The Indo-Iranian cakri-type.
According to the standard view, the cakri-type typically makes agent nouns "mit meist adjektivischer Verwendung" and (sometimes) intensive semantics (AiG 11,2: 291ff.). Although these correspond morphologically to the weak stem of the perfect (Barschel 1986: 307), semantically they are much closer to an imperfective present stem. There are thirty-[one.sub.1] formations of this type in the Rigveda. The better-attested items are distributed evenly across all ten books (see Table 1 in the appendix). The Avestan evidence confirms that this was an Indo-Iranian type. The literature is somewhat confused on what kind of nominalization this type is: essentially agent nouns, which came to be used as adjectives (AiG 11,2: 291ff.), or deverbal adjectival formations (Leumann 1942: 22 n. 1, Barschel 1986: 305). Barschel (p. 307) points out the morphological affiliation of the type to the perfect stem, but fails to notice that it does not always show the semantics expected of a deverbal formation from the perfect stem. He furthermore acknowledges that the slight preference for initial accent that the type displays--twenty of thirty-one stems are accented on the reduplicated syllable--is unexpected under the assumption that the perfect stem was the derivational basis. Barschel (p. 306) proceeds to argue that oxytone accentuation was original to these stems, and that the tendency towards initial accent is part of a broader pattern of innovated accentuation among the Vedic i-stems (citing examples such as cakri-: acakri-: jaghni-; nijaghni-: sthuri-, asthuri, etc., in which the compound form may have preserved the older oxytone accentuation). Since there is no functional difference between the cakri-forms with initial accent and the suffix-accented forms, there is no reason not to follow Barschel's analysis with respect to the accentuation. However, I disagree with his conclusion that the entire type had the perfect stem as its derivational basis. In the following, I will show that the core forms of the type do not have the semantics expected of a perfect stem derivative, and that they behave syntactically like participles rather than real agent nouns, in that they assign accusative case to their direct objects and can be modified by adverbs. They differ from other participles, however, in that they are not integrated into the paradigm of a particular tense/aspect stem.
[section] (2.) A noticeable feature of the cakri-type is its ability to take accusative and dative objects if the corresponding verbal root is transitive, as well as adverbial modifiers. In its simplex attestations, cakri- 'making, doing' (kr 'do, make, act', 3p1. perf. cakrur) is mostly used pred-icatively and takes accusative objects:
RV 9.88.4ab: Indro na yo maHd karmani cakrir, hanta vrtranam asi soma purbhin't Like Indra, who accomplishes great deeds, you. Soma, are the slayer of enemies, smashing fortresses.
There is a clear contrast between karmani cakrir "(repeatedly) doing deeds" and the immediately following hanta vrtranam "destroyer of enemies." The latter is a "real" agent noun with a suffix-accented -tar- and genitive rather than an accusative complement (but note that there is also a root-accented type that does take accusative objects; see Tichy 1995 on this suffix in general). (2) The same syntactic behavior is found in the instances where cakri- has incorporated a preverb or adverb-like modifier:
RV 6.24.5ab: anyad adya kdrvaram anyad u svo, 'sac ca san muhur acakrir indrah One deed today and another one tomorrow; (thus) Indra instantly turns that which is not into that which is. (3)
This example shows both incorporation of the preverb a and accusative case assignment in a double accusative structure. (4) Similarly, jagmi- 'going' (gam 'go, come', 3p1. perf. jagmur) can optionally take an adverbial accusative (5) (Richtungsakkusativ or accusative of goal) specifying the destination of the verbal action:
RV 2.23.11a: ananudo vrsabho jagmir ahavam nistapta satrum pranasu sasahih/asi satya maya brahmanas pate An unyielding bull, approaching the fight, (6) burning down the enemy, victorious in the battles --you are the true avenger of offenses, Brahmakias Pati!
jaghni- 'beating, slaying' (han 'beat, slay, kill', 3p1.perf. jaghnur) provides a clear instance of the modification of this type by an adverb:
RV 9.53.2: aya nijaghnir ojasa, rathasamge dhdne hitel stava abibhyusa hrda Mit diesem (Liede) will ich mit Kraft zuschlagend im Wagenkampf bei ausgesetztem Preise furchtlosen Herzens lobsingen. (Geldner, RV III: 38)
Here, nijaghni- modified by the adverbial instrumental ojasa 'with strength', a clear indication that we are not dealing with a true agent noun, since these are never modified by adverbs, but with a participial form. Compare the use of ojasa with a finite verbal form of han in RV 1.80.2: yena vrtram nir adbhyo, jaghantha vajrinn ojasa. .. "through which you, o cudgel-carrier, have expelled Vrtra from the waters with might." jaghni- is furthermore attested with an accusative object in RV 9.61.20.
In RV 6.23.4 we find the use of dadi- 'giving' (da 'give, donate', 3p1. perf. dadtur; but note that in this case we also find a morphologically corresponding reduplicated present stem), babhri- 'carrying, bearing' (bhr 'carry, bear'), and papi--'drinking' (pet 'drink') with accusative objects:
RV 6.23.4: ganteyanti savana haribhyam, babhrir vajram papih somam dadir gah Coming to so many soma-pressings with his pair of flame-colored horses, carrying a cudgel, drinking soma, giving cows, ... For dadi- we also find three cases of adverbial modification, e.g.: RV 2.24.13cd: viludvesa anu vasa rnam adadih, sa ha vaji samithe brahmanas patih Steadfast in his hatred, (7) taking his dues according to his wish, (8) Brahmaanas Pati is the one who wins the prize in the contest.
Here, anu vasa "according to (his) wish" is a subject-oriented adverbial phrase modifying adadi- (cf. oldenberg 1909: 209). A further indication of the participial status of this form is the incorporation of the preverb [ss]'to' from the underlying syntagma ci da 'take (for oneself)' and the accusative object.
The nine attestations of papuri- and its variant papri- reflect formations to three different Indo-Iranian roots: * parH 'fill' (< PIE * [pleh.sub.1]/[pelh.sub.1]), *par 'help (across), save' (< PIE * per), and * parH 'give, allot' (< PIE * [perh.sub.3], cf. Gk. [??]). Four instances belonging to the last root behave like participles with respect to case, e. g.:
RV 6.50.13ab: uta syd devah savita bhdgo no, 'pam napad avatu paprih Auch der Gott Savitri, Bhaga, Apam ,naben Spendende, sllen uns ihre Gunt schenken. (Ge1dner, RV II: 153)
In RV 4.23.3 papuri- furthermore takes a dative recipient in the phrase papurim jaritre' "(habitually) giving to the singer." If this were an agent nominal, we would expect papurim jaritruh
vavri- 'cover, shell' (vr 'cover, surround, restrain', 3p1. perf. vavrur) is clearly a cakri-type formation that must have been lexicalized with the meaning 'shell' or 'hiding place' (< *'that which covers, covering', see Tichy 1995). We therefore do not expect to find evidence for "verbal" usage (accusative case, adverbial modification, etc.); this is in fact the case, except for the following passage:
RV 1.54.10cd: abhim indro nadyo vavrina hitd, visva anustah pravanesu jighnate Indra bekampft alle von dem Einsperrer der Fliisse (9) gemachten Anstalten in den Strom-gefallen. (Geldner, RV 1: 71)
Geldner takes vavri- to have the meaning 'restraining, restrainer' in this passage, which would then provide an instance of the original meaning of this formation, in contrast to the lexicalized meaning 'shell'.
One more case should be mentioned: 'anasi- 'reaching' (as/ams'reach') in vyanasi- 'reaching, penetrating' (3p1. perf. anasur). The paradigmatic differentiation of the root (see [LIV.sup.2]: 282f.: * [[h.sub.2]nek]) and its perfect stem in Vedic indicate that this must be a comparatively late formation. Kummel (2000: 284ff.) posits two synchronic perfect stems, (RV-YV) and anams-lanaf- (RV+). The latter prevails already in the Rigveda and is presumably the younger formation. Since almost all of the other reduplicated i-stems are synchronically aligned with the weak stem of the corresponding perfect, we can safely assume that vyanasi-is an inner-Vedic formation based on the weak stem of prevalent anams-lanas-. The fact that it is attested with an accusative object implies that this syntactic behavior was perceived as a core property of the type, even for new formations:
RV 3.49.3ab: sahava prtsu taranir narva, vyanasi rodasi mehanavan Mighty in the fights like a traversing runner, pervading (10) both worlds, full of generosity ...
The other instances of reduplicated i-stems with participial behavior are dadhi- 'placing, creating' in RV 10.46.1c, sasni- 'winning' (with both an accusative object and modification by the temporal adverb dive-dive 'daily' in RV 9.61.20), and sasri- 'running', which takes a locative "argument" in RV 10.99.4ab gosu (. . .) pradhanyasu sasrih "running for the cows that constitute the prize." Finite formations of the root sr also take locatival arguments/adverbials, the syntagm meaning 'running because of/for/in order to reach (sth.)'. As in the case of jagmi-, the locative is probably to be analyzed as an event-delimiting adverbial rather than a real argument, but it is instructive that the verbal adjective preserves the event structure of the verbal root it is derived from.
To summarize, of the Rigvedic reduplicated i-stems roughly a third are attested with structural (i.e., accusative) objects and/or adverbial modifiers (see Table 2 in the appendix). These syntactic properties indicate that they are not agent nouns but deverbal nominalizations comparable in syntactic behavior to English "ACC-ing" (11) I I nominalizations and Vedic active participles. In the few cases where we find genitive rather than accusative objects, we are dealing with substantivizations of such formations--this holds, for example, for 'sasahi- 'victor' in RV 10.166.1a or tutuji- 'inciter' in RV 10.22.3a. Note that in the majority of attestations, the internal argument of the verb underlying the nominalizations remains unexpressed, so that in predicative use these cases are ambiguous between adjectival and substantival use.
[section]3. In this section I discuss the temporal and aspectual behavior of the cakri-type in comparison to the corresponding perfect stem. If there is a synchronic derivational link between the reduplicated i-stems and the weak stem of the perfect, one would expect the resulting nominalizations to have the same verbal semantics as the corresponding perfect stems. To be more precise, we expect them to be roughly equivalent to active perfect participles. In the following I will show that this only holds for a subset of the attested Rigvedic formations, which I will refer to as the "form-meaning match" type. On the other hand, there is a significant subgroup whose verbal semantics do not correspond to that of the associated perfect stem. I will call this group the "form-meaning mismatch" group. Finally, there are forms that do not formally correspond to a synchronic perfect stem or that are ambiguous.
[section]3.1 I will start by discussing the cakri-formations that correspond to what is expected of a perfect participle, i.e., the group in which there is a formal and functional correspondence between the two categories. These are the following:
(1) Form-meaning match
anasi- 'reaching': 3p1.perf. anasur 'reach (up to), own sth.' (pres.); 'have reached sth.' (perf.)
tatrpi- 'satisfying': 3pl.perf tatrpur (AVS) 'are satisfied', perf.ptcp. tatrpana- 'satisfied'
tutuji- 'hurrying': perf.ptcp. tutujana- 'hurrying'
[dadhryi-.sup.12] 'brave': perf. stem dadhrs- 'be brave', perf.ptcp. dadhrsvams- 'brave'
yuyudhi-lydyudhi- 'pugnacious': perf. stem yuyudh-l*[yriyudh-.sup.13] 'are fighting' (pres.), have fought' (perf.)
sasri- 'running': perf. stem sasrur '(have) run', perf.ptcp. sasrvams- 'having run; running'
sasahi- 'victorious': pert stem sasah- 'be victorious', sasahvams- 'victorious'
Two more stems presumably also belong here: jajni- 'knowing' (on which see Oldenberg 1912: 274; EWA I: 599ff.; Tichy 1995: 280) and vavri- 'cover'. The former is attested only once, in RV 10.71.9, in the nominative plural aprajafnayall, which supports the interpretation 'ignorant, inexpert(ly)', thus from jna 'know'. However, later attestations of jajna- seem to mean 'sprout, grow' (e.g., jdjni bijam '(well)-sprouting seeds', TS VII 5.20) and probably belong to jan 'beget' (Tichy 1995: 67). The obvious explanation is that we are dealing with two formally identical formations from two different roots, presumably not belonging to the same period (RV vs. TS), so AiG 11,2: 292. The Rigvedic instance of corresponds exactly to the meaning of the perfect participle jajnivams- in RV 3.2.11 (Kummel 2000: 206).
In the case of vavri- the corresponding perfect participle has an imperfective, present-like meaning (cf. Kummel 2000: 460f.) in the same contexts as vavri- in the Indra myth, e.g., RV 4.16.7a apo vrtram vavrivamsam parahan "you beat away Vrtra, who was obstructing the waters" (cf. RV 1.54.10 nadya vavritja "(by) the one restraining/obstructing the rivers" cited above).
As is immediately clear from these forms, the cases in which the reduplicated i-stem has the same verbal semantics as the formally corresponding perfect stem are those in which the perfect stem is actually used as a present stem--these are the so-called "stative perfects" (naktostativ in Kummel's terminology). In these cases, there is no reason to believe that the derivational basis was anything other than the synchronic perfect stem.
[section]3.2 There are furthermore a number of reduplicated i-adjectives whose verbal semantics are not the same as those of the formally corresponding perfect (participle); that is, the correspondence between the two categories is only formal. It is striking that these are usually also attested with structural case objects and adverbial modification, and a number of them are also found in Avestan (see below). This is also the class where the clearest instances of the iterative and intensive Aktionsart posited by Wackernagel (AiG 11,2: 291) and Tichy (1995: 236ff.) are found.
Starting with cdkri- 'making', this form usually characterizes an imperfective verbal action, often with habitual, durative, or iterative connotations, e.g., RV 9.88.4 (cited above) maha karmani cdkrih "accomplishing great deeds" (something that Indra does habitually). Another clear instance of a repeated, habitual action is RV 6.24.5 also already mentioned above:
RV 6.24.5ab: anyad adya karvaram anyad u svo, 'sac ca san muhur acakrir indrah One deed today and another one tomorrow; (thus) Indra instantly turns that which is not into that which is.
The object distributive use underlined by anyad--anyad makes it clear that repeated actions are involved: these are atelic with respect to the point of narration. Contrast with this the use of the perfect active participle of kr:
RV 6.17.13: eva ta visva cakrvamsam indram, maham ugram ajuryam sahodam/suviram tva svayudham suvajram, a brahma navyam avase vavrtyat So moge dich, der dies alles getan hat, den gro[ss]en, gewaltigen, alterlosen, siegverleihenden Indra, dich, den Tapferen, mit schoner Waffe, mit schoner Keule, das neue Kraftlied zur Gunster-weisung herbringen." (Geldner, RV II: 114)
The verses preceding this passage describe Indra's heroic deeds, and the choice of the perfect participle (ta visva cakrvamsam) expresses the completion of these deeds.
As Kummel (2000: 137ff.) notes, the perfect of kr is one of the prime examples for ver-gangenheitsbezogenen use of the perfect stem (both in the indicative and the participle) and one of the best-attested resultative perfects in the Rigveda. But neither the aspectual nor the Aktionsart behavior of the perfect stem of this root corresponds to that of cakri-, despite the superficial morphological equivalence.
The situation is similar for jdgmi- 'going'. In RV 2.23.11 (see above), Brahmapaspati is compared to a bull, with jagmi- describing a habitual activity of the bull, in coordination with the likewise habitual agent nominal nis-taptar- (see Tichy 1995: 245). Again, note the contrast with the synchronic perfect active participle:
RV 3.38.6cd: apayam atra manasa jaganvan, vrate gandharvam api vayukesan Ich sah, im Geiste dorthin gegangen, auch die Gandharven, deren Haare der Wind sind, in eurem Dienste. (Geldner, RV I: 380)
The perfect participle describes a completed action in contrast to the use of the imperfect in the main clause. Although Kummel (2000: 155ff.) points out that the perfect of gam is often used as a naktostativ perfect ('having arrived'[right now] 'being there'), which is imperfective, the use of jagmi- with an accusative object does not correspond to this function either ('going towards' vs. 'being at'). Iterative use of jagmi- occurs in the following passage:
RV 1.89.7ab: prsadasva marutah prsnimatarah, subhamyavano vidathesu jdgmayah Die Marut mit scheckigen Rossen, die Sohne der Mutter Prisni, die prunkvoll ausfahrenden, die gern zu den weisen Reden kommen, ... (Geldner, RV I: 114)
The plural locative goal requires a reading where the Maruts arrive repeatedly at different vidathas.
jaghni- 'beating' is used in RV 9.61.20 to designate an inherent characteristic trait of soma: RV 9.61.20: jaghnir vrtram amitriyam, sasnir vajam dive-dive / gosa u asvasd asi Slaying the hostile Vrtra, winning the prize day after day, you are the winner of cows and horses
All three descriptive characteristics (slaying Vrtra, winning prizes, winning cows and horses) are clearly generic properties of soma. The iterative-habitual semantics are strengthened by the adverbial phrase dive-dive and anchored in the present (... asi; see Tichy 1995: 240). This contrasts with the resultative, completive use of the synchronic perfect participle, an example of which is given in the following passage:
RV 4.18.7cd: mamaitan putro mahata vadhena, vrtram jaghanvam asrjad vi sindhun Mein Sohn hat diese Flusse laufen lassen, nachdem er mit der gro[ss]en Waffe den Vritra erschlagen hatte. (Geldner, RV I: 442)
The perfect participle is again perfective with respect to the verbal action of the main clause.
In RV 6.23.4 (see above), the three reduplicated i-stems in babhri'r vdjram papih soman dadir gah characterize habitual actions performed by Indra. As Tichy (1995: 237) points out, the reduplicated i-stems in this passage display the same syntactic behavior and are used in similar contexts as the root-accented agent nouns in -tar- (e.g., datar- '(habitual) giver, donor', etc.), which, according to her analysis, are likewise used to designate the agents of repeated, habitual actions. The perfect participles of pa 'drink' and bhr 'carry', on the other hand, have different semantics (for da see below). The perfect indicative of pa is always resultative (Kummel 2000: 308f.); the participle always designates a perfective action, as in RV 2.11.10 where the completion of the soma-drinking (papivan sutasya) is a precondition for the verbal action of the main clause:
RV 2.11.10cd: ni mayino danavasya maya, apadayat papivan sutasya Er brachte die Zaubereien des zauberischen Danava zu Fall, nachdem er Soma getrunken hatte. (Geldner, RV I: 288)
The same holds for the perfect stem of bhr (Kummel 2000: 338ff.). Only the perfect middle participle is attested in the Rigveda, but again we see an opposition to the habitual, imperfective semantics of babhrl-:
RV 3.1.8ab: babhranah suno sahaso vy adyaud, dadhanah sukra rabhasa vapumsi Ausgetragen bist du Sohn der Kraft aufgeleuchtet, lichte, grelle Farbenpracht annehmend. (Geldner, RV I: 333)
The perfect participle designates an action that is completed by the time of the action of the main clause (vy adyaut).
Like jagmi-, yapi- 'going, hurrying' (ya 'drive, move') formally belongs to a resultative perfect (Kummel 2000: 409). The finite forms mean 'to have driven, gone (up to) somewhere'. Contrasting with this is the imperfective, durative use of yayi-:
RV 5.73.7ab: ugro yarn kakuho yayih, srnve yamesu samtanih Your strong draft horse is moving; its clatter is heard on the tracks.
The use of the present middle srnve indicates that the movement is ongoing at the time the noise is heard (srnve ... samtanih). The perfect participle yayivams- is attested only once in a difficult passage (RV 9.15.6, see Oldenberg 1912: 157f.), where it seems to have the same resultative sense as the finite stem (yayivam' ati "having gone through").
A less certain example might be provided by vivici- 'distinguishing' ([vic.sup.14] 'separate, sieve, shake'), attested twice (RV 5.8.3 and 8.50.6). The corresponding perfect stem is attested only once in the Rigveda, in the nominative singular of the perfect participle vivikvams- in RV 3.57.1, where it is used as a resultative ('having chosen', Kummel 2000: 496). In RV 8.50.6, on the other hand, vivici- is used to describe a general, habitual characteristic of Indra and does not correspond to the use of the perfect participle.
RV 8.50.6ab: pra viram ugram vivicim dhanasprtam, vihutim radhaso mahah Den gewaltigen, wahlerischen Helden, den Schatzegewinner, der gro[ss]er Gabe machtig ist, (lobe ich). (Geldner, RV II: 372)
Since there are very few examples it is difficult to draw any conclusions about the semantic relationship between the participle and the i-stem, but the two could potentially instantiate another case of a form-meaning mismatch.
In the case of sasni- 'winning' (san `win'), one passage favoring a durative, iterative reading has already been given (RV 9.61.20). An iterative interpretation is also suggested by the following passage, in which Indra and Agni are described as sasni with respect to different prizes (vajesu) and sacrifical actions (karmasu), implying repeated actions:
RV 8.38.1: yajnasya hi stha rtvtja, sasni vajesu karmasul Indragni tasya bodhatam You two are the priests of the sacrifice, repeatedly winning (with respect to) the prizes and the actions (of the sacrifice). Be mindful of this, Indra and Agni!
In the passages where sasni- designates the habitual winner, its use is somewhat closer to the use of the synchronic perfect participle sasavams-, which usually means 'having won', hence 'victorious one, winner' (Kummel 2000: 551ff.). But even in cases in which the meaning of sasavams- is close to that of the stative perfect participle sasahvams- 'victorious', it is still different from stisni- 'repeatedly winning' above.
Summing up, the stems discussed in this section do not correspond to their synchronic perfect stem in their verbal semantics. The differences between the (non-finite) perfect and the corresponding i-adjective are summarized in the following table:
(2) Resultative perfects and reduplicated i-stems Perfect participle Reduplicated i-stem Aspect serfective imperfective Tense past, present sometimes present Aktionsart unmarked iterative, durative, habitual, sometimes intensive
[section]3.3 There are a number of reduplicated i-stems for which no synchronic perfect stem is attested or which have a reduplicated verbal stem different from the perfect as their derivational basis; that is, there is no formal correspondence between the two categories. This holds for cacali- 'swaying, staggering', in avicacali- 'not swaying', e.g., RV 10.173.1 dhruvds tisthavicacalih "stand firm without swaying!" (the second Rigvedic attestation is in 10.173.2), made from the root cal, a variant of car 'move'. There is no perfect attested for this root, and both its phonological make-up and its occurrence within the Rigveda (twice in the tenth book) point to a relatively late formation (cf. [section]5.2 below on the long-vowel reduplicated i-stems). The first two verses of the hymn, which contain the two attestations of avicacali-, are paralleled by AV 6.87.
jaguri- in RV 10.108.1b dare hy adhva jagurih paracaih, which Geldner translates as "Der Weg so weit in die Ferne ist ja aufreibend," may be derived from the roo *gr 'be/make heavy, tired' ([??], the root of Ved. gravan- 'stone'), gr 'devour' (Renou, EVP 16: 161), or may even belong to glayati 'is exhausted' (Werba 1997: 403; KEWA III: 699), which also has the right root shape ([??] vel sim.). In either case, there is no synchronic reduplicated verbal formation attested for either *gr or gla (the latter has a post-Vedic perfect) that could be suspected as a derivational basis. On the other hand, gr 'devour' (< PIE [??]) has a Vedic perfect jagera; a pre-Indic formation *ja-grTH-i- formally based on this stem would indeed give jaguri-. adhva jdgurih would then have to be translated as "devouring path" and would be another instance of a form-meaning mismatch between the i-stem and the perfect stem, which is always resultative (Kummel 2000: 162f.). (15) The root jr 'grow old' favored by Oldenberg (1912: 331) can be excluded on formal and semantic grounds.
For jaghri- 'dripping, splashing' (ghr 'drip', see Grassmann, WB: 464; Geldner, RV I: 223f.) there is no perfect stem attested in the Rigveda and the reduplicated present jigharti does not formally correspond to the i-stem (for a possible cognate in Avestan see below). The form is a hapax in the Rigveda, attested only in RV 1.162.15a mokha bhrajanty abhi vikta jaghrih "may the glowing dripping pot not fall over." Renou (EVP 16: 86) suggests a connection with the root of gharma- 'heat'; both 'sprinkling, dripping' and 'becoming hot' would be fitting modifiers of the pot in this passage. Both roots are an if, so there is no easy way to decide between them based on this passage alone. (16)
In the case of taturi- 'conquering, victorious' (tr 'conquer, cross (over), overcome'), the i-stem does not formally correspond to the attested perfect tataraltitirus; it looks like an old formation with the expected development of an Indo-Iranian laryngeal, as if from *ta-trH-V ([??], cf. EWA I: 629ff.), cf. papuri- and jaguri-. That the etymological connection with rt. was still clear is shown by RV 6.68.7d prd sadyd dyumna tirate taturih "augmenting (its) radiance instantly (as one who is) conquering" (cf. Renou, EVP 5: 98; 7: 80). The semantics of the perfect participle and the reduplicated i-stem, attested five times in the Rigveda (always without expression of its internal argument), partly overlap: titirvams- is used in the vocative meaning '(habitual) conqueror', like taturi- in RV 4.39.2, 6.22.2, and 6.24.2. However, the perfect participle is usually completive-resultative (Kummel 2000: 214), which taturi- never is. This form could therefore be interpreted as belonging to the form-meaning mismatch group, albeit without the formal correspondence to the synchronic perfect stem.
There is no synchronic perfect attested for the root of yuyuvi--'keeping away, restraining sth./sbdy.' (yu 'separate, keep away, restrain'), but there is evidence that the reduplicated present yuyoti actually continues an older perfect stem yayav-[left arrow] yuyav--(Jamison 1983a: 174f.; Kummel 2000: 401 ff.) from which the i-stem may have been derived. On the other hand, the stem seen in the 3sg.inj. viyuyot is equally likely to have provided the derivational basis. This form is usually analyzed as a causative aorist (e.g., Hoffmann 1967: 90), but it is unclear whether this was the synchronic function of the stem (Kummel 2000: 404). At any rate, the figura etymologica in the following passage shows that yuyuvi--was interpreted as derivationally related to yu:
RV 5.50.3cd are visvam pathestham, dviso yuyotu yuyuvih Let the expeller expel everyone who is standing in the way, the enemies, far from here! (17)
There is also no clear derivational basis for vavahi--'moving quickly' (vah 'go, [??] a wagon, lead') in the Rigveda. The corresponding perfect stem has undergone [??] remodelling (uvaha/uhur), but the form was clearly not derived from the expected [??] of the perfect stem ([??]) either. Moreover, the synchronic perfect is always completive-resultative (Kummel 2000: 490), whereas veivahi--is imperfective, possibly with intensive connotations:
RV 9.9.6ab: abhi vahnir amartyah, saptd pasyati vavahih Das unsterbliche Wagenro[ss] uberschaut raschfahrend die sieben Strome. (Geldner, RV III: 16)
Structurally and semantically, vavahi- resembles a "type I" intensive stem (see Schaefer 1994: 25) like papaj- 'become solid, stand still', nanad- 'roar', etc., which is undoubtedly what influenced Geldner's translation of the word. The form is most likely based on other cdkri-formations with a long reduplication syllable (sasahi-, etc.). I will return to these below.
The root of susvi-'pressing (soma)' (su 'press') makes a perfect susava, 3p1. susuvur. The 3p1. suavati in RV 2.16.5 indicates that the use of this stem as a stative perfect led to the (later regular) remodelling of the "present" perfect to a reduplicated present (Kummel 2000: 557). Formally, the latter might be the derivational basis of susvi-. Both the active participle susuvams- and the middle participle susvana- are used in a resultative sense, which can become stative in the meaning 'having pressed soma (and now keeping it ready)' (Kummel, ibid.). On the other hand, sasvi- designates someone who habitually presses soma. This is particularly clear in RV 4.25.7 where suavi- is opposed to the (negated) present participle of su-, sunvant-:
The hapax sisvi- 'growing' (su 'swell, grow (strong)') is attested in the compound susisvi-'growing well' (RV 1.65.4). The corresponding perfect stem differs formally (3p1. susuvur, participle susuvams-) and semantically from the i-stem in that the perfect is used as a stative present ('be strong') while sisvi is durative and telic. The i-reduplication and the unexpected weak stem allomorph -sv- make it likely that the derivational basis of this form was the u-stem sisu- 'child' (< 'growing (one)'), which is itself morphologically problematic (EWA II: 641).
RV 4.25.7: na revata panina sakhyam indro, 'sunvata sutapah sam grnitel/ asya vedah khdati hanti nagnam, vi susvaye paktaye kevalo bhut Indra schlie[??]t keine Freundschaft mit dem reichen Knauser, der keinen Soma auspre[??]t, er der Somatrinker. Er zwackt ihm die Habe ab, erschlagt ihn splitternackt. Einzig fur den Somapressenden, fur den Kochenden (18) ist er zu haben. (Geldner, RV I: 453)
This use is easily understandable if the remodelled "present" susvati was the derivational basis of the i-stem.
[section]3.4 Finally, there are two cases in which either the perfect stem or a reduplicated present stem could formally be the derivational basis of the attested i-stem, making these formations ambiguous: dadi--'giving' (da 'give') could formally belong to either the present dadati, 3p1. dadati, or to the perfect dadau, 3p1. dadati, or Kummel (2000: 240ff.) points out that the perfect is resultative-completive, but also that the perfect participle dadvams--has become lexicalized in the meaning 'donor', which is also the use of dadi- in a number of instances (RV 1.15.10, 1.110.7, 2.37.2, 8.21.6, etc.). In other cases, repeated actions at the point of narration are implied, e.g.:
RV 1.81.7ab: made-made hi no dadir, yuthd gavam Rjukratuh During every intoxication the right-minded one gives us herds of cows.
The adverb made-made underlines that these are repeated single actions, a feature that does not square well with the use of the perfect stem. However, dadi--is equally likely to have been derived from the reduplicated present stem and therefore does not necessarily represent a form-meaning mismatch.
As in the case of dadi-, the root of dadhi--'placing' (dha) also makes both a reduplicated present (dadhau, 3p1. dadhati) and a perfect (dadhau, 3p1. dadhur). The perfect is always resultative-completive (Kummel 2000: 270), but the active participle is not attested in the Vedas. The accent of dadhi--might be interpreted as pointing to the present stem, but since the core forms of the reduplicated i-stems are barytone (cakri-, jagmi-, jaghni-, etc.) this is not necessarily a strong argument for a present-stem derivative.
[section]3.5 To sum up this survey of the Vedic material, we have seen that there is a core set of forms that take structural accusative case objects or adverbial "objects" and are modified by adverbs (for a summary of the syntactic properties see Table 2 in the appendix). This means that this nominalization type is at least synchronically not a "real" agent nominalization. Agent nominals typically do not assign structural case and can only be modified by adjectives. Syntactically, the type patterns instead with deverbal participles and so-called "mixed nominalizations" of the English ACC-ing gerund type, which can assign structural case, combine with overt tense and aspect marking, and be modified by adverbs. (19) The cakri-formations differ from other Vedic participles, however, in that they are not integrated into the averbo of a particular tense-aspect stem, a point that will be taken up below.
Concerning the semantics of the type, I have argued that there is a subgroup of forms that are formally associated with the weak stem of the perfect, but do not behave functionally like a participial formation of the synchronic perfect stem. This is the group discussed in [section]3.2, where I argued that the reduplicated i-stem has imperfective, present-tense-like meaning and that its Aktionsart is iterative (e.g., RV 6.24.5, 9.61.20, 1.81.7), durative (e.g., RV 5.73.7), or habitual (e.g., RV 4.25.7, 6.23.4, 9.88.4), as opposed to the corresponding perfect participle, which is resultative-completive with respect to aspect and unmarked for Aktions-art. Instances of intensive meaning (AiG II, 2: 291) are arguably found in derivatives of verbs of motion, e.g., jagmi--in RV 2.23.11, yayi--in RV 5.37.7, tutuji--in RV 10.35.6 (see Tichy 1995: 241). etc.
Note that instances of iterative and/or habitual use are also found for some of the i-stems that formally correspond to stative perfects or arguably not to perfect stems at all, as discussed above. I quote the following example and translation from Tichy (1995: 240f.):
RV 8.46.15: dadi reknas tanve dadir vasu, dadir vajesu purhuta vajinam/nunam atha Du Vielgerufener, der immer wieder unserem Leib Besitztum schenkt, der immer weider Gut schenkt, der in den Wettrennen immer wieder ein sieggewohntes (Rennpferd) schenkt--auch jetzt.
The temporal adverbial phrase nancam atha emphasizes that these repeated actions extend to the speaker's present. Such instances confirm that this type was perceived as marked for a special Aktionsart--iterative, habitual, or intensive--independently of the semantics of the formally corresponding perfect stem.
In the next section, I will discuss the Avestan forms corresponding to the Vedic reduplicated i-stems.
[section]4. Although the cakri-type is much more sparsely attested in Avestan, it displays the core features that have now been established for its Vedic equivalent. The exact cognate of Ved. cakri--is the OAv. hapax caxri--(cf. YAv. 3p1.perf. caixrara). It is attested in a somewhat difficult passage:
Y.34.7: kuvra toi aredia mazda yoi vayheus vaedena manayho senghus raexana aspancit sadracit caxaiio usauru
Most translations (Bartholomae AIW: 576; Insler 1975: 223; Humbach 1991: I: 141; Kellens-Pirart 1991: III: 117) agree that caxri governs a double accusative in this passage, i.e., "turn/make X into Y," with different interpretations of the constituencies of X and Y. Skjaervo (2002: 52 n. 73), on the other hand, takes only sanghus to be the direct object of caxraiio: "Where are those heavenly arbiters, o Mazda, who by the possession of good thought, / ever and again make (caxraiio) the 'statements'! censures(?) (sanghus): '(These) * pittances (raexana) are non-life-giving indeed (aspancit), grievous indeed (sadracit): " If Skjaervo is correct, the passage does not provide a double accusative parallel with Vedic. It does, however, parallel Vedic cakri--in that it takes an accusative object (sanghus), comparable to the use of cakri--in, e.g., RV 9.88.4a maha karmani cakrir "(repeatedly) performing great deeds" or RV 3.16.4a cdkrir yo visva bhavand "who (constantly) makes/creates all beings." We can therefore conclude that the adjectival use of cakri-lcaxri-, reminiscent of that of an (iterative/intensive) imperfective participle which can assign structural case, is inherited from Proto-Indo-Iranian.
The superlative vijaymista 'going in all directions the most' presupposes a * jaymi--cognate with Ved. jagmi--and formally corresponds to the synchronic perfect stem seen in the lsg. perfect optative jaymiiqm and the feminine perfect participle jaymidi-. It is attested with an accusative object in Yt.1.2 and Yt.1.4 (tat instead of kat in Yt.1.4).
Yt.1.4: tat vispahe ayhaus astuuato mana asti vijaymistam ... this [=Ahura Mazda's name] is what out of the whole material world spreads out most to the thoughts. (20)
A similarly built superlative is the Young Avestan form jaynista- 'who slays most', attested in Yt.11.3, Yt.12.7, 8, and 171.7 (always with an accusative object), which presupposes a *jayni- cognate with Ved. jaghni-:
Yt.11.3: sraoso asiio driyum vrato.tamo ho varadraja drujam jaynisto It is the rewarding Sraosa who best protects the poor one, who, overcoming obstacles, beats down the Lie the most. Yt.12.7: rasnuuo taiium nijaynista 0 Rasnu, best smiter of the thief! (or, most often smiting the thief)
In all three passages we have a Vedic-Avestan parallel in the use of this type with an accusative object, as well as a parallel in the preverb selection between Av. nijaynista-: Ved. nijaghni- (cf. RV 9.53.2 above).
As an anonymous reviewer has pointed out, there is an alternative analysis of [degrees]jaymista-and jaynista- as superlatives of athematic stems derived from reduplicated verbal formations, presupposing a segmentation [degrees]jaym-ista- and jayn-ista-. Superlatives to verbal roots are abundant in Vedic (e.g., yajistha- 'who sacrifices most', madistha- 'most intoxicating', gamistha-'who arrives most (as a helper)', etc.) and Avestan (e.g., bairista- 'who carries the most', [??] 'who knows the most', etc.) and are occasionally found in other Indo-European languages as well, e.g., Greek [??] 'best' (= Av. bairigta-; [??] 'bring'). Furthermore, we occasionally find such formations built to morphologically characterized tense-aspect stems, e.g., Ved. parsistha- 'best at helping across' (s-aorist of pr), YAv. yui[ss]ista- 'best at fighting' (pres.st. darista 'best at supporting' (pres.st. daraiia-); cf. AiG II,2: 446f., AIW: 739. However, these are few and relatively young formations, whose synchronic derivational basis is clear. For a segmentation [degrees]jaym-ista- and jayn-ista- to be credible, we would expect also to find the corresponding Indo-Iranian (athematic?) reduplicated verbal stems *ja-gm- and *ja-ghn-. Indo-Iranian *gam `go' made no reduplicated verbal stem other than the perfect, (21) and the same holds for its Proto-Indo-European ancestor *guem, judging from the evidence from the other branches (cf. [LIV.sup.2]: 209f.). This means that in this case the existence of the parallel Vedic formation jcigmi- makes the assumption of an inherited Indo-Iranian reduplicated *i-stem much less costly. That the superlative of such a stem would end in IIr. *-ista- is trivial. Note that this superlative furthermore contrasts with the superlative of the synchronic perfect participle, for example in Yt.11.9 yo asahe jaymustamo "who has best reached order," where the allomorph -tama- rather than -ista- is selected (as expected), quite apart from the by now familiar semantic difference between [degrees]jaym-ista- and jaymus-tama-. In this case, assuming a nominal base *jaymi- seems much less problematic than assuming a reduplicated verbal stem *jaym- distinct from the perfect.
As for jayn-ista-, Avestan actually does have a synchronic verbal stem jayn- distinct from the perfect. This is attested twice, in Yt.13.45 (ni.jagnante) and Yt.13.105 (auua.jagnat). I follow Garcia Ramon (1998) in assuming that Varna! forms an equation with Greek [??] and continues an inherited reduplicated aorist. The present [degrees]jaynante is a back-formation to this, and the same probably holds for the Vedic reduplicated present jighnate. As Garcia Ramon shows, the stem (PIIr.) *ja-ghn- had iterative-intensive semantics at least in Iranian, and it is conceivable that in this case the derivational basis of ni.jaynista- was the reduplicated verbal stem in the syntagm ni.ja.yn- of Yt.13.45 rather than a nominal stem (Ph.) *jayni-. However, the fact that we have already established the equations Ved. cakri-: Av. caxri- and Ved. jagmi-: Av. *jaymi- makes a *jayni- besides Ved. jaghni- very likely, and this is also how Hoffmann (1956: 15 = 1976: 396) interprets these superlatives. Besides, a thematic base would have been more likely to select the allomorph -tama-. (22) I therefore conclude that Ved. jaghni-: Av. *jayni- in all likelihood constitute another word equation.
Young Avestan probably had a stem [degrees]jayri- in the hapax span.jayrim (name of a demon) in V.19.40. It is tempting to analyze the name as a verbal governing compound with a root noun span- (cf. spanta- life-giving, prosperous') as the first compound member and a second compound member cognate with Ved. jaghri- 'sprinkling'. This is the analysis of Humbach (1959: II: 64 and 1991: II: 172, followed by Kellens 1974: 154), who translates the name as 'spattering prosperity'. The Daeuua in question is therefore thought to be "destroying prosperity by scattering water." While a stem *jayriia- is theoretically also possible, it is highly unlikely that such a stem would occur as the second compound member in this kind of compound. Proto-Indo-European *-iio- is well attested in second compound members of so-called derivational compounds ("Risch-compounds") made from nominal and prepositional phrases, e.g., Ved. dasa-masya- 'ten-monthly', Gk. [??] 'from the same womb', [??] 'in the sea', etc. (cf. AiG III: 106ff., Malzahn 2010). On the other hand, if [degrees]jayrim is indeed derived from the Proto-Indo-Iranian root *ghar 'drip, sprinkle', we are dealing here with a verbal governing compound. In Vedic, -ya- is furthermore highly productive as a gerundive suffix, but formations to morphologically characterized verbal stems are rare (AiG II,2: 794f., e.g., Ved. carkrtya- `to be praised'). While a (transitive and intransitive) present-participle-like use of ya-formations is possible in Vedic (AiG 11,2: 801f.), it is rarer than the default gerundive meaning ('to be x-ed'), and the same seems to be true in Avestan (cf. Av. [degrees]kairiia- 'to be done', karsiia- 'to be plowed', isiia- 'to be desired', etc.). That a compound with an Hr. *ya-stem as its second compound member would have the verbal governing compound-like meaning 'spattering prosperity' is highly unlikely. It seems that we can therefore posit a stem *jayri- for Avestan, cognate with Vedic jaghri-, based on what we know about Indo-European compound formation.
In V.18.65 we find the compound azro.daiaim, the second compound member of which could in principle be cognate with Ved. dadhi-:
V.18.65: [??] ... yava va vahrkqm azro.dai[ss]im gaevqm auui frapataiti More deserving to be killed ... than the she-wolf who gives chase and attacks the herd.
Bartholomae (AIW: 229) equates the second compound member with Ved. dadhi- (dha), but da is in principle equally possible. As for the first compound member, he suggests a comparison with Gk. [??] 'hunt' (Gk. [??], Av. az-) and translates the compound as "Jagd machend, auf Jagdbeute, Raub ausgehend," fitting the description of the she-wolf in this passage. However, a first compound member Av. azra-: Ved. ajra- 'open fields' would also work (thus Geldner 1882: 51f.; Mayrhofer 1985: 167). As for the second compound member, Tremblay (1998: 114), following Mayrhofer (ibid.), points out the possibility of a reduplicated formation of the root *da 'trace, track down' posited by Narten (1963). This would make the wolf 'the one roaming the open fields', an equally fitting epithet. Thus [??] may well be a reduplicated i-adjective to an Iranian root *da, but it seems impossible to decide which one is correct.
The Young Avestan nom.pl. dadaraiio belongs to dar 'hold, sustain', reflecting an older *da-d(h)r-i/ai- with subsequent a-epenthesis in the second syllable. (23) It is attested twice in N.96:
N.96 (2): yezi asparano vastrahe ai[ss]iiastam dadaraiio... If they possess a full set of clothes (to wear with) the sacred girdle... (24) N. 96 (3): yezi aat asparano vastrahe *ai[ss]iiastam dadaraiio... But if they do not have a full set of clothes (to wear with) the sacred girdle...
It is easy to see how the meaning 'owners, possessors' could develop out of a participial formation meaning '(habitually, constantly) holding'. The text is not very reliable, but if this interpretation is correct, this form could provide an interesting counterpart to Ved. dadhrvi-'firm, supporting' formed from the same Indo-Iranian root *dhar. While dad(a)ri- has the imperfective, iterative semantics often associated with reduplicated i-stems (from which the habitual agent meaning could easily develop), Ved. dadhrvi- with its stative semantics is more firmly embedded in the perfect system of the same root, virtually suppleting the perfect active participle in -vams-/-us-. This case can be compared to Av. jarauru-Ned. jagrvi- (25) 'awake, vigilant' (perf. jagara 'is awake'), where the semantics of the "u-i-participle" clearly indicate its derivation from the perfect stem, and Ved. 'shining' (di), where the corresponding perfect also has the semantics of a stative present. The pair Av. dadaraiio vs. Ved. dadhrvi-, in which only the latter corresponds semantically to the synchronic perfect stem, confirms the intuition that the derivational basis of the reduplicated i-adjectives was not the synchronic perfect stem.
0Av. manarois, genitive singular of a stem manari-, probably reflects a dissimilated *ma-mri-, but the passage in which it is attested is problematic:
Y.48.10: kada mazda manarois naro visante When, o Wise One, will (some) honorable persons take up their positions side by side with the reciter? (Humbach 1991: I: 178)
Humbach (1991: II: 203) interprets the form as derived from mar- 'remember', thus 'the one who constantly remembers, memorizes'. (26) Narten (1986: 277) points out that a derivation from mar- 'die' is equally possible, in which case the passage would have to be translated "When ... will the heroes take up position at the side of the mortal?" In that case we would have one more Vedic-Avestan cakri-type word equation, since AV 8.2.26 has a-mamri-'immortal' (contrast resultative mamrvams- 'having died' (RV+); see Hopkins 1893: 28).
Avestan also preserves a cognate of Ved. vavri- in its older meaning '(constantly) covering' in YAv. [degrees]vaoiri- 'skin, covering', attested in the genitive plural ham.vaoirinqm 'with skin/cream' (said of milk) and us.vaoirinam 'without skin/cream' in V.5.52. The passage gives instructions for the treatment of women after a stillbirth, including types of milk (paiiah-) they are allowed to drink. If [degrees]vaoiri- is indeed to be derived from var 'cover', the context would support a meaning '(milk) covering' [right arrow] 'milk skin' (thus EWA II: 513 and Tremblay 1998: 114. following Bartholomae 1896: 2601.
P 45 contains the form afrauuaocis 'not proclaiming' < *a-fra-ua-uc-i-, where [degrees]uuaoci-could reflect a reduplicated i-stem to the root vac 'speak'. (27) This corresponds to the weak stem of the perfect attested in the OAv. 1 pl. vaoxama , YAv. 3sg.act. vauuaca, mid. vaoce, whose use at least in Old Avestan was almost present-like (Kummel 2000: 659). Note that the i-stem corresponds to the syntagm fra+vac 'speak forth; proclaim' attested elsewhere, e.g., Y.65.9 frauuauuaca 'has proclaimed', Y.19.11 frauuaoce 'has been proclaimed', etc. This formation would then correspond to the Vedic formations discussed in [section]3.2, in which the reduplicated i-stem corresponds both formally and semantically to the attested synchronic perfect stem.
The form [??] in the compound [??] (F0.25a) 'causing panic' vel sim. could reflect a reduplicated i-adjective in its first compound member. Tremblay (1998) suggests a connection with the root of Goth. daufs 'deaf' and Gk. [??] '(create) smoke' and hence a perfect *[??] as the derivational basis of the Avestan form. However, this root is not otherwise attested in Indo-Iranian and the perfect stem is purely conjectural. This makes [??] a very uncertain, though not completely impossible, (inner-Iranian) instance of a reduplicated i-adjective.
There is one more possible Avestan example, namely the Young Avestan genitive plural a-iriricinam 'not leaving behind':
Y.65.7: yo no airiricinqm iririxsaite gaevanqm ... who wishes to abandon the herds of us who are not abandoning. (cf. AIW: 190)
The second instance (also in Y.65.7) has tanunqm 'bodies' (gen.) instead of gaevanam. As for [degrees]iririci-, the context suggests a reduplicated formation of the root iric (raek) 'leave behind, abandon' ([right arrow](pre-)PIIr. */i-likui-). In both instances, the form is associated with the synchronic desiderative stem iririxs- ([right arrow] [??]), but this cannot be the derivational basis because of the missing desiderative suffix in the i-adjective. Even though the perfect stem corresponding to the Ved. 3sg.perf. rireca 'has left' (3sg.opt. riricyat) is unattested in Avestan, it presumably underlies this formation. This means that [degrees]iririci- could be a Proto-Indo-Iranian or inner-Iranian instance of this type.
To conclude this survey of Avestan, we have found four secure word equations with Vedic (caxri-, *jaymi-, *jayni-, and [degrees]vaoiri-), one less secure one ([degrees]jayri-), five probable Iranian-only instances of the type (manari-, [degrees]iririci-, [??], dadari-, and 'uuaoci-), and one very insecure one ([??]). Especially noteworthy is the fact that in the core cases (Ved. cakri-: Av. caxri-; Ved. jagmi-: Av. *jaymi-; Ved. jaghni-: Av. *jayni-) we find instances of the same syntactic behavior, namely, accusative assignment.
As for the contrast with the perfect stem, here too we find instances of the formal-functional mismatch described above for Vedic. Passages such as Yt.10.71 nae[delta]a maniiete jaynuua (28) "and he does not think that he has slain" and Y.22.3 imqmca uruuarqm barasmanqm jaymusimca ratufritim "(I bring) both this Barasma-plant and the Ratu-satisfaction which has arrived" show that the same semantic contrast between the resultative perfect active participle and the imperfective (iterative-intensive) i-adjective must be posited for Avestan, could support such an emendation. The reading mana 'my', on the other hand, poses problems for the translation. I have therefore decided to follow Bartholomae.
I would like to thank Jay Jasanoff, Jeremy Rau, Prods Oktor Skj[??]rv[empty set], Melanie Malzahn, Dieter Gunkel, and Hannes Fellner for commenting on earlier versions of this paper and Emma Shoucair for proofreading it. I am also very grateful to Stephanie Jamison and two anonymous reviewers for many helpful suggestions. All remaining errors are my own even though the material is much sparser than in Vedic. We can therefore conclude that the reduplicated i-adjectives are syntactically participial formations in both Vedic and Avestan that can assign structural case to the internal argument of the underlying verb and take adverbial modifiers, unlike true agent nominalizations or typical verbal adjectives. This is confirmed by the fact that the cakri-type patterns with root-accented tar-nominalizations in Vedic, which also behave like participles in that they take accusative and dative complements, combine with preverbs, and are often used predicatively denoting a habitual agent (Tichy 1995: 237; cf. the discussion of RV 6.23.4 above). The reduplicated i-stems differ both in Vedic and Avestan from "real" synchronic participles in that they are not formally part of any particular tense-aspect stem's averbo.
On the semantic side, we have seen that the reduplicated i-stems behave like imperfective present participles, very often with iterative, intensive, or habitual semantics. In a number of cases, there is a clear contrast between a resultative-completive perfect stem and an imperfective (iterative or habitual) reduplicated i-stem. Although the Avestan evidence is sparse, it confirms the overall picture in the relevant cases, which means that the type is at least of common Indo-Iranian origin.
Based on these findings, I will argue in the next section that although synchronically associated with the perfect stem, the cakri-type cannot have its derivational roots there. I will also show that there is no other synchronic verbal category that could be considered the derivational basis.
[section]5. Three possible verbal categories present themselves a priori as candidates for the derivational basis of the cakri-type: the perfect because of the formal equivalence, the intensive because of the semantic equivalence with at least a subgroup of the cakri-formations, and the (a-)reduplicated present because of the formal equivalence and the non-perfective (though not necessarily iterative or intensive) semantics.
[section]5.1 I have argued so far that for a subgroup of reduplicated Vedic i-stems a derivational basis other than the perfect must be sought. These are the form-meaning mismatch forms (several of which have counterparts in Avestan), e.g.:
cakri- 'habitually/repeatedly making, creating': cakrvams- 'having made, done' jagmi- 'often going; going quickly': jaganvams- 'having reached (one's destination)' jaghni- 'repeatedly/habitually beating': jaghanvams- 'having slain' papi- 'drinking (right now or habitually, e.g., soma)': papivams- 'having drunk' sasni- 'habitually winning': sasavams- 'having won'
On the other hand, we have a group of forms that do correspond semantically to their associated (usually stative present) perfects. This is the "form-meaning match" group discussed in [section]3.1, e.g., tatrpana- 'satisfied': tatrpi- 'satisfying'; sasahvams- 'victorious': sasahi- 'id.'; tutujana- 'hurrying': tutuji- 'id.', etc.
One possible way of deriving the cakri-type from the synchronic perfect stem would be to assume that its starting point was the "match" group. (29) This would easily explain the imperfective semantics, since these perfect stems are functionally presents, and the iterative, habitual, and intensive semantics could have developed out of the present-like use in combination with the particular semantics of the roots in question (e.g., the "intensive" use of verbs of motion, etc.). Taking stative perfects as their basis, these reduplicated i-formations might then have been extended to resultative-completive perfect stems and kept their present-participle-like meaning, even though this did not match that of the verbal bases any more. In this scenario, the "match" formations of [section]3.1 would be the starting point of the type and the "mismatch" formations in [section]3.2 its later extensions.
However, as I have argued above there is evidence that the core forms of the reduplicated i-sterns are actually associated with resultative perfects (cakri-, jdgmi-, jaghni-, etc.), precisely the "form-meaning mismatch" type. The "mismatch" formations are much better attested in the Rigveda than the "match" group (with the exception of sasahi- and the formally ambiguous dadi-), and they have clear cognates in Avestan that have the same syntactic properties (accusative objects, preverb selection) as well as the same semantic properties (durative, iterative, or habitual Aktionsart). The only Vedic-Avestan pair for a reduplicated i-stem of the "match" group is vavri-l[degrees]vaoiri-, which at least in Vedic was lexicalized very early (but see the discussion of RV 1.54.10 above).
Furthermore, recall that even when i-adjectives are actually derived from the perfect stem, iterative or intensive connotations are not necessarily part of the semantics of the resulting formations. This is evident from the Vedic u-/vi-stems already mentioned, e.g.jagrvi- 'wary, alert', perf. jagara; dadhrvi- 'firm, able to support load', perf. dadhara, etc., which do not have the same semantic spectrum as the cakri-type.
This supports the conclusion that the oldest forms of the cakri-type are located in the "form-meaning mismatch" group and that the type can therefore not easily be explained in terms of the properties of the synchronic perfect (stative or otherwise). A different starting point should be sought.
[section]5.2 The Proto-Indo-European intensive as canonized in [LIV.sup.2] is generally assumed to have been of the basic shape [C.sub.1]e[C.sub.2]-[C.sub.1]o/[empty set][C.sub.2]-E, with "full" (or "heavy") reduplication and o/0-ablaut of the root vowel. (30) The situation in Vedic is somewhat more complicated. Schaefer (1994) distinguishes four formally different types of intensive formations. The rather marginal type I, which is characterized by accented long-vowel reduplication (e.g., papaje 'stops, becomes still'), structurally resembles the long-vowel reduplicated perfects (except for the accentuation). There are a number of stems whose classification has traditionally been difficult and that seem to vacillate synchronically between intensives and perfects, suggesting that the perfect/intensive stem distinction was not entirely clear-cut for these cases (see Lubotsky 1997). Some of these roots also make cakri-formations: for tuj, variants with intensive stem accent and with perfect stem accent are found both for the perfect middle participle and for the reduplicated i-adjective. Another potential perfect/intensive overlap form is sasahi- 'winning repeatedly', one of the most frequently attested forms of the cakri-type. (31) This means that for two well-attested formations of the type that show long-vowel reduplication, a formal overlap between the synchronic perfect stem and a type I intensive stem could conceivably have existed at some stage preceding the attested language. An i-adjective derived from such an intensive stem could thus be reinterpreted as having been derived from the synchronic perfect stem. We could assume that at some point it was not clear to speakers that the formal derivational basis of these adjectives was the intensive rather than the perfect stem (type I intensives were rather marginal, after all), leading to an analogical process that substituted the perfect stem as the derivational basis:
(3) tutuj-: tutuji- SaSah-: sasahi- cakr-: x, x = cakri-
There are, however, important arguments against deriving the cakri-type from the synchronic intensive stem in the way sketched out in (3). First, most of the well-attested forms of the cakri-type do not have long-vowel reduplication and are matched by Avestan cognates with the same syntactic behavior that also lack long-vowel reduplication (i.e., the "form-meaning mismatch" group). In fact, the long-vowel reduplicated i-stems are marginal in the Rigveda (except for sasahi-) and are rarely attested with structural case objects or adverbials, the syntactic hallmarks of the type.
Second, in most cases the long vowel in the initial syllable of a reduplicated i-stem was clearly taken over from the synchronic perfect stem. A number of perfects of the "match" group had both long- and short-reduplicated forms in their paradigms, (32) which facilitated the intrusion of long reduplication into the corresponding cakri-type formation. This is the case for tutuji-Itutujana- vs. the weak perfect stem tutuj-, as well as sasahi- (perfect stem sasah-/sasah-/sasah- 33) and the pair yuyudhi-/yuyudhi- (perfect stem yuyudh-l*yuyudh-, see Kummel 2000: 413), where both variants were preserved in the i-adjective. The fluctuation between short- and long-vowel reduplication in the perfect is morphologically conditioned, but in some cases has a diachronic phonological basis, as in the case of Ved. jagdra < *[h.sub.1]ge-[h.sub.1]gor-e or yuyudhi- <*Hiu-Hiudh-[left arrow]Hie-Hiudh-. Long-vowel reduplication is furthermore the rule in the weak stem of certain roots of the (Mr.) shape *vV(R)C, i.e., vah, van, vart, etc.,34 and was analogically extended to other perfects with a light root syllable (Kummel 2000: 21ff.). For most forms of the "form-meaning match" type, the long-vowel reduplication can be traced back to the corresponding perfect stem (this holds for [degrees]anasi-, tatrpi-, tutuji-, yuyudhi-, sasahi-, and maybe dadhrsi-); it then became the preferred pattern for trisyllabic reduplicated i-stems made to roots with a root vowel -a-, -i-, or -u- in the weak stem (cf. cacali-, vavahi-, yuyuvi-). This means that the long-vowel reduplicated i-adjectives are an inner-Vedic development linked with a particular group of perfect stems and do not presuppose intensive stems of the same shape. Furthermore, as we have seen, there are only a few roots where the perfect/intensive stem overlap might have occurred in the first place--not exactly solid ground for the analogy scenario proposed in (3). All in all, it looks as though we must discard the intensive stem as the ancestral category to the aikri-type. (35)
[section]5.3 There is one more possibility, namely deriving the cakri-type from the reduplicated present stem. Synchronically, four of the reduplicated i-adjectives are associated with a reduplicated present stem, namely dadi- 'giving': 3sg.pres. dadati, plural stem dad-; dadhi- 'placing': 3sg.pres. dadhati, 3p1. dadhati; papri- 'helping across': 3sg.pres. piparti, 3p1. piprati (the accent on the reduplicated syllable and the preserved *e-grade in the corresponding cakri-formation could indicate that this present once had PIE *e-/PIIr. *a-reduplication); and susvi- 'pressing': 3pl.pres. susvati, although this is an inner-Vedic innovation (cf. the discussion above).
This approach has several drawbacks: The reduplicated i-stems that could fall into this category do not have equivalents in Avestan (the root of cannot be determined), and, apart from dadi- and susvi-, they are rare in the Rigveda. Furthermore, for three out of the four (dad!-, dadhi-, and papri-) both the present and the perfect stem could be the derivational basis on purely formal grounds (cf. the discussion in [section]3.4). These forms hardly seem an ideal starting point for the type in Vedic, let alone in Avestan.
[section]6. To conclude, there is no convincing verbal derivational basis for the cakri-type formations at the synchronic level in Indo-Iranian. The synchronic association of the reduplicated i-stems with the weak perfect stem must be secondary and based solely on formal identity--which in turn explains why we so often find a "form-meaning mismatch" between the two.
Of course, this finding relegates the problem of the original derivational basis of reduplicated i-stems to the pre-Proto-Indo-Iranian stage. The deeper prehistory of this type cannot be discussed here, but one might speculate that it was originally either deverbal to a pre-Proto-Indo-lranian reduplicated tense/aspect category that was modified or lost in the two branches, or denominal to other reduplicated stems, for instance of the type Ved. cakra-, Gk. 'wheel'; YAv. ba[ss]ra-, Lith. bebras 'beaver'; Gk. rcbtkoc 'garment', etc. (see Oet-finger 2012 for more examples).
On the other hand, it is easy to find parallels for the use of -i- as a participle-like suffix in Vedic. Vedic has a number of unreduplicated deverbal i-stems in the second member of compounds (cf. AiG II,2: 294ff.), and there are a few simplex instances as well. These stems usually appear in verbal governing compounds and in composition with preverbs or adverbial elements. They vary between synchronic full- and zero-grade of the root vowel, with a predilection for the first. Examples of verbal governing compounds of this type include pad-grbh-i- 'grabbing the feet' (PN, grbh 'grab'), go-ddr-i- `splitting cows' (out of rock, said of Indra, dr 'split'), saho-bhar-i- 'bringing strength' (bhr 'carry'), pact-raks-i- 'protecting cattle' (raks 'protect'), and vrsti-van-i- 'winning rain' (van 'win'). In composition with preverbs and adverbs we find, e.g., dur-grbh-i- 'difficult to grab', d-tan-i- 'stretching out (to)' (tan 'span, stretch'), 'procuring through sacrifice; herbeiopfern' (yaj 'sacrifice'), and mahi-svdn-i- 'resounding mightily' (svan 'sound'). The latter group closely resembles the cakri-type in that it also assigns accusative case to its internal argument (if expressed):
RV 2.1.10d: tvam visiksur asi yajnam atanih You are willing to help out, spreading out the saerifice. (36)
The deverbal use of the suffix is a largely inner-Indic development; deverbal i-stems are rare in Avestan. Tremblay (1998: 103) interprets this as an archaic trait of the Iranian branch, which has mostly preserved the (older) denominal use of the suffix. There are some signs of verbal (or at least ambiguous) use, however: uz.daez-i- 'wall, dam' (daez 'pile up') is reminiscent of the Vedic formations cited above, but could of course also be denominal to uz.daeza- 'pile of earth', and baoi[delta]i- 'smell, fragrance' could be a derivative of the verbal root bao[delta] 'sense, perceive' (cf. KEWA II: 449f. and Tremblay 1998: 90f.) or denominal to bao[delta]a- 'smell'. Equations such as Ved. cakri-: Av. caxri-; Ved. Av. vaoiri-, etc., sug-gest that the verbal use of this suffix was at least incipient in Avestan, if not as productive as in Vedic.
The suffix -i- was therefore established as a way of deriving adjectival formations from verbal roots at least in Vedic, in particular in compounds. (37) It is instructive to compare this to the development of the adjectival a- and u-stems within Indic, both of which were originally denominal suffixes that were extended to deriving deverbal adjectives from morphologically characterized tense/aspect stems:
(4) Deverbal a-stems (see AiG II,2: 83ff.):
Intensive: vevija- 'driving': vevijana-; reriha- 'licking': rerihana-; -namnama-'bending': namnamiti; a-dardira- 'crushing': adardiruh, etc.
Future: janisya- (ep.) 'who shall be born': janisyate; bhavisya- (ep.) 'who will be(come)': bhavisyati, etc.
(5) Deverbal u-stems (cf. AiG II,2: 468ff., Rau 1998):
Desiderative: jigisu- 'wanting to be victorious': jigisati; didhisu- 'wanting to obtain': didhisati, etc.
Present: bhindu- 'splitting': bhindati; mandu- 'joyous': mandati; kridu- 'dancing': kridati, etc.
The difference between these formations and the cakri-type is that whereas the former nave clear synchronic derivational bases to which they correspond semantically, no such basis can be established for the cakri-type (at least not for the "mismatch" group, [section]3.2). The i-suffix itself, however, may have had a history similar to the adjectival a- and u-stems, starting out as a denominal suffix and then spreading to verbal roots and eventually tense/aspect stems.
[section]7. To summarize, I have argued that based on their verbal semantics the Indo-Iranian reduplicated adjectives of the type cakri- cannot easily be derived from the synchronic perfect stem, even though they are formally associated with it. This is especially evident for the group of forms attested both in Vedic and Avestan that are synchronically associated with resultative-completive perfect stems (the "form-meaning mismatch" group). Semantically, cakri--formations are imperfective, often with iterative, habitual, or intensive meaning. They are used to designate habitual agents, contrasting with the use of the perfect participles of the corresponding perfects. Since there is no synchronic verbal category that could have provided the derivational basis, the type may ultimately have its roots in a pre-Proto-Indo-Iranian linguistic stage.
The following tables summarize the distributive facts of the cakri--type in the Rigveda. Table 1 gives the total number of attestations as distributed across the ten books (ordered from the most to the least frequent ones). In both tables I have subsumed both papuri- and papuri-under one entry each, despite the fact that they may in fact go back to different roots (see the discussion in [section]2).
Table 1. Frequency of attestation by book (simplex and compound forms) Form I II III IV V VI VII VIII IX X Total dadi- 5 5 1 1 7 1 1 21 Sasahi- 4 2 2 5 1 5 19 vavri- 5 1 2 3 14 susvi- 1 5 1 3 1 11 cakri- 2 1 2 1 1 1 2 10 sasni- 2 1 1 1 2 3 10 jagmi- 2 1 1 1 2 1 8 tuturi- 1 1 5 tutuji- 1 1 2 1 5 Yayi- 2 2 1 5 papuri- 2 1 1 4 anasi- 1 2 3 papri- 2 1 1 1 5 cacoli- 1 2 jaghni- 2 2 dadhrsi- 1 1 2 babhri- 1 1 2 Yuyudhi- 1 1 2 vtvici- 1 1 2 jaguri- 1 1 jaghri- 1 1 jajni- 1 1 tatrpi- 1 1 tutuji- 1 1 dadhi- 1 I papi- 1 1 yuyudhi- 1 1 yuyuvi- 1 1 vavahi- 1 1 sisvi- 1 1 sasri- 1 1 Total 29 13 8 11 10 14 4 10 14 24 144
Table 2 sums up the syntactic properties of the type. [empty set] means that no object is expressed, acc. = accusative object, etc. The last two rows indicate whether the form in question is modified by an adverb or an adjective. This distinction, however, is a less fine-grained diagnostic for distinguishing between agent nouns and verbal participles than the case of genitive vs. accusative objects. Note that the "direct object" -like adverbials of jagmi- are listed under "acc.," whereas sasri- is treated as intransitive with a locative adverbial.
Table 2. Case assignment: simplex and compound forms Form [empty acc. ace. & dat. acc. & gen. Total adv. adj. set] acc. dat. dadi- 7 10 1 3 21 3 2 Sasahi- 17 2 19 (38) vavri- 11 1 1 14 1 susvi- 11 11 1 cakri- 6 1 1 1 1 10 1 sasni- 8 2 10 2 1 jagmi- 4 2 1 7 2 tuturi- 5 5 tutuji- 5 5 1 Yayi- 5 5 papuri- 3 1 4 anasi- 2 1 3 papri- 3 2 5 cacoli- 2 2 jaghni- 1 1 2 1 dadhrsi- 2 2 babhri- 1 1 2 yuyudhi- 2 2 vtvici- 2 jaguri- 1 1 jaghri- 1 1 jajni- 1 1 tatrpi- 1 1 tutuji- 1 1 dadhi- 1 1 Papi- 1 1 yuyudhi- 1 1 yuyuvi- I 1 Vavahi- 1 1 sisvi- 1 1 sasri- 1 1 1 Total 102 21 1 3 4 6 143 7 9
AiG II, 2 = Albert Debrunner. 1954. Altindische Grammatik, vol. II, 2: Die Nominalsuffixe. Gottingen: Vandenhoeck & Ruprecht.
AiG III = Jakob Wackernagel and Albert Debrunner. 1930. Altindische Grammatik. vol. III: Nominal-flexion, Zahlwort, Pronomen. Gottingen: Vandenhoeck & Ruprecht.
AIW = Christian Bartholomae. 1904. Altiranisches Worterbuch. Strassburg: Trubner.
Alexiadou, Artemis, and Monika Rathert. 2010. The Syntax of Nominalizations across Languages and Frameworks. Berlin: De Gruyter.
Arnold, E. Vernon. 1905. Vedic Metre in Its Historical Development. Cambridge: Cambridge Univ. Press.
Baker, Mark C., and Nadya Vinokurova. 2009. On Agent Nominalizations and Why They Are Not Like Event Nominalizations. Language 85: 517-56.
Barschel, Bernd. 1986. Zur Vorgeschichte der altindischen reduplizierten Verbaladjektive auf -i. In Sanskrit and World Culture: Proceedings of the Fourth World Sanskrit Conference of the International Association of Sanskrit Studies, ed. Wolfgang Morgenroth. Pp. 305-10. Schriften zur Geschichte und Kultur des alten Orients, vol. 18. Berlin: Akademie Verlag.
Bartholomae, Christian. 1896. Arica X. Indogermanische Forschungen 9: 252-83.
EWA I = Manfred Mayrhofer. 1992. Etymologisches Worterbuch des Altindoarischen, vol.I. Heidelberg: Universitatsverlag Winter.
EWA II = Manfred Mayrhofer. 1996. Etymologisches Worterbuch des Altindoarischen, vol. II. Heidelberg: Universitatsverlag Winter.
Garcia Ramon, Jose Luis. 1998. Indogermanisch * [g.sup.uh]en 'wiederholt schlagen, toten'. In Mir Curad:
Studies in Honor of Calvert Watkins, ed. Jay Jasanoff et al. Pp. 139-54. Innsbrucker Beitrage zur Sprachwissenschaft, vol. 92. Innsbruck: Institut fur Sprachwissenschaft der Universitat Innsbruck. Geldner, Karl F. 1882. Studien zum Avesta. Strassburg: Trubner.
--. 1886-96. Avesta: The Sacred Book of the Parsis. 3 vols. Stuttgart: Kohlhammer.
Geldner, RV = Karl F. Geldner. 1951-57. Der Rig-Veda aus dem Sanskrit ins Deutsche ubersetzt. 3 vols. Harvard Oriental Series, vols. 33-35. Cambridge, Mass.: Harvard Univ. Press.
Grassmann, WB = Hermann Grassmann. 1872-75. Worterbuch zum Rig-Veda. Leipzig. 6th ed., ed. Maria Kozianka. Wiesbaden: Harrassowitz, 1996.
Gunkel, Dieter. 2010. Studies in Greek and Vedic Prosody, Morphology, and Meter. Ph.D. diss., Univ. of California, Los Angeles.
Hoffmann, Karl. 1956. Notizen zu Wackernagel-Debrunner, Altindische Grammatik II 2. Munchener Studien zur Sprachwissenschaft 8: 5-24.
--.1967. Der Injunktiv im Veda: Eine synchronische Funktionsuntersuchung. Heidelberg: Carl Winter Universitatsverlag.
--.1976. Aufsatze zur Indoiranistik, vol. II, ed. J. Narten. Wiesbaden: Reichert.
Hopkins, Edward W. 1893. Vedic Reduplication of Nouns and Adjectives. American Journal of Philology 14: 1-40+138.
Humbach, Helmut. 1959. Die Gathas des Zarathustra. Heidelberg: Universitatsverlag Winter.
--.1991. The Gathas of Zarathushtra and the Other Old Avestan Texts. In collaboration with Josef Elfenbein and Prods 0. Skjaervo. Heidelberg: Universitatsverlag Winter.
Insler, Stanley. 1975. The Gathas of Zarathustra. Acta Iranica, vol. 8. Tehran-Liege: Bibliotheque Pahlavi / Leiden: Brill.
Jamison, Stephanie W. 1983a. Function and Form in the -aya-Formations of the Rig Veda and the Atharva Veda. Gottingen: Vandenhoeck & Ruprecht.
--. 1983b. Two Problems in the Inflection of the Vedic Intensive. Munchener Studien zur Sprachwissenschaft 42: 41-73.
Kellens, Jean. 1974. Les noms-racines de l'Avesta. Wiesbaden: Reichert.
--. 1984. Le verbe avestique. Wiesbaden: Reichert.
Kellens, Jean, and Eric Pirart. 1991. Les textes vieil-avestiques, vol. 3: Commentaire. Wiesbaden: Reichert.
KEWA II = Manfred Mayrhofer. 1963. Kurzgefasstes etymologisches Worterbuch des Altindischen, vol. II: D--M. Heidelberg: Universitatsverlag Winter.
KEWA III = Manfred Mayrhofer. 1976. Kurzgefasstes etymologisches Worterbuch des Altindischen, vol. III: Y--H. Heidelberg: Universitatsverlag Winter.
Klingenschmitt, Gert. 1982. Das altarmenische Verbum. Wiesbaden: Reichert.
Kotwal, Firoze M., and Philip G. Kreyenbroek, eds. 2009. The Herbedestan and Nerangestan, vol. IV: Nerangestan, Fragard 3. Paris: Association pour l'avancement des etudes iraniennes.
Krisch, Thomas. 1996. Zur Genese und Funktion der altindischen Perfekta mit langem Reduplikationsvokal. Innsbrucker Beitrage zur Sprachwissenschaft, vol. 87. Innsbruck: Institut fur Sprachwissenschaft der Universitat Innsbruck.
Kuiper, F. B. J. 1939. Indoiranica 19: Jaw. vyusa Ha[delta].N. 2, 7 und apa.asavan Yt. 19,84. Acta Orientalia 17: 51-63.
--. 1961. Zur kompositionellen Kurzung im Sanskrit. Die Sprache 7: 14-31.
Kummel, Martin J. 2000. Das Perfekt im Indoiranischen: Eine Untersuchung der Form und Funktion einer ererbten Kategorie des Verbums und ihrer Weiterentwicklung in den altindoiranischen Sprachen. Wiesbaden: Reichert.
Leumann. Manu. 1942. Idg. sk im Altindischen und im Litauischen. Indogermanische Forschungen 58: 1-26.
--. 1952. Morphologische Neuerungen im altindischen Verbalsystem. Mededelingen der Koninklijke Nederlandse Akademie van Wetenschappen, Afd. Letterkunde 15/3. Amsterdam: Noord-Hollandsche Uitgevers Maatschappij.
Levin, Beth, and Malka Rappaport Hovav. 2005. Argument Realization. Cambridge: Cambridge Univ. Press.
[LIV.sup.2] = Helmut Rix. 2001. Lexikon der indogermanischen Verben. Bearbeitet von Martin Kummel, Thomas Zehnder, Reiner Lipp, und Brigitte Schirmer. 2nd ed. Wiesbaden: Reichert.
Lubotsky, Alexander. 1997. Remarks on the Vedic Intensive. JAOS 117: 558-64.
Malzahn, Melanie. 2010. All Indo-European Compounds Are Derived from a Common Origin: New Evidence for a Darwinian View of IE Nominal Compounding. In Proceedings of the 21st Annual UCLA Indo-European Conference, ed. Stephanie W. Jamison, H. Craig Melehert, and Brent Vine. Pp. 183-87. Bremen: Hempen.
Mayrhofer, Manfred. 1985. *[h.sub.l]d-ti- "Speise, Futter" im Iranischen. Munchener Studien zur Sprachwis-senschaft 45: 165-69.
Narten, Johanna. 1963. Ved. abhidasati. Zeitschrift fur vergleichende Sprachforschung 78: 56-63.
--. 1981. Vedisch lelaya 'zittert'. Die Sprache 27: 1-21.
--.1986. Der Yasna Haptanhaiti. Wiesbaden: Reichert.
Oettinger, Norbert. 2012. Das Verhaltnis von nominaler und verbaler Reduplikation im Indoger-manischen und Anatolischen. In The Indo-European Verb: Proceedings of the Conference of the Society for Indo-European Studies, Los Angeles 13-15 September 2010, ed. H. Craig Melchert. PD. 241-46. Wiesbaden: Reichert.
Oldenberg, Hermann. 1909. Rgveda: Textkritische und exegetische Noten. Erstes bis sechstes Buch. Abhandlungen der kg1. Gesellschaft der Wissenschaften zu Gottingen, Philologisch-historische Klassse, vol. 11.5. Berlin: Weidmann. [Rpt. Gottingen: Kraus, 1970.]
--. 1912. Rgveda: Textkritische und exegetische Noten. Siebentes bis zehntes Buch. Abhand- lungen der kgl. Gesellschaft der Wissenschaften zu Gottingen, Philologisch-historische Klassse, vol. 13.3. Berlin: Weidmann. [Rpt. Gottingen: Kraus, 1970.]
Rau, Jeremy. 1998. PIE * uoidu-/* ueidu- and Its Derivatives. Die Sprache 40: 133-60.
Renou, EVP = Louis Renou. 1955-69. Etudes vediques et panineenes. 17 vols. Paris: de Boccard.
Schaefer, Christiane. 1994. Das Intensivum im Vedischen. Historische Sprachforschung Erganzungsheft, vol. 37. Gottingen: Vandenhoeck & Ruprecht.
Skjaerv [empty set], Prods Oktor. 2002. Praise and Blame in the Avesta: The Poet-Sacrificer and His Duties. Jerusalem Studies in Arabic and Islam 26: 29-67.
Tichy, Eva. 1995. Die Nomina Agentis auf -tar- im Vedischen. Heidelberg: Universitatsverlag Winter.
Tremblay, Xavier. 1998. Etudes sur les noms suffixaux athematiques de I'Avesta. Diss., Ecole Pratique des Haute Etudes.
Werba, Chlodwig H. 1997. Verba IndoArica. Die primaren und sekundaren Wurzeln der Sanskrit-Sprache, pt. 1: Radices Primariae. Vienna: Verlag der Osterreichischen Akademie der Wissenschaften.
Whitney, William D. 1885. The Roots, Verb-Forms, and Primary Derivatives of the Sanskrit Language. Leipzig: Breitkopf and Hartel.
Wolff, Fritz. 1910. Avesta: Die heiligen Bucher der Parsen. Strassburg: Trubner.
(1.) Barschel does not mention cacali- and sisvi- study, presumably because they do not contribute to the question of the accentuation of the type. He also posits two homophonous stems sasni-, where I assume only one, and includes didivi-, which I exclude because of its composite suffix *-u-i- (although the formation is of course related). I also exclude the form sanasi- 'profitable, rewarding' (15x in the RV), which Barschel (1986: 305 n. 4, following Leumann 1942: 22 n. 1) tentatively interprets as a metathesized form of * sasani- (san `win'). The root actually makes a reduplicated i-adjective sasni-, which must go back to the laryngeal-loss variant (cf. papri- vs. papuri-). At any rate, even without laryngeal loss, a long reduplication vowel would be unexpected. As I argue below, the cases where we find cakri-type formations with a long reduplication vowel are regularly those in which the associated perfect stem also has long-vowel reduplication at least in parts of its averbo (tutuj-, sgsah-, yuyudh-), which is not the case for san. There is also reason to believe that these were inner-Indic formations, so that an alleged * sasani-would be highly unlikely to be an archaism (as proposed by Leumann, ibid.).
(2.) Geldner (RV) actually varies between translating instances of this type as real agent nouns (e. g., "Vollbringer gro[ss]er Taten" in this passage), participles, and finite relative clauses, independent of their syntactic behavior; Renou (EVP) prefers finite relative clauses. In what follows I will translate the nominal complements with the same case as in Vedic unless this makes the result grammatically awkward. Translations are my own unless otherwise indicated.
(3.) Cf. Geldner, RV II: 123.
(4.) Oldenberg (1909: 383) glosses acakrih, as `herbeischaffend', implying a translation "procuring the unreal and the real," but the double accusative clearly fits better (thus also Grassmann, WB: 172).
(5.) See Levin and Rappaport Hovav 2005: 105 for arguments that such "pseudo-objects" are not actually arguments of the verb, but event delimiting adverbial phrases. However, they do stress that there seems to be a cross-linguistic connection between telicity in unaccusative verbs and accusative case. I will therefore group the instances of Ved. jaami- and Av. * jaymi- with other verbal adjectives taking accusative objects, acknowledging that there is a structural difference between their arguments and those of gam.
(6.) "... gern in den Streit ziehend" (Geldner, RV I: 304); "... qui vas (droit) au defi" (Renou, EVP IS: 54).
(7.) Or "die Unnachgiebigen, (...) Trotzigen hassend" (Grassmann, WB: 1315).
(8.) "... nach Wunsch die Schuld einziehend," Geldner, RV I: 308, cf. also Renou, EVP 15: 59.
(9.) Renou (EVP 17: 20) apparently takes nadyah to be a genitive singular ("bloqueur de la riviere"). 1 follow Grassmann (WB: 706f.) in which case nadyo vavrina is better translated as "(by) the one restraining the waters"; Geldner's use of the genitive here suggests a substantival agent noun.
(10.) Renou (EVP 17: 89) translates vyanasi as perfective, presumably based on the finite 3sg. pertf. vyanase 'has reached' in RV 9.86.15b, which may well be the underlying syntagm of this form. However, RV 3.49.3 is concerned with generic qualities of Indra, for which the imperfective "durchdringend" of Geldner (RV I: 389) and Grassmann (WB: 1360) fits better.
(11.) I.e., gerunds with accusative objects. as in baking the cake, etc.
(12.) The length of the reduplication syllable is assured by RV 2.16.7b which has dadhrtsih in the cadence of a Jagati verse. This could reflect the influence of the perfect stem allomorph with long reduplication (which, however, is not attested until the Atharvaveda; see Krisch 1996: 31 n. 64; Kummel 2000: 271). However, since the trisyllabic i-stems clearly prefer the stem structure H(eavy) L(ight) (cf. tatrp-, sasah-, tutuj-, etc.; see [section]5.2 below), an original *dadhrtsi- could have been remodelled very early, independently of the corresponding perfect stem.
(13.) Kummel 2000: 413.
(14.) See Oldenberg 1909: 318, but vyac (Grassman, WB: 1293) cannot be completely excluded. The latter has the occasional reduplicated present forms based on the stative perfect stem already in the Rigveda (e.g., 3du.pres. viviktas, etc.; see Kummel 2000: 509ff.), which could have provided the derivational basis--in that case we are not dealing with a "mismatch" form.
(15.) I am grateful to Stephanie Jamison for bringing this possibility to my attention.
(16.) Collapsing these roots (as hinted at by Whitney 1885: 43) is implicitly rejected by EWA I: 512f. and [??]: 197; 219f. The connection with ghra 'smell' (cf. Geldner, RV I: 223 n. 15b) can be excluded on formal grounds (expected *jaghuri-), but more securely on semantic grounds (ghra and its Indo-European relatives are always transitive, see [??]: 221).
(17.) See Renou. EVP 5: 30.
(18.) Better with Oldenberg (1909: 290) "fur Kochopfer."
(19.) See Baker and Vinokurova 2009 and Alexiadou and Rathert 2010 on the typology of "mixed nominalizations."
(20.) This translation is based on Bartholomae's emendation of mana to +mano, (AIW: 1438); cf. also Wolff's translation (1910). According to Geldner's edition (1886-96), however, there is no variation in the manuscripts that could support such an emendation. The reading man 'my', on the other hand, poses problems for the translation. I have therefore decided to follow Bartholomae.
(21.) The synchronic intensive stem ganigam- and the late reduplicated aorist jigamat can hardly be relevant here.
(22.) Cf. AiG II,2: 596f. for Vedic, but this can obviously not be predicted with absolute certainty since the distribution of -tama-l-tama- and -istha-l-ista- was beginning to fluctuate.
(23.) Whether the length of the reduplicated syllable is original (or real, for that matter) is difficult to decide (see Kellens 1984: 407f. for a discussion of the problem), but the comparison with Vedic (dadhrvi-, 3sg.perf.act. dadhAra vs. mid. dadhre, etc.) suggests that both long- and short-vowel reduplication was available. Avestan itself has both a 3sg.act. da[??]ara (YAv.) and a 3sg.mid. dadre (0Av.).
(24.) This and the following translation are from Kotwal and Kreyenbroek 2009: 53.
(25.) IIr. *-u-, *-i-, and the composite suffix '-u-i- were used to form verbal adjectives to (morphologically) characterized verbal stems. See Rau 1998 for a discussion of this association.
(26.) Cf. also [??] 2002: 52 n. 72: '*blamer ('*memorizer'?)'. The translation 'blamer' is based on a segmentation *mam-ri- and comparison with Gk. [??] 'reproach, blame'.
(27.) The form a-fra-cicis 'not instructing', also attested in P 45, should be restored to +a-fra-cicisis according to Tremblay (1998: 114). thus reflecting a reduplicated i-stem of the root caes 'assign'. However, the text is unreliable as it is, and I will therefore not count this as an instance of the cakri-type in Avestan.
(28.) For expected *jayanuua, which is restored by Kuiper (1939: 52) in this passage, although a generalization of the weak stem jayn- (cf. Ved. jaghnus-, etc.) cannot be completely excluded (Kummel 2000: 629).
(30.) On the intensive in general see Schaefer 1994; cf. also Narten 1981 and Jamison 1983b for arguments that this formation was already part of the proto-language.
(31.) Klingenschmitt (1982: 192) argues that the long reduplication syllable originates in the perfect participle of the root, namely sahvams- from *se-sgh-uos- with compensatory lengthening following cluster simplification. This would explain the origin of the syllable length and is compatible with assuming a reanalysis of certain forms as belonging to a putative type I intensive stem.
(32.) On long reduplication in the perfect in general see Krisch 1996; see also Gunkel 2010: 92ff. for a recent discussion.
(33.) Short-vowel reduplication may actually be metrically preferable for this stem in a number of cases which have been transmitted with long-vowel reduplication; cf. Arnold 1905: 129; Kihnmel 2000: 569ff.
(34.) Cf. Leumann 1952: 14. I am grateful to Dieter Gunkel for bringing this to my attention and for helpful discussion of these forms.
(35.) An alternative "intensive stem scenario" could be based on a suggestion by Kellens (1984: 195 n. 4), who analyzes the Avestan reduplicated stem ja[??]na- (ganljan 'slay') as dissimilated from the intensive stem *jan[??]na-. One could therefore imagine that *ja[??]ni- ([degrees]ja[??]nista-) was derived directly from this intensive stem and that the pattern spread from there. However, the phonological development assumed by Kellens is not tenable (Garcia Ramon 1998: 150; see the discussion in [section]4 above for a different analysis of ja[??]na-), nor is the assumption that this form alone was the starting point of the whole type. Furthermore, the Vedic intensive stem of the same root is non-dissimilated janghan-, and the genitive singular of the active present participle, janghnatas (RV 9.66.25), shows that the sequence VNCNV was at least synchronically tolerated.
(36.) Geldner, RV 1: 277: "du bist der Prufer, wenn du das Opfer ausrichtest."
(37.) An anonymous reviewer has pointed out that the existence of absolutives of the type abhi-gurya, pari-tapya, etc., might be interpreted as evidence that the i-stems used in composition with preverbs were actually nomina actionis and cannot be compared to the cakri-type. However, there is an alternative analysis of these forms, namely as grammaticalized instrumentals of deverbal ya-abstracts (this is also the view of AiG II, 2: 788, where the parallels between the ya-abstracts and the absolutives are discussed in more detail). In Proto-Indo-European terms, this means parsing the ending as *-io-[h.sub.1] rather than *-i-[eh.sub.1]. Furthermore, given what we know about the functions of PIE nominal *-i-, use as a verbal-abstract-forming suffix would not preclude use as an agentive suffix--both functions are attested across the Indo-European branches.
(38.) In RV 2.23.3d mrdhas could also be acc.pl.
(39.) Without RV 10.106.8d, which is unclear.
|Printer friendly Cite/link Email Feedback|
|Publication:||The Journal of the American Oriental Society|
|Date:||Apr 1, 2013|
|Previous Article:||The fructification of the tale of a tree: the Parijataharana in the Harivamsa and its appendices.|
|Next Article:||The Jinshin Rebellion and the politics of historical narrative in early Japan.|