CC BY
20
Andrej V. Sideltsev
Notes on a recent book by Prof. Jay H. Jasanoff
Jay H. Jasanoff, Hittite and the Indo-European Verb. Oxford University Press, 2003. XIV + 270 p.
Being primarily interested in synchronic accentual-ablaut classes of Middle Hittite verb, I do not aim to tackle the complex Indo-European reconstructions of Prof. Jasanoff's book. However, I am absolutely confident that only meticulous philological expertise can serve the basis for any comparisons1.
1. p. 19: «According to Melchert (1997), some pairs of the latter type [the cases in which a single Hittite verb is attested with two stems, one corresponding to an IE present and the other to an aorist [...], for example [... ] parkiya-: park- 'rise'], both in Hittite and in Luvian, show the remains of a synchronic distribution in which the stem in *-ye/o- was confined to the present indicative while the shorter stem was used in the preterite and imperative. If confirmed by further philological research, this pattern would virtually guarantee the existence of the standard PIE aspectual system for some stage of Proto-Anatolian.
This assumption can be proved to be false: the scanty OS data (according to Kassian, Glossary in: Anatolian Languages, Canberra, 2002) do not support it: both tarhu- and tarh- are used each 1x as pres. and never as pret.; both karpiye/a- and karp-are used as pres. (correspondingly 10x and 4x), only karpiye/a-is attested frgm. 1x in the pret.; both karassiye/a- and karas- are used as pres. — each 1x. More voluminous MS data contradict it.
2. p. 36 fn. 22: «The only respect in which wewakk- differs from a canonical perfect is in the position of the accent, which, to judge from the preserved -e-, must have been on the reduplication syllable. As Craig Melchert points out to me, however, the accent was probably still on the root at the time of Eichner's second lenition rule, which would otherwise have taken *wewakki to *wewaki, with laxing or voicing of -kk- to -k- [...]. Another probable example of
'As a rule, MS forms are not explicitly marked.
Труды ИЛИ РАН. Том II, -часть 1. СПб., 2006. 3^3
Notes on a recent book by Jay H. Jasanoff
an old perfect with leftward displacement of the accent is mema/ i-'say' [...]».
118: «The stem of mema/i- is thus to be set up as *memoH-i- / *memH-i- or *meimoH-i- / *meimH-i- (*moimoH-i- / *moimH-i-), the latter with intensive reduplication. Whatever its immediate perform, the accent came to be fixed on the reduplication syllable in Hittite, exactly as in the perfect wewakki».
I would rather suggest that instead of postulating an unmotivated accent shift for wewakk-, it is better to assume analogical spread of -kk- from the forms of 1 sg. pres. wewakk-hi, 1 pl. pres. wewakk-weni, 1 pl. pret. wewakk-wen, 2 pl. pres. wewakk-teni, 2 pl. pret., wewakk-ten, 1 sg. pret. wewakk-hun, 3 sg. pret. (if old) wewakk-td, etc. I dare say there are enough forms to provide trivial analogical leveling without resorting to a special accent shift.
In this case both wewakk- and mema/i- clearly display present-type accent, not perfect. See Oettinger, SHV2, XXIII-XXV.
129, 132: «The parent language clearly had athematic presents with both i- and e-reduplication.»
130: i-reduplication had h 2e-mode of inflection and e-reduplication had mi-mode of inflection.
131: «In Hittite the paradigm of *mimn- was analogically reconstituted on the basis of the 3 pl. mimmanzi, leading to a stem mimma-, with 'thematic' -a- [... ]»
In view of a more realistic assessment of mema - and wewakk -(see above) it is hard to uphold the proposed reconstruction of h2e-mode of inflection for i-reduplication and mi-mode of inflection for e-reduplication. At least on the basis of Hittite data.
3. 37: «[...] the oldest 3 sg. preterite of wewakki is [...] wewakta, with the mi-conjugation ending -t(a) (KUB 43.23, 'mit-telhethitisch oder spat althethitisch') . The form is of crucial importance for all of Jasanoff's reconstruction.
But the form is not as unique as Jasanoff wants us to think. As is well known, the original 3 sg. pret. hi-conjugation ending -s tended to be replaced (or at least influenced by) 3 sg. pret. mi-conjugation ending -t(a). Our data show that it must have started early enough. OS: mark- 'cut (up)' markta 2x.
Also 87 fn. 71: «Not to be confused with the retention of *-t in wewakta is the secondary replacement of -s by -t(a) elsewhere
A. V. Sideltsev
I can see no possible reason how wewakta can't be synchroni-cally confused with markta!
4. 65: «[...] the earliest instance of a hi-conjugation form occurs in a Neo-Hittite copy of an Old Hittite text. Significantly, the hi-form in question is the archaic hapax 3 sg. malli (KUB 7.1 II 1) [...].» Also Prof. Jasanoff cites malluwanzi, malluwar (in NS texts) «to the bare root mall-».
However, this assessment is clearly misleading. The earliest hi-conjugation form is actually m[all]ai in MS (CTH 330.2.A) KBo 15.35+ 15.33 Vs. I 11. Indeed, (CTH 390.A) KUB 7.1+ is an extremely late copy (13 b.c. according to Neu, Fs Neumann, 218 and Kühne, Fs Otten2!) containing very many scribal errors — see Kronasser, Sprache, Bd. 7, 1961, passim; Bd. 8 Hf. 1, 1962. There is no consensus on the date of the original composition — e.g., Rieken, StBoT 44, 141, 205, 217, 369 considers it to be mh./jh. Thus the form clearly lacks the alleged inner Hittite antiquity.
Cf., just among others, Tremblay, AfO, Bd. 46-47, 1999-2000, 225 who chooses to restore on the basis of philological evidence stem [maL-\ and a perform *mj(±h2)-yenti > IE *mlye-.
The fact that an athematic form replaces a thematic one is rare, but paralleled by e.g., MH/NS was-ta for expected wass-e-t. See Oettinger, SHV2, 300; Lehrman, IH Redux, 130.
Unfortunately such assessment of philological data seriously weakens validity of this and many other diachronical claims of the book.
5. Passim and esp. 74-75: Add Kümmel, Zur o-Stufe im idg. Verbalsystem // Akten der Tagung «Indo-European Word Formation — Inventory and Analysis», Kopenhagen, 20.-21. Oktober 2000. This paper was was published after the book was com-pleted2, but it is improbable, that Kümmel's handout which has been widely circulating in Europe was unavailable to Jasanoff. Anyway, rival points of view are largely ignored.
6. Actually after refuting historically acrostatic verbs of §akk-/ sekk-type (see 7. in the present review) Jasanoff bases his recon-
2M. Kümmel. Zur o-Stufe in der idg. Verbalstammbildung. In: IndoEuro-pean Word Formation , Proceedings of the Conference held at the University of Copenhagen, October 20th — 22nd 2000, ed. by James Clackson and Birgit Anette Olsen. [Copenhagen Studies in Indo-European 2]. Kopenhagen: Musem Tusculanum Press, 2004. S. 139-158.
Notes on a recent book by Jay H. Jasanoff
struction entirely on what he calls molo-presents (77-79: mall(a)-, kank-/kank-, padd(a)-, ispant-/ispant-, wars-, probably ans-, ark-/ ark-, arr(a)-, harra-, iskar-/iskar-, ispar-/ispar-, istap-/istapp-, mark-, mald-/mald-). He does not choose to let the reader guess that there is no indication in historical Hittite that they have ever really had e-vocalism: even when there is a phonological explanation of root vocalism from * e, the forms can as safely be traced back to R (where R is a sonorant).
73-74: «But here an important methodological caveat is in order. Since the a : a ablaut pattern clearly enjoyed a period of productivity in Hittite, it is neither necessary nor desirable to trace every case of a plural or middle with a-vocalism to a phonological source .
I have to modify this statement in one little way: it is highly desirable to have at least one clear-cut and totally unambiguous case. In my opinion the only approximation to this paradigm is ispant-/ispant- and only in the way Craig Melchert has mentioned pers. com. in June 2003: there are no plene spellings of participle *ispant-ant, implying [ispnt-ant-]. Even here mall(a)- and padd(a)-are not probative as participles may simply be thematic! Jasanoff leaves the non-Anatolian reader just wonder on the point.
It should probably be also noted that even participial lack of plene spellings is not probative for ispant-/ispant-type. The reasons for this are as follows.
General assessment of participle spellings leads to two main conclusions: (1) the use/lack of plene spellings mostly depends on scribal usage (e.g., OS iskar-ant- 8x vs MS iskar-ant- 3x / iskar-ant- 10x; kun-ant- 4x(100%) vs kur-ant- 4x (67%) /kur-ant-2x (33%) or huy-ant- 33x (100%) vs miy-ant- 2x (100%); tarn-ant- 10x, zinn-ant- 1x, suhh-ant- 1x (each — 100%) on the one hand vs ishuw-ant- 10x (71%) / ishuw-ant- 4x (29%) on the other, although verbs of each pare belong to the same morphono-logical type); (2) quite often the number of forms attested is too small for an objective assessment3 (e.g., sas-ant- 1x, no *sas-ant-).
It follows from these two points that if the data are not statistically considerable, lack of certain forms may be just inciden-
3A well known fact by itself. See, e.g., Melchert, AHP, 27: plene writings are not obligatory. Mind that plene spellings dominate in a clear minority of verbs (26%).
A. V. Sideltsev
tal. This has a direct bearing on sipant-/sipant- type: we can say nothing about the accent in sipant-ant- because all the available evidence is limited to (CTH 147) KUB 14.1 + Rs. I 78 si-pa-an-ta-an-za, (CTH 479.3) KBo 8.91+ Rs. 29 si-pa-an-da-an-ma with totally no OS data. The same refers to maldant- ((CTH 330.2.A) KBo 15.33+ Rs. III16 ma-al-ta-an, no OS data), markant- (no OS or MS forms), arkant- (2x?: (CTH 215) KBo 32.201 u. Rd. 33" (-f ]ar-ga-an-te-es; (CTH 663) KBo 30.106+ Vs. II 9' ar-kân-ta, OS 1x non-plene). kankant- is attested slightly more often (3x: (CTH 324.1.A) KUB 17.10 Rs. IV 28 kan-kan-za, (CTH 325.C) KUB 33.29+ Rs. IV 24' kân-kan-z[a, (CTH 683.1) KUB 55.43 Vs. I 17 ga-an-kan-te-es, also OS 3x non-plene)4.
Even though individual verbs of holokinetic verb types may be spelt without plene, every holokinetic verb type (pai-/pi-, ishuwa-/ ishuwa-, ed-/ ad-, ar-/ar-) is spelt plene at least several times as a class. But even in this respect presumed sipant-/sipânt-type is attested in MS more than three times as rarely (8x) as the smallest 'plene-spelt' types ishuwa-/ishuwa- (26x) ar-/ara (28x). So, the number of attestations even of the whole type is not statistically relevant either.
7. 79-81: new assessment of sakk-/sekk-type: the starting point for Hittite weak stem is pres. pl. sakk- and pret. pl. sekk-. pres. pl. sakk,- was secondarily built to the original pret. (<aorist) pl. sekk-created on the analogy of mi-conjugation verbs like es-/ as-. Pres. pl. sekk-forms are attributed to «the partial spread of e-vocalism» from pret. pl.
Quite understandably, absolute chronology of forms is a very reliable indicator of primary and secondary status. But in view of sâkkanzi (see below) OS sakteni 1x (MS saktâni 1x) has to be secondary in comparison to MS sâekteni 2x just as OS daweni 1x is secondary in comparison to MS tumeni 2x and OS tumeni 4x, OS tumâeni 1 x .
As far as «the partial spread of e-vocalism» is concerned, there is only one clear analogical expansion of sakk-/sekk-type onto ar-/ ar --type in MS texts (none in OS): erteni vs artâni, ârer vs tarer
4Mind that this very verb cannot go back directly to any acrostatic
*kénk-ont-.
Notes on a recent book by Jay H. Jasanoff
(see Oettinger, SHV2, 48 Anm. 36, 403-4). That the case is clearly analogical is shown by aranzi 5x.
MS arer is attested only in (CTH 199) HKM 47 Rs. 55, a letter from Tapigga (Masat) archive, notorious for its very advanced (practically NH!) usage. This circumstance together with the general scarcity of examples in MS texts assures only one thing — that ae weak stem expansion was basically post MH with the very beginning of the analogy at the end of MH period!
I do not think it is incidental that the only MS proper form with clear ae weak stem expansion is er-teni. I am inclined to think that it was created alongside ar-teni because there existed double 2 pl. pres. form sak-teni / sek-teni. I.e. it was sak-teni/ sek-teni variation that started expansion of pl. weak stem ae vocalism even though originally the double forms are due to a different (contrary) analogy (see 3.1).
NS influence of kamp-/karep-type onto ar-/ ar-type in the verbs ak(k)-, han- and ar-/ar- is also clearly secondary (see already HED, A, 111).
NS forms of ak(k)- and han- are limited to ek-er and hen-er contra older OS, MS ak-er5 and NS han-er, han-er6. NS er-stem of ar-/ar- is much more widespread (see ibid.): 3 pl. pret. (er-er, er-er vs ar-er), 1 pl. pres. (er-weni) and pret. (er-wen vs ar-wen),
2 pl. pres. (er-teni vs ar-teni, ar-teni, ar-teni) and pret. (ar-ten, ar-ten). But even here in all the forms e-vocalism is attested alongside regular a-forms (I interpret lack of ar-weni as incidental). Only
3 pl. pres. was consistently NS ar-anzi (ar-anzi, ar-anzi), MS han-anzi 3x, han-ya-anzi7 2x and NS akk-anzi.
Thus there are two points of difference between secondary e-stem expansion and §akk-/sekk-type. First, clear secondary analogy (basically NH) after §akk-/sekk-type (ar-/ar-, ak(k)-, han-) is separated from OS sak-teni and OS, MS asaskae- by all OH and most of MH periods — a couple of centuries at the very least!
Moreover, synchronically §akk-/sekk- behaves completely like other verbs of its class (e.g., karap-/karaep-) in as far as it attests pl. e vocalism. It is true that we have no OS forms of sekk-
5See HED, A, and HW2, A s.v.
6No OS or MS pl. pret. forms are attested for this verb contra Oettinger, SHV2, 52.
7Both forms are actually analogical.
A. V. Sideltsev
due to the limited text corpus. So the near complete (presumed) generalization of sékk- is a clear assumption. But the MS and NS data8 are so unanimous in pl. sekk-forms that it is hard to avoid this assumption.
Another point of difference between secondary e-stem expansion and sakk-/sekk-type is that akk-, han-, ar - never show levelling of e-stem onto the singular even in NS texts9. Neither do they attest 3 pl. pres. with é-vocalism10.
81: «The match between the distribution of es-, epp-, ses-, ed- vs as-, app-, ad- in the plural of the mi-conjugation, and the distribution of sekk-, hes- vs sakk-, hass- in the plural of the hi-conjugation, is surely not accidental.
Actually, the match is first and foremost incomplete. Jasanoff does mention the non-existence of *sakk-anzi, but does not give due account of it.
In view of the fact that ed-/ad- and (?) ekw-/akw- have completely replaced the supposed strong stem in 3 pl. pres., I find it particularly probative that sakk-/sékk-type has not: sékkanzi 4x, asésanzi 12x, taméssanzi 2x; ? OS k]arépanzi 1x (see Kassian, Glossary, 104). Actually the only form displaying analogical a-vocalism is tamassánzi11 1x.
It is also absolutely assured that 3 pl. pres. is not normally influenced by 3 pl. pret.12: akw-anzi 106x (OS 25+ 7 frgm.x)
8See Justus, Mat. Heth. Thes. Lf.10, 1981, Nr. 7, 4-13, CHD, S, s.v.
9Both sakk-/sékk- (see CHD, S, 21-22, Oettinger, SHV2, 55, Sommer, Falkenstein, HAB, 101 Anm. 2), karap-/karép- and asas-/asés- display singular forms with é-vocalism. For damass-/daméss- and karap-/karép- there are even MS forms: (CTH 209) KBo 18.69 Rs. 12 [t]a-mi-is-z[i, (CTH 832) KBo 9.114 13' ga-ri- << ra,- >> pa,-as. Cf. HED, K, 72, Kronasser, Sprache, 7, 167.
10Naturally, we may assume that analogy from karap-/karép- first (prehis-torically) worked for sakk-/sékk- (completely with the exception of 2 pl. pres.), then (starting in late MH and mostly NH) for ar-/aré (incompletely) and only in NH for ak(k)- and han- (sporadically). This scenario is not impossible by itself, but I know of no similar 'intermittent' analogical processes, very unevenly distributed over time. Normally, we would expect the reverse direction (i.e. sakk-/sékk- as the chronologically last, not first form).
11 The only verb that contrasts 3 pl. pres. and pret. is dar-ánzi 5x (OS 3x), tér-ér 2+ 1 frgmx (1x plene). 3 pl. pret. é-vocalism is here analogical after the suppletive form of 2 pl. pret./impv. OS té-ttén 1x, MS té-tté[n 1x.
12Only the verbs with both 3 pl. pres. and pret. attested.
Notes on a recent book by Jay H. Jasanoff
— ekw-er- 1x; app-anzi 65x (OS 15+ 3 frgm.x), app-anzi 2x
— epp-er 4+ 1 frgm.x (OS 2x); OS asanzi 'be' 5+ 1 frgm.x (no unambiguous MS forms!) — es-er 4+ 1 frgm.x (OS 3+ 1 frgm.x); ad-anzi 41 x (OS 7x) — et-er 1x; uw-anzi 2+ 1?x (OS 1x) — aw-er 2x; hu(y)-anzi 33x (OS 13x), ? huy-an[zi?] 1x — h[uw]a-er 1x, huwa-er 2x (1x plene); kun-anzi 8x (OS 3x) — kwen-er 8x; kur-anzi 1x — [k]wer-er 1x; d-anzi 136x (OS 28x), d-anzi 5+ 1?x — da-er 8x (OS 2x), da-er 1x; ti(y)-anzi 237x (OS 46x, tiyanti 1x) — day-er 14x, day-er 3+ 1frgm.x, da-er 5x, day-er 6x. da(y)er 2x13.
This fact falsifies Jasanoff's claim that 3 pl. pres. ssekkanzi is analogical after 3 pl. pret. — the analogy never existed.
3 pl. pres. is the last form to be analogically influenced in paradigms with 2 pl. impv. generalisation as is shown by ne-anzi in the paradigm of nai-/ne-. This nai-type generalization tendency at an earlier stage is represented by sporadic spread of strong stem onto 1 and 2 (not 3!) pl. pres.14: aumani 1x vs u(m)meni 5x, ed]wani and ekwani vs OS adweni 4+ 2 frgmx, akuweni 6x; nahwani vs regular hass(u)weni, OS daweni 1x, probably huekwani. 3 pl. pres. is never affected in MS and OS. There are some rare cases of further analogy (pointed out to me by Craig Melchert) only in NS: sesanzi for sasanzi along with sesueni; kuennanzi.
The only paradigms where 3 pl. pres. and pret. are identical are those with 3 sg. pres. generalization (i.e. strong stem — has-anzi 1x instead of hass-anzi 9x after has-i)! Thus sakk-/ sekk-type is oddly unparalleled by any other. So e-vocalism has to be original in pl. of both pres. and pret. of sakk-/sekk-type.
The analogy is even more difficult to believe if we follow Jasanoff's scenario: a new pres. form was built for pret. *sekk-wen *sek-ten *sekk-er due to the proportion (p. 85) *as-weni *as-teni *as-anti : *es-wen *es-ten *es-er. If even the verbs with original e-vocalism in 3 pl. pres. have completely replaced it with a-vocalism (see ed-/ade above), then why is (presumably analogical) sseekkanzi the only attested form in MS texts while every
13da(y)er 2x are in frgm. contexts and can belong to both da- and dai-. However, there are no clear dayer forms from da- contra Oettinger, SHV2, 75 Anm. 59 w. ref.
142 pl. pres. and OS 1 pl. pres. reveal the same tendency for da-/da-type.
A. V. Sideltsev
conceivable analogy pushes hard the form *sakk-anzi (never attested!)? All I know about Hittite verbal vocalism makes me expect this form, not ssak-teni. Mind that NH analogies did not yet work in MH period for 3 pl. pres. (contra Jasanoff)! So, the only scenario easily reconcilable with the data is as follows: the ssakk-/ssekk-type, whatever its ultimate origin, did not come into being by the same process as ed-/ad-. Even if we follow Jasanoff in reconstructing only pret. of ssakk-/ssekk-type, it formed its pres. forms at a time preceding the spread of accented full-grade vocalism onto pl. pret. and analogical creation of ed-/ade in lieu of *ed-/*ed-. Thus all the paradigm was strong-stem ssakk- and weak-stem sseekk- (both in pres. and pret. pl.!). Then, after the spread of accented full-grade vocalism onto pl. pret. and analogical creation of ed-/ade in lieu of *ed-/* ed- there appeared analogical forms of sakk,e.
The grounds for this and only this development was the retention of sekkanzi.
151: «h2e-conjugation root aorists with *o : *e ablaut», only pret., then presents were back formed of the preterites.
152: In the plural [. . . ] the new present plural in *-weni, *- (t)teni, *-enti took zero-grade root vocalism and shifted the accent to the endings. [... ] The model for the zero-grade end-accented plural forms was supplied by the prehistoric mi-conjugation pattern: * eswen : *asweni, epwen : apweni».
First and foremost there is again no unambiguous evidence for this acrostatic paradigm in historical Hittite for da-/da-verbs. Rather, there is clear evidence against it: this is the type that has a secondary e-grade stem — MS erteni. A very disquieting fact about the whole reconstruction is MS arten 'come' 1x. Actually, the grade and vocalism of this and any other stem of pl. pret. and imper. coincide with those of sg. while, according to Jasanoff's analysis, they are expected to differ. Another major restructuring? In this light there is only one hypothesis that conforms to the actual data:
The pres. of da-/da-, au(s)-/u-/au-, pai-/pi-15 types goes back to IE holokinetic ablaut (strong-stem sg. — zero-grade weak-stem pl.). Strong stem was secondarily generalized onto pl. pret. start-
15 Contra Hart, Fs Gurney, 57, because not just pai - 'give' has zero-grade in pl. pret., but several types.
Notes on a recent book by Jay H. Jasanoff
ing from 2 pl. impv.16 that was strong stem in IE (it coincided with 2 pl. pret. in Hittite). This starting point was characteristic in IE only of root present, root aorist and n-infix present. MS phonetically regular continuants of the primary distribution are esten 'be' 19+ 2 frgm.x and epten 6+ 1 frgm.x. Then strong stem spread onto pl. pret. of -^¿-conjugation verbs17 (mostly < perfects) by the proportion epzi — epten : aki — x. Among -hi-conj. verbs pai-/pi-, halzai- and tarna-types18 attest strong stem in pl. pret. only sporadically.
Second, although Jasanoff's scenario for sakk-/sékk- and da-/ da- is identical, the result of the presumed development is different — if stem distribution of da-/ da- clearly matches that of ed-/ ad-', stem distribution of sakk-/sekk- is synchronically different (see above).
I have to repeat again that da-/ da-type has no synchronic trace of acrostatic ablaut.
8. 85: «[...] there are no comparable ending-accented mi-conjugation forms of the type *asweni/*asteni, *appweni/ *apteni, or *atweni/*azteni in historical Hittite. This state of affairs can only be secondary, a reflection of the fact that [... ] ablauting mi-verbs gave up their accentual mobility», also p. 152.
Even if most generously assessed, this claim cannot be proved. First the number of 1 and 2 pl. pres. forms is scarce: only [ajppwéni 1x and apténi 1x are attested in MS texts alongside ept[eni (?) 1x and ept[an]i. The only form attested for ed-/ad-is édjwani 1x, but its root vowel is in lacuna. No relevant forms are attested for es-/as- 'be'.
There is a well known fact that plene writing is never obligatory, acknowledged by Jasanoff himself on p. 71 fn. 13. Can one really be sure that 2x usage of a verb is enough to say that plene spelling is never attested?
16See Oettinger, SHV, 67, 97, 141, 406; Gs Pedersen, 320 Anm. 64. Cf. Hart, Fs Gurney, 59-60; (?) Rosenkranz, HuI, 220; Justus, Mat. heth. Thes., Lf. 10, Nr. 7, 14-15. Cf. Oettinger, SHV, 113.
17Besides 1 pl. forms from section I.0, there are the following 2 pl. pret./impv. forms: akten 1x, arten 'come' 1x, áusten 8x, eten 'sit' 1x, haláitten 1x, máisten 1x, dátten 11 x (7x plene, OS 1x plene), dáisten 2+ 1 frgm.x (OS 2x), wásten 'buy' 1x.
18Lack of strong stem in pl. pret. of sipant- may be due to chance as plene spellings are very rare in closed syllables.
A. V. Sideltsev
Moreover, there seem to be other factors regulating the use of plene spellings in 1 and 2 pl. pres. and pret. Just to illustrate the complexity of the problem which is far from being purely phonological let us have a look at MS 1 pl. pret. The introduction of strong stem into 1 pl. pret. led to the situation when -wen was stressed only in the verbs of pai-/pi-type19. These forms never show plene spellings. Plene spellings of 1 pl. pret. ending occur only in the unstressed position20 : accented stem: ku-e-u-e-en, la-a-u-e-en, [d]a-i-u-e-en; up-pa-u-e-en; (CTH 199) HBM # 75 Vs. 10' i-ya-u-e-en; KuT 49 Vs. 11 u-wa-u-e-en; (CTH 138.1) KUB 23.77+ Vs. 41' [i-]si-ya-ah-hu-e-en21. Thus we deal with generalization of the vocalism of accented allomorph into the unaccented position22.
It is important to note that the phonetically regular 1 pl. pret. ending should be *-wan. As it is not attested, one has to suppose a factor that blocked the phonetic development — a tendency to keep vocalism identical in pres. and pret. with the orientation at the most common variant -weni. Consequently lack of a vocalism in pl. pret. is in any case secondary and analogical23.
Moreover, if lack of ending plene spellings leads Jasanoff to conclude that the ablauting mi-verbs gave up their accentual mobility, why cannot we interprete the fact that these same verbs never (at least not in OS and MS texts) show plene spellings of root vocalism in the same way, i.e. as a proof for not accented root vowel?
Under the present circumstances I do feel that it is much better to continue to believe that mi-conjugation forms of the type *asweni /* asteni were ending accented in accordance with basic Hittite accentuation rules.
9. 84 fn. 58: édwani (in MS actually only éd]wani 1x; fully attested only in OH/NS) does not directly represent the IE acro-static present ed-/ed- (see LIV1 s.v.) contra Melchert, AHP, 138,
19Even here pédumen and unnumen are preverb stressed (Oettinger, SHV2, 501, StMed 7, 225; Melchert, SHHP, 91, 128, 141; Kimball, HHP, 150-1. Cf. Melchert, SHHP, 91, Oettinger, Gs Kronasser, 166-7).
20Thus already (?) Melchert, SHHP, 128. See generally Kimball, HHP, 62. 1 pl. pres. is different: ending accented U/um-me-e-ni, du-m[e-]e-ni vs root accented ha-a-nu-mi-e-ni and preverb accented u-du-me-e-ni.
21 Following Melchert, SHHP, 128.
22See generally Kimball, HHP, 129; Melchert, SHHP, 108.
23Cf. Eichner, Flexion, 79; Melchert, AHP, 138; StMed 7, 191 n. 4.
Notes on a recent book by Jay H. Jasanoff
also 71. OH/NS edwani is not probative for OH data as there is always the possibility with copies that it is a scribe's incorrect transformation of the original. Moreover, OS adweni is attested many times (4+ 2 frgm.x as different from OS 1x, OH/MS 2x daweni) and has to be considered the normal/standard OS form. There are no OS strong-stem 2 and 3 pl. pres. for the word, either. It is pretty strange that we should find a precious archaism in NS with lots of 'innovative' OS forms. Given frequent expansion of full-grade vocalism onto 1, 2 pl. pres. already in OS and esp. MS texts, I emphatically reject the archaic character of edwani.
10. 94: «As far as the non-Hittite Anatolian languages are concerned, the attested forms of this verb point, if anything, to an otherwise unknown root of the form *bhyeH-, with zero-grade *bhiH-»
This assessment actually independently supports my assessment of the form pi-weni (see Sideltsev, Gs Forrer, 583). In this analysis pi-weni is not analogical after (unattested!) verb forms of the same class24, but an absolutely regular phonetic development.
11. 108-109, 114: ispai- < acrostatic Narten type ('type II') sg. *speh2-i-h2e(i) pl. *speh2-i-me-.
113: [. . . ] in Hittite [. . . ] the distinction between types I and II was lost [...]».
This reconstruction finds totally no support in Hittite where ispai- is identical to dai-/ti(y)- ('type I' of Jasanoff).
12. 114 fn. 56: «Note that i-presents show the same apophonic arrangements as ablauting mi-verbs in Hittite: the plural of the present has zero grade [. . . ], while the plural of the preterite, at least in the more conservative cases, has e-grade [...]. In ispai-, as in many other individual examples, the contrast between the present and preterite plural was given up.»
Actually, pl. strong-stem forms of isspai-type are attested (if sporadically): 2 pl. impv. paisten 'give' 3x. I tend to think that such forms should rather be interpreted as sporadic expansion of full-grade strong-stem pret. pl. The reason for this is simple: isspai-type is on the whole conservative and resistant to innovations, as is shown by irregular a-zero-grade in pl. pres. while full-grade
24Cf. (implausible) Kimball, Fs Watkins, passim (who ignores pi-weni). I was not aware of this paper while writing the article for Gs Forrer. Cf. Oettinger, Fs Rasmussen, 402
A. V. Sideltsev
strong-stem pret. pl. seems to enjoy productivity in historical Hittite. I do not see any motivation why a certain type should first give up a fully productive full-grade strong-stem pret. pl. and then sporadically acquire it again.
13. 118: «The plural forms of [nai-] are thus historically based on the weak stem *ne- < *neiH-, which surfaces in 3 pl. ne(y)anzi [...]», also 174, 183-184, 188-191, 196-197, 219-220.
197: In addition, as in the case of the non-sigmatized h2e-conjugation aorists *doh3-/*doh3-, [... ] a back-formed present was created in Anatolian [. . . ] The vocalism of the middle forms and the non-transparency of the ablaut pattern *nai- : *ne- were probably responsible for the maintenance of full grade in the 3 pl. pres. act. na(y)anzi; the theoretically 'regular' treatment would have been *niyanzi, with the same hyperweak zero grade as in other back-formed presents .
183 fn. 19: «[...] the -ai- of nai,shut is obviously an analogical import from indicative forms like 3 sg. pret. nais»
184: «the transitive and active si-imperative *nesi was responsible for the creation of the oppositional middle neshut and, via the same mechanism, the plural forms naisten < *nesten and nais-tani, -ssteni .
This analysis may be grounded from IE point of view, but not from Hittite one. It raises more questions than gives answers:
1. Why was all the pl. pres. and pret. analogically changed, but not 3 pl. pres.?
2. Why was the original 2 pl. pret. analogically changed too?
The only form that actually corresponds to the original acro-static paradigm outside sg. is 3 pl. pres. According to Jasanoff, this is the form least expected to retain anything.
While it may be true that ne-anzi retained its original ablaut due to its non-transparency, the question is why just 3 pl. pres. and none other form (including the original pret.) keeps to the old vocalism.
This question is answered by my treatment of 3 pl. pres. as the form most reluctant to innovation.
If we analyze the form neanzi as both analogical and continuing directly the original ablaut, this back-formed present cannot have
Notes on a recent book by Jay H. Jasanoff
been formed in the way ed-/ad-type was formed, i.e. with zero-grade vs pret. full grade. The only solution is as follows: if pres. of nai- has to be back-formed from pret., it could only originally have the same grade and type of stem as pret. This is exactly the development of sakk-/sekk-. The influence of es-/ as-type with weak stem in pl. pres. and strong stem in pl. pret. must have followed the process.
Actually, all pret. and imper. inflection of nai- has to be analogical to the pattern of epp-/appe, namely it borrowed its proportion (strong stem in sg. and pl. pret., weak stem in pl. pres.) as nai- is nothing but the strong stem of sg., having no connection with the old weak stem expected in an acrostatic paradigm of the type o : e. Judging by synchronic distribution, there is totally no clue of archaic pret. and imper. inflection for nai-. In this connection it is all the more important that 2 pl. pret. and imper. do not show ablaut type proper to aorist (as Jasanoff wants to believe), they show the ablaut type proper to perfect! It is another indication that Jasanoff's aorist theory must be wrong.
14. 120: «au-/u- 'see' [...] is a treasure trove of archaic apo-phonic and accentual features (cf. 1 pl. pres. umeni vs. pret. aumen, 3 pl. pres. uwanzi vs. pret. auir, 3 sg. mid. impv. uwaru) [. . . ] .
I fail to understand why the forms which are productive inner-Hittite creations should be regarded as archaic.
To sum it all up: the reconstruction of Jasanoff is elegant and very erudite, but it is not based on internal Hittite evidence in the points outlined above. It interprets the evidence in a way that is neither compulsory, nor required by synchronic data.
