The Typology of Amredita Compounds in the Rgveda
Tamara DITRICH
University of Ljubljana t.ditrich@gmail.com
Abstract
Amredita compounds in the Rgveda are considered to be a type of coordinative nominal construction, closely related to dvandva compounds. This article investigates amredita compounds against the background of other coordinative nominal constructions from Rgveda 1.1-1.50 which were analysed and compared with their parallel attestations in other mandalas of the Rgveda. The first fifty hymns of the Rgveda form an organic whole: they belong to the middle period in the Rgvedic chronology and address a rich variety of deities, providing a substantial amount of material to address the typological problems of amredita compounds. The article overviews the problems related to the typology of amredita compounds, their analysis in the Rgvedapadapatha and the AstadhyayT, examines all amredita compounds attested in Rgveda 1.1-1.50 and compares them with dvandva compounds, noting their interrelations, similarities and differences and consequently, identifies some of their typological features.
Keywords
Vedic Linguistics, Rgvedic Exegesis, Typology of Nominal Compounds, Vedic Compounds, Amredita word-groups
Izvleček
Amredita zloženke se v vedskem kot tudi v klasičnem sanskrtu ponavadi klasificirajo kot poseben tip koordinativnih nominalnih zvez, ki so v tesnem sorodstvu z dvandva zloženkami. Prispevek raziskuje amredita zloženke v okviru vseh koordinativnih nominalnih zvez, ki so zabeležene v rgvedskih himnah 1.1-1.50 in jih analizira skupaj z njihovimi vzporednimi zabeležbami v vseh mandalah Rgvede. Prvih petdeset himen Rgvede predstavlja organsko celoto: v rgvedski kronologiji pripadajo srednjemu obdobju in so posvečene številnim različnim bogovom ter nudijo dovolj obsežno gradivo za raziskovanje tipoloških vprašanj povezanih z amredita zloženkami in drugimi koordinativnimi nominalnimi zvezami. Prispevek obravnava analizo amredita zloženk v Rgvedapadapathi in AstadhyayT, analizira vse amredita zloženke iz prvih petdesetih himn Rgvede, jih primerja z dvandva zloženkami in raziskuje njihove medsebojne podobnosti in razlike ter ugotavlja njihov razvoj in tipološke značilnosti.
Ključne besede
vedska lingvistika, eksegeza Rgvede, tipologija zloženk, vedske zloženke, amredita besedne skupine
Acta Linguistica Asiatica, Vol. 1, No. 1, 2011. ISSN: 2232-3317 http://revije.ff.uni-lj.si/ala/index
This article investigates amredita compounds in the Rgveda which have been viewed as one type of coordinative nominal construction. Materials for this research are drawn from the first fifty hymns of the Rgveda: all coordinative nominal constructions attested in these hymns were identified, analysed and compared with their parallel attestations in other mandalas of the Rgveda. The first fifty hymns of the Rgveda form an organic whole: they presumably belong to the middle period in the Rgvedic chronology and address a variety of deities, providing a substantial amount of material for investigation of coordinative nominal constructions, including amredita compounds. The article investigates the problems related to the typology and development of amredita compounds and especially their relationship with dvandva compounds.1
Amredita2 compounds, also called amredita word-groups, are comprised of an inflected form (usually a substantive, or less commonly an adjective, pronoun, adverb, preposition or numeral) which is repeated, giving the compound an intensive, distributive or iterative meaning (e.g. dive-dive "every day"). They are productive in the Rgveda, especially those comprised of two nouns. Before the typological questions related to amredita compounds are addressed here, their analysis in the earliest Rgvedic exegetical text, the Padapatha, will be discussed and the approach to amredita word-groups in the most important old Indian grammar, Panini's Astadhyayi, will be examined.
1. Analysis of Amredita Compounds in the Rgvedapadapatha
The Rgvedapadapatha ("Rgveda-word-text") gives all the words of the Rgvedasamhitapatha ("Rgveda-continuous-text") in a separated form, unaffected by the rules of euphonic combination or sandhi. As the earliest exegesis of the Rgveda — the first known commentary on the samhita text, dated around the end of the Brahmana period — its main purpose is the accurate preservation of the Rgveda during oral transmission. The text also clarifies the meaning of words and seems to be the earliest recorded grammatical analysis of the Rgveda (Jha, 1992, p. 1). In the Rgvedapadapatha the sandhis are dissolved and two kinds of pauses are used to keep separate meaningful units, i.e. they mark morpheme-boundaries or word-boundaries: a long pause (danda)3 always follows a pada ("word") — it separates the constituents of a sentence — and a shorter one (avagraha)4 separates the constituents of a word. Components of a compound are separated by an avagraha: this indicates that the components are analysed in the padapatha as constituents of internal pada.
1	The research leading to the results in this paper has received funding from the Seventh Framework Programme [FP7/2007-2013] under PIRG02-GA-2007-224432.
2	The term amredita means literally "repetition, reduplication" (a-mred- "to repeat"; see MW 147).
3	Danda is transliterated with a single space.
4	Avagraha is transliterated with a short hyphen - .
Nominal compounds are not always analysed in the Rgveda-padapatha, i.e. in some circumstances they do not have their constituents separated by an avagraha; for example, in compounds that have two accents and the first component in an inflected form (e.g. RV 1.90.8: Sp vanaspatir, Pp vanaspatih); in compounds that are proper names (e.g. RV 3.53.9: Sp visvamitro, Pp visvamitrah); in dvandvas signifying deities (devatadvandvas); (e.g. RV 1.93.8: Sp agnisoma, Pp agnisoma); in dvandvas with both components accented and in the dual (e.g. RV 4.6.7: Sp matarapitara, Pp matarapitara); and in some other types of compounds (Jha, 1992, p. 169-176).
Most nominal compounds in the Rgveda consist of two components or, rarely, of three, but never more than three. Having a single accent seems to be the first criterion for a word to be considered a compound in the padapatha: if a compound has two accents, its components are not separated by avagraha. The second criterion for a word to be considered a compound is the use of the uninflected stem of the first component. However, a word is often analysed in the padapatha as a compound as long as it has, with some exceptions, one accent, although the first component may be in an inflected form. This is the case with amredita compounds which have their components separated in the padapatha since they fulfill the first criterion (i.e. they have a single accent) although the first component is always in an inflected form (e.g. RV 1.12.2: Sp agnimagnim, Pp agnim-agnim).
In Rgveda 1.1-1.50, twelve amredita compounds are attested which can be grouped in the following way:
•	eight amreditas comprising two nouns (divedive, dyavidyavi, tunjetunje, sutesute, agnimagnim, vise vise, yogeyoge, vajevaje);
•	one amredita comprising two numerals (ekamekam);
•	one comprising two pronouns (tattad); and
•	two comprising prepositions (parapara and prapra).
All amreditas listed above are analysed as compounds in the padapatha, i.e. their components are separated by avagraha (e.g. Pp dive-dive). These amreditas are all, with one exception (i.e. Sp agnimagnim, Pp agnim-agnim), comprised of non-theonyms. Similarly, in all other hymns of the Rgveda almost all amreditas are formed from non-theonyms: only two deities in the entire Rgveda are addressed in amredita compounds, i.e. Agni with three attestations of agnimagnim (1.12.2, 6.15.6, 8.60.17) and Indra with one attestation indramindram (8.12.19). All amredita compounds from Rgveda 1.1-1.50 as well as their attestation in other mandalas of the Rgveda have their components separated by avagraha. Furthermore, all amredita compounds in the Rgveda — altogether 113 are listed by Collitz (1882, pp. 295-297), having usually only one or a few attestations each in the entire text — were examined and found to be consistently analysed in the padapatha, irrespective as to whether they are comprised
5 In the Samaveda the components of devatadvandva compounds are separated.
of theonyms or non-theonyms. This fact also supports the assumption that having a single accent seems to be the first criterion for a word to be considered a compound in the padapätha.
In comparison, examination of dvandva compounds in the Rgvedapadapätha indicates different patterns. Dvandva compounds comprising non-theonyms are analysed as compounds in the padapätha only when the first component is, without any ambiguity, in a stem form (e.g. RV 1.45.2: Sp tmyastrimsatam, Pp tmyah-trimsatam) whereas those comprised of theonyms (which are by far the most frequent coordinative construction for theonyms in the Rgveda) are, as a rule, never analysed in the padapätha: no devatädvandvas, having constituents in juxtaposition or in tmesi, have their constituents separated by avagraha. Although two theonyms (Indra and Vayu, Indra and Agni) occur in dvandva compounds which have a single accent and the first constituent in a stem form—the two criteria that are required for the compound to be analysed in the padapätha—their components are still not separated by avagraha (e.g. Sp: 1.14.3ab indraväyü, Pp: indraväyü iti; Sp: 1.14.3ab indraväyü, Pp: indraväyü iti.) (Ditrich, 2009). This indicates that special criteria apply for dual theonyms: they are not analysed on the syntactic or semantic level but only on the phonetic or morphophonemic level (i.e. sandhi, replacement of singular endings by dual endings). Thus dvandva compounds cannot be viewed as a single category but a clear distinction has to be made between those comprised of theonyms and non-theonyms (Ditrich, 2007). On the other hand, ämredita groups comprising theonyms are always analysed as compounds in the padapätha; this may indicate that iterated theonyms, expressing repetition, developed later, in analogy with reiterated non-theonyms (Delbrück, 1893, p. 143). The analysis of all ämredita compounds in the padapätha indicates that having one accent is the most important criterion for a word to be considered a compound in the padapätha; this principle applies for ämreditas because they consist almost entirely of non-theonyms but not for dvandvas which comprise mostly theonyms and follow different principles.
2. Analysis of Ämredita Compounds in the Astädhyäyi
Although Panini's Astädhyäyi, the first and most important traditional Sanskrit grammar, probably from the sixth century BCE, describes the language of the late Vedic period, it also provides numerous rules for specific features of the old Vedic language and it seems that Panini was well acquainted with Vedic texts, including the Rgveda. It is also quite certain that the author of the Rgvedapadapätha antedated Panini (Ditrich, 2009); so it is curious that in the analysis of nominal compounds the concept of pada ("word") in the Rgvedapadapätha is different from the Paninian concept. In the Astädhyäyi nominal compounds (samäsa) are generally treated as single words, derived by combining syntactically and semantically connected case-inflected
words (padas)6 which in the derivational process have had internal case endings deleted (P 2.4.71) — unless specified otherwise. Furthermore, there is a great difference in the analysis of amreditas in the two texts: Panini does not treat amredita word-groups as compounds. However, in the first pada of the eighth adhyaya he does provide rules for doubling whole syntactic items under various conditions. He defines the term amredita in P 8.1.2: tasya paramamreditam "of that which is repeated the letter [word] is called amredita" and in P 8.1.3 (anudattam ca) he says that the amredita word is not accented. In P 8.1.4 (nityavipsayoh) he assigns to amredita word-groups two meanings: nitya "always, again and again" and vipsa "distributiveness".7
Modern scholars have attempted to give various explanations as to why Panini does not include amredita groups among compounds. Joshi and Roodbergen (1974, xii-xiv) point out that amredita groups only rarely show the basic characteristic of compounds — the deletion of case endings and those amreditas that have case endings deleted are treated by Panini as bahuvrihi or karmadharaya compounds (P 8.1.98.1.15). Furthermore, they argue that amredita groups consist not only of inflected words but also of finite verbs, and Panini does not allow compounds formed by finite verbs. The main reason for their exclusion from nominal composition by Panini is, they believe, that the meaning of amredita groups is not the result of composition but of repetition itself (Joshi & Roodbergen, 1974, pp. 8-9). Cardona thinks that the reason for the exclusion of amredita groups from nominal composition by Panini is to be found in the structural system of the Astadhyayi.8 It is quite certain that Panini knew Sakalya's Rgvedapadapatha, in which amredita groups are treated as compounds. The question as to why Panini did not adopt Sakalya's procedure has not been convincingly answered yet.
The Astadhyayi accounts for all of amredita groups that occur in Rgveda 1.1-1.50: sutras 8.1.1-8.1.15 describe (or prescribe) their derivation and accent. For Rgvedic amreditas, there is no difference in the treatment of reiterated theonyms and non-theonyms in the Astadhyayi: in both cases the required rules for their derivation and accent are accounted for. In comparison, in derivation of dvandvas there are quite a few features that Panini does not account for although most rules required are given in the Astadhyayi. For example, the rules for the order of constituents do not account for the variation in the order of constituents in two pairs of theonyms, such as in usasanakta / naktosasa and dyavaprthivi / prthividyava; the substitution prescribed for the final vowel of the first constituent (P 6.3.26) has at least one exception, i.e. indravayu (possibly also indragni) which is not notified by Panini. Although
6	A pada is defined in P 1.4.14: suptinantam padam "a pada is [that which] ends in sup [case-ending] or tin [verb-ending]"; in nominal composition only padas with the case-endings are involved (P 4.1.1; 1.2.45; 1.2.46; 1.4.14).
7	Cardona (1996, 67) periphrases this rule: "a syntactic item is repeated on condition that repetition of an action or pervasion of a thing by a property or an action is to be conveyed; e.g. grhe grhe 'in each and every house', pibapiba 'drink again and again.'"
8	Cardona (1996, 67-72) shows that P 1.2.64 (sarupanamakasesa ekavibhaktau) does not allow derivation of compounds having identical nominal bases.
derivational rules for dvandva compounds comprising non-theonyms are provided, the derivation of mätaräpitarä, commonly used as an epithet for Heaven and Earth, is not accounted for in the Astadhyayi (Ditrich, 2009).
To summarise, Panini does not treat ämredita word-groups as compounds; however, he does provide rules for doubling whole syntactic items without any distinction between reiterated theonyms and non-theonyms. This fact further supports the hypothesis being developed in this article, that iterated theonyms, expressing repetition of an activity, evolved later on, in analogy with reiterated non-theonyms and consequently — unlike dvandva compounds — they do not display any specific features, neither in the Rgvedapadapatha nor in the Astädhyäyi.
3. Typology and Development of Ämredita Compounds
In ämredita compounds (or word-groups) an inflected form is repeated, thus giving the group an intensive, distributive or iterative meaning. The prior component of this construction retains its own independent accent while the other component is not accented (e.g. divé-dive 'every day'); however, when the constituents are in tmesi both retain their accents (e.g. RV 5.52.17 sapta me sapta).9 Several modern scholars classify them as a separate type, so-called iterative compounds; they define them as compounds that express iteration in time (e.g. divé-dive "every day"), distribution in space (visé-vise "in every house"), frequency and succession (e.g. agnim-agnim "Agni again and again"; yaßäsya-yafiasya "of every sacrifice") or intensity (e.g. dhiya-dhiyä "with increasingly repeated thought") (Delbrück, 1900, pp. 141-142; Renou, 1952, pp. 123-124). Others classify ämredita word groups as a subtype of dvandva compounds and discuss them in the sections dealing with copulative compounds. (Elizarenkova, 1987, p. 235; Whitney 1964, p. 488). Wackernagel (1957, p. 147) treats ämredita groups in his grammar under the compound section although he says that they are not proper compounds; he believes that they may develop into proper compounds in three ways: by deletion of the case ending of the first constituent, by deletion of the case endings of both constituents, or by transformation of the ämredita into an adjective. Renou (1961, p. 121) thinks that it is rather difficult to draw a line between mere word repetition and a compound; he uses for ämreditas a term "faux composés" and classifies them as iterative compounds under a section titled "Composés Anormaux". However, the treatment of ämredita word-groups as compounds is supported by the accentuation pattern as well as by the close relationship between ämreditas and dvandvas pointing to the same semantics of both types (e.g. ämredita group divé-dive "every day" and a dvandva compound nakta-divam "day and night").
By far the most common ämredita compounds in the Rgveda are formed from two substantives (e.g. divé-dive "day by day"). There are also several attestations of
9 P 8.1.3: anudättam ca [ämreditam 2].
amredita groups consisting of two pronouns (e.g. tvâm-tvam "you and (again) you" ), adjectives (e.g. pânyam-panyam ... somam "Soma who is again and again to be praised"), pronominal adjectives (e.g. anyâm-anyam "one after another") or adverbs (e.g. punar-punar "again and again"), only a few occurrences of iterated numerals (e.g. dva-dva "two and two"), prepositions (e.g. prâ-pra "further and further, ever more ') and one attestation of a repeated finite verb (piba-piba "drink again and again") (Collitz, 1882, pp. 287-298; Wackernagel, 1957, pp. 143-146). Amredita word-groups were originally in the singular but the plural meaning of the repetition led to the development of plural forms that occur already in the Rgveda (e.g. RV 5.52.17 ékam-eka sata daduh "they have given a hundred each"). The transition from iterative compounds to regular compounds started in the later Vedic language; e.g. from RV 8.68.14 dva-dva "two and two", to Maitrayani Samhita dvan-dvâm "in pairs" and finally to Taittirîya Samhita dvandvâ- "pair" (Macdonell, 1910, p. 155).
Amredita word groups identified in Rgveda 1.1-1.50 are as follows:
Nouns
1.	divédive: "day by day", has the largest number of attestations of all amreditas in the Rgveda (47 attestations). The compound comprises two substantives in D. Sg., used with L. Sg. meaning. The nominal stem may be diva n. or divm.; however, it seems that the frequency of the locative ending -e of -a stems led to the usage of -e also for the consonant stem div- (Wackernagel, 1957, p. 146).
2.	dyâvidyavi: "day by day"; 2 attestations, two substantives in L. Sg.
3.	tunjétunje: uncertain meaning; 1 attestation, two substantives in L. Sg.
4.	sutésute: "in every libation"; 3 attestations, two substantives in L. Sg.
5.	agnimagnim: "Agni again and again"; 3 attestations, two theonyms in A .Sg.
6.	visévise: "in every house"; attestations, two substantives in D. Sg., used with L. Sg. meaning.
7.	yogeyoge: "in every deed"; 1 attestation, two substantives in L. Sg.
8.	vajevaje: "in every attempt for price"; 4 attestations, two substantives in L. Sg. Pronouns and Pronominal adjectives
1.	ékamekam: "one by one"; 6 attestations, two numerals: in A. Sg. (1.20.7, 5.52.17, 8.70.14), in N. Sg. m. (3.29.15, 5.61.1), in N. Sg. f. (1.123.8).
2.	tâttad: "any"; 4 attestations, two pronouns: in A. Sg. n. used as adjectives (1.46.12, 1.155.4, 8.39.4, 10.23.5).
Verbal prepositions
1.	pârapara: "further and further away"; 1 attestation.
2.	prâpra: "forward and forward"; 12 attestations.
The distribution of these compounds in the ten mandalas of the Rgveda may give some indication of the development of amreditas, presuming the generally accepted relative chronology of the Rgveda which situates the family books (mandalas 2-7) and mandala 10 at opposite ends of the chronological spectrum, and evaluates the duration of the composition of the entire Rgveda up to seven hundred years (Witzel, 1997, pp. 257-345). The distribution of the amredita compounds identified in Rgveda 1.1-1.50 in the ten mandalas is as follows:
Table 1: Distribution of amreditas in Rgveda 1.1-1.50
mandala	1	2	3	4	5	6	7	8	9	10	total
divedive	8	4	8	3	2	5	2	7	5	3	47
dyâvidyavi	2	0	0	0	0	0	0	0	0	0	2
tunjétunje	1	0	0	0	0	0	0	0	0	0	1
sutésute	1	0	1	0	0	1	0	0	0	0	3
agnimagnim	1	0	0	0	0	1	0	1	0	0	3
visé vise	1	0	0	2	1	0	0	0	0	1	5
y—geyoge	1	0	0	0	0	0	0	0	0	0	1
vajevaje	2	0	0	0	0	1	1	0	0	0	4
ékamekam	2	0	1	0	2	0	0	1	0	0	6
tâttad	2	0	0	0	0	0	0	1	0	1	4
pârapara	1	0	0	0	0	0	0	0	0	0	1
prâpra	4	0	1	0	2	1	2	1	1	0	12
Only divedive occurs frequently enough to indicate its distribution: this ämredita compound is found more often in the family books than in the younger layers of the Rgveda, having the smallest number of attestations in the latest mandala 10. It is evident from Table 1 that most frequently used iterative compounds are formed from two substantives (e.g. dive-dive "day by day"), usually occurring in the locative. This pattern of distribution does not support the hypothesis by Collitz (1882, pp. 287-298) that iteration of substantives is a later development but it seems that it may have developed at an early stage of the Indo-Aryan period. As noted before, theonyms rarely occur in ämredita compounds: apart from Agni who is attested in the sample investigated in this article, only one more deity is addressed in this construction, i.e. Indra. In the entire Rgveda they have only a few attestations: agnimagnim has three attestations (1.12.2, 6.15.6, 8.60.17) and indramindram only one (8.12.19), most of them occur in the middle chronological layer of the Rgveda and seem to express repetition of an activity; their pattern of distribution indicates that they may have developed in analogy with reiterated non-theonyms.
In comparison, the distribution of dvandva compounds in the ten mandalas of the Rgveda does not reflect the relative chronology of the Rgveda which is widely
accepted by modern scholarship (Ditrich, 2006). Dvandva compounds in the Rgveda comprise almost entirely theonyms which occur throughout the text without any marked differences among the ten mandalas. Names of deities that are addressed in pairs occur in a variety of coordinative nominal constructions (i.e. dvandva compounds, asyndeta, elliptic duals and syntagms constructed with coordinative particles) which follow specific paradigms. The specific distribution of theonyms reflects the Vedic ideas about the great significance and the magical power of divine names: they do not conform to the historical development of Rgvedic language, but follow special, well-established paradigms instead (Ditrich, 2007). On the other hand, the distribution of coordinative constructions comprising non-theonyms reflects the widely accepted relative chronology of the Rgveda; non-theonyms usually occur in a particular coordinative construction casually, with very few attestations, and seem to be more evenly distributed among the ten mandalas. Research of dvandva compounds provides evidence that in an investigation of coordinative nominal constructions a distinction has to be made between theonyms and non-theonyms (Ditrich, 2006; 2007).
Consequently, it is evident that in the Rgveda, dvandva compounds comprise almost exclusively theonyms and that the rather closely related ämredita compounds consist nearly entirely of non-theonyms. The historical development of the two types is rather elusive and may be interrelated; both types seem to have Indo-European origins followed by specific developments at the Indo-Aryan stage. The so-called iterative compounds are attested in several Indo-European languages but only in Sanskrit can they be formed from any part of speech. Those consisting of two substantives seem to be of Indo-Iranian origin: apart from Sanskrit they occur only in Avestan: e.g. nmäne-nmäne, visi-visi "in every house, in every clan". Repeated adjectives and adverbs are attested in Old Greek (e.g. nXsov nXsov ) and Latin (e.g. magis magis), however, these groups seem to express mainly coordination and are not really iterative compounds (Delbrück, 1900, pp. 144-145). Pronouns are iterated in several Indo-European languages, i.e. in Avestan (e.g. kanhe kanhe), Old Greek (e.g. rigrig), Latin (e.g. quisquis), Slavonic (e.g. Serb. kad-kad). Numerals are iterated in Vedic and Old Greek but attestations in other Indo-European languages seem uncertain. Iteration of prepositions is known also in Old Greek (e.g. nponpo) but iteration of finite verbs occurs only in Sanskrit. Delbrück (1900, pp. 149-153) argues that iteration of pronouns, prepositions and numerals in distributive sense may be of Indo-European origin whereas the iteration of nouns developed in the Indo-Aryan period. Collitz (1882, pp. 287-298) believes in the Indo-European origin of iterative compounds but only of those that iterate pronouns, adverbs and verbal prepositions. More recently, Dressler and Dunkel revisited the Indo-European history of iterative compounds. Dressler (1968, pp. 39-46) focuses on iterative compounds of Vedic type dive-dive: he shows that the most common case for iteration of two substantives is the locative which he believes to be the oldest, of Indo-European origin. On the other hand, Dunkel (1981, pp. 214-231) argues that iteration had started in Indo-European with preverbs, then spread to other adverbs and later to nouns in adverbial function; only in Vedic was the process generalized to nouns in grammatical function (subject, direct object,
possessive genitive). The origin and development of amredita groups seems uncertain; the only iterative compound ever ascribed to proto Indo-European is preverbal * pro pro that can be reconstructed in Vedic Homeric and Latin and, as shown by Dunkel, also in Hittite which probably reflects the earliest stage of evolution of Indo-European iteration of preverbs (Dunkel, 1981, pp. 214-231).
In the Rgveda, as evidenced also on Table 1 above, most frequently used iterative compounds are formed from two substantives (e.g. dive-dive "day by day"), usually occurring in the locative which confirms Dressler's (1968, pp. 39-46) argument. The diachronic analysis of amredita word-groups examined in this article indicates that they all originate in the Indo-Aryan period. Although iterated nouns seem to be of Indo-Iranian origin (with a few attestations in Avestan), there is only one Rgvedic amredita compound that has a parallel compound attested in Avestan, i.e. RV vise-vise, Av. visi-visi. Apart from Avestan, the Rgvedic amreditas examined here have no attested parallels in any other Indo-European language. Similarly, the iterated pronouns (tattad, RV 1.46.12ab) and numerals (ekamekam, RV 1.20.7c) have no parallels in other Indo-European language groups. Of the two iterated verbal prepositions examined here, i.e. parapara (RV 1.38.6ab) and prapra (RV 1.40.7cd), only prapra has several parallel forms attested in Vedic, Homeric Greek and Latin and, as shown by Dunkel (1981, pp. 214-231), also in Hittite.
To summarise: the most frequently used iterative compounds in the Rgveda are formed from two nouns and, as shown above, seem to be of Indo-Iranian origin but have developed and became very productive only in the Indo-Aryan period. Amredita groups comprising other parts of speech are rare and, though some scholars believe them to be of Indo-European origin, they have, apart from prapra, no parallels in any other language group. Theonyms rarely occur in amredita compounds: only two deities are addressed in this construction, Agni and Indra; these amredita groups seem to express repetition of an activity and have developed later, well into the Indo-Aryan period, in analogy with reiterated non-theonyms.
There is certainly a strong link between dvandva compounds and amredita word-groups. Salus (1963, pp. 551-554) developed a hypothesis that dvandvas originated from amredita groups — from those which were distributive in nature and in which one of the two identical parts was replaced by a different word (e.g. devamdevam "the god and again the god" > *manusyadevam "man and god"). Although semantically there seems to be, in Salus's words, "not too great a leap" from one type to the other he does not provide sufficient evidence to support his hypothesis (Salus, 1963, p. 553). Most of the amredita word groups examined in this article express repetition and distribution and indicate a semantic link with dvandva compounds. However, the examined material does not give any evidence for Salus's hypothesis that dvandvas originate from amredita groups — namely from those which were distributive in nature and in which one of the two identical parts was replaced by a different word: among the material examined here no dvandva compound occurs that would have one component also attested in an amredita word group. The only exceptions are the two
theonyms, Agni and Indra which occur both in amreditas (agnimagnim and indramindram) and in numerous dvandvas. However, as shown above, their rare occurrence, their distribution in the ten mandalas, and their analysis in the Rgvedapadapatha indicate that they cannot represent an ancient link between iterative and dvandva compounds but rather to the contrary: iterative compounds comprised of theonyms seem to have developed later on, originating from analogy with amredita word group comprised of non-theonyms as it has been argued in this article.
4. Conclusion
In this article all amredita compounds attested in the first fifty hymns of the Rgveda were identified and their analysis in the Rgvedapadapatha and the Astadhyayi, examined and compared with dvandva compounds attested in those hymns. In the Rgveda-padapatha, all amredita compounds are perceived as compounds i.e. the components are separated by avagraha, which suggests that in the padapatha a word is considered to be a compound as long as it has one accent although the first component may be in an inflected form. This principle applies for amreditas but not for dvandvas which comprise mostly theonyms and follow different principles. Theonyms very rarely occur in amredita compounds: only two deities are addressed in this construction, Agni and Indra, and these amredita groupsare always analysed in the padapatha.
Most rules required for derivation of dvandvas are given in the Astadhyayi although there are some features that Panini does not account for. Unlike dvandvas, Panini does not treat amredita word-groups as compounds; however, he does provide rules for doubling whole syntactic items under various conditions. There is no difference in the treatment of reiterated theonyms and non-theonyms in the Astadhyayi; the required rules for their derivation and accent are accounted for. This fact supports the hypothesis that iterated theonyms, expressing repetition of an activity, developed later and are consequently rare in the Rgveda and—unlike devatadvandvas—do not display any specific features in the Astadhyayi.
Among amredita word groups examined it is only divedive that occurs frequently enough to display its distribution: it occurs slightly more often in the family books than in the younger layers of the Rgveda, having the smallest number of attestations in the latest mandala 10. The most frequently used iterative compounds in the Rgveda are formed from two nouns, usually in the locative. Most of the amredita word groups examined express repetition and distribution and indicate a semantic link with dvandva compounds. The most frequently used iterative compounds in the Rgveda are formed from two nouns; they seem to be of Indo-Iranian origin but have developed and become very productive only in the Indo-Aryan period. Amredita groups comprising other parts of speech are rare and, though some scholars believe them to be of Indo-European origin, they have, apart from prapra, no parallels in any other language group.
It was been argued that all types of dvandva compounds and other coordinative constructions signifying dual theonyms display specific grammatical and stylistic features; the reason for this seems to lie in Vedic ideas about the magical power of divine names (Ditrich, 2009). In the light of this argument, it may be presumed that since amredita compounds comprised of theonyms are extremely rare in the Rgveda — unlike dvandvas and other coordinative constructions consisting of theonyms and do not display any distinct feature, i.e. neither in the Rgveda-padapatha nor in the Astadhyayi nor in their distribution in the ten mandalas, it seems that they developed later, well into the Indo-Aryan period, by analogy with reiterated non-theonyms. Due to their later development amreditas comprising theonyms are rare in the Rgveda. It can be concluded that neither dvandvas nor amreditas can be examined as single categories but a distinction has to be made between non-theonyms and theonyms.
Abbreviations
A.	accusative
D.	dative
Du.	dual
f.	feminine
G.	genitive
L.	locative
m.	masculine
n.	neuter
N.	nominative
P	Panini's grammar, the Astadhyayi
Pp	Padapatha
RV	Rgveda
Sg.	singular
Sp	Samhitapatha
-	avagraha
References
Bloomfield, L. (1963). Language (Delhi: Motilal Banarsidass).
Cardona, G. (1996). "Ämredita compounds?" In Veda-vyäkarana-vyäkhyäna: Festschrift Paul Thieme zum 90. Geburtstag am 18. März 1995, ed. by H. Schmidt and A. Wezler. Studien zur Indologie und Iranistik 20. Reinbek: Verlag für Orientalische Fachpublikationen, 6772.
Collitz, H. (1882). "Ueber eine besondere Art vedischer Composita." Verhandlungen des
fünften internationalen Orientalisten-Congresses: gehalten zu Berlin im September 1881, vol. II, 2. Berlin: A. Asher & Co. Weidmannsche Buchhandlung, 287-298.
Delbrück, B. (1893). Vergleichende Syntax der indogermanischen Sprachen. Theil 3. Grundriss der vergleichenden Grammatik der indogermanischen Sprachen, by K. Brugmann and B. Delbrück. Band 3. Strassburg: Karl J. Trübner.
Ditrich, T. (2009). "Chronological Relationship between the Rgvedasamhita, the
Rgvedapadapatha, and the Astadhyayi Revisited". Paper presented at the 14th World Sanskrit Conference, Kyoto University, Japan, 01-05 September 2009. [forthcoming]
Ditrich, T. (2007). "Stylistic Analysis of Coordinative Nominal Constructions for Dual Deities in the Rgveda". In Proceedings of the 13th World Sanskrit Conference, Edinburgh, July 2006. Delhi: Motilal Banarsidass. [forthcoming]
Ditrich, T. (2006). "Dvandva Compounds and the Chronology of the Rgveda". Crossroads 1 (Autumn 2006), 26-35 (Website: http://uq.edu.au/crossroads/).
Dressler, W. (1968). "Ved. dive-dive und die idg. Iterativkomposita." In Studien zur
Sprachwissenschaft und Kulturkunde: Gedenkschrift für Wilhelm Brandenstein (18981967), ed. by M. Mayrhofer. Innsbrucker Beiträge zur Kulturwissenschaft, Band 14. Innsbruck: AMŒ, 39-46.
Dunkel, G. E. (1981). "Ämredita and iteration of preverbs in Vedic and Hittite." Zeitschrift fur vergleichende Sprachforschung begründet von Adalbert Kuhn 95, 214-231.
Elizarenkova, T. IA. (1987). Vediiskii iazyk. Moskva: Nauka.
Gonda, J. (1975). Vedic Literature: Samhitas and Brahmanas. A History of Indian Literature, vol. I, fasc. 1. Wiesbaden: Otto Harrassowitz.
Jha, V. N. (1992). A Linguistic Analysis of the Rgveda padapatha. Sri Garib Dass Oriental Series No. 142, Pre-Paninian Grammatical Traditions (Part 1). Delhi: Sri Satguru Publications.
Joshi, S. D. & Roodbergen, J. A. F. (eds., trans.). (1974). Vyakarana-Mahabhasya:
Bahuvrihidvandvahnika (P. 2.2.23-2.2.38), Publications of the Centre of Advanced Study in Sanskrit, Class C, No. 9 (Poona: University of Poona).
Joshi, S. D. & Roodbergen, J. A. F. (1993). The Astadhyayi of Panini: With Translation and Explanatory Notes, vol. 2 (1.2.1-1.2.73). New Delhi: Sahitya Akademi.
Macdonell, A. A. (1910). Vedic Grammar. Strassburg: Karl Trübner.
Oldenberg, H. (1888). Die Hymnen des Rigveda. Band I: Metrische und textgeschichtliche Prolegomena. Berlin: Wilhelm Hertz.
Renou, L. (1952). Grammaire de la langue védique. Collection "Les langues du monde", vol. 9. Paris: IAC.
Renou, L. (1961). Grammaire sanscrite. Paris: Adrien-Maisonneuve.
Salus, P. H. (1963). "On the Origin of the Dvandva." Orbis 12 (2), 551-554.
Wackernagel, J. (1957). Altindische Grammatik: Einleitung zur Wortlehre,
Nominalkomposition. Band II,1, 2nd ed. Göttingen: Vandenhoeck & Ruprecht.
Whitney, W. D. (1964). Sanskrit Grammar. 2nd ed. Cambridge, MA: Harvard University Press.
Witzel, M. (1997). "The Development of the Vedic Canon and its Schools: The Social and Political Milieu." In Inside the Texts, Beyond the Texts, ed. by Michael Witzel Harvard Oriental Series, Opera Minora, vol. 2. Cambridge, MA: Department of Sanskrit and Indian Studies, Harvard University, 257-345.