slavic, the PBS forms would have led to:

    primary secondary pl -me, -m -me, -mo -te -te du *-v *-va *-t -ta The attested 1du form -v was reshaped from *-va after the 1du personal pronoun v (the same development explains yet another attested Slavic 1pl form: -my (Polish, Sorbian, Kashubian), besides -m (OCS, Russian, Bielorussian), -me (Czech, Slovak, Slovene, Bulgarian) and -mo (Ukrainian, Serbo-Croatian). Note that in some, but not all, cases an alternative explanation for the endings -me and -mo is that the vowel reflects an irregular strong jer, i.e. that the forms simply go back to OCS -m.

    Third person dual The third person dual is problematic. Theoretically, and in the light of what we said above about the

    other dual endings, we would have expected:

    **- **-ku These would have resulted in:

    *-(j)n-t *-(j)h3-t

    Just as we saw in the 3pl, there is no trace of the ergative ending *-(j)h3, which only appears in the

    thematic nominal dual ending *--(i)ku > *-h3. The absolutive ending can be presumed to have

    merged with the 3pl (*-nt).

    The actually attested 3rd person dual endings are:

  • primary: *-ts (Slav. -te, Ved. -tas)

    *-tm (ToB -tem, Grk. -ton) secondary: *-t (Slav. -te ?) *-th2m (Ved. -tm, Grk. -tn)

    All of these forms are likely to be imports from the middle system, where, as will be discussed below, a 3rd person dual subject could carry the endings *-h2tom, *-h2tor,*-h2te, *-h2todh, *-h2toj, *-h2tah2m, *-h2tah2r or *-h2tah2, depending on the beneficiary of the action. It is likely that the forms *-h2tm > *-tm (3du x 1sg), *-h2t > *-t (3du x 3sg), and *-h2th2m > *-th2m (3du x 1du) were adopted to provide distinct 3du endings to the mi-conjugation. The form *-ts besides *-t is then a calque on the 2pl primary ~ secondary pair *-t ~ *-ts. Some middle forms are also found in the 2du, namely Ved. secondary -tam, and Greek primary and secondary -ton. Both are formally identical to primary 3du *-tom, but we would have expected 2du middle forms to show a dental aspirate (*-(h2)th2om: 2du x 1sg). Perhaps the simple dental is due to the analogical influence of secondary 2du *-th2. The Greek use of -tn as a primary 2du ending is then, well, secondary.

    Summary This concludes the review of the mi-conjugation endings. In summary, we can set up the following paradigm for the earliest reconstructible phase of pre-PIE: transitive intransitive 1. *-mu *-mi 2. *-tu [*-ti] 3. *- *-ta 1. *-mutu *-mun 2. *-tutu *-tun 3. [*-tu] *-n-ta 1. *-muku *-mu 2. *-tuku *-tu 3. [*-ku] *--ta There is no actual evidence for the forms in square brackets. Subsequent developments led to the following system, which is reconstructible for PIE (i.e. the common stage just before the break-up of the family into Anatolian and non-Anatolian branches): primary secondary 1. *-m(w)i *-m 2. *-s(w)i *-s 3. *-ti *-t 1. *-ms(w)i, *-wni *-ms(w), *-wn 2. *-t,*-tni *-t, *-tr 3. *-nti *-nt 1. *-wni *-wh2, *-wn 2. *-tni *-th2, *-tr 3. *-nti *-nt

  • The Hittite endings 1. -mi -un 2. -si -s 3. -zzi -t 1. -weni, -wani -wen, -wan 2. -tteni, -ttani -tten, -ttan 3. -anzi -ir The endings -wani, -ttani etc. reflect the 1 and 2pl endings when unstressed. The past tense 3pl ending -ir (< *-r) is originally a hi-conjugation ending.

    The Tocharian endings

    Tocharian B < secondary Tocharian A < primary 1. -u, -m *-mw -m *-mi 2. -t(o) (*-sw) -t (*-si) 3. -m (*-t) -s *-ti 1. -m(o) *-m -ms *-msi 2. -cer (*-tr) -c *-t 3. -m *-nt -ic *-nti After a vowel, *-mw gives *-w in Toch. B. The second and third person singular forms were reduced to zero, and supplemented with *tu and *nu, respectively. Second plural *-cer is from *-tr by way of *-tr, with the long vowel probably analogical after hi-conjugation 3pl *-r. Tocharian A reflects the primary endings, except perhaps for the 2sg ending -t, where we might have expected primary *-si to have given *-s, and to have fallen together with the 3sg ending. The ending was then probably replaced by secondary *-t < *-tu.

    The Vedic endings 1. -mi -m 2. -si -s 3. -ti -t 1. -masi, -mas -ma, -m 2. -tha(na) -ta(na) 3. -anti -ant 1. -vas -va, -v 2. -thas -tam 3. -tas -tm The -na that appears in the variant 2pl ending -t(h)ana originates in the imperative. It is most likely derived from the question tag *ne not. The first person plural and dual forms show variation in the length of the vowel: the length is only original in the dual form.

  • The Greek endings 1. -mi -n 2. -s, -i -s 3. -si, -ti - 1. -men (-mes) -men (-mes) 2. -te -te 3. -()si -(sa)n 2. -ton -ton 3. -ton -tn The 2nd and 3rd person singular endings, and the 3pl, have been complicated by the loss of intervocalic -s-, and the assibilation of -ti to -si (the 3pl past tense ending -san is from the s-aorist). A stem ending in a consonant is *es- to be: 1. *es-mi > eimi 2. *e(s)-si > essi (Hom.), ei 3. *es-ti > esti A stem ending in a vowel is for instance *ei- to go: 1. *ei-mi > eimi 2. *ei-si > ei 3. *ei-ti > eisi More often, the 2sg is extended with -s from the past tense, and sometimes the 2nd and 3rd person forms merge by repeated loss of -s-: 1. tithmi 2. *tithsi > tithei, titheis, tiths 3. *tithti > tithsi > tithei, titheis The present thematic endings in the singular are: 1. - 2. -eis 3. -ei where we might have expected 1. - 2. *-ei 3. *-esi The development in the 2sg of *-ei > -eis is unsurprising. However, in the 3sg, although the loss of -s- from assibilated -tj- is attested in the athematic forms, the total lack of a thematic form -esi suggests that we are dealing with a different kind of development:

  • I II III

    1. - -on - -on - -on

    2. -ei -es -eis -es -eis -es

    3. -esi -e -esi -e -ei -e

    In stage I, sound law eliminated -s- and created a new -s- from -assibilated -tj-. In stage II, -s was copied analogically from the past to the present 2sg form. In stage III, the proportional analogy -es : -eis :: -e : X is resolved as X = -ei.

    The Balto-Slavic endings Proto-Balto-Slavic

    athematic thematic present past present past 1. *-mi *-maj *-m *- *-m 2. *-si *-saj *-s *-eswi *-s 3. *-ti *-t *-t *-eti *-et *-t 1. *-mes *-me *-omes *-me 2. *-tes *-te *-etes *-te 3. *-enti *-ent *-ent *-onti *-ont *-ent 1. *-vas *-v *-ovas *-v 2. *-tas *-t *-etas *-t 3. *-tes *-te *-etes *-te

    The variant 1 and 2sg forms are from the middle paradigm (originally *-h2aj and *-th2aj). The third person present forms without *-i were originally used in modal function, presumably from injunctive and/or subjunctive forms.

    Lithuanian Slavic present / past present past athem. them. athem. them. athem. them. 1. -mi, -mie- -u, -uo- -m - - 2. -(s)i, -(s)ie- -i, -ie- -si -e,-ei - -e 3. -t(i) -a -t, -t, - -et, -et, -e - -e 1. -me, -m- -ame, -am- -m, -me, -mo -em, etc. -om, etc. 2. -te, -t- -ate, -at- -te -ete -te -ete 3. -t, -t, - -t, -t, - - - 1. -va, -v- -ava, -av- -v -ev -ov

    2. -ta, -t- -ata, -at- -ta -eta -ta -eta 3. -te -ete -te -ete

    Lithuanian 2sg - arose in the athematic paradigm as a reanalysis of 2sg es- you are. The thematic vowel has been regularized to -a- everywhere. Slavic has generalized the thematic vowel -e- in the present, although in the past the -o- appears at the expected positions.

  • The Latin endings 1. -m 2. -s 3. -d, -t 1. -mus 2. -tis 3. -unt

    Latin has given up the distinction between primary and secondary endings, although Old Latin still

    distinguished 3rd person *-t > -d from *-ti > -t, as does Sabellian in the 3pl (*-nt > -ns vs. *-nti > -nt).

  • The hi-conjugation endings

    Elsewhere, I have developed the theory that the hi-conjugation originally distinguished three

    paradigms: intransitive, transitive with singular object, and transitive with plural object. The plurality

    of the absolutive was marked by reduplication of the verbal root:

    intr tr/sg tr/pl sg. *bhar- *bhar- *bhi-bhar- pl. *bhi-bhar- *bhar- *bhi-bhar- Iterative (neo-imperfective) forms were derived from hi-conjugation verbs by lengthening the first (stressed) syllable of the verb: intr tr/sg tr/pl sg. *bhr- *bhr- *bhi-bhar- pl. *bhi-bhar- *bhr- *bhi-bhar- The six resulting paradigms eventually gave rise to the PIE verbal categories of:

    I. old perfective: reduplicated present; root aorist; reduplicated aorist II. neo-imperfective: perfect; mol-verbs; Narten-verbs

    Additionally, mi-conjugation verbs could form a neo-perfective by adding the hi-conjugation ending *-s to the verbal root (the sigmatic aorist). Despite the fact that most of these verbal categories, except for the perfect and, in Hittite, the mol-verbs, eventually lost their hi-conjugation endings, a number of phenomena still point to the hi-conjugation roots of the six verbal categories mentioned above:

    1. All of them, and no others2, have the 3pl ending -ur in Vedic. 2. The non-reduplicated forms (root-aorist and mol-verbs) originally had 5:1 root Ablaut

    (i.e. only the 3pl has end-stress) 3. Roots ending in a laryngeal show *-k- < *-HH- in the first person singular of the root

    aorist (Latin ici, fci; Greek , , , Toch. tk-).

    First person singular The ending was originally *h2, which was extended with *-e to give *-h2a. The question is: what was

    the function of *-e? I had until now assumed that it was added to all non-syllabic hi-conjugation

    endings (i.e. all except the 3pl), but I must now raise two objections against this. The first is the fact

    that -e was not added to non-syllabic 3sg *-s in the past tense. The other is that the Hittite past tense

    1sg ending -hhun (< *-h2m) also seems to lack the extension. The conclusion must be that the *-e, like

    mi-conjugation *-i, was originally a present tense marker:

    *-h2 *-h2a

    After the loss of the hi-conjugation past tense in the old perfective forms, the *-e became a free variant, and a new system of distinguishing past from present could arise, calqued on the mi-conjugation convention:

    2 Except for the s-aorist.

  • *-h2a *-h2ai The ending *-h2ai is not only attested in Hittite (-hhi, older -hhe), but also in the Latin perfect ending -i and the sole Slavic perfect form vd I know.

    Second person singular The second person is similar to the first person: *-th2 *-th2a *-th2ai

    For roots ending in a dental, the ending could be reinterpreted as *-sth2, which could then further spread to non-dental roots (e.g. Greek -stha, Toch. -st(a)).

    Third person singular The ending must at first have been zero, after which a split occurred between the preterit with added *-s, and the present without it: *- *-s *-e *-ei The origin of this *-s is either the old third person pronoun **su (which otherwise only survives in the oblique as the reflexive pronoun *swe, *sewe), or the nominative of the demonstrative pronoun *so, *to-. That the *-s is a third person marker is obvious (it also appears in the 3pl ending *-er-s). That it is also a preterit marker is highly likely, given its presence in the Hittite hi-conjugation preterit and its absence both from the Hittite hi-conjugation present and from the Extra-Anatolian perfect (which was in origin a neo-imperfective present tense). But whether all hi-conjugation preterit forms carried the *-s is hard to tell, given the fragmentary state in which the hi-conjugation is attested. One problem in particular is the reduplicated aorist, which, despite being a preterit form, has largely gone over to the thematic conjugation, suggesting a third person *-e, rather than *-s.

    First person plural If we compare the 1 and 2pl hi-conjugation endings with the middle endings, we can see that the former lack the element *-dh(w)- which is present in the middle. The most elegant way to resolve the discrepancy is to assume that *-dh(w) was lost in absolute final position: *-mdhw > *-m *-me The ending *-me was subsequently adopted by the mi-conjugation as a preterit ending, as we saw above.

    Second person plural The loss of *-dh(w) in final position led to a zero ending in this form (Ved. -a). *-dhw > *- *-e

  • For roots ending in a dental, the ending could be reinterpreted as *-s, which could then further spread to non-dental roots, as in Toch. *-s(), Hitt. -sten(i), the latter augmented with the mi-conjugation ending -ten(i).

    Third person plural *-an *-er-s *-er(-e) *-r *-er *-rs *-r In the preterit forms, we again have *-s. It should be obvious that, unlike what we saw in the mi-conjugation with the third person ending *-t, the addition of *-s was late, after the soundlaw *-n > *-r. In forms without reduplication, the ending *-r(s) was the only ending to carry the stress, and the root had zero grade. Originally reduplicated forms had stress on the reduplication syllable, and the 3pl ending was eventually reduced to *-r(s) there.

    Summary 1. *-h2 *-h2a 2. *-th2 (*-sth2) *-th2a (*-sth2a) 3. *-s *-e 1. *-m *-me 2. *- (*-s) *-e (*-se) 3. *-r, *-rs *-er, *-r(e) The loss of the hi-conjugation endings in the root aorist and allied forms in favour of mi-conjugation endings (*-m, *-s, *-t) may in part have been due to phonetic weakness or loss of final *-h2 after a consonant in the 1sg. In Hittite, the mi-conjugation ending -m was added to remedy the awkwardness of the laryngeal ending (-hhun). It is possible that the whole ending was prone to be replaced with -m, after which the rest of the paradigm followed suit. The root aorist has mi-conjugation endings even in Hittite. Only the third person plural remained strongly in place, and even spread to the whole of the mi-conjugation in the Hittite preterit.

    Hittite forms 1. -hhi -hhun 2. -tti -(s)ta 3. -i -s 1. -weni -wen 2. -tteni -(s)ten 3. -anzi -ir Old Hittite still has -hhe, -tte and -e (from *-h2ai, *-th2ai and *-ei) in the present singular. The plural forms have been replaced by the mi-conjugation endings. In the preterit, -hhun has added -m from the mi-conjugation. The 1 and 2pl forms have again been taken over from the mi-conjugation, The 3pl ending *-rs, on the other hand, has spread to the mi-conjugation preterit.

  • Tocharian forms 1. *-w 2. *-st 3. *-s 1. *-m(s) 2. *-s 3. *-r, *-r The first person has added *-m- (*-mh2a > -w). The second persons have -s- from the dental verbs. The third plural form is either an import from middle *-ro, or *-r augmented with *-ont (> -r) and *-nt (> -r).

    Vedic forms 1. -a 2. -tha 3. -a 1. -ma 2. -a 3. -ur The 2pl -a is original, while elsewhere the form has largely been replaced with *-te. The third plural ending -ur is from *-rs, generalized to all 3pl forms (besides the (plu-)perfect, also in the reduplicated imperfect, the root aorist, the optative past, and the s-aorist). Avestan mostly has *-r (once *-rs).

    Greek forms 1. -(h)a, -ka 2. -(h)as, -kas, -(s)tha 3. -(h)e, -ke 1. -(h)amen, -kamen 2. -(h)ate, -kate, -(s)te 3. -(h)si,-ksi < *-a-nti First person *-h2a has led to the k-perfect (roots ending in a laryngeal) and the h-perfect (roots ending in a stop). In a pure (albeit unreduplicated) perfect like oida I know, the second persons are oistha and oiste.

    Latin forms 1. -i 2. -is-ti 3. -it 1. -i-mus 2. -is-tis 3. -re, -er-unt

  • The Latin perfect represents a compromise between the simple perfect and a periphrastic form, made with the locative of the perfect participle (*-usi, haplologically reduced to *-ui- before the *s of the copula) and forms of the verb to be. We have: perfect periphrastic 1. *-h2ai m-i *-uzi-zm *am-verum 2. *-th2ai *m-ti *-uzi-s *am-vis 3. *-e *m-i *-uzi-st *am-vist 1. *-me *m-im *-uzi-zmos *am-verimus 2. *-te *m-te *-uzi-stes am-vistis 3. *-ri m-re *-uzi-znt am-verunt The resulting forms are: 1. m-i am-vi 2. m-isti am-visti 3. m-it am-vit 1. m-imus am-vimus 2. m-istis am-vistis 3. m-re am-verunt The presence of variant forms in *-sth2ai and *-ste in the second person perfect may have contributed to the merger.

    The stative conjugation The mi-conjugation endings are clearly derived from agglutinated personal pronouns. But where do the endings of the hi-conjugation come from? If we compare the endings of the hi-conjugation with the Afro-Asiatic stative and perfective, and with the Uralic stative and intransitive, a number of parallels emerge.

    Semitic Uralic PIE

    1. -ku -k -x

    2. -ta, -ka - -tx

    3. - - -

    1. -nu -tmk -mdhw

    2. -tunu, -kunu -ttk -dhw

    3. -u -t -an

    We can hypothesize that the hi-conjugation endings derive from an old stative, i.e. from postfixed forms of the verb to be. The actual forms of this copula seem to be derived from a root **ku-, with prefix-conjugation forms: Nostratic Pre-PIE Early PIE 1. *-a -ku(-a) *-ka *-x 2. *-a t-ku-a *-tka *-tx 3. *-a *-a *- 1. *-at m-kun(-u) *-mtku *-mdhw 2. *-at t-kun(-u) *-ttku *-dhw 3. *-an *-n *-n

  • The forms of the copula were added to nouns and verbal nouns in the absolutive: sg. *-a and plural *-at (> *-an in the Auslaut). In the 1pl, expected *-tmku was metathesized to *-mtku. In the Auslaut, *-tku gave *-txw, and then *-dhw in the 1 and 2pl.

  • The middle A reconstruction of the PIE middle endings is notoriously difficult. What seems clear to me is that the endings consist of three elements:

    1. a personal ending that is clearly cognate with the hi-conjugation endings: 1sg *-h2- 2sg *-th2- 3sg *--

    *-t- 1pl *-mVdh- 2pl *-dhw- 3pl *-r-

    *-nt- 1du *-h2m-; *-wVdh- 2du *-h2th2- 3du *-h2t-

    2. a connecting vowel *-o- or *-e-

    3. an additional ending: *-m, *-h2m *-r, *-h2r *-, *-h2 *-dh *-j

    The additional endings in the attested middle forms have sometimes been pressed into service as tense markers:

    *-r marks the present in Hittite, Tocharian and Italo-Celtic *-j marks the present in Indo-Iranian and Greek, but occurs in the preterit in Tocharian *-m occurs in the preterit in Indo-Iranian and Greek *-dh marks the preterit in Hittite *- marks the preterit in Indo-Iranian, Greek and Tocharian

    The distribution suggests that tense marking was not the original function of these endings. A possible clue to the original function of these endings can be inferred from the fact that the ending *-r seems to avoid the second persons (e.g. in Latin, and partially in Old Irish and Hittite), while *-m seems to avoid the first persons in Indo-Iranian and partially in Greek. This opens up the possibility that the additional endings were in origin person markers denoting the indirect object, i.e. the person for the benefit of which the action was performed. This is a superset of the basic meaning of the middle in Proto-Indo-European, which is an action performed for the benefit of the actor (reflexive, reciprocal, etc.). If so, we can reinterpret the additional endings as dative person markers:

    1. *-m 2. *-n > *-r 3. *-

  • 1. *-m 2. *-dh 3. *-j

    1. *-h2m 2. *-h2r 3. *-h2

    The original middle grid would have been:

    1sg/pl 2sg 3sg 2pl 3pl 1du 2du 3du

    1sg *-xr *-x *-xdh *-xj *-xxr *-xx

    2sg *-txm *-tx *-txj *-txxm *-txx

    3sg *-m *-r *- *-dh *-j *-xm *-xr *-x

    *-tm *-tr *-t *-tdh *-tj *-txm *-txr *-tx

    1pl *-m dhr *-m dh *-m sdh *-m dhi *-m dhxr *-m dhx

    2pl *-dhwm *-dhw *-dhwj *-dhwxm *-dhwx

    3pl *-rm *-rn *-r *-rdh *-rj *-rxm *-rxr *-rx

    *-ntm *-ntr *-nt *-ntdh *-ntj *-ntxm *-ntxr *-ntx

    1du *-xmr *-xm *-xmdh *-xmj *-xmxr *-xmx

    2du *-xtxm *-xtx *-xtxj *-xtxxm *-xtxx

    3du *-xtm *-xtr *-xt *-xtdh *-xtj *-xtxm *-xtxr *-xtx

    In the 1pl, the expected form *-mdhw-- was metathesized to *-m(w)dh- or *-m(w)adh-, perhaps

    under the influence of the mi- and hi-conjugation endings *-masw,*-mwan, *-ma.

    The connecting vowel *-- becomes *-- when stressed, *-e- (before voiceless) or *-o- (before

    voiced) when unstressed.

    Hittite 1. -hha(ri), -hhahari -hhat(i), -hhahat(i) 2. -tta(ri), -ttati -tta(ti) 3. -tta(ri) -tta(ti) -a(ri) -a(ti) 1. -wasta(ri), -wastati -wastat(i) 2. -ttuma(ri) -ttumat(i) 3. -anta(ri) -antat(i) Hittite forms with apparent zero additional ending, may in fact contain *-r, which was dropped in unstressed position. Both present and preterit forms may show additional *-i, which only in later Hittite became confined to the present (after the mi-conjugation present marker -i). The 1sg ending variant -hhahari (and secondarily -hhahat) is possibly from 1sg x 2du *-h2ah2r. In the present tense, *-r did originally not occur in the 2sg/pl forms, as well as in the 1pl form, which originates from 1pl x 2pl *-mwadh-dh > *-wast(a). It is indeed the case that the 2sg and 1pl show the aberrant forms -ttati and -wastati in the present. The 2pl has -ttumat in the imperative. The forms -ttari, -wastari and -ttumari were formed secondarily,, when -ri had become firmly established as the present tense marker in the middle.

  • The 2pl form has a fortis ending, as if from *-t(u)wo- instead of expected *-dh(u)wo-. Undoubtedly, this is under the influence of fortis -tten(i), -steni in the 2pl elsewhere.

    Tocharian 1. *-mr *- mj, *-wj 2. *-tr *-tj 3. *-tr *-t 1. *-mtr *-mt 2. *-tr (A: *-cr) *-t (*-c) 3. *-ntr *-nt In the present tense, Tocharian has generalized *-r, even in the second persons. The preterit has *-aj in the 1 and 2sg. The first person has inserted *-m- from the mi-conjugation (*-m-h2ar and *-m-h2aj). The 2pl has in part adopted active endings *-te > -c(r) instead of *-dhwe > -t(r).

    Vedic 1. -e -i, -a 2. -se -ths 3. -te -ta -e -a 1. -mahe -mahi 2. -dhve -dhvam 3. -nte -nta -re -ra(n), -ram 1. -vahe -vahi 2. *-the *-thm 3. *-te *-tm In the present, all forms end in *-oj (*-aj). In the past, different strategies are involved. Second singular *-saj is reshaped after *-th2aj. The *-th2- is preserved in the past tense form -ths, which likely derives from 2sg x 3du *-th2ah2 plus added -s. The 2 and 3du forms are -it(h)e, -it(h)m in the athematic paradigms, -et(h)e, -et(h)m in the thematic ones. This suggests the presence of a vocalized laryngeal, deriving from the original 2du and 3du subject markers *-xtx- and *-xt-. First person -i in -i, -mahi, -vahi can be derived from the 1pl x 3pl form *-medhi or 1pl x 3du *-medhh2. The optative has expected -a from *-h2a. Another possibility is that 1sg -i is an import from hi-conjugation secondary *-h2. We have the marker *-m in preterit 2pl -dhvam, 3pl -ram, 2du -ethm and 3du -etm. Preterit -ran could be from *-ro + -nt. Another, perhaps more remote, possibility is that it derives directly from 3pl x 2sg *-r--n, with *-n > *-r blocked after *-r-.

  • Greek 1. -maj -mn 2. -saj -so 3. -toj, -taj -to 1. -me(s)tha -me(s)tha 2. -sthe -sthe 3. -ntoj -nto 2. -sthon -sthon 3. -sthon -sthn In the 1sg and 2sg, *-m- and *-s- have been introduced from the mi-conjugation. The 1sg preterit form might, like the Ved. 2sg preterit form, reflect a form with 3du dative agreement *-h2ah2, with added *-m. The 1pl variant -mestha, like Hitt. -wasta, seems to be derived from 1pl x 2pl *-mesdh. The regular form -metha reflects 1pl x 3du -medhh2. In the 2pl *-(s)dhwe > -sthe, as well as in the 2/3du forms, we have -s- from the dental stems. The 2pl form, as in Tocharian, additionally shows the regular unstressed vowel *-e instead of stressed or analogical *-o, as is found almost everywhere else, except in the neighbourhood of *-h2. 2/3du -sthon, -sthn carry the dative marker *-m.

    Latin 1. -or *- + -(a)r 2. -re, -ris *-so 3. -tur *-tor 1. -mur *-mor 2. -mini *-dhwom-oj 3. -ntur *-ntor The marker *-r appears everywhere, except in the second person. The singular form -re (secondarily -ris) is from *-so, for original *-th2o. The origin of the 2pl form is enigmatic. I would derive it from *-dhwom + -oj, by way of *-bimi > -mini.

    Old Irish 1. -ur, -or 2. -der, -ther 3. -dar, -thar -ar 1. -mar, -mor 2. -th 3. -tar, -dar The marker -r has been generalized everywhere except in the 2pl. The 3sg distinguishes forms with *-t (-dar < *-tor) and without it (-ar < *-or).

  • Armenian 1. -a-y *-ah2-h2aj 2. -a-r *-ah2-th2a 3. -a-w *-ah2-to 1. -a-k` *-ah2-mesw 2. -a-ruk`, -a-yk` *-ah2-dhwesw 3. -a-n *-ah2-nto In Armenian, the middle endings survive in the aorist passive, where they are added to a suffix -a-. The second person ending -r probably reflects lenited *-th2a > -dha, with regular development of -dh- to -r(-). Original -dhw- in the 2pl gave -a-ruk`, with added -k` after the other plural endings.

    Gothic 1. -da = 3sg 2. -za *-sV 3. -da *-tV 1. -nda *-mdhV 2. -nda = 1/3pl 3. -nda *-ntV Gothic still retains a residual set of middle endings. The plural forms have all merged (1 and 3pl regularly, the 2pl presumably analogically).

    Middle endings in the active dual As we saw above, a number of middle endings were pressed into service as dual active endings: -h2tom 3du present -tom -h2tor 3du perfect Vedic -tur -h2te 3du present -te(s) -h2tah2m 3du past -tm