Amreditas and related constellations in the Rigveda.
Semantically, amreditas were recognized already by Panini as signalling durative and distributive values, i.e., the notions of undelimited continuation ('over and over, again and again') and singulatim repetition ('one by one, one after the other'), which, when applied to all members of a set, may attain universality ('every'). (3) But, as we shall see, the semantics of this category as a whole shows nuanced variation depending on the part of speech class of the individual amreditas.
[section]1. Amreditas are by no means rare in the Rigveda. I have located 132 different forms occurring 291 times, not counting another twelve exact pada-repetitions. (4) This figure represents, to my knowledge, the first complete tabulation of these forms. (5) Amreditas are fashioned from forms belonging to six different parts of speech: nominals (including both nouns and adjectives) (94 forms, 211x + 5 repetitions), pronominals (15 forms, 31x + 5 repetitions), adverbials (8 forms, 17x[.sup.6] + 1 repetition), preverbs (7 forms, 20x), numerals (7 forms, 11x), and verbs (1 form, 1x + 1 repetition). Because these forms have never been listed in their entirety, I provide a complete listing in Appendix 1. Where nominal types (taken in the broadest sense to include nouns, adjectives, pronouns, and numerals) are concerned, all case forms but the vocative are represented. Moreover, with the single exception of the adjectival amredita navya-navyah, all nominal amreditas are morphologically singular. A partial exception to this is the (improper) numeral amredita ekam-eka, where the second member has been assimilated to a following term (cf. [section]15). Nevertheless, numeral amreditas such as dva-dva (2x), panca-panca (1x), and sapta-sapta (2x) do occur, the latter two indifferent to morphological number and the first obligatorily dual. Similarly, personal pronoun amreditas include first person plural forms vayam-vayam (1x) and asman-asman (1x). A single example of the pronominal neuter plural ta-ta is also attested. The fact that non-singular amreditas are limited to numerals, pronouns, and adjectives suggests that amreditas associated with these parts of speech are to be interpreted at least in part differently from those involving nouns, a suspicion which will be confirmed in what follows.
In [section][section]2-7 we shall provide a rapid general survey of the Rigvedic amreditas along part-of-speech lines, focusing on formal and lexical features. We shall then turn to a more detailed investigation of these forms, again by grammatical category, in which semantics will play a central role ([section][section]8-16). Following this we shall investigate various constellations which, while not themselves amreditas, seem to be formally related to this type ([section][section]17-22).
[section]2. The order of relative frequency of nominal case forms represented among noun and adjective amreditas in the Rigveda is: locative (29 forms, 64x + 1 repetition), accusative (32 forms, 59x + 3 repetitions), (7) dative (4 forms, 54x), (8) genitive (10 forms, 13x), instrumental (8 forms, 9x), nominative (7 forms, 7x + 1 repetition), (9) and ablative (4 forms, 4x). In addition, one form (samit-samit: 1x) is morphologically nominative but semantically instrumental. This form will be given special treatment in [section]9, as will vise-vise and dive-dive, two forms which, despite their dative appearance, have been treated normally or frequently as locative in the literature.
[section]3. Moving on to pronominals, we note that the most frequent word appearing in amreditas is anya-(4 forms, 10 occurrences), followed by the ta-pronoun (3 forms, 9 occurrences + 1 repetition), the ya-stem (4 forms, 8 occurrences + 4 repetitions), personal pronouns (3 forms, 3 occurrences), and forms of ayam (1 form, 1 occurrence). (10) On the semantics of these, cf. [section]12.
[section]4. Under adverbials we include amreditas formed from pure adverbs, such as iha (8x + 1 repetition), punah (2x), etc. as well as most adverbially employed nominative/accusative neuter forms, such as idam (2x), bhuyah (1x), and sasvat (1x). For more on these forms, cf. [section]13.
[section]5. Preverb amreditas show little that is remarkable from a formal perspective other than the predominance of pra-pra which, while one of seven different forms, accounts for sixty percent of the total preverb amreditas (12/20). In fact, the only other preverb amredita to occur more than once is upopa (3x). The semantics of preverb amreditas will be discussed in [section]14.
[section]6. Similarly, numeral amreditas are dominated by forms of eka- (4 out of 7 forms, 6 out of 11 occurrences), although dva-dva and sapta-sapta each occurs twice. The only other attested numeral amredita is panca-panca (1x). These forms are treated in more detail in [section]15.
[section]7. Finally, the only verbal amredita treated as such in the padapatha text is the imperative piba-piba (piba-pibed) (1x + 1 repetition). However, it appears arbitrary that stuhi stuhi (stuhi stuhid) (VIII.1.30a) should not be treated in the same way. (11) The semantics of these forms is discussed in [section][section]16 and 17.
[section]8. Having discussed in very general terms the formal and lexical aspects of the Rigvedic amreditas, let us now turn to the semantics of these forms. One would of course expect such constructions to possess iconic semantics, plurality or iteration of form being matched by a sememe of plurality or iteration. This is indeed the case. The grammars and handbooks emphasize the intensive, distributive, and repetitional meaning of these forms (e.g., Whitney 1889: 488). To these Brugmann adds the notions of "Verschiedenheit" (idam-idam 'here and there, in different places') and "Eindringlichkeit" (iheha 'precisely here') (1906: 56-57). However, the relevant nuance of the iterative semantic feature differs in subtle ways from word-class to word-class. In the case of nouns this nuance may be said to involve unlimited but usually individuated repetition. (12) This absence of delimitation easily passes over into a notion of universality or totality. This is illustrated in (1a-g), where we provide an example of an amredita involving each case form represented, in order of frequency:
(1) a. V.43.2cd pita mata madhuvacah suhasta / bhare-bhare no yasasav avistam "Let father (sc. heaven) (and) mother (sc. earth), she of sweet speech, of good hands, the glorious ones aid us in battle after battle / in every battle."
b. VII.74.1d visam-visam hi gachathah "For you two go unto clan after clan / unto every clan."
c. X.37.8 mahi jyotir bibhratam tva vicaksana / bhasvantam caksuse-caksuse mayah // ... / vayam jivah prati pasyema surya "Thee, O broadly-seeing one, the one possessing radiance, the one bringing great light, a balm for every eye ... might we behold while alive, O Surya."
d. III.31.8ab satah-satah pratimanam purobhur / visva veda janima ... "The equal of every existing thing, the one standing at the front, he (viz. Indra) knows all the races."
e. IX.65.2 pavamana ruca-ruca / devo devebhyas pari / visva vasuny a visa "O purified (soma), with radiance after radiance enter into all the good things(?), as a heavenly one (coming forth) from the heavenly ones."
f. VII.90.2d jato-jato jayate vajy asya "Every son of his is born a victor in battle."
g. VI.19.13ab vayam ta ebhih puruhuta sakhyaih / satroh-satror uttara it syama "May we be higher than every rival through these friendships of thine, O often invoked one."
The above passages illustrate normal, nonproblematic employments of amreditas to convey the notions of universality or totality. We now turn to particular instances showing less frequent or unusual values. These are seen particularly among accusative amreditas. Thus, the single clear instance of distributive value among nominal amreditas is the following:
(2) X.15.11ab agnisvattah pitara eha gachata / sadah-sadah sadata supranitayah "Sweetened by the fire, come hither, O fathers. Sit ye each in his own seat, ye who provide good leading."
Here the fathers are not being summoned to sit without further qualification in every seat, but rather each is invited to what must be thought of as a place individually prepared for him.
A special case of an accusative amredita is janam-janam. The sense 'folk after folk' easily comes to mean 'every man, everyone' (3a), and this in turn, under the scope of negation, may mean 'no man, nobody' (3b):
(3) a. V.15.4ab mateva yad bharase paprathano / janam-janam dhayase caksase ca "When, spreading out, thou dost bear like a mother every man for nourishment and seeing ..."
b. X.91.2c janam-janam janyo nati manyate "The one belonging to men (sc. Agni) despises no man."
The most important special circumstance among nominal amreditas is when the form is a proper name or clearly refers to a singular, unique individual. For in such instances the meaning cannot be 'every ...'. Cf. the following passage:
(4) VIII.12.19ab devam-devam vo (a)vasa / indram-indram grnisani "Over and over again the heavenly Indra will I sing with praise for aid for all of you."
The translation we have provided here seems to be the one that most straightforwardly suggests itself, based on the meanings of the amreditas seen so far. There is only one Indra, and so the notion of unlimited repetition ('Indra after Indra') cannot be distributed over a totality of different Indras but must signal an indefinitely repeated object of an action: "Indra over and over, Indra always." Less certain, perhaps, is the possibility that the amredita conveys a delimiting and focusing value ("Indra alone," "precisely Indra," "that very Indra") (cf. below on pronominal amreditas). What makes this passage particularly interesting is the cooccurrence of the amredita devam-devam, which provides additional focus and specification, as if the phrase devam indram were subject to a rule of "amreditization" to devam-devam indram-indram. With its idiomatic phrase immer nur, German is perhaps better able than English to capture the nuance of these amreditas, and it is precisely this phrase that is used by Geldner (1951) in translating both: "immer nur den Gott, immer nur Indra soll man loben."
Of a similar nature, but perhaps more equivocal than indram-indram is agnim-agnim. As deity, Agni is of course unique, but there are many fires, so that in its naturalistic value, "fire after fire/every fire" would be possible. However, the three agni-amreditas occurring in the Rigveda all refer to the deity Agni. It is therefore likely that the sense is "Agni over and over again," especially with the verb hu 'call':
(5) a. I.12.2 agnim-agnim havimabhih / sada havanta vispatim / ... purupriyam "Agni over and over again do they call constantly with their invocations, the master of the clan.... the very dear one (or: the one dear to many)."
b. VI.15.6ab agnim-agnim vah samidha duvasyata / priyam-priyam vo atithim grnisani "Worship your Agni over and over again with the kindling log; sing with praise the ever dear one, your guest." (13) (Cf. VIII.60.17ab agnim-agnim vo adhrigum / huvema vrktabarhisah "Agni over and over again, the adhrigu(?), would we call for you, having [our] barthis twisted out.")
In (a) the notion of indefinite repetition is reinforced by the adverb sada 'constantly', whereas priyam-priyam in (b) is a nuanced variant of purupriyam (a).
The only other unusual instance among nominal amreditas is the following, in which pade-pade seems to have acquired an idiomatic value:
(6) V.41.15a pade-pade me jarima ni dhayi "Let old age be secured/deposited for me step-by-step." (14)
In this passage the singer is most likely praying for a gradual easing into old age rather than sudden decrepitude. The literal and normal sense of the amredita pade-pade is seen in the following passage:
(7) IX.73.4cd asya spaso na ni misanti bhurnayah / pade-pade pasinah santi setavah "Of this one (viz. Varuna) the spies, eager (for action), do not blink. In place (lit., footprint) after place/in every place there are traps possessing fetters."
[section]9. Having completed our survey of nominal amreditas on a semantic basis, we come now to three particular forms which require further comment: samit-samit, vise-vise, and dive-dive. The first of these occurs a single time in the following passage:
(8) III.4.1ab samit-samit sumana bodhy asme / suca-suca sumatim rasi vasvah "With kindling log after kindling log become well-minded among us, with every flame grant (us) the favor of the Vasu (god)."
Here the contrast as well as the parallelism with pada b guarantees the instrumental value of samit, formally a nominative. The example is vaguely reminiscent of the type navyasa vacah 'with a new speech', although this latter represents a truncation of navyasa vacasa at the end of a pada. The most likely reason for the absence of *samidha-samidha is that trisyllabic amreditas are rare in the Rigveda. (15) Instructive in this regard is yajna-yajna (2x). The regular simplex instrumental of yajna- in the Rigveda is of course yajnena, which occurs 13x. However, the latter form never occurs as the basis of the instrumental amredita of yajna- but is replaced by the archaic, otherwise moribund yajna. In each instance yajna-yajna stands in parisyllabic relationship with a line-initial amredita in the following pada, and this in turn helps to produce parisyllabic vertical cola in the two lines: I.168.1ab yajna-yajna vah samana tuturvanir / dhiyam-dhiyam vo devaya u dadhidhve and VI.48.1ab yajna-yajna vo agnaye / gira-gira ca daksase. Note the similar vertical parisyllabism produced by the amreditas samit-samit and suca-suca at the beginning of adjacent padas in III.4.1.
The form vise-vise appears to be nonproblematic. Morphologically this can only be a dative to the root noun vis- 'clan'. But Wackernagel (1905: 146), Brugmann (1906: 56), Macdonell (1916: 282, n. 2), and Dressler (1968) consider this form to possess locative value and to represent a replacement, on metrical grounds, of a more original *visi-visi. Significantly, none of these scholars examines the passages in which vise-vise occurs. In fact, the five occurrences of vise-vise can all be interpreted as datives and are translated as such in all instances by Geldner. Of the following passages, (a) and (b) rather clearly show this value. It is somewhat less clear in (c):
(9) a. IV.7.4 asum dutam vivasvatah / ... // a jabhruh ketum ayavo / bhrgavanam vise-vise "The swift messenger of Vivasvat, ... the associate of the Bhrgu's have the Ayu's brought hither as a beacon to/for clan after clan/every clan."
b. V.8.5ab tvam agne pururupo vise-vise / vayo dadhasi pratnatha purustuta "O Agni, thou of many forms dost create strength for every clan, as of old, O often-praised one."
c. I.27.10 jarabodha tad vividdhi / vise-vise yajniyaya / stomam rudraya drsikam "O early-waking one, bring that about: a praise worthy of being seen to Rudra, the one worthy of worship to/for clan after clan/every clan."
Each of these passages shows a context where a dative is completely at home in Rigvedic syntax: after a verb of bringing to express the recipient (a), following a verb of creation to express the one or group for whose benefit the created thing has been made (b). In (c) the sense of yajniya-approaches that of a gerundive, a category which may take a benefactive dative object in the Rigveda (cf. I.33.2d ya stotrbhyo havyo asti yaman "who is to be invoked for the singers in his course"). In fact, the first two of these contexts are encountered in four of the eight separate (i.e., nonrepeated) occurrences of the simplex dative vise:
(10) a. VII.33.10d agastyo yat tva visa ajabhara "when Agastya brought thee hither to/for the clan." (16)
b. I.93.8d (=VII.82.1b) vise janaya mahi sarma yachatam "Extend great protection to the clan, to the folk." (Cf. IX.86.15a so asya vise mahi sarma yachati.)
c. V.48.4d ratnam dadhati bharahutaye vise "(when) he creates a treasure for the clan having a battle-shout."
Example (10a) shows exactly the same verb (a bhr) as (9a), whereas the nearly synonymous yam appears in I.93.8d and IX.86.15a (b). Again, dha in (c) shows the same value as in (9b).
The two remaining instances of vise-vise both occur in the pada-final phrase vibhvam vise-vise. The meaning of vibhu- in the Rigveda is "very great, outstanding, marvelous." Hence, the value of vise-vise in the following passages is similar to that seen in (9c) above:
(11) IV.7.1 ayam iha prathamo dhayi dhatrbhir / hota yajistho adhvaresv idyah // yam apnavano bhrgavo virurucur / vanesu citram vibhvam vise-vise "This one has been set here as the first one by the establishers (of the worship), the most worshipful Hotar, the one to be called in the ceremonies; whom Apnavana (and) the Bhrgu's lit up, gleaming in the forests, marvelous/a marvel to/for every clan." (Cf. also X.40. 1c prataryavanam vibhvam vise-vise "driving early in the morning, a marvel ...") (17)
Another, rather weaker argument in favor of taking vise-vise as a dative is that the simplex locative visi is found only 2 times (+ 1 repetition) in the Rigveda, whereas the simplex dative vise is found 8 times (+ 1 repetition). (18) The rare occurrence of visi is probably owing to the fact that this word possessed a somewhat more abstract meaning than grhe and dame, words that occur, respectively, 26 times and 41 times in the Rigveda, and which were better suited for indicating a concrete locale. If the bards wished, on metrical and/or semantic grounds, to avoid *visi-visi, they could and did employ instead either grhe-grhe (4x) or dame-dame (6x). Both of these are among the most frequent of the Rigvedic amreditas. (19)
The scholars who interpret vise-vise as a locative uniformly apply the same analysis to dive-dive. According to their views dive-dive is a pseudo-locative which has replaced the metrically unfavorable tetrabrach locative *divi-divi. It would be nice if we could apply the same elementary procedure of merely inspecting the passages in which dive-dive occurs to prove that this form, like vise-vise, must be a dative; but unfortunately, we cannot. In all 47 of its occurrences dive-dive means simply "day after day/every day" with a completely static meaning of the sort which one would expect in a temporal locative. Even if we were to consider a reconstructed dative value (e.g., "Grant our desire for every [succeeding] day" vel sim.), we would expect at least some Rigvedic passages to reflect this value; (20) but none do. Speaking against a dative amredita is also the fact that none of the other Indo-European temporal constructions involving iteration show datives (cf. Appendix 2), as well as the extreme infrequency of dative amreditas other than vise-vise and dive-dive (cf. n. 7). Indeed, the dative itself is rarely employed in adverbial value in the Rigveda. (21)
On the other side of the ledger, the utter pervasiveness of dive-dive (it occurs in every Mandala as follows: [I.sup.8], [II.sup.4], [III.sup.8], [IV.sup.3], [V.sup.2], [VI.sup.5], [VII.sup.2], [VIII.sup.7], [IX.sup.5], [X.sup.3]) and the fact that it is four times as frequent as the next most common amredita means that it belongs to the very core of the Rigvedic poetic language and may well represent, together with perhaps pra-pra and yad-yad (cf. Appendix 2), one of the starting points for the whole formation. Even its metrical distribution is classically archaic: it comprises the line-final cadence thirty times (its remaining seventeen occurrences are all line-initial), and its formulaic status could certainly have led to semantic fading and stereotypical usage in the value 'every day'. True, aside from ahar-ahah (6x), itself likely to be archaic, other nominal amreditas indicating time units in the Rigveda are either locative or instrumental; but all are late: masi-masi (1x: X.52.3), dyavi-dyavi (2x: I.4.1, 25.1), ahnahna (1x: X.37.9). It is therefore exceedingly difficult not to see dive-dive as an archaism and its case value as a constructio difficilior. Also speaking against a replacement of *divi-divi by dive-dive is that surely the oldest form of the locative of dyauh is dyavi, (22) and an old locative amredita dyavi-dyavi would have completely obviated the metrical difficulties necessitating a replacement of *divi-divi. To argue that dyavi only means 'heaven' will not work, because the same is true of the simplex dive (15x). (23) One would therefore have to claim that dyavi was already moribund and being replaced by divi at the earliest, formative period of the Rigvedic poetic tradition. But there is no internal Rigvedic evidence to suggest that dive-dive is any younger than dyavi.
On balance, I believe that the evidence slightly favors the hypothesis that dive-dive is a real dative. To turn the metrical argument of Wackernagel et al. on its ear, I believe that it was precisely because of its external form that dive-dive was able to maintain itself in a semantic value that may have struck the Rigvedic poets as locatival. If we examine the locative amreditas of the Rigveda, we note that of twenty-nine such forms all but seven are disyllabic a-stems having the shape Xe-Xe (cf. Appendix 1). The only exceptions are karman-karman/karmani-karmani (3x), janman-janman (1x + 1 repetition), parvani-parvani (1x), masi-masi (1x), yamani-yamani (1x), and dyavi-dyavi (2x). In terms of actual occurrences, therefore, 55 of the 64 instances of locative amreditas are a-stem forms. The resemblance of dive-dive to those other forms could easily have contributed to its retention, even though it was not itself etymologically a locative. Evidence that perhaps some circles of Rigvedic bards were unhappy with dive-dive in apparently locative value is the occurrence of dyavi-dyavi in the two late passages I.4.1c juhumasi dyavi-dyavi and I.25.1c minimasi dyavi-dyavi. In these cases neither dive-dive nor, of course, the metrically impossible *divi-divi would have yielded the apparently desirable iambic tetrameter of these passages as we have them, and therefore the bards resorted to archaism, creating an amredita out of a metrically favorable form which was probably moribund at the time they were composing. Yet another aborted attempt to improve upon dive-dive is the hapax formation ahardivi with adverbial first member found at IX.86.41(a)b (sa bhandana ud iyarti prajavatir/) visvayur visvah subhara ahardivi "He raises up day by day/every day for an entire lifetime all the praiseworthy things consisting of offspring, the easily borne things." This bizarre form is not, of course, an amredita, (24) but it does manage to incorporate the locative divi in association with the nominative-accusative ahah, treated as a stem and no doubt based on its occurrence in the true amredita ahar-ahah. (25)
[section]10. Having surveyed in very general terms, by part of speech category, the kinds of amreditas occurring in the Rigveda and having discussed some particular forms which either are or have been treated in the literature as non-straightforward, it will be useful here to discuss some salient characteristics of the rhetorical and poetic employment of amreditas by the Rigvedic bards. Chief among these is the very significant degree to which amreditas tend to cluster. Leaving aside the exceedingly numerous instances in which amreditas occur within different stanzas of the same hymn, there are thirty intrastanzaic amredita sequences in the Rigveda, representing sixty-seven amreditas or twenty-three percent of the total. These include twenty-four two-term sequences, five three-term sequences, and one four-term sequence. Two-term sequences are seen in (4), (5b), and (8) above. Sequences of three amreditas are seen in (12a,b) and the single occurrence of a four-term sequence is seen in (12c):
(12) a. V.53.11 sardham-sardham va esam / vratam-vratam ganam-ganam susastibhih / anu kramema dhitibhih "Every troop of yours here, every flock, every retinue would we accompany with praises, with (pious) thoughts."
b. VIII.27.13a-c devam-devam vo (a)vase / devam-devam abhistaye // devam-devam huvema vajasataye "We would call you, heavenly one after heavenly one, for aid, heavenly one after heavenly one for succour, heavenly one after heavenly one for the attainment of booty."
c. X.91.2 sa darsatasrir atithir grhe-grhe / vane-vane sisriye takvavir iva // janam-janam janyo nati manyate / visa a kseti visyo visam-visam "That beautifully appearing one is the guest in home after home/in every home. He has fixed himself in wood after wood/in every wood like a hunter (in the forest). The one belonging to men (sc. Agni) despises no man. The one belonging to the clans dwells among the clans, clan after clan."
In the second of these passages the Amredita-Haufung takes the form of the rhetorically anaphoric employment of devam-devam at the beginning of each pada, whereas in (c) the amreditas are positioned chiastically within each distich: ... A/B ... (contact chiasm) in the opening distich and C ... / ... D (distant chiasm) in the second distich.
In yet another pattern the stanza-initial amredita vamam-vamam is continued in the second distich by the anaphoric employment of the simplex vamam:
(13) IV.30.24 vamam-vamam ta adure / devo dadatv aryama // vamam pusa vamam bhago / vamam devah karulati "Every desirable thing let the heavenly Aryaman give thee, O Aduri, the desirable thing Pusan, the desirable thing Bhaga, the desirable thing (let) the gapped-tooth heavenly one (give thee)."
In addition to the rhetorical employment of sequences of amreditas just noted, the bards seem to have embellished the lexical and phonological repetition associated with amreditas with other forms of lexical and phonological repetition within the stanza. Cf. the following passages:
(14) a. III.4.1 samit-samit sumana bodhy asme / suca-suca sumatim rasi vasvah // a deva devan yajathaya vaksi / sakha sakhin sumana yaksy agne "With kindling log after kindling log become well-minded to us, with every flame grant (us) the favor of the Vasu (god). Hither, O heavenly one, convey the heavenly ones for worship. As a well-minded friend, O Agni, worship the friends."
b. VI.48.1 yajna-yajna vo agnaye / gira-gira ca daksase // pra-pra vayam amrtam jatave-dasam / priyam mitram na samsisam "With worship after worship and with song after song (do we wish) to render honor to Agni for all of you. Over and over will we praise the immortal Jatavedas like a dear friend." (On vayam ... samsisam cf. Geldner ad loc.; on pra-pra, cf. [section]14.)
c. X.76.8cd vamam-vamam vo divyaya dhamne / vasu-vasu vah parthivaya sunvate "(Let) every desirable thing of yours (be) for the heavenly abode; let every good thing of yours (be) for the earthly (soma)presser."
(a) shows the virtual commutation of amreditas and polyptoton (case variation of the same lexical item) as distich organizing principles (amreditas in ab, polyptoton [deva devan ... / sakha sakhin ...] in cd). In (b) three different amreditas serve as an anaphoric stanza organizing principle (padas a-c) supported by line-final rhyme and assonance (ab) or rhyme together with wholesale homoioteleuton (c,d). (26) Finally, (c) is remarkable for its word-forword isosyllabic responsion and syntactic and semantic parallelism. For a gayatri stanza in which an amredita (ruca-ruca) in pada a is complemented by polyptoton (devo devebhyah) in b and v-alliteration (visva vasuny a visa) in c, cf. IX.65.2(=1e) above.
[section]11. It is necessary now to look briefly at adjectival amreditas in order to characterize the special semantic nuance of these forms. Substantivized adjectives show, of course, the same values as nouns. Cf. VIII.27.13a-c devam-devam (12b), III.31.8a satah-satah (1d), and X.76.8cd vamam-vamam ... / vasu-vasu ... (14c); and in several instances we have seen, no doubt as a poetic device, that an adjectival amredita modifies adjectivally or appositionally a proper noun amredita: VIII.12.19ab devam-devam ... / indram-indram ... (4) and VI.15.6ab agnim-agnim ... / priyam-priyam ... (5b). Otherwise, however, adjectival amreditas seem to convey an adverbialized universal value ("ever" rather than "every"). A good example of the distinction between substantivized and non-substantivized adjective amreditas is the following pair of passages:
(15) a. VIII.93.28a bhadram-bhadram na a bhara "Bear hither to us every auspicious thing."
b. I.123.13b bhadram-bhadram kratum asmasu dhehi "Set among us ever-auspicious determination."
Note that, unlike a nominal amredita, an adjectival one may be plural in agreement with a plural head noun:
(16) III.55.16a.c a dhenavo dhunayantam asisvih/ ... // navya-navya yuvatayo bhavantih "Let the cows without offspring (viz. the rain-clouds) roar hither ... becoming everanew youthful."
In two other instances the amredita seems to reinforce a comparative or superlative value of the simplex adjective: (27)
(17) a. V.77.2d purvah-purvo yajamano vaniyan "The earlier the worshipper (performs the worship), the more he gains." (So Geldner; but possibly "The worshipper who is ever-earlier gains more").
b. I.123.4d agram-agram id bhajate vasunam "She (sc. the dawn) possesses as her portion only the very best of goods."
Finally, in one instance it appears that the nonsubstantivized adjectival amredita must be taken in the same way as its substantivized version:
(18) IV.57.7cd sa nah payasvati duham / uttaram-uttaram samam "Let that one, full of milk, give milk to us for every later year (i.e., for every year to come)."
[section]12. A special discussion of pronominal amreditas is necessitated, once again, by the particular semantics and, in part, the grammaticalization of such forms. As indicated in Appendix 1, pronominal amreditas fall into six morphological/lexical groups: anya- (10x), (sa/)ta- (9x + 1 repetition), ya- (8x + 4 repetitions), ayam (1x), and personal pronouns (3x). In the case of anya-the value of the amredita is in all instances but one "one after the other." As subjects these always take a dual verb:
(19) a. I.62.8cd krsnebhir aktosa rusadbhir / vapurbhir a carato anyanya "Night with black, dawn with ruddy forms, the two wander hither one after the other."
b. VI.47.15cd padav iva praharann anyam-anyam / krnoti purvam aparam sacibhih "Like one putting forth his two feet one after the other, he makes the one at the fore into one who lags behind with his powers."
c. VI.47.16ab srnve vira ugram-ugram damayann / anyam-anyam atineniyamanah "The hero is famed as taming every mighty one, as repeatedly leading one after the other beyond (obstacles)."
Whereas (a) and (b) clearly illustrate the notion 'one after the other' in association with duality, it would be possible to interpret anyam-anyam in (c) as meaning 'everybody' in the category of Indra's friends. In that case the pronominal amredita would possess the same universal value as the substantivized amredita ugram-ugram of the prior pada. (28)
The only anya- amredita not showing the value just illustrated conveys a distributive meaning:
(20) II.24.5cd ayatanta carato anyad-anyad id / ya cakara vayuna brahmanas patih "Not striving (with each other) the two wander each into another (sphere) (according to) the demarcative spheres which Brahmanaspati made."
The distributive value of the amredita here is assured if we assume, with Oldenberg (1909) and Ludwig (1881-83) (cf. Geldner ad loc.), that the dual subject of this passage is suryamasa 'sun and moon'. Each of these is master of its own independent realm. Hence, the value cannot be 'one (sphere) after another/the other'. The delimiting function of id seems also to play a role here in circumscribing the realm of each and thereby clarifying the value of the amredita. Indeed, id appears with particular frequency (16x) following amreditas in the Rigveda. In nine instances it follows a pronoun, where it normally possesses delimiting value, as here. The same is true of its occurrence with the substantivized adjective amredita agram-agram at I.123.4d (=17b above). In other instances, however, particularly in its use with preverbs (3x), numerals (1x), and verbs (1x + the closely related VIII.1.30, cf.  and  below), it possesses an emphasizing value. (29) Such instances will be noted as they occur.
Another set of amreditas showing a special, clearly grammaticalized value involves relative pronouns. As is apparent from a consideration of forms and/or constructions like Hittite kuis kuis, Lat, quisquis, and Gk. o[sigma][tau][iota][zeta] 'whoever', the interrogative/indefinite/relative pronouns of the early Indo-European languages may be doubled or, as in Greek, adjacently placed and univerbated, in order to create an indefinite relative pronoun. The iterated type is widespread among the languages of the world (cf. Haspelmath 1997: 179-80). The same is true in Vedic, with the result that we find four separate amredita forms showing this value:
(21) a. VIII.61.6d yad-yad yami tad a bhara "Whatsoever I beseech (of thee), bear that hither."
b. II.25.1cd jatena jatam ati sa pra sarsrte / yam-yam yujam krnute brahmanas patih "That one overtakes the one born (of another) with (his own) offspring, whomever Brahmanaspati makes into his ally." (1d=2d-5d)
c. VI.75.6ab rathe tisthan nayati vajinah puro / yatra-yatra kamayate susarathih "Standing upon the chariot, the good charioteer leads the coursers forward wherever he wishes."
d. IV.54.5cd yatha-yatha patayanto viyemira / evaiva tasthuh savitah savaya te "In whatever manner they have spread themselves out in flight, in that same way they have come to a standstill at thy impelling, O Savitar."
Note that, with the exception of (c), all of these passages show correlative structures involving either a simplex sa/ta- pronoun (i.e., not an amredita) or a doubled correlative (evaiva) which is, however, not an amredita, at least as the text has been transmitted to us. In a single instance we do, however, find an amredita as correlative:
(22) VIII.39.4ab tad-tad agnir vayo dadhe / yatha-yatha krpanyati "Agni grants that very strength of whatever sort/of exactly that sort which (the singer) desires."
In this passage the value of tad-tad seems to be strongly specifying ("that very"), and yatha-yatha may be taken to convey a similar value or to be indefinitizing. In the remaining yatha-yatha passages a strongly specifying value seems to yield a slightly better rendition than a generalizing indefinite value:
(23) a. IV.19.10 pra te purvani karanani vipra / avidvan aha viduse karamsi // yatha-yatha vrsnyani svagurta / apamsi rajan naryavivesih "Thy previous deeds did the one precisely knowing the(se) deeds speak forth to the initiate, O inspired poet, exactly as/in whatever way(?) thou didst accomplish the self-praising virile powers, the manly works, O king."
b. X.111.1ab manisinah pra bharadhvam manisam / yatha-yatha matayah santi nrnam "O possessors of (poetic) wisdom, bring forth your (poetic) wisdom exactly as/in whatever way(?) the thoughts of the men (i.e., the worshippers) are (i.e., precisely convey their thoughts through your poetry)."
c. X.100.4a-c indro asme sumana astu visvaha / raja somah suvitasyadhy etu nah // yatha-yatha mitradhitani samdadhuh "Let Indra be well-minded to us at all times; let King Soma be solicitous for our well-being, just as (if) they [viz. the heavenly ones in general, including Indra and Soma] had made friendship-pacts (with us)."
In (a) the key to our preferred rendition is the a of avidvan, which seems to convey the idea of intimate or precise knowledge on the part of the knowing one and therefore reinforces the specifying value of the amredita. In (b) the context suggests that the poets are being beseeched to convey perfectly the thoughts of the worshippers. Finally, if our interpretation of the subject of samdadhuh in (c) is allowed to stand, (30) then the specifying value of yatha-yatha seems inescapable.
The same specifying value that we saw in VIII.39.4ab (=22 above) is seen in the remaining instances of amreditas involving the (sa/)ta-pronoun. Of nine such examples (not counting repetitions), six show correlative structures. Some additional examples of this sort are (24a-c):
(24) a. I.162.19cd ya te gatranam rtutha krnomi / ta-ta pindanam pra juhomy agnau "Which of thy body parts (= as many ... as) I prepare in the proper way, so many of (those) lumps do I pour forth in the fire."
b. I.132.6bc yo nah prtanyad apa tam-tam id dhatam / vajrena tam-tam id dhatam "Who(ever) will fight us, smite that very one off, with (thy) cudgel smite that very one."
c. VIII.68.7 tam-tam id radhase maha / indram codami pitaye // yah purvyam anustutim / ise krstinam nrtuh "That one alone, Indra do I impel for great granting of success, for drinking, who (possesses) the first praise, (who) is master of the folk, the (lead) dancer."
In both (b) and (c) here the delimiting value of id must also be considered in assessing the value of the amredita (cf. discussion following  above). The same is true of the single amredita form of ayam which goes here (on idam-idam see below):
(25) VI.42.4ab asma-asma id andhaso / (a)dhvaryo pra bhara sutam "To him (viz. Indra) alone bring forth the pressed (drink) of the (soma) stalk, O Adhvaryu."
The three instances of amreditas involving personal pronouns all seem to show an iterative-durative value ("we/you/us always") augmented, in some instances, with a sense of exclusiveness (with or without id):
(26) a. X.96.5 tvam-tvam aharyatha upastutah / purvebhir indra harikesa yajvabhih // tvam haryasi tava visvam ukthyam / asami radho harijata haryatam "Thou (alone) hast ever been desired, O golden-haired Indra, praised by the previous worshippers. Thou dost desire (praise); thine is everything praiseworthy, complete, desirable granting of success, O gold-born one."
b. X.22.12cd vayam-vayam ta asam / sumne syama vajrivah "May we ever be in the good favor of these (aids) of thine, O cudgel-bearer."
c. IV.32.4 vayam indra tve saca / vayam tvabhi nonumah / asman-asman id ud ava "We are together with thee, O Indra; we repeatedly shout unto thee. Aid us alone always!"
As is so frequently the case, the first and third of these passages incorporate the pronominal amredita into a larger structure of pronominal anaphora. Of interest in (c) as well is the juxtaposition of the intensive verb nonumah with its iterative value and the amredita, creating a complementary set in which the repetition of the singer's shouts is answered by the god's repeated acts of aid (asman-asman ... ava).
[section]13. The eight different adverbial amreditas show temporal, spatial, quantitative, and qualitative notions which break down into six semantic values: universal quantification, unlimited iteration, distribution, monotonic increase, heightened degree, and specification. The most straightforward of these are instances in which the amredita indicates some time period and is functionally equivalent to a universal quantifier + noun:
(27) VIII.61.17 adyadya svah-sva / indra trasva pare ca nah // visva ca no jaritrn satpate aha / diva naktam ca raksisah "Every today, every tomorrow, and beyond rescue us, O Indra; and protect our singers throughout all the days, day and night, O master of the seat."
Here the notion inherent in the phrase adyadya svah-svah ... pare ca is repeated in the phrase visva ... aha with universal quantifier in c and is yet further refined in the collocation diva naktam ca in d.
Two other adverbial amreditas involve terms which signify repetition in their simplex forms. In each case the value of the amredita is once again that of constant and unlimited iteration:
(28) a. III.5.7d punah-punar matara navyasi kah "He makes his two parents ever again new."
b. III.36.1a-c imam u su prabhrtim sataye dhah / sasvac-chasvad utibhir yadamanah // sute-sute vavrdhe vardhanebhih "Designate this offering well for attainment, uniting ever again with (thy) aids. He has strengthened himself through his means of strengthening at every soma (pressing)."
Note how, in (b), the iterative value of the adverbial amredita is reinforced by the nominal amredita sute-sute of the following distich. Despite the fact that the two are in different clauses, they contribute mutually to what may be thought of as an iterative disposition on the part of Indra: what he does, he does not once but repeatedly. Thus, from the optimistic perspective of the singer, to have gained Indra once is to have gained him for all time.
In one instance the amredita seems to convey a sense of constant, monotonic increase:
(29) VI.28.2cd bhuyo-bhuyo rayim id asya vardhayann / abhinne khilye ni dadhati devayum "Increasing ever more/more and more his wealth, he sets down the man desirous of the heavenly ones in unbroken pastureland."
In all of the above passages the sense of unlimited repetition seems manifest, whether or not the base term itself indicates repetition (28a,b). In a single instance, however, the sense is that of heightened degree and may be rendered in English as 'very':
(30) III.31.20d maksu-maksu krnuhi gojito nah "Quickly, quickly/Very quickly/Quam celerrime make us conquerors of cows."
There remain just two adverbial amreditas, idam-idam and iheha, both spatial in meaning and approximately synonymous, in which the value appears to be distributive relative to the verb: 'here and there, this way and that'. (31) But the distribution does not appear to extend to a universal "everywhere":
(31) a. VII.59.1 yam trayadhva idam-idam / devaso yam ca nayatha // tasmai ... / ... sarma yachata "Which one ye rescue and which one ye lead here and there/this way and that, O heavenly ones, to that one ... extend protection." (32)
b. IV.43.7ab (=44.7ab) iheha yad vam samana paprkse / seyam asme sumatir vajaratna "Since I have strengthened you two equally (with praise) here and there, so let (your) good favor be among us, O you having treasure as booty."
In the second of these instances Geldner explains that the poets are travelling singers who therefore have offered their praise in various places. (33)
Two other occurrences of iheha show different values, however. In (32a) the notion 'here and there' does appear to approach a universal 'everywhere', whereas in (b) the value is exclusively specifying ('just here', as opposed to other sacrificial grounds): (34)
(32) a. V.30.10ab sam atra gavo (a)bhito (a)navanta / iheha vatsair viyuta yad asan "The cows lowed together on all sides there, since they were separated here and there/everywhere from their calves."
b. VII.59.11 iheha vah ... / ... / yajnam maruta a vrne "I wish (to entice) you to (this) worship just here, O Maruts."
iheha is also the only amredita in the Rigveda which itself enters into a longer compound structure, the bahuvrihi ihehamatara:
(33) VI.59.2cd samano vam janita bhratara yuvam / yamav ihehamatara "You two (viz. Indra and Agni) have the same begetter, you are brothers, twins whose mothers are here and there (i.e., one is here, the other there)."
[section]14. Moving now to preverbal amreditas, we note once again four values: iterative, intensive, elative, and completive. (35) Examples of iterative amreditas are the following:
(34) a. V.58.5ab ara ived acarama aheva / pra-pra jayante akava mahobhih "Like wheel-spokes unending, like days they are born again and again, undiminished in their greatness."
b. IX.110.5ab abhy-abhi hi sravasa tatarditha / utsam na kam cij janapanam aksitam "For with (thy) fame thou hast bored out over and over, as it were, some kind of inexhaustible well from which men drink."
c. VIII.74.9 sa dyumnair dyumnini brhad / upopa sravasi sravah / dadhita vrtraturye "May this (poem), scintillating with splendor, heap up over and over lofty fame upon fame in the battle against the obstacle."
In (a) the similes of unending wheel-spokes and constantly returning days reinforce the iterative value of the amredita + verb complex, whereas in (c) the iteration is implicit in the polyptotic phrase sravasi sravah 'fame upon fame'. Cf. also VI.48.1 (=14b above), a classic case of Amredita-Haufung in which yajna-yajna ... / gira-gira ... of ab clarify and reinforce the iterative value of pra-pra ... / ... samsisam in cd.
Intensive amreditas include the following:
(35) a. IV.21.9cd ka te nisattih kim u no mamatsi / kim nod-ud u harsase datava u "(Of) what (sort) (will) thy rest (be), and wilt thou not exhilarate thyself? And why dost thou not bristle up, up for giving?" (or: "Why art thou not extremely eager to give?")
b. X.191.1ab sam-sam id yuvase vrsann / agne visvany arya a "All together, O Agni, thou dost take for thyself all (the goods) of the stranger, O bull."
As is often the case with amreditas, the intensive value of sam-sam in (b) is strengthened by the following particle id, and also by visvani of the following pada.
In one instance upopa is indeterminate as to iterative or intensive value:
(36) VIII.51.7cd upopen nu maghavan bhuya in nu te / danam devasya prcyate "Over and over again/Ever more now, O liberal one, the gift of thee, the heavenly one, is further/greatly (bhuyah) strengthened."
As noted by Dunkel (1981: 215), the value of the amredita, whether intensive or iterative, is reinforced by the phrase bhuya in nu echoing upopen nu.
An elative value is seen in the following instances. What I mean by this is that the spatial notion of the preverb (in each passage directional) is extended to an extreme. Only the preverbs apa, pra, and possibly para are represented in this category:
(37) a. V.34.3cd apapa sakras tatanustim uhati / ... maghava yah kavasakhah "The mighty liberal one pushes far off the ... (?) who brings low (his) friend."
b. VII.6.3c pra-pra tan dasyunr agnir vivaya "Agni drove those dasyu's far forth."
c. VII.8.4ab pra-prayam agnir bharatasya srnve / vi yat suryo na rocate brhad bhah "This fire of Bharata is widely famed when (its) lofty light shines out like the sun."
Both (a) and (b) illustrate the elative values of apapa and pra-pra in association with a verb of pushing or driving, where the spatial sense is concretely directional. In (c) the value, in association with srnve, is static but nevertheless elative.
A passage in which the amredita is ambiguous between an intensive and an elative value is the following:
(38) I.38.6ab mo su nah para-para / nirrtir durhana vadhit "Let not baneful destruction smite us far into the distance/utterly destroy us."
As noted by Dunkel (1981: 219, n. 20), the occurrence of the verbal lexeme para + vadhin the Atharvaveda and the Taittiriya Samhita in the sense 'kill' (= simplex vadh-) renders otiose Geldner's statement that the meaning of para + vadh - would not fit the passage here. But if para + vadh-does mean simply 'kill', then para-para + vadh- would signal an intensive version of this value: 'utterly destroy'. However, the possibility also remains that this passage shows an inversion of the formula badh-nirrtim paracaih: I.24.9c badhasva dure nirrtim paracaih and VI.74.2c are badhetham nirrtim paracaih (cf. also X.59.1[-4]d parataram su nirrtir jihitam). In such instances a deity is beseeched to press off destruction into the distance/far away (or the singer wishes that destruction will of its own accord recede far away). It is therefore possible that in the present passage the formula is refashioned such that destruction is enjoined not to smite the singers far into the distance, with para-para effectively equaling paracaih or parataram.
In two passages the elative value of pra-pra yields, in association with as and stha, a sense of preeminence that approaches superlativity:
(39) a. I.150.3c pra-pret te agne vanusah syama "May we be preeminent/at the very fore/before all as thy worshippers, O Agni."
b. I.40.7cd pra-pra dasvan pastyabhir asthita / antarvavat ksayam dadhe "The worshipper has become preeminent in regard to dwellings; he receives a dwelling pregnant (with goods)." (36)
There remain two instances of the amredita pra-pra that show values related to but different from those of iteration, intensiveness, and elativity demonstrated above. The first of these involves an ellipsed verb of motion and appears to possess the force not so much of iteration as durativity. This results in part from the intransitivity and near stativity of the underlying verb, which is not conducive to repeated action but rather denotes constancy of position:
(40) I.129.8a-c pra-pra vo asme svayasobhir uti / parivarga indro durmatinam / dariman durmatinam "(Let) your Indra (go) ever at the fore among us with his self-glorious aids in the overcoming of hostilities, in the splitting of hostilities."
It is possible that this passage is not to be understood as containing a sentence with ellipsed full verb (like etu) but rather as similar to (39a), in which case we have a nominal sentence with pra-pra signaling preeminence ("Let Indra be preeminent among us in the overcoming ...").
In the final passage to be discussed here pra-pra seems to present a sense of completion:
(41) V.5.5 devir dvaro vi srayadhvam / suprayana na utaye / pra-pra yajnam prnitana "Open up, ye heavenly doors, offering us good access to aid. Make the worship completely fulfilled."
Here again, the relationship to the elative value ("make ... very full") is clear.
[section]15. Numeral amreditas take us back morphologically to the nominal type, but they show some differences, in part owing to their special status as numerals, in part as the result of what appear to be semi-grammaticalizations of numeral amreditas as a type. Thus, on the analogy of a nominal amredita such as vaje-vaje in VII.38.8a vaje-vaje (a)vata vajino nah "Aid us in battle after battle/in every battle, O coursers," we would expect the numeral amredita eka- -eka- to mean 'one after one/one after the other/every one (=everyone)'. This is indeed what we find; moreover, as with the nominal types, the sense of individuation is particularly pronounced:
(42) a. I.123.8 sadrsir adya sadrsir id u svo / dirgham sacante varunasya dhama // anavadyas trimsatam yojanany / ekaika kratum pari yanti sadyah "Having the same appearance today and having the same appearance tomorrow, they follow the long law of Varuna. Faultless, they (complete) the thirty course lengths; one after the other they encompass their intention (i.e., intended path) in one day."
b. III.29.15cd dyumnavad brahma kusikasa erira / eka-eko dame agnim sam idhire "The Kusika's have instituted the splendorous formulation. Every/Each one (of them) has kindled/One after the other they have kindled the fire in the house."
c. I.20.7 te no ratnani dhattana / trir a saptani sunvate / ekam-ekam susastibhih "(As such) create wealth for us, thrice in the amount of seven, for the presser, one for one/one each in return for the praises."
In (a) the reference is to the dawns conceived either as consisting of a troop of individuals or in reference to various dawns appearing from day to day. In (b) the sense is individualizing in reference to the Kusika's. Finally, (c) illustrates the sense of distributive individuation in which one ratnam is mapped onto each susastih.
Similar to (42c) in that the amredita represents an attribute of the object is the following passage:
(43) VIII.70.14 bhuribhih samaha rsibhir / barhismadbhi stavisyase // yad ittham ekam-ekam ic / chara vatsan paradadah "Thou wilt be somehow praised by many rishis whose barhis has been prepared, if thou wilt thus give out calves, only one each (to the singers), O Sara."
This passage represents a sarcastic reference to Sara, a stingy patron who gives only one cow as daksina to each of the rishis who has performed his service. The limitative sense is signaled by id, whereas ekam-ekam is distributive over the number of rishis active in the worship. Hence, as in (42c) it can modify a plural head vatsan.
The final eka- -eka- passage requiring discussion is the following:
(44) V.52.17ab sapta me sapta sakina / ekam-eka sata daduh "The seven-after-seven (37) mighty ones have given me hundreds (of cows), one (hundred) each."
This is the only Rigvedic amredita showing any variation in the form of the individual members. Wackernagel (1930: 395) says of this form: "ekam-eka mit Angleichung an das unmittelbar folgende sata, dessen Attribut es ist." This interpretation maintains the distributive value of the eka- -eka- amredita seen in (43). At the same time, the "broken" subject amredita sapta ... sapta (for more on this structure, see [section]20 below) forms a nice rhetorical counterpoise to the amredita object attribute: each sapta is mapped onto an ekam (satam), the total result being sata.
The remaining numeral amreditas are dva-dva, panca-panca, and sapta-sapta. These seem to possess the values 'two-by-two/two each', etc. Hence, we may speak of a semigrammaticalization of such notions as "pairwise" and the like:
(45) a. X.48.6a-c aham etan chasvasato dva-dva / indram ye vajram yudhaye (a)krnvata // ahvayamanan ava hanmanahanam "I (have smitten) these snorting ones two-by-two, who have incited Indra (and) his cudgel to battle. With a smiting I have smitten down those provoking (me)."
b. III.55.18a-c virasya nu svasvyam janasah / pra nu vocama ... // solha yuktah panca-panca vahanti "We shall proclaim now the good team of horses of the hero (viz. Indra), O men ... They convey (him) here yoked in six rows of five each."
c. X.75.1a-c pra su va apo mahimanam uttamam / karur vocati sadane vivasvatah // pra sapta-sapta tredha hi cakramuh "The singer will proclaim well your highest greatness, O waters, in the seat of Vivasvat. For they have stridden forth in three (directions), seven (in) each."
Note the co-occurrence of the amredita with a numeral multiplicative in -dha in both (b) and (c). Presumably, the amredita specifies the size of each entity to be multiplied by the multiplicative, not as a lump sum but rather as a set of repetitions of individuated entities. Thus, solha yuktah panca-panca does not mean merely "thirty" as a total number but rather "thirty" as a grouping of five entities replicated six times. Similarly, sapta-sapta tredha refers to a grouping of seven rivers replicated three times (in the east, the center, and the west, cf. Maurer 1986: 203). However, in the following passage it is difficult to see what kind of an entity composed of seven sapta-sapta represents:
(46) X.55.3 a rodasi aprnad ota madhyam / panca devan rtusah sapta-sapta // catustrimsata purudha vi caste / sarupena jyotisa vivratena "He (viz. Indra) has filled up the two worlds and the middle space; the five (times/grouped in) seven heavenly ones according to their times. He looks out many times with his thirty-four lights having the same form (but) following different laws."
That panca devan ... sapta-sapta is meant to indicate roundly thirty-five is clear from the following more precise catustrimsata. Sayana saw the problem and attempted to explain the amredita as designating the totality of those things in the world occurring seven in number (Geldner ad loc.). How, ultimately, we are to understand sapta-sapta in this instance is presently unclear to me. (38)
[section]16. There remains only the unique verbal amredita piba-piba (id):
(47) II.11.11a (=X.22.15a) piba-pibed indra sura somam "Drink over and over/Drink, drink the soma, O hero Indra."
If we are to understand the verbal amredita as showing the same iterative semantic value as that seen in the case of many of the other amredita types, then the first of our suggested renderings would appear to be appropriate. However, the very nature of the imperative (N.B. the only verbal category represented) (39) renders it particularly subject to affective doubling (cf. English, where one frequently hears "drink, drink!" said to somebody in order to urge him/her to imbibe some liquid), and the paucity of the data does not allow an immediate resolution of this question. (40)
Surely, however, the most obvious question surrounding verbal amreditas is why they should be practically nonexistent. Why, despite the obvious potential for a construction such as *srnavac-chrnavad dhavam "he will hear over and over again (our) call" to signal iteration, did the Rigvedic bards not avail themselves of this possibility? The answer, I believe, lies in the functional overlap of the amredita and another seemingly unrelated category of Sanskrit grammar: the intensive. One of the important conclusions of Schaefer in her study of the Vedic intensive (1994) is that the semantic value of this formation in the Rigveda is primarily iterative. In n. 28 above we noted laconically that the intensive form atineniyamanah of VI.47.16ab (=19c) complemented the amredita anyam-anyam ("repeatedly leading one after the other beyond [obstacles]"). Moreover, in our discussion of IV.32.4 (=26c) we pointed out that the intensive nonumah possessed iterative value and stood in complementary relationship with the amredita object asman-asman of the following verb ud ava. Other passages showing intensive verb forms together with amreditas are I.131.5, 181.4; II.25.1 (=21b); III.53.8; IV.15.6; VI.36.5; and VIII.55.4, 91.2. Moreover, Schaefer (1994: 197-98) points out the sequence III.56.6ab.7a trir a divah savitar varyani / dive-diva a suva trir no ahnah ... /// trir a divah savita sosaviti "Thrice hither by day, O Savitar, day after day, impel hither desirable things, thrice to us by day ... / Thrice hither by day does Savitar impel (repeatedly) ..." Here a simplex verb + amredita adverb in 6 essentially commutes with an intensive verb + simplex adverb in 7. (41) Thus, the Rigvedic bards apparently felt no compulsion to resort to a second derivational pattern in order to secure a semantic value already available to them through verb morphology. It is therefore no accident that neither pa 'drink' nor stu 'praise', the only two verbs forming actual or putative amreditas in the Rigveda, form intensives in the Vedic language. A subsidiary reason for the absence of verbal amreditas might well have been that most such forms would have been trisyllabic, some even quadrisyllabic (*havamahe-havamahe) and would have been avoided on those grounds alone (cf. n. 15 above). In any event, of course, verbal iteration could have been signaled by employing a simplex, non-intensive verb together with an adverb such as sada (cf. I.12.2 [=5a]) or sasvat or, for that matter, an adverbial amredita such as dive-dive.
[section]17. Although verbal amreditas are nearly non-existent in the Rigveda, one finds a number of verbal constellations which, while not themselves amreditas, approach this category in form. First among these is the collocation stuhi stuhid in the following passage:
(48) VIII.1.30ab stuhi stuhid ete gha te / mamhisthaso maghonam "Praise over and over./ Praise, praise! These are indeed the most liberal of the liberal ones to thee."
It will be immediately apparent that there is no observable difference, contextually or semantically, between the amredita piba-pibed and the juxtaposed iteration stuhi stuhid. (42) The fact that both verb structures are followed by id (rendering the second ambiguous on accentual grounds), not, of course, in its delimiting but rather in its emphasizing value, further attests to the identity of context in these two passages.
I believe it is fair to say that the hesitation between amredita and iteration in the case of juxtaposed verbs is the result of the narrow, almost invisible line between affective iteration on the level of form and content-based iteration on the level of meaning, a point discussed relative to piba-piba (id) above. Whether or not we agree with it in a given instance, the decision on how to understand these various collocations was made by the redactors of the Rigvedic text.
Very similar to the above is the following passage:
(49) X.14.7ab prehi prehi pathibhih purvyebhir / yatra nah purve pitarah pareyuh "Go forth, go forth along the earlier paths where our previous fathers have gone forth."
Here the iterated structure is a Preverb + V complex. The iteration obviously enhances the remarkable p-alliteration which informs this distich.
Moving yet a further step away from the verbal amredita proper, we find the following passage in which the juxtaposed verbs are construed with different, antonymous preverbs. Hence, the content-based meaning cannot be associated with iteration of an identical action:
(50) X.130.1 yo yajno visvatas tantubhis tatah / ... // ime vayanti pitaro ya ayayuh / pra vayapa vayety asate tate "The worship which has been extended on all sides with its threads ..., (that) do these fathers weave, who have come hither. They sit on the extended (seat of the worship) (and say,) 'Weave this way, weave that way.'"
[section]18. What is true of the imperative, the most affective form of the verb, is true as well, mutatis mutandis, of the vocative, the most affective form of the noun. The chief mutandum in this case, however, is that a call to an individual is not to be understood as iterative ("O X over and over/O X and again X"); and this is probably the reason that the vocative is the one case that is not represented among nominal amreditas. There is, however, one example of juxtaposed vocative iteration in the Rigveda:
(51) X.146.1a-c aranyany aranyany / asau ya preva nasyasi // katha gramam na prchasi "Aranyani, Aranyani, thou there who dost, as it were, disappear, how is it that thou do not ask about the village?"
In this passage, the fearful call for protection of one who must sojourn alone in the forest at night, the affective element is particularly striking. (43)
The semantic and contextual conditions associated with the occurrence of amreditas and adjacent word iterations lead us to conclude that the Rigvedic amredita is an iconic but not an affective formation. Its primary associated semantic feature is that of iteration; but where the formal iteration is associated primarily with affective value an amredita is not likely to occur. Hence, the total absence of vocative amreditas and the near-absence of imperative ones in the Rigveda. This consideration apparently, then, settles the issue, posed above, as to whether piba-piba is iterative or affective, especially in view of the absence of an attested intensive for this verb in the Rigveda. Further evidence for this conclusion is provided by two occurrences of non-amredita noun juxtapositions in the nominative case which happen to appear in the same stanza:
(52) X.115.9 iti tvagne vrstihavyasya putra / upastutasa rsayo (a)vocan // tams ca pahi grnatas ca surin / vasad vasal ity urdhvaso anaksan / namo nama ity urdhvaso anaksan "Thus, O Agni, have the sons of Vrstihavya, the Upastuta's, (and) the rishis spoken to thee. Protect those singers and the lords. (Saying) 'Halloo! Halloo!' they have approached upright; (saying) 'Reverence! Reverence!' they have approached upright."
If we compare these repetitions with the noun amreditas seen in [section]8 above, the difference is immediately obvious: these are ritual shouts, those are syntactically integrated case forms with universal semantic force. Put another way, these are affective collocations, those are not. Hence, those are amreditas, these are not. (44)
[section]19. In our discussion of pronominal amreditas ([section]12) we noted that some of these forms, particularly those involving relative pronouns, represent grammaticalizations functioning as indefinite pronouns. The same is true of non-amredita juxtapositions of the interrogative/indefinite pronoun, especially in the construction ka- ka- cid:
(53) a. VIII.102.20 yad agne kani kani cid / a te daruni dadhmasi / ta jusasva yavisthya "If we set out any wooden logs for thee, O Agni, enjoy those, O youngest one." b. VIII.73.5 yad adya karhi karhi cic / chusruyatam imam havam / anti sad bhutu vam avah "If you two should hear this call somewhere or other today, let your aid be present."
A similar phenomenon that may be brought under this same heading is the employment of evaiva in correlation with yatha-yatha (21d above). But correlation with an amredita is not a necessary condition for evaiva to appear. Cf. the following:
(54) X.44.6cd.7a-c na ye sekur yajniyam navam aruham / irmaiva te ny avisanta ... /// evaivapag apare santu dudhyo / (a)sva yesam duryuja ayuyujre / ittha ye prag upare santi davane "The ones who were not able to mount the ship of the worship, those have finished their day (standing) still ... / In the very same way let those having bad thoughts in later days remain behind, whose ill-yoked horses have been yoked. So (also) (on the other hand,) (let those go forward) who in the future are forthcoming in giving."
Here it appears that the force of evaiva is to affirm a very strong parallelism across the stanza boundary, similarity of fate and continuity of idea being thus emphasized through iteration across what is normally a significant discourse breaking point. As was the case in (49)-(52) (cf. n. 44), this passage is once again from the Popular Rigveda.
[section]20. In [section]15 above we characterized the collocation sapta me sapta in V.52.17a (=44) as a "broken" amredita. Diachronically, this must represent a pre-univerbation stage of the amredita, when Wackernagel's Law allowed the enclitic me to gravitate to second position between two independent words which were later to be treated as one with a consolidated accentual pattern. Evidence for this is the fact that the broken structure does not differ in meaning from the other numeral amreditas seen in [section]15. This particular hymn must be quite early, as evidenced also by the unique ekam-eka, whose amredita form has here not yet been fixed as invariant. Related to the broken amredita is the following, this time from consecutive stanzas of a late hymn:
(55) X.119.9.10 hantaham prthivim imam / ni dadhaniha veha va / kuvit somasyapam iti /// osam it prthivim aham / janghananiha veha va / kuvit ... "Smash! I will set down this earth here or there. Can't you see that I'm drunk with soma? / Straightway for sure I will smite the earth again and again here or there. Can't you see ...?"
That this case is different from sapta me sapta may be appreciated by recognizing the distinction in meaning between iha veha va here and the eight occurrences of iheha. The latter, with its primitive conjunctive iterative value 'here and there/everywhere', is the foundation of the present collocation with its alternative iterative value. Iha veha va is therefore not an archaism but rather an innovation within the late Tenth Mandala.
[section]21. The remaining two juxtaposed word-iterations in the Rigveda are the following:
(56) a. I.17.8 indravaruna nu nu vam / sisasantisu dhisv a / asmabhyam sarma yachatam "O Indra (and) Varuna, (precisely) now, when the thoughts are desiring to win you two, extend protection to us." b. IV.18.3b na nanu gany anu nu gamani "I will not not go along (that path). Now I will go along (that path)."
The first of these collocations is not altogether clear, because nu may possess either a temporal or a discourse continuative value ("now then"). It is therefore in principle possible that we have two different nu's here. Alternatively, and more likely in this case, both could be temporal, and the force of the collocation would be to emphasize the critical urgency of the present in the gods' response. Cf. Ovid, Fasti 2.745 nunc, nunc properate, puellae! (Wills 1996: 108). The second example, on the other hand, has nothing to do with the phenomena we have discussed in this study but is rather a litotic affirmation of a positive resolution by negation of its opposite: na (nanu gani), a value which is insured by the immediately adjacent affirmative clause anu nu gamani.
[section]22. We complete our discussion of amredita-like constellations by considering four forms which differ from all those previously discussed in that they are neither full-word juxtapositions nor compounds consisting of two full words. Rather, they are descriptively single word reduplications of disyllabic sequences having the overall shapes CVCaCVCa-. In each instance the value of the word is intensive or iterative. All are adjectives, and all belong, once again, to the Popular Rigveda:
(57) a. X.103.1 asuh sisano vrsabho na bhimo / ghanaghanah ksobhanas carsaninam // ... / satam sena ajayat sakam indrah "Swift, sharpening (his weapon) like a fearsome bull (his horns), smiting over and over, the bestirer of the people ..., Indra defeated a hundred armies together."
b. X.85.11 rksamabhyam abhihitau / gavau te samanav itah // srotram te cakre astam / divi panthas caracarah "Harnessed with stanza (and) song, thy two cows go united (as a pair). The two wheels were thy hearing mechanism, the path stretche(s/d) continuously/on and on/move(s/d) back and forth in heaven."
c. I.164.48 dvadasa pradhayas cakram ekam / trini nabhyani ka u tac ciketa // tasmin sakam trisata na sankavo / (a)rpitah sastir na calacalasah "Twelve (are the) wheel-rims, one (is) the wheel, three (are) the naves. And who has perceived that? On it are set together three hundred sixty (days) like pegs. They do not move around."
d. X.119.12ab aham asmi mahamaho / (a)bhinabhyam udisitah "I am very powerful; unto the clouds have I mounted."
Except for the last of these, an iterated form of the readily accessible adjective maha- (with apparent metrical lengthening to avoid a tetrabrach), the precise formation of these items is not at all clear. (45) However, they all appear to be akin to the types we have been discussing throughout this study, attesting once again to the iconic association of plurality of form and plurality (especially iterativity and intensity) of meaning. (46)
THE AMREDITAS OF THE RIGVEDA
I. Nouns and Adjectives (Including Proper Nouns and Participles)
agnim-agnim I.12.2a; VI.15.6a; VIII.60.17a
ahar-ahah I.123.9d; II.30.1c; VIII.24.24c, 26.12c; X.52.2c.3c
uktha-ukthe VII.26.2a; X.45.10b
karman-karman I.102.6b; X.28.7b
grham-grham I.123.4a, 124.11d; VIII.91.2b
grhe-grhe I.28.5a, 71.4b; V.11.4b; X.91.2a
janam-janam V.15.4b; X.91.2c
jane-jane V.65.2d, 67.4b
dame-dame I.128.4a; IV.7.3d; V.1.5c, 6.8d; VI.74.1c; VII.15.2b
dive-dive I.1.3b.7a, 31.7b, 34.7a, 89.1d, 123.4b, 136.3b.c; II.9.5b, 30.2d.11d, 34.7b; III.4.2b, 29.2c.14c, 30.12b, 35.3d, 51.1d, 52.8c, 56.6b; IV.8.7a; 15.6c, 54.6b; V.20.4b, 49.1d; VI.15.2d, 30.2c, 32.5d, 47.21a, 71.6b; VII.32.19a; 98.2b; VIII.4.10c, 12.28b, 15.6c, 31.4b, 45.12a, 53.2b, 98.8c; IX.1.5b, 61.20b, 75.4d, 101.6d, 107.19b; X.37.7c, 87.22c, 92.8d
devam-devam I.26.6b; VII.79.5a; VIII.12.19a(=27.13a), 27.13b.c
devasya-devasya X.1.5c, 136.4c
dyavi-dyavi I.4.1c, 25.1c
dhanam-dhanam II.23.13b; IX.85.8d; X.156.1c
dhiyam-dhiyam I.168.1b; VI.49.8d
navyam-navyam I.159.4c; X.96.11b
pade-pade V.41.15a; IX.73.4d
panyam-panyam VIII.2.25a, 74.10d
parus-paruh I.162.18d; X.97.12b
bhadram-bhadram I.123.13b; VIII.93.28a
bhare-bhare I.100.2b, 102.4b; V.43.2d; VII.32.24d, 82.9a; X.67.9d
made-made I.81.7a; VIII.13.7c; X.120.4b
yajna-yajna I.168.1a; VI.48.1a
yajne-yajne I.136.1e; V.5.9c; VIII.59.1c; X.93.2a
yuge-yuge I.139.8d; III.26.3b; VI.8.5a, 15.8b, 36.5d; X.94.12a
rane-rane I.74.3c; VI.16.15c
rupam-rupam III.53.8a; VI.47.18a
vane-vane V.11.6b; X.91.2b
vayo-vayah V.15.4c; VIII.55.4b
vaje-vaje I.30.7b, 138.3g; VI.61.12c; VII.38.8a
vamam-vamam IV.30.24a; X.76.8c
visam-visam VII.74.1d; X.43.6a, 84.4b, 91.2d
vise-vise I.27.10b; IV.7.1d.4d; V.8.5a; X.40.1c
viso-visah VI.49.2a; VIII.74.1a
sadah-sadah X.15.11b, 76.1d
sute-sute I.9.10a; III.36.1c; VI.45.28a
II. Pronouns (Including Relative Adverbials)
anyad-anyad II.24.5c; III.38.7c
anyam-anyam V.44.11c; VI.47.15c.16b; X.117.5d
anyanya I.62.8d, 95.1b, 113.3b
tat-tad I.46.12a, 155.4a(=X.23.5c); VIII.39.4a
tam-tam I.132.6b.c; VI.42.3d; VIII.68.7a; X.125.5c
yatha-yatha IV.19.10c, 54.5c; VIII.39.4b; X.100.4c, 111.1b
idam-idam VII.59.1a; VIII.21.9a
iheha I.181.4a; III.60.1a; IV.43.7a(=44.7a); V.30.10b, 47.5d; VI.59.2d [ihehamatara]; VII.59.11a; X.131.2c
punah-punah I.92.10a; III.5.7d
upopa I.126.7a; VIII.51.7c, 74.9b
pra-pra I.40.7c, 129.8a; 138.1a, 150.3c; III.9.3c; V.5.5c, 58.5b; VI.48.1c; VII.6.3c, 8.4a; VIII.69.1a; IX.9.2a
V. Numerals (Including eka-)
eka-ekah III.29.15d; V.61.1b
ekam-ekam I.20.7c; VIII.70.14c
dva-dva VIII.68.14a; X.48.6a
sapta-sapta X.55.3b, 75.1c
AMREDITAS AND PROTO-INDO-EUROPEAN GRAMMAR
In this study I have eschewed a treatment of amreditas from an Indo-European perspective. The reason for this is that unless we hold to the strict Sanskrit definition of the type as a compound with accent on the first member alone, we are left with a universal iconic type of word juxtaposition which exists probably in every language that has ever been spoken. Thus, to cite the one non-Indo-European language I know well, Biblical Hebrew shows such juxtapositions as 'is 'is 'every man', yom yom 'day after day, every day', sana sana 'year after year, every year', l[TEXT NOT REPRODUCIBLE IN ASCII]dor dor 'for future ages' (lit. 'for generation [after] generation'), babboqer babboqer 'every morning'. Haspelmath (1997: 180-81) provides examples from other non-Indo-European languages: Tagalog taon-taon 'every year', Manggarai leso leso 'every day', Hungarian ki ki 'everyone' (ki 'who'). Note that such iterations are particularly likely to signal countable entities that recur in human cognition, especially periods of time and human individuals. The sense is overwhelmingly that of individuated universality or totality.
It is in this context that we must view such Indo-European structures as Mycenean Greek weteiwetei, Hittite MU.KAM-ti MU.KAM-ti, Classical Armenian ami ami 'year after year, every year', Cypriote Greek amatiamati, Hitt. UD-at UD-at, Cl. Arm. awur awur 'day after day, every day'. Cf. Dunkel 1981: 214, n. 2 and Dressler 1968: 39-42. The latter terms such collocations epanadiplosis, but I would prefer to reserve this designation for rhetorical rather than these (in effect) grammaticalized syntactic juxtapositions, the sort seen, for example, in Lucretius, De Rerum Natura 3.894 Iam iam no domus accipiet te laeta neque uxor, Keats, Ode to a Nightingale 31 "Away! away! for I will fly to thee," Coleridge, The Rime of the Ancient Mariner, 167-68 "Alone, alone, all, all, alone, / Alone on a wide, wide sea." That the locative is so often represented (Sanskrit, Greek, Hittite, possibly Classical Armenian (47)) is a result of the fact that most of the iterated juxtapositions we have been discussing refer to periods of time within which people live and act. I do not believe that much if anything should be made of this in regard to the issue of a putative original reconstructible case for amreditas. (48)
The situation is different, on the other hand, for Greek [TEXT NOT REPRODUCIBLE IN ASCII] 'entirely', a linguistic form equivalent to the Sanskrit amredita in its univerbation, accent, and sandhi. Even though the Vedic language has no amredita cognate with this form, I find these features to be far more telling in any consideration of the possibility of reconstructing amreditas for Proto-Indo-European than the universal word-juxtapositions mentioned above. Furthermore, the intensive value of the Greek form equals that of some Rigvedic adverbial amreditas such as maksu-maksu. The same is true of the [TEXT NOT REPRODUCIBLE IN ASCII] of Homeric [TEXT NOT REPRODUCIBLE IN ASCII] (Od. 17.525) 'rolled on and on', despite the ambiguity concerning the accent of this form relative to Rigvedic pra-pra; and to this we may add Hittite para para discussed by Dunkel (1981: 216-18). Although not univerbated, this form provides a tertium comparationis, at least for the juxtaposition, if not the amredita.
In sum, there can be little doubt that in Proto-Indo-European one could form the juxtaposition *pro pro in order to create an elative or intensive value associated with iteration of verbal action ('forth and forth, more and more, ever more'). Similarly, we may assume that at least one way of forming an indefinite relative pronoun was to juxtapose the relative or interrogative: *[k.sup.w]o/i- [k.sup.w]o/i-, *yo- yo-. But these should not be thought of as amreditas as that term is used in Sanskrit grammar. Similarly, as part of a universal tendency in human language, one could iterate and juxtapose words for units of time: day, year, month, etc. But that such sequences represented amreditas in the sense in which that term is employed in Sanskrit grammar is unlikely. Over time, however, juxtaposed forms could be univerbated and consolidated under a single accent. This is seen clearly in Greek [TEXT NOT REPRODUCIBLE IN ASCII] and is possibly implicit in Homeric [TEXT NOT REPRODUCIBLE IN ASCII]. (49) But the full development of the type as a compound is a feature only of Sanskrit.
An earlier version of this paper was presented orally at the 210th meeting of the American Oriental Society, Portland. Oregon, March 12-15, 2000. I am particularly grateful to Stephanie Jamison and George Cardona for exceedingly helpful remarks made to me at that time and in subsequent communications (both e-mail and per litteras) on the Rigvedic intensive and on Panini's treatment of amreditas, respectively. Some of these will be acknowledged separately in the text and the footnotes. Naturally, I take full responsibility for any errors that may appear here.
1. E.g., Whitney 1889: 488 (last section but one under copulative compounds, headed "Repeated Words"), Macdonell 1916: 281-82 (last compound type discussed), Renou 1952: 123-24 (small print under heading "classement des composes"). An exception is Delbruck 1888: 51-55, who places his discussion of amreditas at the very beginning of his discussion of "aus Satzteilen gebildeten Gruppen" (51).
2. In fact, amreditas are not considered compounds at all by Panini but merely iterated words; and the term amredita refers only to the second of the repeated words, which is low pitched (Astadhyayi, 8.1.1-8.1.3, Bohtlingk 1887: 428).
3. The relevant rule is 8.1.4 nityavipsayoh. Here nitya clearly refers to repetition. vipsa (vi + ap- 'pervade'), literally "the wish of an agent to pervade some object with a particular attribute or quality," is illustrated in the scholiastic tradition by sentences such as gramo gramo ramaniyah "village after village/every village is pleasant" and purusah puruso nidhanam upaiti "every person goes unto death" (Bohtlingk 1887: 429). This latter value is to be understood as distributive over each of the objects or things designated in the repetition. Bohtlingk, without further explication, translates nityavipsayoh as "wenn man ein 'Immer' oder ein 'Jedes' ausdrucken will" (ibid.).
4. Not included in this number is the form nana 'variously', which has become relexicalized in the Rigveda but which is no doubt an original amredita. Cf. n. 33.
5. Hitherto the most complete listing, and the one which forms the basis of most of the discussions in the handbooks, was Collitz 1882. However, Collitz's collection fails to list 19 different amredita forms and contains other omissions and misprints.
6. Including the first member of the bahuvrihi ihehamatara at VI.59.2. Cf. [section]13.
7. Including ahar-ahah (6x), an original accusative of extent normally having the adverbial value 'day after day/every day'.
8. Including dive-dive (47x) and vise-vise (5x). Aside from these two forms the dative is represented by only caksuse-caksuse and sirsne-sirsne, each of which occurs only once.
9. The form vamam-vamam appears once as nominative and once as accusative.
10. We have treated the two occurrences of idam-idam as adverbials.
11. Indeed, Lubotsky (1997) treats stuhi stuhid as an amredita (stuhi-stuhi).
12. In what follows the individuation associated with the semantics of amreditas will often be captured by translations of the sort 'X after X', where this produces an idiomatic English phrase.
13. Our differing translations of agnim-agnim ... priyam-priyam and devam-devam ... indram-indram are based on the variant semantic roles of devam and priyam. The former signifies a permanent status, the latter describes a quality. Hence, it would be redundant to term Indra "ever a heavenly one," but Agni may be characterized as "ever dear." The translation "I will praise" of grnisani in VIII.12.19 (= 4) is based on Geldner's translation "will ich unterlegen" of the identically formed upastrnisani at VI.44.6. On the other hand, in VI.15.6 (= 5b) the parallelism with duvasyata of pada a leads us to translate grnisani here as second-person plural imperative.
14. Or, more concretely, if also more obscurely, "Let old age be deposited in my every step/in every step for me." Stephanie Jamison raises the possibility (per litteras) that pada-here could refer to the ritual ground.
15. Only 12 out of the 291 different occurrences of amreditas are trisyllabic, and two-thirds of these appear in the late Mandalas I (2x) and X (6x). Seven are found in trimeter verses, and in four of these the first member of the amredita begins following the caesura and occupies the entire trisyllabic break with the second member beginning the cadence: I.94.4b citayantah parvana-parvana vayam, I.101.4b ya aritah karmani-karmani sthirah, X.37.8b bhasvantam caksuse-caksuse mayah, X.37.9d ahnahna no vasyasa-vasyasod ihi. A fifth shows the first member in the opening, while the second forms the break: VI.27.3c na radhaso-radhaso nutanasya. In two instances, however, representing consecutive padas of the same hymn, the general rule proscribing a word-internal caesura is broken: X.1.5bc yajnasya-yajnasya ketum rusantam // pratyardhim devasya-devasya mahna. In view of X.136.4c munir devasya-devasya it is probably not necessary to assume, with Bartholomae 1907: 167, an original text yajnayajnasya ... // ... devadevasya for X.1.5bc. Rather, inferior craftsmanship in a late period of Rigvedic versecomposition should be entertained. But III.4.1 belongs to an earlier period when, we may presume, verse-opening *samidha-samidha would have been proscribed. It is possible that a form such as the adverbially employed neuter accusative (=nominative) ahar-ahah provided an inexact parallel which allowed the poet here, under the pressure of rhetorical responsion and isosyllabism, to avoid the ill-fitting *samidha-samidha by substituting a nominative singular samit-samit, albeit of a feminine noun.
16. On the interpretation of visa as dative, cf. Oldenberg 1912: 33, who is followed by Geldner ad loc. The padapatha transmits visah.
17. Geldner (ad X.40.1c) takes vibhu- to mean "eigentlich 'genugend, hinreichend'" and translates vibhvam vise-vise in both these passages as "der fur jeden Clan zu haben ist."
18. Although there are seventeen instances in the Rigveda in which an amredita is more frequent than its simplex, fifteen show a difference of no more than two occurrences between the two. Most of these (11x) show a single amredita beside no occurrences of the simplex (gatre-gatre, tunje-tunje, navya-navyah [nom. pl. fem.], nidamnidam, nithe-nithe, parvana-parvana, parvani-parvani, praye-praye, lomno-lomnah, sirsne-sirsne, and suca-suca), two show a difference of one with at least one occurence of the simplex (parus-paruh [2:1], yuge-yuge [6:5]), and two show a frequency difference of two (panyam-panyam [2:0], yajna-yajna [2:0]; on this latter form cf. our discussion in text above). The only two cases of greater predominance of the amredita are dive-dive (47:15), on which see below in the text, and, admittedly, visam-visam (4:1). In the former case the simplex dive means only 'heaven'. vise-vise would show a normal relationship vis-a-vis vise (5:8) but an abnormal one relative to visi (5:2).
19. It is sometimes implied by juxtaposition, if not outright alleged, that the interpretation of vise-vise as a locative is supported by Avestan, which shows the sequence nmane nmane visi visi "in every house, in every district" at Videvdat V.10 (cf. Wackernagel 1905: 146, Brugmann 1906: 56, and Dressler 1968: 43). Dressler even makes the Avestan into a kind of clinic on Vedic vise-vise. He notes first that the best manuscript tradition reads visi vise and nearly all others show vise vise. Dressler reports a letter from Karl Hoffmann dated 14 November 1967 in which the late, great Erlangen scholar interpreted vise as "eine banale Verschreibung" for either visi or vise. Dressler then states that vise vise as lectio facilior (in context beside nmane nmane) must have replaced a more original visi visi attested only in the "wertlose Handschrift" K 2, thus paralleling what he conceives to be the origin of Rigvedic vise-vise. The trouble with all this is that irrespective of whether we read visi visi or vise vise, the meaning of the Avestan iteration is absolutely certain, by virtue of its juxtaposition with the unambiguous nmane nmane. But none of the five occurrences of Rigvedic vise-vise offers any such broad external contextual hints. The decision must be made on a close reading of the passages alone, and when this is done the "problem" surrounding this form evaporates: it looks like a dative, behaves syntactically like a dative, and translates like a dative. Can there be any further doubt about its case status?
20. As a case in point, cf. the collocations anu dyun (17x) and the amredita ahar-ahah (6x). Both are in origin accusatives of extent which are sometimes translated by Geldner "Tag fur Tag," the same rendition which he regularly employs for dive-dive. However, both (especially the first) retain their original values in some passages: IV.4.8d asme ksatrani dharayer anu dyun "Mayest thou (sc. Agni) maintain (thy) dominion among us for all time" and II.30.1c ahar-ahar yaty aktur apam "The gleaming light of the waters drives day after day." In neither of these instances are we to think of a separate daily event. Rather, Agni's dominion is to be constant and unending. Similarly, the waters are said to stop their movement neither for Savitar nor for Indra, but to move constantly throughout time.
21. The frequently cited temporal dative aparaya at VI.33.5a nunam na indraparaya ca syah "Mayest thou be ours, O Indra, now and for the future" does not, as indicated in our translation, possess the same kind of static temporal value as does dive-dive.
22. This form occurs in a restricted metrical distribution which is itself indicative of archaism. Thus, 11 of the 12 occurrences of dyavi are in pada-final position in the phrases adhi dyavi (4x), upa dyavi (5x), and the imitative (diday)ati/(diday)asi dyavi (2x). Its only non-pada-final appearance is in the moved phrase upa dyavi at VI.52.13b ye antarikse ya upa dyavi stha. These and similar determinations have been rendered almost effortless now thanks to the indispensable Concordance of Lubotsky (1997).
23. This is not the place to go into the vexed issue of the distribution of the values 'heaven' and 'day' in the *dyews paradigm. The number of languages in which *dyews and its derivatives show both meanings, including Latin, where the meanings are assigned to different forms resulting from the split of a single original paradigm, is sufficient to demonstrate that this word could originally signal either meaning. As far as the Rigveda is concerned, the form divi (130x) is employed both in the meaning "heaven" (116x) and "day" (14x, usually in the phrase parye divi/divi parye "on the decisive day"). Dressler is therefore wrong in claiming that divi means only 'in heaven' (1968: 45). The instrumental diva (10x), on the other hand, is used only to mean "heaven," and a specific adverbial form diva (26x) with shifted accent is employed to mean 'by day', 23 times in overt association with naktam 'by night'. It is likely, therefore, that the shifted accent of diva was occasioned by the barytone accent of the latter word. Moreover, as noted in the text, in the cases of dyavi and dive even simplex and amredita differ in meaning. Significant, no doubt, in both instances, is the fact that a meaning "for/in heaven after heaven, for/in every heaven" would be excluded on semantic and pragmatic grounds. Although, to be sure, there are numerous references in the Rigveda to three heavens, this would still not satisfy the most common sense of the Rigvedic amreditas, which is that of unlimited repetition.
24. By its pattern of attestation (only IX.86.41b and AV V.6.21) this form shows itself to be innovative rather than archaic. In arguing for the latter, Dressler (1968: 45) cites the non-occurrence of divi 'am Tage' and the rarity of the simplex ahah in the meaning 'am Tage' as militating against a recent composition. But the first of these propositions has been shown above (n. 23) to be incorrect. The second point is technically correct: of 8 occurrences of ahah only one possesses the value 'am Tage' (III.48.2ab yaj jayathas tad ahah ... / amsoh piyusam apibo giristham "When thou wast born, on that day ... thou didst drink the cream of the stalk situated on the mountain"). However, ahardivi could only have been created on the model of other amreditas; and of the 132 different amredita forms in the Rigveda, only dive-dive (47x), pra-pra (12x), and iheha (8x) occur more frequently than ahar-ahah (6x: [I.sup.1], [II.sup.1], [VIII.sup.2], [X.sup.2]). This latter is therefore well ensconced in the language of the Rigveda, and its first two syllables provided a familiar anacrusis for a new poetic creation of the late Rigvedic period.
25. In addition to treating dive-dive as an innovation (for *divi-divi) or a dative retention, there is one further possibility, taught to me by my teacher, Stanley Insler, in 1969. That is to understand dive-dive as originally a sandhi development of the phrase *divaz dive "from day (abl.) to day." If this is truly the source of this form, then it is of course not originally an amredita but has accidentally come to resemble and be treated as one. The problems with this analysis, as I see it, are twofold. First, the phrase "from day to day" is good colloquial English, but I am unaware of other ablative plus dative locutions in the sense 'every' in Vedic. Second, the cross-linguistically common pattern of iteration to indicate universality and individuated totality, particularly in temporal value (cf. Appendix 2), which stands at the basis of the Rigvedic amredita, is a primitive construct in human language and does not require our positing some secondary development in order for it to be instantiated.
26. In Klein 2002. I define this term as repetition of one or more word-final linguistic signs (i.e., [suffixes and] endings). By this usage I oppose homoioteleuton to rhyme, taken to be pure sound repetition at the end of a word without morphological identity of the repeated stretches. Thus, in VI.48.1cd (=14b) ... vayam amrtam jatavedasam / priyam mitram na samsisam the four medial words in-am (amrtam jatavedasam / priyam mitram) illustrate homoioteleuton, the first and last (vayam ... samsisam) rhyme both with the medial set and with each other. So deeply embedded are both rhyme and homoioteleuton in the textuality of the Rigveda that they may on occasion, as here, contribute to a mismatch between the 1st person plural pronoun vayam and the 1st person singular verb samsisam. The actual mechanism of this mismatch was no doubt anacolouthon, as the poet changed the construction in the long stretch of text intervening between subject and verb. Cf. such English parallels as "I--we should (all) be thankful that ...," not infrequently heard in oral communications.
27. On the other hand, there is no instance in the Rigveda of the value described in Panini's sutra 8.1.12 prakare gunavacanasya, whereby an adjectival amredita indicates that its head noun possesses a quality only to a limited extent, e.g., patu patu 'somewhat sharp'.
28. Another important signal of iteration in this passage is the intensive participle atineniyamanah. For more on the relationship between (verbal) amreditas and intensives, cf. [section]16.
29. Also emphatic is the collocation panyam-panyam id at VIII.2.25 panyam-panyam it sotara / a dhavata madyaya / somam ... "O pressers, wash off(?) the ever-praiseworthy soma for exhilaration."
30. Probably because the verb is not dual but plural, Geldner translates samdadhuh impersonally: "Ganz so wie man Bundnisse schliesst." However, a more precise similitude with the prayer expressed in the opening distich would be achieved if Indra and Soma were being beseeched to treat the rishi and his patrons in the very manner that they would treat their contractual allies. But the plural verb suggests that the friendship pact is meant to include other deities as well. I assume a surface reduction or brachyology of the sequence *yatha-yatha yad ... or the like.
31. That is, every instance of some action X occurs either 'here' or 'there', with the additional proviso that at least one instance of the action will occur in each place. Cf. English "now this way, now that," "one time this way, one/another time that."
32. The other instance of idam-idam in the Rigveda possibly shows the same value but may function grammatically as a pronominal adjective: VIII.21.9ab yo na idam-idam pura / pra vasya aninaya tam u va stuse "Which one has previously led us forth unto this and that better thing, that one do I praise for you" (so Geldner, but possibly "Which one has previously led us forth here and there unto what is better ...").
33. It is here that it is profitable to speak of the adverb nana 'variously, separately'. This form easily admits of some original amredita-like sense such as "now here, now there" (cf. iheha), "now one, now the other" (cf. anyam-anyam), e.g., II.12.8cd samanam cid ratham atasthivamsa / nana havete ... "Even the two mounting the same chariot (i.e., charioteer and warrior) call (him) (each) separately." By its accentuation as well this form betrays its origins as a (relexicalized) amredita. An original na-na 'man after man' (so Mayrhofer 1992: 34-35) would account for the relexicalization, because the original nom. sg. *na of the 'man'-word had ceased to be understood. Semantically, however, this derivation leaves something to be desired, because, as we have shown, an amredita built to the word for 'man' ought to mean 'man after man/every man/everyone', but the basic meaning of nana, both in the Rigveda and in the Avesta (Gathic nana) seems to be 'separately, variously', in some passages 'in various places'. I therefore believe that a better starting point would be (a)na-ana 'in this way, in that way' with aphaeresis of the first member. Ana is attested 5X in the Rigveda in the value 'therefore, thus, in this way'. It is an original instrumental sg. masc. and neut. of ayam (cf. fem. instr. sg. anaya), which has become extraparadigmatically specialized in adverbial value and replaced in paradigmatic usage by ena. This analysis will also work for Gathic Avestan, which attests ana in paradigmatic value as instrumental sg. of aii[TEXT NOT REPRODUCIBLE IN ASCII]m (cf. Yasna 28.5 ana ma[theta]ra "through this prayer"). It might even be possible to derive the apparent local value of nana in a passage such as III.54.6c nana cakrate sadanam yatha veh "They (viz. heaven and earth) have made their seat in various places ([congruent to] one here, one there) like a bird" from the local value of the (paradigmatic) instrumental in a phrase such as *ana patha "along this path ([congruent to] here)." Cf. VIII.5.39a makir ena patha gad "Nobody should go along this path." The infrequently invoked but cross-linguistically common process of aphaeresis has recently been exploited to good advantage by Joshua Katz in his study of Indo-European pronouns (1998a,b). Invoking this process here affords us an explanation of nana semantically in line with what we expect of an adverbial pronominal amredita. Cf. the similar semantic values of iheha and idam-idam. For a parallel phenomenon, also in a pronoun, under exactly the same prosodic conditions which we hypothesize for *anana, cf. Biblical Hebrew nahnu (5X) beside normal 'anahnu 'we'. For further details, see Klein MS.
34. I am indebted to Stephanie Jamison for this interpretation.
35. Panini treats amreditas involving pra, sam, upa, and ud as "expletive" (padapurane), i.e., as simply used to fill out the line (Astadhyayi 8.1.6, cf. Bohtlingk 1887: 429). This employment is not evidenced in the text of the Rigveda. Moreover, at least in the case of pra, the non-expletive employment of iterated forms of this preverb is much older. Cf. Appendix 2.
36. Our translation of pastyabhih here is owing to Brereton 1981: 95 n. 45.
37. This seems much better than Renou's "sept (fois) sept" (1962: 29) in that it maintains the normal value of the (here "broken") amredita. In AV XIII.1.3 the Maruts are characterized as trisaptasah 'thrice-seven', and the same adjective is applied to them in RV I.133.6 (where, however, their name does not appear). Moreover, in TS II.2.11.1 the Maruts are said to be saptaganah 'whose troops consist of seven'. Thus, the meaning 'seven after seven' (= 'troop after troop') would seem to yield good sense for the "broken" amredita here.
38. One possibility: panca is somehow reduced from pancadha in the sequence *pancadha devan. Note the presence of purudha in the very next pada. Brachyology or a combination of proximal and distant haplology?
39. Within the Sanskrit scholiastic tradition verbs, both imperatives and non-imperatives, are among those forms cited to illustrate Paninian sutras dealing with amreditas; e.g., bhunkte bhunkte 'eats and eats', pacati pacati 'cooks over and over', and lunihi lunihi 'keep cutting' (ad 8.1.3 and 8.1.4, cf. Renou 1966: 352 and Bohtlingk 1887: 429). The first two, not being imperatives, represent a type of amredita that does not occur in the Rigveda.
40. For arguments in favor of an iterative reading, cf. [section]18 below.
41. I thank Stephanie Jamison for this reference.
42. The possibility of faulty textual transmission should of course always be borne in mind. Cf. n. 11. There is a surfeit of terminology for such adjacent iteration in traditional Western rhetoric. Thus, from Greek one finds pallilogia, anadiplosis, and epanalepsis; from Latin, geminatio, duplicatio, and iteratio. Cf. Wills 1996: 45 n. 1.
43. That instances of this sort must have been commonplace at all stages of Sanskrit (as, indeed, in all languages everywhere) is confirmed by Panini's rule 8.1.8, which provides for sentence-initial vocative amreditas in the expression of jealousy, honoring, anger, scorn, or threats, i.e., in a whole range of affective employments. We must therefore reckon with a difference in the accentual status of the repeated item in Panini's language and the language of the Rigveda.
44. It is also probably not chance that the non-amredita iterations of (49)-(52) are all from the late Tenth Mandala, the Popular Rigveda, the style and subject matter of which are closest to that of the Atharvaveda. As a product of lower religious circles and more popular forms of religion (including charms and curses), the Atharvaveda together with the Popular Rigveda may be expected to show a greater range and predominance of affective usage than other parts of the Rigveda.
45. These cannot be said simply to be built to the intensive stem, for in the case of han- that stem is either janghan- or ghanighnat- (ppl.), ca[r.sup.(i)]-forms only a carcurya- in the Rigveda (carcariti is Atharvavedic), and the intensive of the variant root ca[l.sup.(i)]-, calcaliti, is attested no earlier than the Maitrayani Samhita (on these patterns of attestation, cf. Schaefer 1994: 111, 203). Similarly, uncertain is esaisya at X.102.11c esaisya cid rathya jayema "May we (also) win with a very swift charioteer" (uttered by the spectators following Mudgalani's victory). This appears to be an iteration of esa- 'swift' feminized in -i (Wackernagel 1905: 147, so Geldner ad loc. who translates "die noch fixer als fix ist"), but (far less likely, to my mind) is taken by both Bohtlingk-Roth (1855-75) and Grassmann (1873) to represent esa- 'desire' + esya-, gerundive of is 'seek, desire' and to mean 'desirable'. The former analysis is supported circumstantially by the appearance of esaisya in the Popular Rigveda together with the other forms of (57).
46. Although this is not the place to pursue in detail largely glottogonic ideas about other, more truncated types of formal plurality such as, say, reduplication, there is sufficient evidence, even in the living semantics of some such forms, to support a similar iconic interpretation. Cf. the intensive, which combines full-bodied form with robust, repeated action: janghanani "I will smash over and over" (Schaefer 1994: 204) vs. jaghana "s/he smote." In addition, forms such as Skt. jigati 'goes' and Greek [TEXT NOT REPRODUCIBLE IN ASCII] 'striding' clearly admit of an iterative exegesis, "I proceed/proceeding step by step." In the Indo-European perfect the link between form and meaning is perhaps still visible in the timeless stative-durative value of that category (partial iteration of form reflecting iconically preservation of state) in its oldest manifestations: Gk. [delta][epsilon]i[delta][omega] 'I am afraid', Skt. bibhaya 'id'.
47. If here, then only diachronically, since synchronically the Classical Armenian locative occurs almost exclusively preceded by the preposition i. A rare exception is giseri 'in the evening', a petrified old locative (not, with Jensen [1959: 179], a genitive; cf. Meillet 1913: 49 and Kunzle 1984: 157), as demonstrated by its concurrent i giseri 'id'.
48. Note in particular that neither of the two oldest temporal amreditas in the Rigveda, dive-dive and aharahah, is a locative. Moreover, such locative temporal amreditas as masi-masi and dyavi-dyavi are late.
49. The same would be true of the Classical Armenian collocation awr awur 'every day' cited without textual reference by Jensen (1959: 46) beside awur awur. The loss of u in this form is regular in an unaccented syllable, which can only mean that awr awur was treated as a single accentual unit at a time when this rule applied. The final accent here is the reverse of that seen in Greek [TEXT NOT REPRODUCIBLE IN ASCII] and the Rigvedic amreditas, because with few exceptions Classical Armenian words are accented on the last syllable.
Aufrecht, Theodor. 1877. Die Hymnen des Rigveda. 2nd ed. Bonn: Adolph Marcus. Rpt. 1968. Wiesbaden: Otto Harrassowitz.
Bartholomae, Christian. 1907. Review of Wackernagel, 1905. Anzeiger fur Indogermanische Sprachund Altertumskunde: Beiblatt zu den Indogermanischen Forschungen 20: 162-72.
Bohtlingk, Otto, ed. 1887. Panini's Grammatik herausgegeben, ubersetzt, erlautert und mit verschiedenen Indices versehen. Leipzig. Rpt. 1964. Hildesheim: Georg Olms Verlagsbuchhandlung.
______, and Rudolph Roth. 1855-75. Sanskrit Worterbuch. St. Petersburg: Kaiserliche Akademie der Wissenschaften. Rpt. 1990. Delhi: Motilal Banarsidass.
Brereton, Joel Peter. 1981. The Rgvedic Adityas. American Oriental Series, vol. 63. New Haven: American Oriental Society.
Brugmann, Karl. 1906. Grundriss der vergleichenden Grammatik der indogermanischen Sprachen (2), II.1. Strassburg: Karl J. Trubner. Rpt. 1967. Berlin: Walter de Gruyter.
Collitz, Hermann. 1882. Ueber eine besondere Art vedischer Composita. Abhandlungen und Vortrage des funften internationalen Orientalisten-Congresses gehalten zu Berlin im September 1881, zweiter Theil, zweite Halfte. Pp. 287-98. Berlin: Weidmannsche Buchhandlung. Rpt. 1968. Nendeln, Liechtenstein: Kraus Reprint.
Delbruck, Berthold. 1888. Altindische Syntax. Halle a.d.S.: Verlag der Buchhandlung des Waisenhauses.
Dressler, Wolfgang. 1968. Ved. dive-dive und die idg. Iterativkomposita. In Studien zur Sprachwissenschaft und Kulturkunde: Gedenkschrift fur Wilhelm Brandenstein, ed. Manfred Mayrhofer. Pp. 39-47. Innsbrucker Beitrage zur Kulturwissenschaft, vol. 14. Innsbruck: AMOE.
Dunkel, George. 1981. Amredita and Iteration of Preverbs in Vedic and Hittite. Zeitschrift fur vergleichende Sprachforschung 95: 214-26.
Geldner, Karl F. 1951. Der Rigveda ... ubersetzt. 3 vols. Harvard Oriental Series, vols. 33-35. Cambridge, Mass.: Harvard Univ. Press. (Cited as Geldner.)
Grassmann, Hermann. 1873. Worterbuch zum Rig-Veda. Leipzig: Otto Harrassowitz. 4., unveranderte Auflage, 1964. Wiesbaden: Otto Harrassowitz.
Haspelmath, Martin. 1997. Indefinite Pronouns. Oxford: Clarendon Press.
Jensen, Hans. 1959. Altarmenische Grammatik. Heidelberg: Carl Winter.
Katz, Joshua T. 1998a. Topics in Indo-European Personal Pronouns. Ph.D. diss., Harvard Univ.
______. 1998b. Archaische keltische Personalpronomina aus indogermanischer Sicht. In Sprache und Kultur der Indogermanen: Akten der X. Fachtagung der Indogermanischen Gesellschaft Innsbruck, 22.-28. September 1966, ed. Wolfgang Meid. Pp. 265-91. Innsbrucker Beitrage zur Sprachwissenschaft, vol. 93. Innsbruck: Institut fur Sprachwissenschaft der Univ. Innsbruck.
Klein, Jared S. 2002. Homoioteleuton in the Rigveda. Indo-European Perspectives, ed. Mark Southern. Pp. 199-222. Journal of Indo-European Studies Monographs, no. 43. Washington, D.C.: Institute for the Study of Man.
______. MS. Nominal and Adverbial Amreditas and the Etymology of Rigvedic nana. Forthcoming in Proceedings of the Third International Vedic Worship, ed. Arlo Griffiths and J. E. M. Houben.
Kunzle, Beda. 1984. Das Altarmenische Evangelium/L'Evangile armenien ancien. Teil II/II[.sup.e] partie: Lexikon/lexique. Europaische Hochschulschriften, Reihe XXI. Linguistik und Indogermanistik, vol. 33. Bern: Peter Lang.
Lubotsky, Alexander. 1997. A Rgvedic Word Concordance, parts I and II. American Oriental Series, vol. 82. New Haven: American Oriental Society.
Ludwig, Alfred. 1881-83. Der Rigveda ... ubersetzt mit Commentar und Einleitung. Band 4-5: Commentar. Prague: Verlag von F. Tempsky. non vidi.
Macdonell, A. A. 1916. A Vedic Grammar for Students. Bombay: Oxford Univ. Press. Rpt. 1966.
Maurer, Walter H. 1986. Pinnacles of India's Past: Selections from the Rgveda. University of Pennsylvania Studies on South Asia, vol. 2. Amsterdam: John Benjamins.
Mayrhofer, Manfred. 1992. Etymologisches Worterbuch des Altindoarischen, II. Band. Heidelberg: Carl Winter.
Meillet, Antoine. 1913. Altarmenisches Elementarbuch. Heidelberg: Carl Winter. Rpt. 1980.
Oldenberg, Hermann, 1909-12. Rgveda: Textkritische und exegetische Noten: Erstes bis sechtes Buch (1909). Abhandlungen der koniglichen Gesellschaft der Wissenschaften zu Gottingen, Phil.-Hist. Klasse, Neue Folge, vol. XI.5; Siebentes bis zehntes Buch (1912) AKGWG, Phil.-Hist. Kl., N.F., vol. XIII.3. Berlin: Weidmannsche Buchhandlung.
Renou, Louis. 1952. Grammaire de la langue vedique. Lyon: IAC.
______. 1962. Etudes vediques et panineennes. Tome X. Les Hymnes aux Marut. Paris: Editions E. de Boccard.
_______. 1966. Le grammaire de Panini, volume II (Adhyaya 5 a 8). Paris: Ecole francaise d'Extreme-Orient.
Schaefer, Christiane. 1994. Das Intensivum im Vedischen. Historische Sprachforschung, Erganzungsheft, vol. 37. Gottingen: Vandenhoeck & Ruprecht.
Wackernagel, Jakob. 1905. Altindische Grammatik, Band II.1: Einleitung zur Wortlehre. Nominalkomposition. Gottingen: Vandenhoeck & Ruprecht. Rpt. 1957.
______, and Albert Debrunner. 1930. Altindische Grammatik, Band III: Nominalflexion--Zahlwort--Pronomen. Gottingen: Vandenhoeck & Ruprecht. Rpt. 1975.
Whitney, William Dwight. 1889. Sanskrit Grammar (2). Cambridge, Mass.: Harvard Univ. Press. Rpt. 1967.
Wills, Jeffrey. 1996. Repetition in Latin Poetry: Figures of Allusion. Oxford: Clarendon Press.
JARED S. KLEIN
UNIVERSITY OF GEORGIA