Appendix B. Dovetailing: Speculations on Mechanics and Origins

Glyconics, among other Greek meters, ‘dovetail’ forward. [1] That is, the word-break at the end of one Glyconic can be skipped and transferred to the position after the first long/short (⏓) of a following Glyconic: Examples abound, as in the Theseus Dithyramb of Bacchylides (18S), line 9: Even such complex meters as Asclepiads, which are internally-expanded Glyconics, [2] are subject to dovetailing. [3] There are actually two stages involved in such instances of Glyconic dovetailing: not only a shift of word-break from …8||1… to …8 1|| but also the specialization of the long/short (⏓) of syllable 8 into long (–).

The latter type of specialization is also {279|280} found in the Indic dodecasyllable known as Jagatī. In Chapter 8, I have argued that an original suture of the standard closing of octosyllables, … 5̆ 6̄ 7̆ 8̄̆, onto a full octosyllable ending in … 8̄̆’ is reflected in the fact that the dodecasyllable has a specialized closing … 8̄’ 9̆ 10̄ 11̆ 12̄̆ rather than … (8̄̆’) 9̆ 10̄ 11̆ 12̄̆. Only the caesura pattern …8||9… permits the preservation of a short (⏑) in syllable 8 of Jagatī. [4] In the cognate Greek dodecasyllable known as the iambic trimeter, there is a parallel metrical regularization of the closing from … ⏓ ⏑ – ⏑ ⏓ to … –⏑ – ⏑ ⏓, but this regularization is also accompanied by the evolution of a caesura pattern …7||8… : Contrast the dovetailing Glyconics discussed above: If we may call this second formula forward dovetailing, then the first formula is backward dovetailing. We must concede, however, that only the second of these two is a process still {280|281} operative on distinct meters. Backward dovetailing is no longer some sort of operative process within the iambic trimeter. Rather, I am simply claiming that it has left behind an indirect trace in the caesura pattern …7|| of this meter.

There is, however, an instance where forward dovetailing is likewise no longer operative. Again we must observe the caesura patterns of the Greek iambic trimeter, as attested in Attic Drama. If backward dovetailing is reflected by the caesura pattern …7||, then forward dovetailing may be reflected by the caesura pattern …5||, which is the functional alternative of …7|| in Greek iambic trimeter. In the related Indic trimeter, the dodecasyllabic Jagatī, we find the same caesura pattern …5||, but here it is the functional alternative of …4||, not of any …7||. Significantly, the Indic sequence 1̄̆ 2̄ 3̆ 4̄ 5̄ || is more rigid in its rhythm than 1̄̆ 2̄ 3̄̆ 4̄ ||, and this relationship suggests that the former is more recent than the latter. [6] Furthermore, the opening sequence ⏓ – ⏓ – is common to both dimeter and trimeter, whereas ⏓ – ⏑ – – is unique to the trimeter. In Chapter 8 I have argued that the Indic dodecasyllable evolved from a conflation of two types of suture: I would now argue that the caesura pattern …4̄|| is a direct product of the second suture, and that the pattern …5̄|| is an indirect product of the same, by way of a process that I propose to call forward dovetailing.

Similarly, I posit a conflation of two types of suture to explain the origins of the iambic trimeter of Attic Drama: The products of non-dovetailing, …4|| and …8||, are not extant in the dodecasyllables of Attic Drama; they are clearly preserved, however, in the dodecasyllables of the Rig-Veda.

A Greek metrical segment such as which may optionally shift into a forward dove-tailing pattern may eventually undergo a split in its evolution, into A typical instance of meter A in Greek is the Hipponacteum, a 9-syllable unit shaped ⏓ ⏓ – ⏑ ⏑ –⏑ – ⏓, which seems to have evolved from the resegmentation of the Glyconic (⏓ ⏓ – ⏑ ⏑ – ⏑ ⏓) that dovetails forward into the next metrical unit. Alkaios uses the plain Hipponacteum for the third part of a four-part stanza, as in 130.34LP: An instance of meter B in Greek is the acephalic Glyconic ⏓ – ⏑ ⏑ – ⏑ ⏓, otherwise known as the Telesilleion (after the poetess Telesilla, [7] who used this unit as a repeated verse) and well known by metricians for its occurrence in the Epidaurian Hymn to the Mother Goddess. [8] To take an example from another genre, consider the Telesilleion in the obscene Skolion 892P, which begins

Whereas a consequence of forward dovetailing from a given meter A to a given meter B is the possible generalization of hypersyllabism for A and acephaly for B, the backward dovetailing of A to B may result in the opposite generalization: catalexis for A and hypersyllabism—this time in the other direction—for B. In fact, the very {283|284} mechanism of catalexis may well have been caused by backward dovetailing. For a contemporary analogue to backward dovetailing—an analogue not necessarily genetic but merely typological—I cite Jakobson’s comments on a certain convention found in Serbo-Croatian epic versification: “A preference for open syllables is shown in a tendency to avoid closed syllables at the end of the line; and in actual delivery, consonants within the line which follow the syllabic phoneme are joined to the next syllable. [Cf. the situation in Greek, which has only three consonants (r n s) that can be word-final, let alone line-final.] Some Guslars generalize this principle of open syllables, and either swallow the final consonants (sometimes even the whole final syllable) or transfer them to the next line” (italics supplied). [9]

In Greek versification, the caesura pattern …7|| of iambic trimeter seems to be an example of backward dovetailing, albeit on the diachronic level. [10] I see a synchronic reflex, however, in the surviving poetic device whereby Greek 8-syllable verses can be truncated into 7-syllable verses via catalexis, with the catalectic octosyllable (= heptasyllable) serving as functional correlate of the plain octosyllable in stanzaic structures. For a familiar example, consider the correlation of octosyllabic Glyconic with heptasyllabic {284|285} Pherecratic. [11] In Indic versification, there is a negative parallelism to the Greek developments. Just as it fails to generalize caesura pattern …7|| in trimeter, Indic versification also fails to have an operative mechanism of converting 8-syllable verses into 7-syllable verses via catalexis. There are only sporadic attestations of residual 7-syllable verses, and these are deemed archaic by the metrician Arnold. [12]

By contrast, Indic has preserved catalexis in trimeter, since there are frequent interrelations of dodecasyllabic Jagatī and hendecasyllabic Triṣṭubh in the Rig-Veda. These interrelations, however, are interpreted differently by Oldenberg, who thinks that the dodecasyllable is derived from the hendecasyllable rather than vice versa. [13] Granted, the relationship between acatalectic and catalectic trimeter verses may have become reinterpreted in Indic versification as an additive process (11 + 1 vs. 11) instead of a subtractive process (12 vs. 12 - 1). Consider the Rig-Vedic meter known as ātijagatī (8.97.13, etc.), which consists of 12 + 1 syllables and which presupposes that the Jagatī dodecasyllable equals a Triṣṭubh hendecasyllable + 1. [14] Nevertheless, it does not follow that the original relationship between the Indic dodecasyllable {285|286} and hendecasyllable was likewise additive. The comparative evidence of the cognate Greek trimeters suggests the opposite, and the Gāthic evidence adduced by Oldenberg [15] is inconclusive. Admittedly, there seem to be no Gāthic dodecasyllables with caesura pattern …4||, only hendecasyllables. [16] On the other hand, there are Gāthic dodecasyllables with caesura pattern …7||, [17] for which there is a direct analogue in the dodecasyllables of Greek iambic trimeter. [18] Since the patterns …4|| in 11 and …7|| in 12 are mutually irreconcilable and since neither of the two can be derived directly from the other, the absence of …4|| in 12 (or of …7|| in 11) would be of no probative value.

I conclude this phase of my argumentation with a synopsis on forward and backward dovetailing:

It bears repeating that besides being synchronically operative in attested Greek Lyric, forward dovetailing is also attested in synchronically inoperative phases. In a given pattern M⏓||^N (where ⏓ designates a final syllabic addition to the meter M and ^ designates an initial syllabic subtraction from meter N), the underlying plain meters M and N can be perceived only so long as M⏓ and ^N remain together with intervening ||; once separated, however, M⏓ and ^N can become reinterpreted as new meters E and F, no longer synchronically connected with plain M and N. True, E = M⏓ via hypersyllabism and F = ^N via acephaly, but these are diachronic rather than synchronic equations. Forward dovetailing is the synchronic mechanism, while hypersyllabism in E and acephaly in F are products of forward dovetailing only after it has ceased to be synchronically operative in a given instance.

I propose that we see an example of meter E in the Hipponacteum, ⏓ ⏓ – ⏑ ⏑ – ⏑ – ⏓, [19] derived from the Glyconic ⏓ ⏓ – ⏑ ⏑ – ⏑ ⏓ (meter M) via hypersyllabism; an example of meter F is the Telesilleion, ⏓ – ⏑ ⏑ – ⏑ ⏓, [20] likewise derived from the Glyconic, {287|288} via acephaly. But now an additional complication is noticeable: where M = N, three rather than two segmentations are possible. Since both the Greek and the Indic evidence indicate that identical dimeters can occur in a series, let us consider what would happen when forward dovetailing is applied: If M = Glyconic, then I would imagine the following original contrast between a plain series of M vs. a dovetailing series: The final segment ⏓ – ⏑ ⏑ – ⏑ ⏓, according to this hypothetical scheme, becomes the Telesilleion, while the Hipponacteum results from the first segment, via generalization of ⏓ ⏓ – ⏑ ⏑ –⏑ – ⏓ over a potential ⏓ ⏓ – ⏑ ⏑ – ⏑ ⏑ ⏓. But the latter type of pattern is still apparently attested as a synchronic combinatory variant of the former in the stanza structure of Alkman’s Partheneion (1 P). Consider this contrast between lines 35 vs. 49 and 77: [21] {288|289} As for the medial segment ⏓ – ⏑ ⏑ – ⏑ ⏓ ⏓, I propose that it may become generalized as either ⏓ – ⏑ ⏑ – ⏑ – ⏓ or ⏓ – ⏑ ⏑ – ⏑ ⏑ ⏓. The first pattern is likewise found as a component in the stanza structure of Alkman’s Partheneion. Consider line 37: This octosyllabic type is well known for the Lesbian poem (Sappho 94D = Adespota 976P) which begins: The alternative generalization which could have stemmed from the medial segment of my hypothetical schema, ⏓ – ⏑ ⏑ – ⏑ ⏑ ⏓, has the shape of a meter known as the Prosodiakon, a basic component in the Dactylo-Epitrite versification of Pindar and Bacchylides. Significantly, the segment ⏓ – ⏑ ⏑ – ⏑ – ⏓ is still attested as a variant of ⏓ – ⏑ ⏑ – ⏑ ⏑ ⏓. For example, the former pattern is {289|290} joined with a preceding Lekythion (–⏑ – ⏓ – ⏑ ⏓) in the stanza structure of Alkman’s Partheneion (lines 36-37): The latter pattern, ⏓ – ⏑ ⏑ – ⏑ – ⏓, is the equivalent of ⏓ – ⏑ ⏑ – ⏑ ⏑ ⏓ in such familiar segments as Pindar’s Pythian 1.1: As is his wont in constructing Dactylo-Epitrites, Pindar here has made a word-bridge where tradition calls for a word-break. In fact, the essence of Dactylo-Epitrite versification is that distinct metrical members have eventually become fused members by way of word-bridging. [22]

The closing of the octosyllabic segment ⏓ – ⏑ ⏑ – ⏑ – ⏓, as Watkins has noticed, [23] is exactly opposite to that of the Glyconic ⏓ ⏓ – ⏑ ⏑ – ⏑ ⏓: that is to say, trochaic … – ⏑ – ⏓ vs. iambic … ⏑ – ⏑ ⏓. It remains now to apply the hypothetical schema {290|291} of sequential forward dovetailing to the plain iambic dimeter in Greek. The obvious result is the trochaic dimeter: The corresponding result in Indic would have implications for trimeter as well as dimeter:
An iambic dimeter sequence 8 + 8 + 8 (= 24), if it is resegmented as 12 + 12 (= 24), would lead to an iambic trimeter of the shape symbolized by bracket A. Bracket B symbolizes a stage transitional to a shift from iambic trimeter to trochaic trimeter, which in turn is symbolized by bracket C. The caesura pattern …5|| is, according to this hypothetical construct, a reflex of such a transitional segment B, with 5 + 7; the latter results from the model of segmentation A, with 4 + 8. This 4 + 8 of segmentation A would be in turn a sequential result of segmentation A’, with 8 + 4. [24] A conflation of the patterns labeled B and C would result in a Triṣṭubh hendecasyllable with caesura {291|292} pattern …4|| or …5|| and with the last five syllables shaped …⏓ – ⏑ – ⏓. The Triṣṭubh sequence …⏓ – ⏑ – ⏓ is identical to the opening of segmentation B, [25] or in other words, to the sequence preceding caesura at …5||.

The unit ⏓ – ⏑ – ⏓ is also attested in Greek as the trimeter opening, that is, the sequence preceding caesura at …5|| in iambic trimeter. I propose that the alternate pattern …4||, well attested in Indic, was once present but has been ousted in the standard Greek iambic trimeter of Attic Drama; the caesura pattern …8||, also well attested in Indic, would likewise have been ousted, while …7|| would have been kept as an alternative to …5||. [26] But there is internal Greek evidence for an earlier state of affairs. In the most archaic iambic trimeters attested, namely the corpus of the so-called Iambographoi (notably Archilochos, Semonides, Hipponax), there are survivals of …4|| and …8|| besides …5|| and …7||. Moreover, there is an important distributional constraint: I cite the following respective examples: {292|293} The symbol  ̮̑ indicates a word-bridge, that is, a constraint against word-break; in the first and second examples above, the  ̮̑ moves the || backward and forward respectively, by one syllable: in other words, backward and forward dovetailing occurs. Taking the posited Indic 4 + 8 and 8 + 4 dodecasyllabic sutures [27] as our comparative frames of reference, we are now in a position to reconstruct the following stages for Greek iambic dodecasyllables: The pattern 4||3||5 is a conflation of 4 + 8 (⏓ – ⏑ – + ⏓ – ⏑ – ⏓ – ⏑ ⏓) and of 8 + 4 via backward dovetailing (⏓ – ⏑ – ⏓ – ⏑ ⏓ + ̮̑⏓ – ⏑ ⏓), while the reverse pattern 5||3||4 is a conflation of 8 + 4 (⏓ – ⏑ – ⏓ – ⏑ ⏓ + ⏓ – ⏑ ⏓) and of 4 + 8 via forward dovetailing (⏓ – ⏑ – + ̮̑⏓ – ⏑ – ⏓ – ⏑ ⏓).

At this point, I must reiterate the basic theory: while the combination 8 + 4 results in backward dovetailing (thus 7||5), the combination 8 + 8 results in forward dovetailing. Such a {293|294} theory, as I have argued, allows for the following series of derivatives from resegmented Glyconic octosyllables: [28] The ocotsyllabic Prosodiakon is in turn subject to another round of resegmentation through forward dovetailing, whence these derivatives:

In sum, the iambic dimeter is to the trochaic dimeter

as the Glyconic is to the Prosodiakon

as the Prosodiakon is to the Hemiepes’

In the formation of the trimeter, Glyconic octosyllables behave differently from the octosyllables of iambic dimeter. Whereas 8 + 4 becomes 8||4 or 7||5 when 8 = iambic dimeter, the 8 + 4 combination may become 9||3 as well as 8||4 or 7||5 when 8 = Glyconic. By contrast, the caesura pattern 9||3 is strictly avoided in the iambic trimeter of the Iambographoi and of Attic Tragedy. [29]

In other words, Glyconics apparently undergo the innovation of extending the forward dovetailing of 8 + 8 combinations to the 8 + 4 combinations, where backward dovetailing seems to be the archaic heritage. For example, in Alkaios 357, there is a series of two Glyconics followed by ⏓ – ⏑ ⏓, with the following final word-break patterns: {295|296} Wherever Pindar combines a Glyconic with a following ⏓ – ⏑ ⏓, the resulting dodecasyllable may show either the pattern 7||5 (archaic backward dovetailing) or 9||3 (innovating forward dovetailing); the arrangement 8||4 is completely suppressed. [30] This distribution is typical of Pindar’s practice. He consistently avoids caesura at the most traditional junctures; instead, the caesura may be displaced one syllable forward or backward—but purposefully with as little predictability as possible. The resulting effect of fusion suits Pindar’s taste, in highlighting the structure of the stanza and shading the structure of the stanza-components.

Once forward dovetailing is extended from 8 + 8 to 8 + 4 combinations, I posit that the 4-segment itself (⏓ – ⏑ –) becomes capable of all {296|297} the rearrangements possible for 8-segments: [31] Here again are the rearrangements possible for the 8-syllable Prosodiakon:

Suture of these latter four with the former four patterns results in a bewildering variety of Dactylo-Epitrite stanzaic components, as employed by Pindar and Bacchylides. From Irigoin’s study [32] of frequencies in word-break and word-bridge, certain metrical types emerge which are conditioned not only by word-break but also, even more essentially, by word-bridge (  ̮̑ ). For example, the pattern is attested in 24 Pindaric lines (Pythian 9 strophe 4, Nemean 9 strophe 1, Nemean 11 epode 3) and 12 Bacchylidean lines (1 strophe 3/4, 1 epode 1/2, 10 strophe 3/4). {297|298}

Out of the 24 Pindaric lines, there is a word-break

Out of the 12 Bacchylidean lines, there is a word-break

If we consider y as the point of suture, then this pattern can be described as D + D. In poetic genres other than Dactylo-Epitrite, we would expect such a suture D + D to be reflected as D||D, that is, by a word-break between the two segments and

And yet, the statistics of Irigoin show that Dactylo-Epitrite versification tends to observe this point y not by word-breaking (||) but by word-bridging (  ̮̑ ). The actual word-break is pushed one syllable to the right (z) or left (x). The point x marks a C + B division, and this C||B pattern of word-breaking should be contrasted with the simultaneous D ̮̑D pattern of word-bridging.

Other examples: the pattern {298|299} is attested in 55 Pindaric lines and 11 Bacchylidean lines.

Out of the Pindaric lines, there is a word-break

Out of the 11 Bacchylidean lines, there is a word-break

Thus points y and x mark the simultaneous patterns d ̮̑D and c||B respectively.

The pattern is attested in 49 Bacchylidean lines. There is a word-break

Thus points y and x mark the simultaneous patterns B ̮̑d and A||b respectively.

I conclude with a list of further Dactylo-Epitrite schemata formulated from Irigoin’s statistical survey of bridge- and break-variations. [33] {299|300}

Earlier versification built from ABCD and abcd does not operate on the principles of both word-bridge and word-break: rather, the latter factor suffices. Among the most notable meters of this earlier genre are One sure reason for their preeminence was their archaic dodecasyllabism. Because of the regular word-break, such metrical types are called asynartetic. Typical of a transitional phase between asynartetic and synartetic phases is the approach of Anakreon to versification with the Enkomiologikon: he alternates C||b with an optional variant D||d: At times patterns latent in Dactylo-Epitrite versification reveal affinity to familiar patterns of other genres. As an example for comparison, let us consider the following two segments from Pindar (Nemean 1.9 and Pythian 9.22): Diachronically, these are a + A and c + A respectively: (When a + B = b||C, then the latter corresponds to the asynartetic Iambelegos.) The A-component, or Prosodiakon (⏓ – ⏑ ⏑ – ⏑ ⏑ ⏓), has a diachronic octosyllabic variant likewise generated from the parent Glyconic; this variant has already been discussed in the context of the Lesbian poem which begins If we designate this octosyllabic type (⏓ – ⏑ ⏑ – ⏑ – ⏓) as A’ vs. the A of the Prosodiakon, then the formula a + A’ yields the famous Alcaic verse (384LP): The caesura pattern b||C’ here corresponds to the plain b||C of the Iambelegos. Even more significant, the formula c + A’ yields the familiar Sapphic hendecasyllable (– ⏑ – ⏓ – ⏑ ⏑ – ⏑ – ⏓), as in line 1 of Sappho 1 LP: The caesura pattern c||A’ frequently results, as in line 2 of Sappho 1: Word-breaks may also occur after syllable 4 (hence d||C’), and word-bridges are not observed with any regularity. Given these data, we are now in a position to offer an explanation for why the Sapphic hendecasyllable does not operate on the inherited trimeter caesura system of …4|| and …5||. The reason, I suggest, is that the Sapphic hendecasyllable is not a catalectic but an acephalic dodecasyllable.