A New Rule in Vedic Metrics

A New Rule in Vedic Metrics GAŠPER BEGUŠ Harvard University In this paper I propose a new rule of Vedic meter. The glides *v and *y are regularly lost before the corresponding high vowels ŭ and ̆ i in Vedic. I argue that the word-initial glides *v and *y before the short vowels ̆ and ̆ still “make position” and that they should be restored for metrical purposes. This means that word-inal syllables of the shape -V̆C should be scanned long if the following syllable begins with a u- or i- that goes back to *vu- or *yi-. This new rule has consequences for the general metrical shape of the Rigveda, as cadences previously scanned as irregular will be repaired to their canonical shape. The rule can also be employed as etymologically decisive for words that can potentially go back to forms with or without an initial glide. §1 INTRODUCTION In Vedic, consonantal *v (or *u̯) is regularly lost before the high back vowel *ŭ (AiG I: 261–62). There were two main sources of the sequence *vŭ in pre-Vedic: the Proto-IndoEuropean sequence *(C)u̯R̥H, which yields *(C)vŭr and subsequently (C)ŭr (e.g., urú- ‘wide, broad’ < *vurú- < PIE *u̯r̥Hú-), 1 and the reduplication syllable of roots with the structure vaC (*vuvaC- > uvaC-, e.g., *vuv̄́ca > uv̄́ca). 2 The loss of glides also occurs in external sandhi. When word-inal -o (*-av) is followed by word-initial ŭ-, *v is lost, resulting in -a ŭ-, e.g., RV 1.2.2a v̄́ya ukthébhir for Pada-Pạ̄ha v̄́yo ukthébhị. If, on the other hand, any other vowel (except for short a-) follows word-inal -o (*-av), *v is preserved and the result is -av V-, e.g., RV 7.92.3b v̄yav ị̣áye for Padapạ̄ha v̄yo ị̣áye (cf. AiG I: 261–62). The few exceptions to the glide loss can easily be explained by analogy. AiG I: 262 lists three such examples: the compound hotr̥-vúrya- ‘election of priest’, the 3rd sg. medial form of the optative vurita, and 3rd pl. forms with the -ur ending, such as babhuvúr. In the irst two cases v was restored on the basis of other forms in the paradigm of the verb √vari I would like to thank Jay Jasanof, Jeremy Rau, Kevin Ryan, Janez Orešnik, and Alenka Šivic-Dular, as well as the audiences of ECIEC 33, the Linguistic Circle at the University of Ljubljana, and the GSAS Workshop in Indo-European and Historical Linguistics at Harvard for their useful comments on earlier versions of this paper. Thanks are also due to the editor, Stephanie Jamison, and two anonymous reviewers. All mistakes are, of course, my own responsibility. 1. The development of PIE *u̯r̥Hú- to Ved. urú- includes several steps: irst, PIE *u̯r̥H regularly develops to Ved. vur and the laryngeal gets lost, yielding *vur-ú-. Thereafter, *v is lost before u, yielding the attested urú-. This is exactly parallel to the development PIE *pl̥h1-ú- > Ved purú-. The only alternative explanation would be to assume that the laryngeal was lost early in *u̯r̥H-ú- and that the *u̯r̥-V sequence resyllabiied to *ur-. The Avestan cognate vouru- (< *u̯r̥Hú-) with initial v-, however, clearly shows that in Proto-Indo-Iranian and in Avestan no resyllabiication occurred. It would be highly unparsimonious to assume that such resyllabiication happened on the way from Proto-Indo-Iranian to Vedic, especially because there is no motivation or independent evidence for such resyllabiication. For a recent treatment of ̆ ir and ŭr sequences, see also Lubotsky 1997. 2. The original reduplication vowel was a, which is conirmed by RV 1.67.8b vav̄ca (attested once in the Rigveda). Reduplication in Vedic, however, undergoes innovation: the reduplication vowel of verbs with u or i in the zero-grade and of verbs with the structure vaC-/yaC- becomes u or i respectively (Kümmel 2000: 21). Parallel to sụvap-, sụup-, the strong and weak forms of the perfect of √vac were *vuvac- / *vu-uc- (yielding uvac- and uc- after the loss of *v and contraction). Journal of the American Oriental Society 135.3 (2015) 541 542 Journal of the American Oriental Society 135.3 (2015) ‘to choose’. In the latter case *v was a hiatus-preventing glide that was inserted in order to avoid the sequence **uu or was analogically transferred from forms with other vowel-initial endings (e.g., babhuváthur). The same is true for the -v-u- outcome in the weak form of the perfect participle in -ụ- to roots in -u /-u,, e.g., sụuv-ụ́as. The glide loss before homorganic vowels targeted not only *v but also *y (or *i̯). The examples of this loss are, however, limited to word-internal position: pre-Vedic *́ray-ị̣ha(< PIE *k̑rei̯H-isto-) yields *́ra-ị̣ha- after the loss of *y before i and inally ́rẹ̣ha- 3 after desyllabiication of i. 4 There is one example where *y may have been lost before i word-initially in the reduplication syllable: the middle participle iȳná-. However, this participle could go back either to the root present *iȳná- (< *ih2-m̥h1nó-, as in Ved. ȳ́mi) or to the reduplicated present *yiȳná- (< *i̯i-i̯h2-m̥h1nó-, as in Grk. δ́ζημαι, διζ́μενος; cf. Garcı́a-Ramón 1999: 88; LIV2: 310–11). There is no formal way to decide between the two options. The purpose of this paper is to show that the lost glides *v and *y were still present before the corresponding high short vowels—or at least accessible to the poets—at the time when books 2–8 of the Rigveda were composed, and that they still “make position,” in the sense that the glides must be restored for metrical purposes and that word-inal syllables of the shape -V̆C must be scanned long if they were followed by *vu- or *yi- (*-V̆C. vu-; *-V̆C. yi-). This new rule repairs at least twenty-six irregular cadences to their canonical structure (thirty-two by a less conservative count). I also show that this new scansion that restores the lost *v and *y repairs irregular openings as well, although their metrical structure is less regular and therefore less relevant. Finally, I argue that the rule can be employed decisively for cases with unclear etymology, such as the participle iȳná-. To my knowledge, no such rule has previously been proposed in the existing literature on Vedic metrics. 5 §2 METRICS The cadences of the trị̣ubh (eleven-syllable verse) and jagat̄ (twelve-syllable verse) are metrically the most regular features in the Rigveda. According to van Nooten and Holland (1994: xvii), 15,855 of 16,450 (or 96.4%) trị̣ubh cadences have the canonical structure − ⏑ − ×. Likewise, 4,863 of 5,011 (or 97.0%) jagat̄ cadences have the canonical structure − ⏑ − ⏑ ×. 6 This leaves room for only 3.6% or 3.0%, respectively, of cadences that have other structures and are therefore “irregular” or non-canonical. Openings in the Rigveda are much less regular than cadences. Nevertheless, there exist some tendencies for metrical regularity even in openings. The structures × ⏑ − ⏑ − and × − − − constitute 72.6% of all openings, with structures such as × − ⏑ and × ⏑ ⏑ ⏑ occurring 3. After the loss of *i̯, the i in the disyllabic sequence *a-i desyllabiied, yielding *-ai̯- and consequently -e- (cf. Lubotsky 1995: 219). Such sequences, however, scan disyllabically in the Rigveda (cf. van Nooten and Holland 1994: iv), which again indicates that *y was present when the hymns were composed. 4. Exceptions to this rule can here too be explained by analogy. As Lubotsky (1995: 219) points out, “all -ayi- sequences actually attested in Vedic must be secondary.” The one prominent counterexample, rayí-, is either explained as analogical or non-contracted because of the accented í (Lubotsky 1995: 221–23). For this explanation and a thorough treatment of the loss of *y before i word-internally, see Lubotsky 1995 and Kobayashi 2014. 5. The only observation with respect to *u̯ that I have found in the literature is in Kümmel’s treatment of the gen. pl. ending (Kümmel 2010: 4). There he notes that ur̄ṇ̄́ in RV 7.73.3a áhema yajñám pathâ̄́m uṝ̣̄́ goes back to *u̯ər̄n̄́ and that in pre-Vedic the preceding syllable -̄m would be long not just because of the long vowel but also by position. However, the strong tendency to avoid the extra-long sequences -V̄C CV- in Vedic metrics renders the example inappropriate for establishing length by position. If anything, we would expect a word after the sequence -V̄C to be vowel-initial. No further examples are discussed there. 6. Cadences of the anụ̣ubh and gāyatr̄ (both eight-syllabic verses) are also regular, but since all relevant examples appear in cadences of trị̣ubh and jagat̄, I omit the former two cadence types from the discussion. 543 BEGUŠ: A New Rule in Vedic Metrics at frequencies as low as 2.1% and 0.4%. The most canonical structure for openings is thus × − × − (Arnold 1905: 193–94). More precisely, the second syllable of the trị̣ubh openings is long in 13,823, or 83.9%, of all cases (cf. van Nooten and Holland 1994: xvii–xviii). The evidence for the proposed rule will only be taken from cadences, since openings are not regular enough to be conclusive. Moreover, lengthening of any syllable in the opening of a verse will automatically yield a better structure. However, if restoration of glides proves to be necessary in cadences, there is no reason not to restore them in openings as well. §3 the data Let us now take a look at words that feature an initial u- that goes back to *v̆- and appears after the sequence -V̆C in cadences of the Rigveda. The most frequently attested such word is the verb urụy- ‘to widen’, formed from the adjective urú- ‘wide, broad’. The adjective goes back to PIE *u̯r̥H-ú- (EWAia, I, 227), which yields pre-Vedic *vurú-. The verb urụyappears eight times in cadences after the sequence -V̆C. In all eight of these instances, the restoration of *v repairs the irregular cadence. In (1), we see that the traditional scansion 7 gives the metrical structure ⏑ ⏑ − ×, which occurs in only 250 (1.5%) trị̣ubh cadences in the Rigveda. Restoring the *v, we get the canonical structure − ⏑ − ×, which accounts for 96.3% of the trị̣ubh cadences. (1) RV 1.58.8c ágne gr̥̣ántam ạ́hasa urụya ágne gr̥̣ántam ạ́haso *vurụya ⏑⏑−× − ⏑ − ×8 On the subject of examples with inal -ạ, Van Nooten and Holland (1994: 582) remark that “[t]he uncommon cadence . . . can be normalized by reading ạ́haso (with long inal vowel) for S. ạ́hasa (Pp. ạ́hasaḥ) before a following vowel.” However, we see that there is no need to invoke a special reading of -ạ, 9 because it is the lost *v that has caused the preceding syllable to be heavy (the sandhi outcome was in fact most probably *-o v- or -az vfor *-as v-). 10 Moreover, as seen in (2) and subsequent examples, the reading with *v should be employed not only after -ạ, but also after any other sequence -V̆C, even -e < *-ai̯. The other seven examples in which the reading of urụ́y- with *v repairs irregular cadences are: (2) RV 1.152.6d ̄s̄́ vív̄sann áditim urụyet ⏑⏑−× ̄s̄́ vív̄sann áditị *vurụyet −⏑−× (3) RV 1.158.4a úpastutir aucathiyám urụyen ⏑⏑−× úpastutir aucathiyạ́ *vurụyen −⏑−× (4) RV 4.2.6d ví́vasm̄t sim agh̄yatá urụya ⏑⏑−× ví́vasm̄t sim agh̄yató *vurụya −⏑−× (5) RV 4.2.11d dítị ca r̄́ sva áditim urụya ⏑⏑−× dítị ca r̄́ sva áditị *vurụya −⏑−× 7. By “traditional scansion” I mean the analyses of Vedic meter in Oldenberg 1888, Arnold 1905, and van Nooten and Holland 1994. 8. The irst line in all examples is from the metrically restored text provided in van Nooten and Holland 1994, with the corresponding metrical structure. The second line is my proposed restoration with the lost *v. 9. For a thorough treatment of -ạ see Malzahn 2001. 10. There are three possible scenarios for how the external sandhi sequence *-o vu- (< *-as vu-) yielded the attested -a u- (instead of **-o u-) after the loss of *v-. First, we could assume that word-inal -o was analogically repaired to -a, which was the regular sandhi outcome of -as before all vowels including u- (excluding short ̆-). Secondly, we could assume that *-as vu- irst yielded *-az vu- and that *-as V- yielded *-az V-. Initial *v- would then have been lost and the two sequences merged to *-az V-, further yielding -a V-. Thirdly, we could assume that *-o u- (from *-as vu-) was resolved to *-av u- and that *v was lost again, yielding the attested -a u-. Journal of the American Oriental Society 135.3 (2015) 544 (6) RV 4.55.5d (7) RV 7.1.15b (8) RV 8.48.5a mitró mitríȳd utá na urụyet mitró mitríȳd utá no *vurụyet sameddh̄́ ram ạ́hasa urụȳ́ t sameddh̄́ ram ạ́haso *vurụȳ́ t imé m̄ pit̄́ yáása urụyávo imé m̄ pit̄́ yááso *vurụyávo ⏑⏑−× −⏑−× ⏑⏑−× −⏑−× ⏑⏑−⏑× −⏑−⏑× One example of the jagat̄ verse is provided in (8). Here, the traditional reading gives the cadence structure ⏑⏑−⏑×, which accounts for 83 (or 1.7%) of all jagat̄ cadences. The restoration of *v, on the other hand, yields the reading −⏑−⏑×, which accounts for 4,863 (or 97.0%) of the jagat̄ cadences. 11 There are furthermore two examples of the adjective urucí- ‘extending far’ from urú- (< *vurú- < PIE *u̯r̥H-ú-) and one example of a compound urukramá- ‘far-stepping’ with the irst member urú-. In all three instances the restoration of *v repairs the irregular cadences. (9) RV 3.57.5b ágne devẹ́u ucyáta urucí ⏑⏑−× ágne devẹ́u ucyáte *vurucí −⏑−× (10) RV 7.45.3c víráyaṃ̄o amátim urucí m ⏑⏑−× víráyaṃ̄o amátị *vurucí m −⏑−× (11) RV 5.87.4a sá cakrame maható nír urukramạ́ ⏑⏑−⏑× sá cakrame maható nír *vurukramạ́ −⏑−⏑× The nouns uloká- ‘free space’ and ulokaḱ̥t- ‘creating free space’ also have urú- as their irst member. These two words go back to a compound *uru-loká- or *ulu-loká-, in which the second syllable was haplologically lost (for an explanation, see Aufrecht 1888: 152; Leumann 1968: 57–58). They are attested eight times in cadences of the Rigveda. In all examples, the restoration of *v repairs irregular cadences to their canonical structure. (12) RV 1.93.6d urụ́ yajñ̄́ ya cakrathur ulokám ⏑⏑−× = RV 7.99.4a urụ́ yajñ̄́ ya cakrathur *vulokám −⏑−× (13) RV 2.30.6d asmín bhayásthe kr̥̣utam ulokám ⏑⏑−× asmín bhayásthe kr̥̣utạ *vulokám −⏑−× (14) RV 6.23.3c kárt̄ vir̄́ ya sụ́vaye ulokạ́ ⏑⏑−× kárt̄ vir̄́ ya sụ́vaye *vulokạ́ −⏑−× (15) RV 6.23.7d urụ́ kr̥dhi tuv̄yatá ulokám ⏑⏑−× urụ́ kr̥dhi tuv̄yató *vulokám −⏑−× (16) RV 6.73.2a ján̄ya cid yá í vata ulokám ⏑⏑−× ján̄ya cid yá í vate *vulokám −⏑−× (17) RV 7.84.2d urụ́ na índrạ kr̥̣avad ulokám ⏑⏑−× urụ́ na índrạ kr̥̣avad *vulokám −⏑−× (18) RV 9.86.21b ayạ́ síndhubhyo abhavad ulokakr̥ ́ t ⏑⏑−⏑× ayạ́ síndhubhyo abhavad *vulokakr̥ ́ t −⏑−⏑× The restored *v repairs irregular cadences not only in derivatives of urú-, but also in other lexical items. There are four examples of the aorist middle participle uṝá- from the verb √vari ‘to choose’ (PIE *u̯elh1, cf. EWAia, II: 511), which goes back to *vuṝá- < *u̯l̥h1m̥h1no-. In all four examples, the reading with *v repairs the irregular cadence. 12 11. The example in (7) is of the virạ̄sthānā type. Such verses have ten syllables and are fairly rare. According to Arnold (1905: 211), they can be restored to the normal trị̣ubh verse by inserting a rest after the fourth syllable. Regardless of how we treat the virạ̄sthānā verse, the restoration of *v repairs the irregular cadence. 12. I remove RV 7.73.3a áhema yajñám pathâ̄́m uṝ̣̄́ from my count because the second vowel in the disyllabic gen. pl. -̄m can be either short or long. 545 BEGUŠ: A New Rule in Vedic Metrics (19) RV 3.19.2c = RV 4.6.3b (20) RV 4.6.4d (21) RV 4.7.8c pradaḳịíd devát̄tim uṝạ́ pradaḳịíd devát̄tị *vuṝạ́ trivị̣í eti pradíva uṝạ́ trivị̣í eti pradívo *vuṝạ́ dutá iyase pradíva uṝó dutá iyase pradívo *vuṝó ⏑⏑−× −⏑−× ⏑⏑−× −⏑−× ⏑⏑−× −⏑−× Moreover, the irst part of the compound ur̄-máthi- ‘killing sheep’ goes back to *vuran< PIE u̯r̥h1en- ‘lamb’ (cf. EWAia, I: 225–26). In its only cadence attestation in the Rigveda, the restoration of *v repairs the irregular cadence: (22) RV 8.66.8a vr̥ ́ ká cid asya v̄rạá ur̄máthir vr̥ ́ ká cid asya v̄rạó *vur̄máthir ⏑⏑−⏑× −⏑−⏑× There are also two examples, both from the root √vac ‘to speak’, where the restored *v is part of the reduplication syllable. The perfect form uv̄ca goes back to *vuv̄ca and the ̆ C. The line in RV 7.98.3b (24) is lost *v closes the preceding syllable with the structure -V exactly identical to the line AV́ 20.87.3b, and in both cases the restoration of *v repairs irregular cadences. (23) RV 2.30.2b prá prá (24) RV 7.98.3b prá = AV́ 20.87.3b prá tạ́ jánitri vidụ́a uv̄ca tạ́ jánitri vidụ́e *vuv̄ca te m̄t̄́ mahim̄́ nam uv̄ca te m̄t̄́ mahim̄́ nạ *vuv̄ca ⏑⏑−× −⏑−× ⏑⏑−× −⏑−× There is only one example in the Rigveda where the restoration of *v does not repair an already irregular cadence. In (25), we have the plain adjective urú- ‘wide, broad’ (< *vurú- < PIE *u̯r̥Hú-). (25) RV 6.50.3a utá dȳv̄pr̥thivi ḳatrám urú 13 utá dȳv̄pr̥thivi ḳatrạ́ *vurú −⏑⏑× −−⏑× The hymn RV 6.50 itself has eight irregular cadences (van Nooten and Holland 1994: 624–25) and the cadence in RV 6.50.3a is irregular with or without restoration of *v. For an explanation of this case, see §4 below. In any event, this example is the only one in the Rigveda where the restoration of *v before the short ̆ does not repair the irregular cadence to its canonical shape. From the discussion above, it is clear that *v was still present before ̆, or at least accessible to the poets in this position, at the time when the hymns were being composed; therefore, the glide should be reconstructed for the purpose of metrical scansion in the Rigveda. It would be reasonable to expect the same situation for *v before the long vowel u. This is, however, not the case: *v before the corresponding long high vowel should not be restored for metrical purposes. In none of the cases of *vu- attested after -V̆C does the restoration of *v repair an irregular scansion. In fact, the reading with *v converts regular cadences into irregular ones. In (28), for example, the noun urmí- ‘wave’ goes back to *vurmí- and PIE *u̯l̥Hmi-. This form is attested twenty-seven times in cadences, and in all the examples the reading with *v yields an irregular cadence, whereas the traditional reading yields the canonical structure. (28) RV 2.16.5a vŕ̥̣̣ạ kó́ạ pavate mádhva urmír vŕ̥̣̣ạ kó́ạ pavate mádhvo *vurmír −⏑−× −−−× 13. Gippert (1997: 69) repairs this line by assuming that the laryngeal causes the irst syllable of urú to be long by position: *uR’V from *u̯r̥Hu-. Note, however, that the irst syllable of this adjective is nowhere else to be read as long. 546 Journal of the American Oriental Society 135.3 (2015) Other examples in which the restoration of *v before a long u converts regular cadences into irregular ones are (a) feminine forms of the adjective urví- (shortened either by analogy or by sound law 14 from *vurví- < PIE *u̯r̥Hu̯ih2-), attested eighteen times, (b) the adjective urdhvá- ‘raised’ (< *vurdhvá- < PIE *u̯r̥Hdhu̯o-; EWAia, I: 244–45, Barnes 2014: 7, Chantraine 1968–1980: 818–19), attested three times, and (c) the noun úṝ- ‘wool’ (< *vúṝ- < PIE *u̯l̥h1neh2-; EWAia, I: 243), attested once. 15 There are generally two possible explanations for the metrical absence of *v before the long vowel u. First, we could assume that the loss of *v before both long and short ŭ was an ongoing sound change at the time when the hymns were composed and that there was variation between the forms with and without initial *v. The poets could thus optionally drop the initial glide in order to prevent two long syllables (− −), which would have occurred if a glide was preceded by -V̆C and followed by u. Note that two consecutive long syllables are not suitable for any kind of cadence in the Rigveda. More probably, however, the loss of *v before the long high vowel u preceded its loss before the short high vowel ̆. The long vowel u was most likely higher than its short counterpart, as there exists a tendency across languages for long high vowels to be higher than their short counterparts (Lehiste 1970: 30–32). By virtue of being higher, the long u was most likely also more rounded (higher vowels are usually more rounded; see Ladefoged and Maddieson 1996: 293). Precisely this higher degree of roundedness could easily have led to an early dissimilation and loss of *u̯ or, even more plausibly, an early complete assimilation of *u̯ to the following vowel u. 16 §4 STATISTICAL TESTS We know that superheavy syllables are strongly disfavored in cadences of the Rigveda (Kobayashi 2004: 30–31). This fact allows us to apply statistical tests to the data to show that the distribution of words with the initial glides *v or *y is not simply due to chance or to the fact that irregular short syllables before minor caesura are common. That is, we can test whether the words above behave as consonant-initial or vowel-initial words. Consonant-initial words (CV-) appear after the minor caesura (ninth syllable) 2,535 times in books 2–8 when the preceding word contains a light inal syllable of the form -V̆C, and 151 times when the preceding syllable is heavy (-V̄C or -VCC). Vowel-initial words (V-), on the other hand, appear 234 times after -V̆C and 527 times after -V̄C or -VCC in books 2–8. 17 The distribution is represented in Table 1: Table 1 -V̆C -V̄C CV- 2,535 151 V- 234 527 14. For discussion, see Lubotsky 1997. 15. There are, further, three attestations of the noun urvá- ‘container’ and one attestation of the imperative form uṛuhi. Both are probably formed from the root √var ‘to enclose’, but the etymology here is unclear. Note that √var is an anị root, which means that the noun urvá- or the verb uṛuhi most likely did not have an initial glide at all. The source for the length of the sequence ur is obscure and needs further investigation. Because of the uncertainties I exclude the two lexical items from the discussion (for etymology, see EWAia, I: 245; II: 512). 16. Note also that *v was followed by another v in two of four words (*vurví-, *vurdhvá-) and by another labial m in additional one word (*vurmí-). The early loss could thus be reinforced by dissimilation of *v away from another glide in the following syllable. 17. Thanks to Kevin Ryan for providing me with this count and generally for his help with the statistics. BEGUŠ: A New Rule in Vedic Metrics 547 We can compare this distribution to the distribution of the words with initial *vu- to test whether they behave like consonant- or vowel-initial words. Table 2 below presents all words beginning with u- that goes back to *vu- and that appears after the ninth syllable of trimeter verse in books 2–8 (the irst column). The second column shows the number of occurrences ̆ C; the third column shows the number of occurrences of such words after the sequence -V after the sequence -V̆C. Table 2 -V̆C -V̄C urụy-18 5 1 uṝá-19 urucí- 4 1 2 0 urukramá- 1 0 ur̄máthi- 1 0 uv̄ca 2 1 uloká-20 6 3 ̆ C/-V̄C (Table 2) The number of occurrences of particular words in the position after -V is compared to the number of occurrences of all words in these positions (Table 1) and the Fisher exact test is used to test for statistical signiicance. The test shows that the distribution of urụy- and uṝá- difers signiicantly from the overall distribution of vowel-initial words but does not difer signiicantly from the distribution of consonant-initial words. urucí-, urukramá-, ur̄máthi-, and uv̄ca are each individually too rare to permit statistical conclusions to be drawn, but taken together, their distributions difer signiicantly from V- onsets, but not from CV- onsets. We can therefore reconstruct with certainty *vurụy-, *vuṝá-, *vurucí-, *vurukramá-, *vur̄máthi-, and *vuv̄ca as consonant-initial words. 21 For the distribution of uloká-, we get slightly diferent results: the post-minor-caesura distribution of this word is signiicantly diferent from that of V-initial words, but also signiicantly diferent from that of CV-initial words. This points to a variation between a consonant-initial and vowel-initial form of this word. It would appear, then, that in the time of the composition of the family books, both variants *vuloká- and *uloká- were present in the language. 22 The explanation for this variation is quite straightforward. We know that Vedic o goes back to pre-Vedic *au̯ (cf. AiG I: 35–36) and that at the time of composition it was probably still a diphthong. Because of the two subsequent glides *u̯ and two subsequent rounded back vowels *u in *vurulau̯ká-, the initial *v was probably undergoing an early dissimilation (even before the short u), but the process was not yet complete. Alternatively, we could assume that variants without initial *v were analogically inluenced by the plain adjective urú- (see immediately below). 18. There is another attestation of urụy- after -V̄C in the eighth book (RV 8.101.4c). However, hymn 8.101 is understood to be a later addition, since it violates the ordering (Oldenberg 1888: 228); accordingly, I have removed it from the count. 19. Note that one of the four verses attested after -V̆C is a repetition, but from two diferent books. 20. Again, there is another attestation of uloká- after -V̄C in 7.33, but because of the lateness of the hymn (Oldenberg 1888: 200) I took it out of the count. 21. Taken together, they are also signiicantly diferent from V-, but not signiicantly diferent from CV-. 22. Note that all examples where uloká- is attested after -V̄C occur in the seventh book. This distribution could indicate the starting point of variation. 548 Journal of the American Oriental Society 135.3 (2015) There is only one word with etymological *v in Vedic that behaves as if it were vowelinitial in the Rigveda: the plain adjective urú-. The following table represents the distribution of urú- in cadences of dimeter verse: the second column shows number of occurrences after -V̆C; the third column number of occurrences after -V̄C. The distribution of urú- is signiicantly diferent from that of CV-, but not signiicantly diferent from that of V-. Table 3 urú- -V̆C -V̄C 0 3(8) The explanation here follows from the fact that *v was lost before the long u earlier than before the short u. This caused the early loss of *v in the feminine forms of the adjective urví- (< *vurví- with long *u which later shortens either analogically or by sound change; cf. Lubotsky 1997). The v-less forms were then analogically transferred to the masculine urú- (< *vurú-). Thus, instead of *vurú- vs. urví- the opposition became urú- vs. urví-. Note that this analogical transfer did not happen in the non-adjectival forms, where the model for analogy was lacking (e.g., *vurụy-). This also explains RV 6.50.3a utá dȳv̄pr̥thivi ḳatrám urú (discussed above under 25), the only case in the Rigveda where the restoration of *v does not repair the irregular cadence: the cadence includes *urú-, which was (due to analogy) a vowel-initial word. 23 We have thus established that *v should be restored for metrical purposes in cadences. There is no reason not to assume the same rule in openings as well. In fact, by reconstructing initial *v, a more regular opening is obtained. Consider the example in (29). (29) RV 7.35.3b ́ạ́ na urucí bhavatu svadh̄́bhị ×⏑⏑− ́ạ́ no *vurucí bhavatu svadh̄́bhị ×−⏑− As already mentioned, the second syllable of the trị̣ubh openings is long in 14,190, or 86.2%, of all cases (van Nooten and Holland 1994: xvii). The reading with *v in (29) converts the short second syllable into a long one and yields a much more common opening structure. Examples from openings, however, cannot be decisive, because openings are not metrically regular enough to produce conclusive evidence and because any lengthening will automatically yield a better structure. However, now that we have independently established the rule for cadences, we can reasonably reconstruct the lost glides in openings as well. The statistical tests presented so far have been based on books 2–8 of the Rigveda. 24 We saw that *v at this point was still present in the language and that it should be restored for metrical purposes. However, if we apply these tests to the same words in books 1, 9, and 10, we see that the words with etymological *v̆- become signiicantly diferent from those beginning in CV-, but not signiicantly diferent from those beginning in V-. This means that on the way to the composition period for the later books, *v was lost and no longer afected the meter. For example, reconstruction of initial *v in the following cadence from book 10 would yield a superheavy syllable, a type that was strongly avoided by the poets. (30) RV 10.40.8b yuvạ́ vidhántạ vidháv̄m urụyathạ yuvạ́ vidhántạ vidháṿ̄ *vurụyathạ −⏑−⏑× −⏑−⏑× 23. As already mentioned, RV 6.50 is a highly irregular hymn, but the restoration of *v in urú- would make the cadence even more irregular by causing the ninth syllable to be long. 24. The distinction between books 2–8 as “old” and books 1, 9, and 10 as “young” is partially supericial, but accurate enough for the purpose of statistical tests. BEGUŠ: A New Rule in Vedic Metrics 549 §5 DIAGNOSTICS The rule proposed above can also be diagnostic for cases with unclear etymology. As already mentioned, the middle participle iȳná- can in principle go back to either *iȳná(root present) or to *yiȳná- (reduplicated present). The following table shows the distribution of iȳná- before -V̆C (second column) and -V̄C (third column) in the eighth syllable. If we look at the distribution of the word in books 2–8, we see that the distribution of iȳná- is signiicantly diferent from that of vowel-initial words and not signiicantly diferent from that of consonant-initial words. 25 Table 4 iȳná-26 -V̆C -V̄C 5 2 This suggests that iȳná- was a consonant-initial word and that it must therefore go back to the reduplicated present *yiȳná-. The restoration of *y in iȳná- additionally repairs the following ive cadences to their canonical structure: (31) RV 2.20.4c sá vásvạ k̄́ mam piparad iȳnó ⏑⏑−× sá vásvạ k̄́ mam piparad *yiȳnó −⏑−× (32) RV 7.17.7b mahó no rátn̄ ví dadha iȳnạ́ ⏑⏑−× mahó no rátn̄ ví dadho *yiȳnạ́ −⏑−× (33) RV 7.25.5b índre sáho devájutam iȳn̄́ ̣ ⏑⏑−× índre sáho devájutạ *yiȳn̄́ ̣ −⏑−× (34) RV 7.38.6b rátnạ devásya savitúr iȳnạ́ ⏑⏑−× = RV 7.52.3b rátnạ devásya savitúr *yiȳnạ́ −⏑−× §6 CONCLUSION In this paper, I propose a new rule of Vedic meter: I contend that the lost glides *v and *y should be restored before the short vowels ̆ and ̆ for metrical purposes. I argue that the glides were still present at the time when books 2–8 were composed and that word-inal syllables of the shape -V̆C should be scanned as long when they are followed by a u- or ithat goes back to *vu- and *yi-. I also argue that *v was lost earlier before the long vowel u than it was before the short vowel ̆ and that it therefore does not afect the meter in these cases. This early loss was perhaps due to the long vowel being more rounded than the short vowel and therefore triggering early dissimilation. The paper also sheds some light on the behavior of urú- and uloká-. urú- is shown to have been analogically remodeled as a vowel-initial word on the basis of the feminine forms; uloká- had variant forms with and without the initial glide due to early dissimilation from the following *u̯ and two rounded vowels *u. This new rule repairs twenty-six irregular cadences in books 2–8 to their canonical shape; it repairs thirty-two irregular cadences if we count all ten books together. The twenty-six repaired examples represent 11.1% of all cases in books 2–8 where -V̆C is followed by V- in 25. The p-value here is marginally non-signiicant (0.0558). Note also that I excluded RV 7.29.1d from the count, although maghávan- most likely ended in a single consonant (from the van-stem rather than the vant-stem; note, however, that some forms show transition from the former to the latter inlection): (i) RV 7.29.1d dádo magh̄́ni maghavann iȳnạ́ − ⏑ − x 26. Note that one of the ive verses attested after -V̆C is a repetition. 550 Journal of the American Oriental Society 135.3 (2015) the ninth syllable. Moreover, restoration of *v before *̆ repairs 54% of all cases where -V̆C is followed by u- in the ninth syllable. references AiG I = Jakob Wackernagel. 1896. Altindische Grammatik, vol. 1: Lautlehre. Göttingen: Vandenhoeck & Ruprecht. Arnold, E. Vernon. 1905. Vedic Metre in Its Historical Development. Cambridge: Univ. Press. Aufrecht, Theodor. 1888. Notizen. Zeitschrift für deutschen morgenländischen Gesellschaft 42: 151–52. Barnes, Timothy. 2014. Artemis ϝορθασια. Handout: 26th Annual UCLA Indo-European Conference, 24–25 September, 2014. University of California, Los Angeles. Chantraine, Pierre. 1968–1980. Dictionnaire étymologique de la langue grecque: Histoire des mots. 4 vols. Paris: Éditions Klincksieck. EWAia = Manfred Mayrhofer. 1992–2011. Etymologisches Wörterbuch des Altindoarischen. 3 vols. Heidelberg: Winter. Garcı́a-Ramón, José Luis. 1999. Griechisch Ζητ́ρ · Ζὲς ̓ν Κ́πρͅ, Vedisch ȳtár- “Rächter” und die Vertretung von *i̯ im Griechischen. In Compositiones Indogermanicae in memoriam Jochem Schindler, ed. H. Eichner, H. Ch. Luschützky, and V. Sadovski. Pp. 77–96. Praha: Enigma Corporation. Gippert, Jost. 1997. Laryngeals and Vedic Metre. In Sound Law and Analogy: Papers in Honor of Robert S. P. Beekes on the Occasion of His 60th Birthday, ed. Alexander Lubotsky. Pp. 63–79. Amsterdam: Rodopi. Kobayashi, Masato. 2004. Historical Phonology of Old Indo-Aryan Consonants. Tokyo: ILCAA. . 2014. Origin and Development of Sanskrit yy. In Vedic and Sanskrit Historical Linguistics: Papers of the 13th World Sanskrit Conference, vol. 3, ed. Jared Klein and Elizabeth Tucker. Pp. 1–23. Delhi: Motilal Banarsidass. Kümmel, Martin. 2000. Das Perfekt im Indoiranischen: Eine Untersuchung der Form und Funktion einer ererbten Kategorie des Verbums und ihrer Weiterentwicklung in den altiranischen Sprachen. Wiesbaden: Reichert. . 2010. Der Genitiv Plural im Indoiranischen: Paper presented at the 31st Deutscher Orientalistentag, Marburg. Handout available at https://www.academia.edu/342153/Der_Genitiv_Plural_ im_Indoiranischen. Ladefoged, Peter, and Ian Maddieson. 1996. The Sounds of the World’s Languages. Oxford: Blackwell. Lehiste, Ilse. 1970. Suprasegmentals. Cambridge: MIT Press. Leumann, Manu. 1968. Zu den lautlichen Fernwirkungen im Altindischen. In Pratid̄nam: Indian, Iranian and Indo-European Studies Presented to Franciscus Bernardus Jacobus Kuiper on His Sixtieth Birthday, ed. G. H. Schokker, J. C. Heesterman, and V. I. Subramoniam. Pp. 53–59. The Hague: Mouton. LIV2 = Helmut Rix et al. 2001. Lexikon der indogermanischen Verben: Die Wurzeln und ihre Primärstammbildungen, 2nd ed. Wiesbaden: Reichert. Lubotsky, Alexander. 1995. Relexes of Intervocalic Laryngeals in Sanskrit. In Kuryłowicz Memorial Volume, pt. 1, ed. W. Smoczýski. Pp. 213–44. Cracow: Universitas. . 1997. The Indo-Iranian Relexes of PIE *CRHUV. In Sound Law and Analogy: Papers in Honor of Robert S. P. Beekes on the Occasion of His 60th Birthday, ed. Alexander Lubotsky. Pp. 139–54. Amsterdam: Rodopi. Malzahn, Melanie. 2001. Sandhiphänomene im Rigveda als Relexe von Archaismen und Dialektismen. PhD diss., University of Vienna. van Nooten, Barend A., and Gary B. Holland, eds. 1994. Rig Veda: A Metrically Restored Text with an Introduction and Notes. Harvard Oriental Series, vol. 50. Cambridge, Mass.: Dept. of Sanskrit and Indian Studies, Harvard University. Oldenberg, Hermann. 1888. Metrische und textgeschichtliche Prolegomena zu einer kritischen Rigveda-Ausgabe. Berlin: Hertz.