The Proto-Indo-European middle

The Proto-Indo-European middle Miguel Carrasquer Vidal, Feb. 2016 Formally, middle forms can be separated into (1) a verbal stem, (2) personal endings, (3) a o e ti g o el, a d a set of additio al e di gs . For an overview of the attested middle forms in Indo-European, see Appendix I. Verbal stem In the middle, the verbal stem shows no reduplication of itself, although middles from reduplicated formations (perfect, reduplicated imperfect, etc.) are reduplicated. The root vocalism normally shows zero grade, with stress on the endings, but i so e ases a rostati iddles , the stem has e-grade and is stressed. Personal endings The normal endings are clearly related to those of the h2e-conjugation: -h2-th2-Ø-mVdh-dhw-rIn the third person, we also find the mi-conjugation endings: -t-, -nt-. In the 1sg and 2sg -m- and -sare occasionally found, but these look like secondary innovations. Linking vowel After the personal ending, we find a linking vowel, which behaves in some respects like the thematic vowel. As in the case of the thematic vowel, we can hypothesize that it was *ó when stressed, and *o or *e when unstressed, depending on the voicing of the following segment. And, as in the case of the thematic vowel, due to the existence of both barytonic and oxytonic forms, the pattern was subsequently subject to analogical reshapings. Additional endings Very characteristic for the middle is the presence of additional endings after the linking vowel. Besides no ending (*-Ø), we find *-m, *-r, *-j, and (in Hittite) *-d(h). The ending *-Ø is attested in the Vedic 3sg past -ta, -a, 3pl past -nta; in the Greek 2nd and 3rd persons past (-so, -to, -(s)the, -nto), in the Latin 2sg -re < *-so, in the Gothic forms -za, -da and -nda, and in Armenian -aw, -aruk` and -an if from *-a-to, *-a-dhwe(sw) and *-a-nto. But the Hittite forms without additional ending (-hha, -tta, -a, -wasta, -ttuma, -anta) may have resulted from loss of *-r in final unstressed position (Yoshida 1990, Kloekhorst 2007). The ending *-m is attested in the Vedic 2du. past -ithā , 3du. past -itā , 2pl. past -dhvam, 3pl. past -ram. Greek has 1sg. past *- ā , 2/3du. past -sthon, and we may also have *-m in the Latin 2pl. ending - i ī if that is from *-dhwom-oj. The ending *-r is attested in the Hittite present middle (all forms except 2pl. +-ttumari), the Tocharian present middle (all forms), and the Latin non-second person forms (-or, -tur, -mur, -ntur). In Old Irish, the 2pl has no -r, and the -r in 2sg. -ther may be analogical (cf. 2sg impv. -the). The ending *-j is found in the Vedic present middle (all forms), the Greek present middle (-maj, -saj, -toj, -ntoj), and the Tocharian past middle (ToB -mai, -tai, -te, -mte, --, -nte). The ending *-d(h) is only found in the Hittite past middle (all forms), but occasionally also occurs in the present tense. We can see that the additional endings have been pressed into service as tense-markers in several Indo-European languages, but this does not appear to be their original function: *-j, which is presentic in Vedic/Greek, is preteritic in Tocharian. The ending *-r is only attested as presentic, and *-Ø, *-m and *-dh only as preteritic, but it is hard to imagine what need there possibly could have been for two or more distinct present-markers, besides two or more preterit-markers. Of course we ha e a si ilar u er of se o dar prese t for ati es *-u-, *-nu-, *-e-, *-je-, *-śe-, *-de-, *-dhe-, *-te-), but none of them show any ambiguities between imperfective and aoristic sense, such as is the case in middle *-oj- which can have both presentic and preteritic force. If we look for other possible original functions of the additio al e di gs , atters get e e orse: why, for instance, five different aspectual markers, or five mood markers, without traces of any clear semantic distinction between them? Dative agreement In fact, I believe there is only one category that comfortably fits the presence of that many different endings, and that is the category of person. This is partially confirmed by the distribution of the additio al endings themselves. The clearest case is the marker *-r, which avoids being attached to second person forms in Italo-Celtic and Hittite. The marker *-m is found only with second and third person forms, with the exception of (secondary) Greek *- ā . This is very reminiscent of the distributional pattern we would expect for a paradigm with dative agreement, as is found in a language such as Basque. In Basque, dative agreement is marked on the (auxiliary) verb, but the resulting grid lacks forms for first person subject performing an action for first person, and likewise for second person x second person (see Appendix II). We can therefore tentatively assign the ending *-r to a second person (singular) indirect object, while *-m would denote a first person indirect object. It stands to reason to assign *-Ø to the third person singular and *-j to the third person plural, as these two endings seem to show no clear restrictions in their distribution. That leaves only *-d(h), which we can tentatively assign to 2pl indirect object agreement, although there is no compelling evidence whatsoever in favour of that. In that case, *-m would mark both the singular and plural of a first person indirect object. There are also forms showing a long vowel or an additional laryngeal: perhaps the Vedic 1st persons -i, -vahi, -mahi, if -i < *-h2, and certainly 2sg -thās, 2du -ithā , 3du -itā . Further, Hitt. 1sg -hhahari, -hhahati, Grk. 1sg *- ā , 3du (act.!) *-tā , and perhaps the enigmatic OIr. 2sg ending -the(r). To explain these forms, we can set up additional dative agreement markers for the dual persons: *-h2m, *-h2r, *-h2. Grid The resulting grid consists of the following elements: ROOT SUBJECT IND.OBJ. 1 2 3 1 2 3 1sg/pl --txā́ -ā́ -tā́ --dhwā́ -rā́ - tā́ e.g. CaRC- (> oxytone), CāRC- (> acrostatic) sg -x, -tx, -Ø ~ -t, pl -mdhw, -dhw, -n > -r ~ -nt sg -ā́ , -ā́r < ā́ ? , -ā́ pl -ā́ , -ā́d(h), -ā́j du -ā́xm, -ā́xr, -ā́x 2sg -xā́r --ā́r -tā́r -mdhwā́r --rā́r - tā́r 3sg -xā́ -txā́ -ā́ -tā́ -mdhwā́ -dhwā́ -rā́ - tā́ 2pl -xā́dh --ā́dh -tā́dh -mdhwā́dh --rā́dh - tā́dh The 1pl subject forms were metathesized: 1 -- ā́dhr - ā́dh - ā́sdh 3pl -xā́j -txā́j -ā́j -tā́j -mdhwā́j -dhwā́j -rā́j - tā́j - ā́dhi 1du --txā́x -ā́x -tā́x --dhwā́x -rā́x - tā́x -- 2du -xā́xr --ā́xr -tā́xr -mdhwā́xr --rā́xr - tā́xr 3du -xā́x -txā́x -ā́x -tā́x -mdhwā́x -dhwā́x -rā́x - tā́x - ā́dhxr - ā́dhx In the acrostatic forms (built on forms with long vowel in the verbal root), the accent was retracted to the root, and the long vowel of the dative agreement marker developed into *e finally and before voiceless consonants, *o elsewhere. The original distribution would have been: 1sg/pl 1 -2 -th2om 3 -om -tom 1 -2 -dhwom 3 -rom -ntom 2sg -h2or --or -tor -modhr --ron (?) -ntor 3sg -h2e -th2e -e -te -modh -dhwe -re -nte 2pl -h2odh --odh -todh -mesdh --rodh -ntodh 3pl -h2oj -th2oj -oj -toj -modhi -dhwoj -roj -ntoj 1du --th2eh2m -eh2m -teh2m --dhweh2m -reh2m -nteh2m 2du -h2eh2r --eh2r -teh2r -modhh2r --reh2r -nteh2r 3du -h2eh2 -th2eh2 -eh2 -teh2 -modhh2 -dhweh2 -reh2 -nteh2 In the zero-grade forms (built on unlengthened verbal roots), the vowel should always have remained as *-ó-. In practice, Ausgleich between the acrostatic and oxytone paradigms took place, and we find unetymological forms like *-h2ar, *-h2aj; * -(t)o, * -ro, * -nto, etc. Because of the presence of dual dative agreement markers, it is likely that dual subject forms were eventually created. They can tentatively be reconstructed as follows: 1 --wodhr -wodh -wesdh -wodhi 2 -h2th2om --h2th2e --h2th2oj 3 -h2tom -h2tor -h2te -h2todh -h2toj --wodhh2r -wodhh2 -h2th2eh2m --h2th2eh2 -h2teh2m -h2teh2r -h2teh2 The third person marker is *-h2-t-, which testifies to the late creation of these forms (after the spread of the middle to the mi-conjugation, see below). Semantics and development It is quite obvious that the endings of the middle are related to the endings of the h2e-conjugation. To explain this, it has been suggested (Jasanoff 2003, Pooth 2014) that the PIE perfect and the Anatolian hi-conjugation are ela oratio s o a origi al proto iddle ith stati e-intransitive or detransitive semantics. The middle (unambiguously reflected both the Anatolian and non-Anatolian branches of PIE) would best retain the original semantics of this Pre-PIE proto iddle . Inherited background I agree that the PIE h2e-conjugation can be traced back to a stative inflection, more specifically from a (verbal) noun suffixed by personal forms of the copula (except in the third person). The original Proto-Nostrati formation was: (verbal) noun in the inactive (*-a, pl. *-at- or *-an) + prefixconjugated forms of a er to e (*-ku-): 1. 2. 3. 1. 2. 3. -a -a -a -at -at -an Ɂa-ku ta-ku-a, (f.) ta-ku-i ma-ku ta-ku-an, (f.) ta-ku-in In Semitic, this resulted in the stative/perfective suffixed conjugation, e.g. the Akkadian stative: 1. 2. 3. 1. 2. 3. -âku -âta, -âti -a, -at -ânu -âtunu, -âtina -û, -â (W.Sem. -tu) (S.Sem. -ka, -ki) (and -âna, -âni) (S.Sem. -kum, -kin) The pl has su stituted the roke plural e di gs -û (m), -â (f). The variation between North Semitic -t- and South Semitic -k- in the second persons can be explained if the original form was *-tk-. In the 1pl, expected *-tmku has been simplified to *-nu(a/i). If we compare the Uralic1 and the PIE stative/perfective forms (as reflected in the h2e-conjugation a d the iddle , e a re o stru t the Norther Nostrati e di gs as follo s: 1. *-ka 2. *-tka 3. *-a Uralic *-k, PIE *-h2 Uralic *-tk > *-nk > *-ŋ > *-n (?), PIE *-th2 Uralic *-ø, PIE *-ø 1. *-tmku 2. *-ttku 3. *-an Uralic *-mmək, PIE *-mtku > *- dγw (with metathesis) Uralic *-ttək, PIE *-tku > *-dγw Uralic *-n, PIE *-ér < *-én Compared to the Afro-Asiatic forms, *-ku-a > *-ka has been generalized in the 1/2 singular, *-ku(n) in the 1/2 plural. The original prefix- o jugatio > i perfe ti e after the rise of the perfective in West-Semitic) was completely lost in Northern Nostratic. 1 And, further afield, Chukchi-Kamchatkan and Eskimo-Aleut. Uralic and Indo-European In Uralic, the inherited stative became the stative/intransitive, with corresponding transitive forms provided by new paradigms formed by addition of possessive personal markers to a verbal noun in the inactive case (sg. *-ø < *-a; pl. Inlaut -t-, Auslaut -j < *-ati): 1. 2. 3. 1. 2. 3. intr. *-k *-n (?) *-ø *-mmək *-ttək *-n, *-t tr. sg. *-mə *-tə *-sa *-mat *-tat *-sat tr.pl. *-t-mə *-t-tə *-j + sa *-t-mat *-t-tat *-j + sat PIE apparently utilized a completely different strategy as compared to Uralic. Instead of assigning the stative forms to the intransitive paradigm, the inherited stative developed (as in West-Semitic and Egyptian) into a perfective, and acquired full-fledged intransitive, transitive singular and transitive plural forms, as well as forms with indirect object agreement. This is similar to a verbal system like that of Basque, for an impression of which see Appendix II. Forms with a plural absolutive (intr.pl, tr. with pl. object) reduplicate the verbal root: sg pl intr. CáCCí-CaCCi-CáC- tr.sg. CáCCáCCaC-‘ tr.pl. Cí-CaCCí-CaCCi-CáC- tr.sg. + dat. CaC- + ā́-DAT CaC- + ā́-DAT B le gthe i g the first o el, eo-i perfe ti e paradig s ere also reated: sg pl Cā́Ć Cī-CaCCī-CáC Cā́CCā́CCāC-‘ ́ Cī-CaĆ Cī-CaCCī-CáC CāC- + ā-́ DAT CāC- + ā-́ DAT For further details, see Carrasquer Vidal 2016. Besides the perfective, a completely new set of verbal roots was created, disjoint from the h2econjugation, with endings based on the agglutination of personal pronouns, originally perhaps both transitive and intransitive (which later merged): 1 2 3 Pre-PIE tr. / intr. *-mu *-mi *-tu (*-ti) (*-u) *-ta PIE tr. / intr. / progressive *-m *-s *-t *-Vmwi, *-Cmi *-Vswi, *-Csi *-ti 1 2 3 *-mu-átu *-tu-átu (*-átu) *-mé(sw) *-té *-ént *-méswi, *-wéni *-té, *-téni *-énti *-mu-án *-tu-án *-án-ta *-wén *-tér This is the mi-conjugation. The creation of a whole new set of verbs might seem unusual, but it is hat ulti atel ust follo fro the di hoto et ee aoristi = h2e-conjugation) and imperfective (= mi-conjugation) verbs that we see in PIE (or at least in LIV2). The origin of the dative agreement forms The fact that the 3pl ending is *-r in the h2e-conjugation middle (by the Auslaut-rule *-n > *-r) suggests that the li ki g o el a d the additio al e di gs ere origi all a si gle ord, hi h as later agglutinated to the personal forms of the verb. We can reconstruct the dative agreement markers as follows: 1. ā́ 2. ā́r (< ā́ ?) 3. ā́ 1. ā́ 2. ā́d h 3. ā́j 1. ā́x 2. ā́xr 3. ā́x I ot a are of a thi g si ilar to this elsewhere. It looks like a pronominal element *ā, pl. *āj followed by the familiar first person marker *-m (and third person *-Ø). The second person endings, however, are mysterious: *-r (or *-n?), *-dh (or *-d?). There is only a vague similarity to the h2econjugation endings (Uralic *-n, PIE *-dhw … From dative to reflexive to middle In the third person of the h2e-conjugation with dative agreement, originally, forms like *X-ó, *X-r-ój ould ha e ea t he X-e(s/d) for her , or the X(ed) for the ith the su je t a d the i dire t object being different entities (like *X-óm he X-es for e . But the for s ould also e i terpreted reciprocally and reflexively: *X-ó he X-e(s/d) for hi self , *X-r-ój the X(ed) with each other . I time, this apparently became the primary use of these forms, and the pattern was copied by the miconjugation, giving rise to reflexive forms with 3rd person *-t-ó and *-nt-ój. From the 3rd person, the reflexive spread to the other persons (the mi-conjugation forms adopting the established h2econjugation endings), and the connection with dative agreement was lost sight of. The refle i e I do/did X for self (etc.) could at some point be expressed, both for h2e-conjugation as for miconjugation verbs, using either of the endings *-h2ar, *-h2a, *-h2adh, *-h2aj, *-h2ah2r or *-h2ah2, and different dialects made different choices. Since the reflexive (middle), unlike the mi-conjugation and the h2e-conjugation, did not have primary or secondary endings, one way in which the variant endings were put to use was to distinguish present tense from past tense. Another way in which the de ris fro the dati e agree e t for s could be used was to provide additional dual endings in the active paradigms (e.g. 2du. *-(h2)th2om, *-(h2)th2eh2m, or 3du. *-(h2)tom, *-(h2)tor, *-(h2)te and *-(h2)teh2m). From the reflexive use, the semantics of the PIE middle, ranging from reflexive/reciprocal through impersonal to passive, follow effortlessly. In summary, the PIE middle arose as a verbal category showing indirect object agreement. Starting in the third person, these forms also acquired reciprocal/reflexive use, and were then also adopted by the mi-conjugation (in the third person). Subsequently, the reflexive spread to the 1/2 persons, and the o e tio ith dati e agree e t as lost. There as o protomiddle with stative-intransitive or detransitive semantics from which both the h2e-conjugation and the middle arose. The middle arose as a pluripersonal agreement form within the context of the h2e-conjugation perfective. Appendix I. Middle forms 2 Vedic Hittite Toch. A Toch. B Greek Latin O Irish Gothic Arm. -ai -i -sai2 -thās -tai, -ai -ta, -a -vahai -vahi -āthai, -ithai -āthā , -ithā -tam3 -athur -ātai, -itai -ātā , -itā -t̄ -atur -mahai -mahi -dhvai -dhvam -ntai, -rai -nta, ran, -ram -̮̮a, - ̮̮ari, -̮̮a̮ari -̮̮at, -̮̮ati, -̮̮a̮ati -tta, -ttati, -ttari -tta, -ttat, -ttati -tta, -ttari, -ttati, -a, -ari -tta, -ttat, -ttati -ār -(w)e -tār -te -tär -t -mar -mai -tar -tai -tär -te -mai - ̄ -sai -so -tai,-toi -to -or -ur -da -ay -re,-ris -ther,-the -za -ar -tur -thir -da -aw -ton -aton -sthon -ton -t̄ -aton -me(s)tha -mur -mir -nda -ak` -sthe - i ī4 -the, -d -nda -ntoi, -ntai -nto -ntur -tir -nda -aruk` -ayk` -an -sthon -tẹ -wašta, -waštari -waštat -ttuma -ttumat -anta, -antari -antat, -antati - tār -mät -cär -c -ntär -nt - tār -mte -tär -t -ntär -nte Cf. also Lith. -ì, Slav. -si < *-saj, employed as general 2 sg. endings. Bold forms are active forms, in my opinion borrowed from the middle to make dual endings. 4 The origin of this form is controversial. I would derive it from *-dhwom + *-oj > *-bi ī > *- i ī > - i ī (or > *-bi ī > - i ī). 3 Appendix II. Basque intransitive, transitive and dative agreement forms To simplify matters, I will only give the present indicative forms of the intransitive auxiliary *da to e , a d the tra siti e auxiliary *du to ha e o l the for s with third person object for the latter). For more details, see Carrasquer Vidal 2004. Intransitive and transitive forms: intr. tr.sg. tr.pl. 1 naiz dut ditut 2m haiz duk dituk 2f dun ditun 3 da du ditu 1 gara dugu ditugu 2 zara duzu dituzu 3 dira dute dituzte The intransitive forms had reduplication in the plural: *ga-di-dá > *gairá > gera ~ gara; *za-di-dá > *zairá > zera ~ zara; *ø-di-dá > dira. Likewise in the plural transitive forms: *ø-da-di-dú- > *diddú- > dituThe 1/2sg intr. were in my opinion originally *ná-da > *nā, *há-da > *hā, which were then extended with the suppletive root *iza (which provides the infinitive/participle izan to e . The absolutive prefixes are: 1 *na-, 2 *ha-, 3 *ø-, 1 *ga-, 2*za-, 3 *ø-. The ergative suffixes are: 1 *-da, 2m *-ga, 2f *-na, 3 *-ø, 1 *-gu, 2 *-zu, 3 *-te (~ *-de). The intransitive forms with dative agreement: intr.dat. 1 2m 2f 1 -natzaik natzain 2 hatzait --3 zait zaik zain 1 -gatzaizkik gatzaizikin 2 zatzaizkit --3 zaizkit zaizkik zaizkin 3 natzaio hatzaio zaio gatzaizkio zatzaizkio zaizkio 1 -hatzaigu zaigu -zatzaizkigu zaizkigu 2 natzaizu -zaizu gatzaizkizu -zaizkizu 3 natzaie hatzaie zaie gatzaizkie zatzaizkie zaizkie The root is suppletive tza- to lie instead of da- to e . The -i- < *-gi- comes from the verb e-gin to do . In the plural this becomes -zki-, from the absolutive plural suffix -z- + -gi-. The dative suffixes are almost identical to the ergative suffixes: 1 *-da, 2m *-ga, 2f *-na, 3 *-o, 1 *-gu, 2 *-zu, 3 *-e, where 3rd person *-o and *-e probably derive from the demonstrative pronoun *hau, pl. *hai-. The transitive forms with dative agreement: tr.3sg.dat. 1 2m 2f 1 -diat dinat 2m didak --2f didan --3 dit dik din 1 -diagu dinagu 2 didazu --3 didate diate dinate tr.3pl.dat. 1 2m 2f 1 -dizkiat dizkinat 2m dizkidak --2f dizkidan --3 dizkit dizkik dizkin 1 -dizkiagu dizkinagu 2 dizkidazu --3 dizkidate dizkiate dizkinate 3 diot diok dion dio diogu diozu diote 3 dizkiot dizkiok dizkion dizkio dizkiogu dizkiozu dizkiote 1 -diguk digun digu -diguzu digute 1 -dizkiguk dizkigun dizkigu -dizkiguzu dizkigute 2 dizut --dizu dizugu -dizute 2 dizkizut --dizkizu dizkizugu -dizkizute 3 diet diek dien die diegu diezu diete 3 dizkiet dizkiek dizkien dizkie dizkiegu dizkiezu dizkiete The stem di- is to be derived from *du-gi-, from du- to ha e a d gi- to do . I the plural, e again have the morpheme -zki-. References Carrasquer Vidal, Miguel, 2004, The Basque verb. Carrasquer Vidal, Miguel, 2016, The prehistory of the h2e-conjugation. Jasanoff, Jay, 2003, Hittite and the Indo-European Verb. Kloekhorst, Alwin, 2007, The Hittite inherited lexicon. LIV2 , 2001, Lexicon der indogermanischen Verben. Die Wurzeln und ihre Primärstammbildungen, unter Leitung von Helmut Rix und der Mitarbeit vieler anderer bearbeitet von Martin Kümmel, Thomas Zehnder, Reiner Lipp, Brigitte Schirmer. Zweite, erweiterte und verbesserte Auflage bearbeitet von Martin Kümmel und Helmut Rix. Pooth, Roland, 2014, The Proto-Indo-European voice system (draft). Yoshida, Kazuhiko, 1990, The Hittite mediopassive endings in -ri.