8. An Inquiry into the Origins of Indic Trimeter

Let us first survey the basic types of dimeter and trimeter which are formally comparable in Greek and Indic. In the case of dimeters, the correspondences are obvious, and the essentials have already been presented in the Introduction. The analogue of the Greek octosyllables is the Indic octosyllable as attested in the Rig-Veda: To repeat, the rhythmical pattern of the closing in the scheme for the Indic verse represents simply the average, [2] not a constant as in the Greek verses.

In the case of trimeters, however, the correspondence between the Greek and Indic types is not obvious in all respects. The analogue of the Greek dodecasyllable is a Rig-Vedic dodecasyllable with two basic alternative shapes: {166|167} There is also a Greek hendecasyllable (as attested in Archilochos 188-192W) [3] which is a derivative of the iambic trimeter via catalexis: Similarly, the Rig-Vedic hendecasyllable may be described as equaling the dodecasyllable minus the last syllable. [4] Like the Rig-Vedic dodecasyllable, it has two basic alternative shapes: Note again that these schemes for Indic dodecasyllables and hendecasyllables represent averages, not constants.

The formal relationship between Greek dimeter and trimeter seems clear: the first consists of double ⏓ – ⏑ – and the second of triple ⏓ – ⏑ –. Given that the dimeter is divisible into an opening and a closing, we may redefine the trimeter as a structure built from the components of the dimeter: or Such a formulation seems to fit the surface facts of Greek metrics: It has yet to be shown, however, whether this formulation also accounts for the facts of Indic metrics, where the formal relationship between dimeter and trimeter is obscure. In order to examine this relationship more closely, we will have to have a more detailed notion of the Indic verse.

The Indic verse functions within the larger framework of a stanza. The major types of Rig-Vedic stanzas are composed of octosyllabic, dodecasyllabic, and hendecasyllabic verses. In the following list of stanzaic types in the Rig-Veda, I symbolize these verses as 8, 12, and 11 respectively. (table1) If we wish to be more specific about Rig-Vedic verses, we must be aware of these types of stanzas. A given type of verse in one particular type of stanza may have a rhythmical behavior different from that of the same type of verse in a different type of stanza. For example, the Gāyatrī octosyllable has a distinct subtype the so-called trochaic Gāyatrī, [6] whereas the Anuṣṭubh octosyllable has no such subtype except in late Vedic hymns. Or again, the Uṣṇih-B dodecasyllable has a word-break after the eighth syllable, whereas no such pattern emerges distinctly in an overall survey of Uṣṇih-A and Jagatī dodecasyllables. In sum, it is convenient to specify the verse-type by prefixing the stanza-type.

We are now ready to consider the formal problem of the relationship between Indic dimeter and trimeter. From the standpoint of {169|170} Indic metrics, the term ‘dimeter’ [7] arises from the fact that the Rig-Vedic octosyllable is divisible into a 4-syllable opening and a 4-syllable closing. [8] Note, however, that there is no obligatory word-break after the 4th syllable. Rather, the primary evidence for a 4 + 4 bipartition is the statistical contrast between the unfixed pattern of the opening (#⏓ ⏓ ⏓ ⏓) and the fixed pattern of the closing (predominantly ⏑ – ⏑ ⏓#). But there are also secondary factors indicating the functional reality of this 4 + 4 division. For example, the dodecasyllable of an Uṣṇih-B stanza (8 8 12) is regularly divisible into an 8-syllable and a 4-syllable segment, on the basis of a word-break after the 8th syllable. Now this 4-syllable segment closing the Uṣṇih stanza (8 8 8 + 4) has the same meter as the closing of the simple octosyllables in the stanza. [9]

To cite another example where the functional reality of a division into opening + closing is apparent, let us consider the metrical etymology of the trochaic Gāyatrī (with verse-closing – ⏑ – ⏓ #) in terms of the simple Gāyatrī (with verse-closing ⏑ – ⏑ ⏓ #). While the opening of the latter has a markedly free pattern, #⏓ ⏓ ⏓ ⏓, the opening of the trochaic Gāyatrī verse shows the most constrained meter of all the 4-syllable {170|171} openings in the Rig-Veda, three-quarters of the attestations being of the type #⏓ – ⏓ –. [10] Notice that the most distinct subtype #⏑ – ⏑ – of this regular opening is the exact reverse of the closing, – ⏑ – ⏓ #. Moreover, this opening subtype is metrically the same as the closing of the simple Gāyatrī verse. Thus there are different levels of reversal in the switch from simple Gāyatrī to trochaic Gāyatrī:

From the standpoint of Indic metrics, the term ‘trimeter’ [11] could be justified by showing that the dodecasyllable and hendecasyllable are built on a 4 + 4 + 4 and 4 + 4 + 3 tripartition. From the comparative evidence of Greek, I have postulated that the structure 4 + 4 + 4 originates from the components of the dimeter: From the evidence of Indic, there are two facts which support such a proposed derivation: (1) there is a regular word-break after the 8th syllable in Uṣṇih-B dodecasyllables; (2) there {171172} is a common word-break after the 4th syllable in all dodecasyllables (and hendecasyllables). But there are two important complications: (1) unlike Uṣṇih-B dodecasyllables, other dodecasyllables have no consistent word-break after the 8th syllable; (2) all dodecasyllables (and hendecasyllables) have a word-break after the 5th syllable if a word-break after the 4th is wanting.

Let us start with the first complication, which seems easier to motivate. Even if Uṣṇih-B were not attested, an original 8 + 4 partition of dodecasyllables seems indicated by the evidence of Jagatī dodecasyllables. Consider the statistical data from Rig-Vedic hymns of the so-called bardic layer—a period so archaic that originality in composition still flourished and actually caused rivalry among ‘bardic’ families, with the result that metrical forms are not only different from those generalized in other hymns but also peculiar to each family. [12] If we survey the hymns of one such family, the Vasiṣṭha-s of Book 7, we find an important idiosyncrasy; in the 12-syllable (and 11-syllable) verses of the Vasiṣṭha hymns, preponderantly Jagatī (and Triṣṭubh), there occurs a word-break after the 8th syllable no less than 57% of the time, vs. 35% in corresponding verses of other hymns. [13] Thus the internal evidence of the Rig-Veda suggests that the Jagatī dodecasyllable {172|173} originally had a consistent word-break after the 8th syllable, but that this archaic pattern was gradually eroded. By contrast, this same archaic pattern could have been retained and generalized in the Uṣṇih-B dodecasyllable.

The Slavic dodecasyllable, in the genre of lamentation, preserves clearly not just a 4 + 8 but also an 8 + 4 partition. Watkins has summarized the parallelisms between Indic and Slavic trimeter as follows: “identity of structure between longer and shorter verse is apparent in Vedic, where the shorter verse, typically ⏓ ⏓ ⏓ ⏓ ⏑ – ⏑ ⏓ … simply lacks the partially regulated internal colon, between the caesura and the cadence which is present in the longer verse. This system is clear in Slavic as well, where in the Serbo-Croatian laments the longer verse has 12 syllables divided into three tetrasyllabic cola; the shorter verse, of 8 syllables and two cola, differs only by the absence of the middle colon.” [14] In Greek too as in Slavic, a dodecasyllabic iambic trimeter has an alternate, the octosyllabic iambic dimeter. In fact, the Greek iambic trimeter and dimeter can still function together as components of the same poem: A notable example is Archilochos 177W: {173|174}

As viable long meters, the dodecasyllabic (and hendecasyllabic) verses of the Rig-Veda seem to conceal the posited 4 + 4 + 4 (and 4 + 4 + 3) tripartition. In Appendix B, this concealment on the ‘left’ (4 + 8) and ‘right’ (8 + 4) will be explained in terms of forward and backward dovetailing, from the Greek standpoint. I will now attempt an explanation from the Indic standpoint. On the ‘right’, with 8 + 4 (or 8 + 3), let us say that the fusion of syllables 9 10 11 12 (or 9 10 11) to 5 6 7 8 has resulted in the regularization of long over short in syllable 8: We are left with a metrically regular unit of five syllables (8 9 10 11 12) instead of four (9 10 11 12). The choice of long over short in syllable 8 might be explained in terms of its purported origin as a verse-final syllable. A verse-final syllable may be long or short by nature and it may function as a long or short depending on the underlying rhythm. Nevertheless, {174|175} it must be long by position, simply because a verse-final pause produces ipso facto a closed and therefore long syllable. Significantly, there is but one metrical condition which allows the frequent occurrence of short quantity for syllable 8 of dodecasyllables (and hendecasyllables): that is, if there is an actual word-break after syllable 8. [15] For example, in the trimeter hymns of Rig-Veda Book 7, there are 64 occurrences of short quantity for syllable 8 when caesura follows. We are not even counting such instances as where this slot is occupied by -e and -o before an a- placed in syllable 9. By contrast, there are only 6 instances of short quantity for syllable 8 when no caesura follows. [16] Since Book (= Maṇḍala) 7 is noted for the frequency of word-breaks after syllable 8 in trimeter, [17] this additional statistical evidence from that hymnic collection is all the more significant. If syllable 8 of trimeter is diachronically verse-final, we would expect reflexes of short quantity to be preserved by precisely this inherited mechanism, namely word-break immediately after 8.

So much for the metrical fixation of the segment 8̄ 9̆ 10̄ 11̆ (12̄̆) on the ‘right’ side of trimeter: we are saying that the closing has been regularized ‘leftward’, from 9̆ 10̄ 11̆ (12̄̆) to 8̄ 9̆ 10̄ 11̆ (12̄̆). On the ‘left’ side, metrical {175|176} fixation seems conditioned by the varying wordbreak patterns and I suggest that the latter type is less archaic, on the basis of the most immediate comparative evidence. Corresponding to the Indic Triṣṭubh stanza of four 11- syllable verses is a parallel Iranian instance, an Avestan stanza of four 11-syllable verses with regular word-break after 4 rather than 5. Theoretically, however, the Avestan state of affairs could be the result of innovation via failure to preserve a given metrical pattern B, even if it in turn were more recent than another given metrical variant A, which was preserved. The Iranian evidence is therefore insufficient, but we do still have recourse to the internal evidence of the Rig-Veda itself. In the trimeter, the segment 1 2 3 4 5 has stricter metrical patterns than 1 2 3 4, and we have already seen that rhythmical constraints in the opening are a mark of relative lateness. [19] Thus it is significant that whereas two-thirds of the openings have the pattern #⏓ – ⏓ – when the caesura comes after syllable 4, an even higher proportion, three-fourths of the openings, have {176|177} this fixed pattern when the caesura comes after syllable 5. Besides this restriction to #⏓ – ⏓ –, there are even further constraints when the caesura comes after 5: #⏓ – ⏓ – tends to be regularized to #⏓ – ⏑ –, and this regularization in turn tends to be connected with length for 5. [20] Even the segment after …5|| is rhythmically more inflexible than after …4||, and the predominant pattern is a rhythm-breaking ⏑ ⏑ for syllables 6 7. [21] In sum, a caesura after 5 is accompanied by a rhythmical pattern which “tends towards an absolutely rigid scheme”: [22]

The pre-caesura portion of the Triṣṭubh pattern is identical with the basic unit of the Indic decasyllable, namely the so-called pentad ⏓ – ⏑ – ⏓. The standard environment of the decasyllable in the Rig-Veda is the Dvipadā Virāj stanza: The word-break regularities of the Dvipadā Virāj reveal a 5 + 5 + 5 + 5 partition into four pentads, underlying the 10 + 10 partition into two decasyllables. For this reason, we may view the pentad, ⏓ – ⏑ – ⏓, as the basic unit of the Dvipadā Virāj stanza. Besides identifying the pentad ⏓ – ⏑ – ⏓ with the pre-caesura portion of the Triṣṭubh pattern it is also possible to describe the entire Dvipadā Virāj decasyllable as being the equivalent of this entire Triṣṭubh hendecasyllable, minus syllable 6: Furthermore, the last five syllables of the Triṣṭubh hendecasyllable are likewise equivalent to the pentad of the Dvipadā Virāj decasyllable, via the subtype ⏑ – ⏑ – ⏓ of ⏓ – ⏑ – ⏓. These comparisons between hendecasyllable and decasyllable imply an affinity between the two which is not just descriptive but also genetic. The most telling sign of such an affinity is that the rhythmical rigidity of the Dvipadā Virāj decasyllable is paralleled by the Triṣṭubh hendecasyllable only when the Triṣṭubh caesura comes after syllable 5 and not after {178|179} syllable 4. [23] Of course, any genetic relationship between decasyllable and hendecasyllable would explain the attested interchangeability, within the same stanza, of Dvipadā Virāj verses with Triṣṭubh verses; Rig-Vedic hymns especially noted for this phenomenon are 7.34 and 7.56. Moreover, these hymns contain Triṣṭubh verses with an especially high proportion of word-breaks after the fifth rather than fourth syllable. [24]

If, then, we posit that the relatively rigid pattern of rhythm in the trimeter sequence 1 2 3 4 5|| (as opposed to the more flexible pattern in the sequence 1 2 3 4||) is to be connected with decasyllables, this question still remains: how in the first place did the opening 1 2 3 4 5|| of the decasyllable become rhythmically inflexible? The factor of symmetry may well be the basic cause: a 5 + 5 partition in leads to the evolution of two pentads from the original opening and closing. In other words, the opening and closing come to function as two separate verses, “and these two exercise a mutual influence which tends toward their complete assimilation.” [25] The results we see in the fixed patterns ⏓ – ⏑ – ⏓ ⏓ – ⏑ – ⏓ of the decasyllable. {179|180} An example showing this split into smaller verse-units consisting of the pentad ⏓ – ⏑ – ⏓ is Hymn 4.10 of the Rig-Veda: its stanzas actually comprise an uneven number of pentads—three—plus one Triṣṭubh verse. This uneven number shows that there are no more distinct openings and distinct closings—just uniform pentads.

Even if we should be prepared to acknowledge that the skipping of syllable 6 from an 11-syllable verse results in a caesura after syllable 5, which in turn results in a 5 + 5 symmetry, we still have to account for the actual skipping. Thus it becomes necessary to reexamine and then possibly rearrange the factors of 11-syllable verse, 10-syllable verse, skipping, symmetry, and caesura, from a synchronic into a diachronic interrelation. One further factor is needed for setting up a revised interrelation: the 12-syllable verse, from which the 11-syllable verse is derived by catalexis. [26] Because of this derivation, the diachronic definition of decasyllables should be in terms of dodecasyllables, not hendecasyllables. With the even-numbered dodecasyllable as frame of reference, the factor of symmetry assumes greater importance. Though it seems like a result on the synchronic level (‘skipping syllable 6 of a hendecasyllable results in the 5 + 5 symmetry of a decasyllable’), symmetry may well be a cause on the diachronic level. In the {180|181} framework of dodecasyllables, the factor of symmetry means the half-and-half relationship of a structure intrinsically divisible into 6 + 6. It also means that the catalexis of a dodecasyllabic verse by skipping syllable 12 may be mirrored in a secondary catalexis, by skipping syllable 6. In other words, a 10-syllable verse may be derived from a 12-syllable verse via double catalexis. There is an overt Indic instance of internal catalexis at the center of a synchronic 6 + 6 division, and that is the so-called Bhārgavā verse: it is of the pattern Whereas the Triṣṭubh hendecasyllabic verse may be defined as a Jagatī dodecasyllabic verse with external catalexis of syllable 12, the Bhārgavā is a Jagatī verse with internal catalexis of 6; [27] and, most important, the Dvipadā Virāj decasyllable is a Jagatī dodecasyllable with internal catalexis of 6 and external catalexis of 12.

Given the archaic character of the 10-syllable verse in the Rig-Veda (it is not even to be found in the subsequent Atharva-Veda), [28] we may postulate decasyllabic influence on hendecasyllables and dodecasyllables. Consider the frequency of opting for caesura after syllable 5 in trimeter, vs. caesura after syllable 4. [29] The converse type of caesura-interchange {181|182} is yet to be considered: the Indic decasyllable with the pattern Here the caesura of trimeter seems to have imposed its pattern upon the decasyllable, with word-break after syllable 4 and the apparent skipping of syllable 5 rather than 6. The synchronic sequence seems to have switched from the conjectured diachronic sequence: from skipping causes new caesura to caesura causes new skipping. Actually, the mechanism of the switch may well be simpler: Since such decasyllables are inherited alternates of trimeters in the hymnic structures of the Rig-Veda, [30] this switch in the caesura patterns of decasyllables may actually reflect closing-regularization from the synchronic point of view. The Bhārgavā verse [31] likewise develops an alternate caesura after syllable 4: the switch here is {182|183}

If the Bhārgavā and Dvipadā Virāj verses stem from an inherent 6 + 6 division in the Indic dodecasyllable, it remains to reconcile this division with the posited 4 + 4 + 4 tripartition of the same dodecasyllable. [32] For such a conflict between 6 + 6 and 4 + 4 + 4, there is an indirect parallel from the comparative evidence of Slavic. As Jakobson has noticed, “the essence of Slavic epic verse is the unresolved tension between the bipartition and the tripartition both of the line and of the colon.” [33]

In the genre of Epic, then, Slavic shows a suppressed and latent opposition between tripartite and bipartite structure. In non-epic dodecasyllables, on the other hand, there is an overt variation: besides regular 4||4||4 there exist {183|184} instances of 6||6. [34] Unlike Slavic, however, Indic fails to generalize an optional caesura after syllable 6 in dodecasyllables. And yet the presence of an Indic 5||5 pattern apparently derived from an unattested *6||6 does not mean that the absence of 6||6 is an anomalous phenomenon. There is an obvious constraint against 6||6 which does not apply to 5||5. Given the mutual assimilation of the two halves in the latter, as attested in the Dvipadā Virāj meter, we would have to expect the same trend in any hypothetical Indic 6||6 as well. And what is tolerable for odd-number constituents 5||5 would be intolerable for even-number constituents *6||6. Among other things, the latter would become subject to further symmetrical division into an unwieldy *3||3||3||3. But it bears repeating and emphasizing that there is an indirect Indic cognate to Slavic 6||6, if indeed the 5||5 of Dvipadā Virāj verse is derivable from potential 6||6 via double catalexis. Greek shows a tendency to avoid this 6||6—a tendency possibly cognate with the Indic. [35]

Unlike the Indic decasyllable, the Slavic decasyllable has no 5 + 5 symmetry: caesura comes regularly after syllable 4, not 5. This avoidance of …5|| is significant for its compliance with a genre-constraint of Slavic versification: caesuras causing a symmetrical division of the verse are reserved for laments, {184|185} while those causing an asymmetrical division are required for epics. [36] Now the genre of the Slavic decasyllable is epic, whence the asymmetrical 4 + 6 bipartition. Compare the uneven division of Greek epic hexameter, or as opposed to the half-and-half division of Greek elegiac pentameter, Furthermore, the designation ἔλεγοc covers not only the elegiac meter but the elegiac genre as well. [37]

Although the caesura type …5|| is not attested in the Slavic decasyllable, it survives at least as a Slavic trimeter pattern, albeit indirectly. First, it is important to understand the relationship of the Indic decasyllable to the Slavic decasyllable. The former may be derived directly from the dodecasyllabic trimeter {185|186} via catalexis of syllables 6 and 12, whence the caesura pattern …5||. But there is also a conflated type of Indic decasyllable, with caesura pattern …4||. [38] This secondary Indic decasyllable is only indirectly derivable from the dodecasyllable, and it is with this conflated type that the Slavic decasyllable seems parallel. The ancestral dodecasyllable itself is preserved by Slavic in the genre of lamentation. Here the tripartition and occasional bipartition are on symmetrical lines, 4 + 4 + 4 and 6 + 6: Slavic laments also have a shorter verse-unit, of the pattern {186|187} This octosyllable is the perfect illustration of a dimeter counterpart to the trimeter dodecasyllable. If, therefore, we reconstruct an original dodecasyllable for Slavic epics (with asymmetrical caesura at …5||) in contrast to the attested dodecasyllable for Slavic laments (with symmetrical caesura at …4||), then the following proportion emerges: is to as is to The last type is precisely the pattern for the Common Slavic short epic verse, attested in Serbo-Croatian, Czech, Polish, Russian, and Bulgarian. [40] As for the long epic verse, the fact that a decasyllable ultimately prevailed over the potential hendecasyllable or dodecasyllable prevented by virtue of its syllable-count the generalization of caesura pattern …5|| and promoted the preservation of caesura pattern …4||: {187|188}

Whereas the genre of lamentation in Slavic preserves a dimeter type with caesura between opening and closing, no such regular word-break exists between the opening and closing of Greek or Indic dimeters. Thus the metrical partition between rhythmically free opening and fixed closing has not been matched by an obligatory phraseological partition. Similarly, Indic shows no generalization of word-break after syllable 7 in trimeter, despite the opportunity afforded by the metrical segmentation into a regular closing pattern – ⏑ – ⏑ ⏓ in dodecasyllables and – ⏑ – ⏓ in hendecasyllables. [42] By contrast, the Greek iambic trimeter of Attic Drama adopts the cognate opportunity, in that a word-break after syllable 7 is obligatory if it is absent after syllable 5. Both these examples, taken from dimeter and trimeter, point to the same principle: metrical segmentation may or may not be accompanied by obligatory phraseological {188|189} segmentation, known as caesura. There is a difference, however, between the metrical segmentation of dimeter into 1 2 3 4 vs. 5 6 7 8 and the metrical segmentation of trimeter into 1 2 3 4 5 6 7 vs. 8 9 10 11 (12). The former is primary, while the latter is secondary in that it can result only from the fusion of two preexisting metrical sequences. In fact, the segmentation 8 9 10 11 (12) is not just secondary but tertiary, since it is the outcome of adding a secondary metrical segment 9 10 11 (12) to another secondary segment, 5 6 7 8. Suture of 9 10 11 (12) to 5 6 7 8 led to obligatory lengthening of 8, whence the new segment 8 9 10 11 (12), based on its fixed rhythmical sequence as opposed to the free sequence preceding it. [43] In the archaic hymns of the bardic family Vasiṣṭha-s, there is actually a tendency toward insertion of word-break along the lines of the secondary rather than tertiary segmentation: that is, …8|| rather than *…7||. [44] In the Serbo-Croatian laments, such a tendency is generalized into a constant: there is a caesura after syllable 8 as well as after 4. [45]

I conclude with a synopsis of the cognate reflexes featured by caesuras in trimeter: