Rhythm without Beat: Prosodically Motivated Grammarisation in Homer

Citation: Use the following persistent identifier: http://nrs.harvard.edu/urn-3:hul.ebook:CHS_BlankenborgR.Rhythm_without_Beat.2014.

4. Audible Punctuation in Prosody

4.0 Introduction

In the preceding chapters I have shown that developing grammatisation in Homer results in grammatical clauses of variable size, strung together into a progressive tendency with the aid of extra-clausal transitional constituents. Parts of the grammatical clauses take the shape of minor phonological phrases, as do the transitional constituents. Developing grammatisation seems to make use of the possibility to form larger scale phonological phrases from the alignment of minor phrases. Grammatisation is thus brought to a higher and larger-scale level of organisation. This description of Homeric syntax lessens the autonomy and paratactic alignment of the metrical-phrase shaped units of Homeric discourse. It actually lessens the compositional importance of the recurring metrical building blocks, though they keep reflecting the underlying structural norms of dactylic meter that are still intact on the surface.

Clause formation in Homeric poetry is in a stage of transition. Where a considerable number of minor-phrase units is grammatically complete within the boundaries of the repetitive metrical phrase (verse, hemistich, or colon), almost an equal number of minor-phrase units is not. These units are not grammatically complete in themselves, but either preparatory or additive to other units. Smaller-scale phonological phrases appear to be reorganised into larger-scale phonological phrases: their independent character is lessened somewhat as they become part of a larger-scale whole. The alignment of smaller-scale phonological phrases is not paratactic anymore—not from the point of view of syntax, and not from that of prosody.

Phonological reorganisation may be at the expense of metrical boundaries, including the verse end. Starting from modern research into phonological organisation and reorganisation, in this chapter I will show that the Homeric epic offers proof for the intuition that the development of grammatisation is facilitated by prosodic reorganisation. While the underlying structural norms of dactylic meter remain intact on the surface, the different combinations of segmental weight, metrical position and word-internal position lead to a type of phrasing, even a kind of rhythmical repetition, that audibly differs from metrical colometry. Homer’s grammatical clauses become audible as phrases in the continuous flow of dactyls.

4.1 From Phonological to Phonetic Clues for Phrasing

As indicated in chapter 1, Bakker combines the notion of metrical cola with cognitive restraints on phrasing, but I showed that grammatisation is not similarly restrained. Syntactical considerations are not the main guide when it comes to metrical colometry. I follow the lead of Rossi 1965 when he suggests—unfortunately he does not more than suggest—that syntactical considerations can only be seen as decisive when meter is “indifferent” and when there is rhythmical “choice”. Choice in this respect means that frequent word end is not just evidenced in phonology. Even when it is, Rossi points at the need to reconsider its realisation. Frequent word end is a phonological reality, but a purely phonological approach [1] to Homeric verse disrupts dactylic repetition as it disregards the necessary phonetic adjustment. A purely phonological approach to Homeric meter also disregards that the maintenance of the structural norms of dactylic meter is on the surface guaranteed by metrical bridges, [2] meant to hinder phonetic disruption to the flow of dactyls. Phonology mirrors meter in that the binary pair of the structural elements thesis and arsis is kept intact. Phonology also mirrors meter in that on the surface the metrical colometry is kept intact; hence the positions of frequent word end. Phonology alone is not conclusive when it comes to sense-phrasing, as evidenced by grammatisation. The demarcative force of phonological boundaries like positions of frequent word end depends on the combination with another type of realisation. This other type of realisation has to do with audible termination: termination of the phonetic word, of the phonological phrase, of the coherent utterance. This realisation has to do, in other words, with audible pause. And that means that we have to look for phonetic clues for audible termination in Homeric hexameters.

4.2 Phonetic Clues

In their Prosody of Greek Speech, Devine and Stephens point out that the relationship between phonology and phonetics reflects the relationship between meter and language. This latter relationship is described as the mapping of language onto temporal patterns of linguistic origin: [3]

The position taken in this work is that metrics is a subdivision of rhythmics in the broad sense of that term. Metrics can be viewed as the central segment of a hierarchy. Above metrics in the hierarchy lies the highly abstract rhythmics that is the ontogenetic basis of metre, and below it lies the surface phonetic rhythmics that are generated in the performance of verse. Metrics is assumed to share a common property with rhythmics: the rules of metrics and the rules of rhythmics belong to one and the same system because they work together towards a single final objective, namely a degree of rhythmic regularity that neither could achieve alone. The relationship of Greek metre to the Greek language is also a topic of considerable disagreement. The view implicit in some modern work is that rhythms of Greek verse are of primarily non-linguistic origin and that versification involves searching through language for phonetic categories that can appropriately be mapped onto these extraneous rhythmic patterns and categories. It is tacitly assumed that while verse has rhythm, language—or at least pitch accented languages—have only intrinsic durations: segments and syllables just have the durations that they have, and those sequences that coincidentally fit the verse pattern are metrical. According to this conception of metre, the durational categories of language do not necessarily correlate with those of verse and are not organized into rhythmical patterns at all (or, if they are, it is into rhythms different from and unrelated in any direct way to the rhythms of verse). This view is not well supported by evidence from other languages.

In accordance with the assumption that the properties of Greek conform to the expectations of general linguistics and psychology, Devine and Stephens state that ‘the position taken in this work is that verse is “merely the language itself, running in its natural grooves”.’ They further assume that ‘the rhythms of Greek verse are simply more highly constrained versions of rhythms already existing in Greek speech.’ The rhythm of Greek speech is then analysed (101) as a temporal pattern comparable to stress feet:

Whatever the importance of literary convention, in principle Greek verse rhythm is born of the rhythm of Greek speech, and the former is consequently a valuable source of information about the latter. Just as an analysis of the distribution of words like Tennessee in English verse will show that in English speech they have primary stress on their final syllable or, under certain conditions, on their initial syllable but never on their medial syllable, so, for instance, a study of the distribution of fourth paeon-shaped words in the trimeter will indicate which rhythmic organizations of that word shape were possible in Greek speech and which were not, and which were usual at a certain speech rate and which less so. Finally, it is our position that however abstract the phonological terms in which metrical rules are stated and interpreted, most metrical rules reflect some form of measurable acoustic correlate at some rate of (prose) speech, not excluding artificially slow rates and styles. Whereas metrical rules can, and often do, abstract away from physically existing distinctions, it is reasonable to ask for empirical confirmation of the converse assumption, namely that some nonconventionalized metrical rules reflect a distinction that has no measurable surface correlate in any type of speech. The extent to which this position can be defended depends critically on the appropriate formulation of metrical rules.

Devine and Stephens note that phonology abstracts away from physically existing distinctions. They point out that the ‘purely metrical’ approach rationalises the timing of audible segmentation of language: [4]

[Phonological and morphophonological rules, especially rules governing the location of word and phrase accent] classify syllables into at most two, or rarely three, categories of syllable weight, according as they contain one, two or, less commonly, three units of rhythmic measurement called morae. If syllables having different structures are assigned to the same class by the phonology in rules relating to rhythm, that must be because they share some intrinsic property relating to rhythm, namely their mora count. […]

The status of the mora is not the same in all languages. In socalled syllable weight languages, light syllables have one mora and heavy syllables have two morae overall. In socalled mora timed languages, heavy syllables also have two morae, but there is a greater tendency for each mora to be timed independently, so that particularly in slow speech vocalic morae and consonantal morae tend to make roughly equal contributions to bimoraic weight. This difference is not grounds for dispensing with the mora in syllable weight languages.

For Devine and Stephens, mora count is a matter of looking at the phonological syllable structure. The count rationalises differences in syllable weight, as it is determined by differences in the duration of syllables’ rime structure. Devine and Stephens are keen, however, to stress the importance of the syllable’s phonetic realisation: [5]

Whereas there has been a tendency in the phonologically influenced tradition to assume that submoraic duration is uninteresting and definitionally irrelevant, in the philological tradition there has been a tendency to make just the opposite assumption, namely that submoraic duration is automatically relevant. In antiquity, the diversity of syllable durations within the categories of heavy and light syllables arising from differences in segmental structure (subcategorical syllable duration) was widely recognized. Dionysius of Halicarnassus has a full discussion in the fifteenth chapter of De Compositione Verborum […].

Despite the ancient distinction between rhythmics and metrics Devine and Stephens (51) stress the modern metrists’ position that assumes a categorisation of quantities, rather than the direct, sensory perception of quantitative relationships, as in music:

The ancient distinction between rhythmics and metrics correlates to a certain extent with the modern distinction between phonetics and phonology. Modern metrists, by contrast, have increasingly come to claim that Greek metre is immediately sensitive to segmentally motivated infracategoric differences in syllable duration.

The phonetic realisation of syllables reflects their submoraic duration: infracategoric differences in syllable duration, motivated by segmentation, so that there may be different “heavy” and “light” syllables. Such differences are not strictly phonologically motivated—let alone demonstrable. What is demonstrable is that the patterning of various “weights”, that differ in the way they are analysed phonologically, produces a very regular realisation of rhythmic alternation in verse. Verse seems to be very restrictive about the usefulness of specific weights for the patterning, especially in the last part of the line. The usefulness of specific syllable weights for the patterning depends on the modification of the actual performance duration of syllables, and hence on the modification by speech production. Devine and Stephens show that the phonetic realisation of syllables in speech reflects their rhythmical intensity. Often, phonetic realisation in speech, especially in other than slow speech, mirrors significant submoraic adjustment.

Devine and Stephens assert that footing (thesis/arsis) in metrical speech is a regularisation of the same processes that produce footing into stress-feet at the level of the word in normal speech. In their 1994 study, Devine and Stephens deal with the issue of rhythm as a timing mechanism for speech (“temporal organisation”), both in prose and in verse. They define the rhythm that arises from the alternating light and heavy syllables grouped into thesis/arsis-feet [6] as a form of “stress”. They recognise that the problem of the terminology “stress” is partly typological. Dynamic stress is the stress that has the physiological implication of a degree of additional energisation of the speech musculature. [7] If Greek prosody is in any way the key to the identification of word stress or sentence stress, it proves to be tempting to assume that word stress is in some way related to the pitch accent, and to the “long” or heavy segment. [8] Devine and Stephens note that, in general, pitch accented languages (like ancient Greek) have a tonal accent and not automatically a dynamic stress accent. [9] Sentence stress, as expressed through energisation of speech musculature, is therefore approached with equal care. Any reference to dynamic stress, both on the level of individual words and the sentence, must take into account that the development of the dynamic stress accent in Greek has been a matter of what Devine and Stephens describe as restructuring. [10] According to this approach the eventual stress accent in Greek is the result of the alignment of pitch with heavy, or lengthened, syllables. In Homer (and many metrical texts after Homer), however, there is no such alignment, or rather, not yet. [11] Though “stress” is gradually being used in the typological literature to imply other exponents of prosodic prominence, Devine and Stephens use it to describe the matrix formation underlying the temporal organisation of Greek speech. Metrical speech is then described as mapping of language onto the temporal patterns of linguistic origin, the rhythm arising from stress feet. [12] In poetry, of course, the submoraic adjustment of prima facie mora count is more restricted than in non-metrical speech. Looking for phonetic clues for phrasing in Homeric prosody means looking for opportunities for submoraic adjustment within the parameters, that is, the restrictions, of metrical formalism.

4.3 Submoraic Adjustment as a Clue for Phrasing

Devine and Stephens argue that all words come into the mind with syllable weights, and that it fits each word or word group to a rhythm of thesis/arsis alternation for speech production. Speech has a number of mechanisms for submoraic modification in order to fit the different weights to this patterning that implies a binary opposition based on quantitative temporal contrast. Submoraic modification is a speech mechanism to adjust the actual performance duration of syllables if the weight-contrast does not automatically provide the necessary durational contrast. That is why the study of patterning in poetry is so very useful for our understanding of the rhythm of Greek speech. To the observation of Devine and Stephens I want to add a point that is important for the study of poetry itself: metrical composition remains sensitive to the rhythmical properties that words and word groups have in normal speech. The step I would like to take from there is that the phrase structure apparent in metrical composition is expressed in terms of the same rhythmical properties that are organised in speech. Hence the modifications of rhythm on the level of the word and word group will also become apparent as metrical composition produces larger scale domains like the phrase.

Modifications in submoraic duration of vowels and consonants are perceptible to the ear in many spoken languages. Devine and Stephens [13] list a number of studies on different languages all providing empirical evidence for audibly noticeable modification of submoraic duration. They stress the need for an empirical answer to the question to what extent differences in syllable duration are systematically relevant for the rules of Greek meter, as ‘there is no reason to believe that poets are not consciously or intuitively aware of such differences and able to exploit them’:

Greek verse is sensitive not only to the intrinsic weight or mora count of a syllable but also to its syllabic context. The distribution of heavy and light syllables in the metres of Greek verse cannot be accounted for without the additional principle that syllables are systematically different according to the position they occupy in the various word shapes in which they are embedded. Greek verse does not treat the syllables of Greek words as sequences of monomoraic and bimoraic units following one another according to an equipollent mora or syllable timing scheme within the domain of the word; rather, there must be some sort of temporal organization below the level of the word domain that is reflected in the metre. [14]

Devine and Stephens follow Dionysius of Halicarnassus when he explicitly recognises the systematic metrical relevance of the two categories of syllable weight and the phonetic reality of multiple durations arising from the differences between syllables in the way they are segmentally structured, and from the position the syllables occupy in the various word shapes. Devine and Stephens list the scholars who have sought to compute syllable weight like Aristides Quintilianus did, and those who explained the metrical restrictions like bridges as due to submoraic differences. Following Wifstrand 1933, they emphasise the nonresolution of the longum in the hexameter. Differences in submoraic duration led Ruijgh 1987 to the identification of six different syllable quantities, West 1970 to seven different quantities, and others even up to twenty. Without fixing the exact number of syllable quantities Devine and Stephens note that:

it was either explicitly claimed or implicit in their assumptions that the segmentally longer types of heavy syllable are preferred in the biceps and the shorter types in anceps, while the longer types of light syllable are preferred in resolutions and the longer type of anceps. It would follow that metrical distribution could be exploited as a valuable source of evidence not merely for moraic but also for submoraic and segmental duration in Greek.

In relation to phonology, phonetics is some sort of submoraic adjustment. In their 1994 study, Devine and Stephens deal with a number of phonological phenomena in metrical surface structure as examples of submoraic adjustment: brevis in longo, phrase-final lengthening, light syllable prolongation on the anceps, heavy syllable subordination on the arsis, heavy syllable prolongation on the thesis, hiatus, shortening, prodelision, elision, and crasis. Of these, all phenomena pointing at vocalic coalescence (shortening, prodelision, elision, and crasis) as well as light syllable prolongation and heavy syllable subordination, are treated as indicators of phonological synizesis, just like phonological synaphy. The others are seen as demarcative. Brevis in longo, final lengthening, heavy syllable prolongation on the thesis, and hiatus are all treated in relation with prepausal location. Where brevis in longo is dealt with at the phonological boundary [15] , and hence as a phonological clue, final lengthening, heavy syllable prolongation on the thesis, and hiatus are valued as phonetic indicators of audible termination.

4.3.1 Final lengthening, heavy syllable prolongation on the thesis, hiatus

In an overview of the literature on final lengthening, Devine and Stephens present final lengthening as ‘a sort of drawling at the end of a group of articulatory events’ and an example of the ‘deceleration that is typical of various types of human motor activity’. Final lengthening is hence a signal of “close to termination of the domain” but not necessarily of demarcation of the domain. Devine and Stephens describe the relationship between final lengthening and demarcation as follows: [16]

The timing of speech is programmed within domains, and these domains are commonly demarcated by a lengthening of terminal elements. It is a general perceptual principle that longer intervals tend to demarcate.

They note that additional lengthening may be considered as an indicator of termination of a domain:

Later segments are lengthened more than earlier segments, and the vowel closing gesture is lengthened more than the vowel opening gesture, which suggests that final lengthening works inwards from the end of the domain.

Additional lengthening is not by itself a correlate of pause, as it is not independent of stress. ‘Final lengthening and lengthening under stress have results tending in the same direction’ but are produced by different mechanisms. Informally,

final lengthening is a sort of drawling at the end of a group of articulatory events, while stress is an actuation of the rhythmic beat. Final lengthening is a feature of the temporal organization of the prosodic domain and not merely a correlate of pause, since it often occurs in the absence of pause. However, the lengthening may be greater before a pause than otherwise. [17]

Additional lengthening provides the speech producer with time to think and plan, but may also show the characteristics of human motor activity:

The primary motivation of final lengthening is presumably related to the overall pattern of acceleration followed by deceleration that is typical of various types of human motor activity […]. In addition to deceleration, final lengthening may perhaps reflect a partial temporal allowance for a “deleted” pause or for a “gapped” constituent.

In general, the larger or higher ranked the domain, the greater the final lengthening: [18]

Lengthening at the end of an utterance is often greater than at the end of a major phrase; lengthening at the end of a major phrase tends to be greater than at the end of a minor phrase; and lengthening at the end of a minor phrase is greater than lengthening at the end of a word. When a word is uttered in isolation, its last syllable is not only word final, but also phrase final, sentence final and prepausal, and consequently undergoes significant final lengthening.

Discussing final lengthening in relation to metrical bridges, their observation concerning the hexameter underlines the possible equasion of final lengthening and demarcation: [19]

Iteration of word end coinciding with metron end or trochaically dividing the metron is also avoided in the hexameter, and the requirement that the medial division be a caesura rather than diaeresis helps to reduce word boundary patterning. Patterned iteration of word boundary replicating caesura or diaeresis is even more strongly avoided. Such rules presuppose that word final syllables are prosodically differentiated from word medial syllables, and demarcation typically has just this function.

Devine and Stephens identify both the minor and the major phonological phrase in ancient Greek. One of the phonetic aspects in the identification of phonological phrases is final lengthening: [20]

Evidence has already been cited that the amount of final lengthening is related to the hierarchy of prosodic domains; evidence has also been cited that in fluent speech prosodic domains encompass a larger span of phonosyntactic substance than in slow speech; it follows from these two premises that final lengthening is not merely physically reduced as speech becomes more fluent but is also adjusted to reflect the progressively more extensive prosodic domains. This argument may be restated in concrete terms as follows. It was found that nonbranching subject noun phrases in simple sentences had less final lengthening than branching subject noun phrases, which suggest that, at normal rates of speech, the nonbranching noun phrase was not processed as a separate phonological phrase but was joined with the following verb phrase into a single phonological phrase, whereas the branching noun phrase was accorded the status of an autonomous phonological phrase. Final lengthening was less at the ends of words that were not the last word of the phonological phrase. To the extent that this branching distinction governs phonological phrasing at normal rates of speech but not at slow rates of speech, it illustrates the mechanism of readjustment.

Applying general linguistic data to ancient Greek, they identify demarcation of the minor phonological phrase on the basis of metrical evidence: [21]

Evidence has already been cited that a minor phrase boundary may be foot internal, and that there is no necessary theoretical conflict between lengthening at the end of the minor phrase and foot structure assignment in the domain of the major phrase. Not only are longer items preferentially located at the end of domains, but phrase final lengthening tends to be proportionately greater on intrinsically or contextually long segments than on short ones: lengthening before voiced consonants in English is greatly exaggerated in prepausal position; […] It follows that any trend to prefer intrinsically longer syllable structure in metrical positions corresponding to the end of the minor phrase can be taken as evidence of phrase final lengthening. By the second half of the third century B.C. a long-term diachronic trend is discernible for increasingly strict regulation of the syllable structure preceding the caesura and diaeresis. The fact that this trend is hard to identify in earlier texts is a reflection of the generally more stringent rules of Hellenistic and later versification.

Devine and Stephens continue by listing the metrical evidence for the strictness of syllabic structure preceding the, in principle, phonological boundaries. When dealing with the major phonological phrase, they start from the concept of pause, and from there include the aspect of final lengthening: [22]

Final lengthening or drawling may be substituted for a true pause. Grammatical pauses occur between prosodic units, generally of the rank of the phrase or higher.

The occurrence and duration of a grammatical pause depend among other factors on the hierarchy of the prosodic domains as it reflects the hierarchy of syntactic structure. […] Inhalation requires more time than simply ceasing phonation, and consequently breathing pauses have greater duration than nonbreathing pauses, approximately twice the duration at most rates of speech; since longer pauses occur at deeper constituent boundaries, respiration occurs preferentially at major syntactic divisions.

After listing some examples they continue:

Speakers prefer to balance the phonological length of constituents in the output, which can introduce a disparity between syntactic and prosodic domains. Subjective listening often leads to perceived pauses when there is no actual pause, merely phrase final lengthening. Both normal and time compressed speech is easier to understand when pauses are inserted periodically without reference to syntactic structure, even in the absence of other phrasal prosodic cues.

In conclusion, Devine and Stephens identify the boundaries of the phonological phrase with the aid of phonetic clues like final lengthening. They do not postulate that final lengthening is as such always a marker for breathing or nonbreathing pause. On the contrary, they make quite clear that final lengthening is evident both in phonology (as maintaining the phrase-internal foot structure of the prosodic domain) and in phonetics (as an indication for audible termination).

In addition to final lengthening, audible termination may also stem from heavy syllable prolongation on the thesis. Such prolongation characterises the thesis of iambic or trochaic meter constituting a monosyllabic foot, as the preceding or following arsis of the same foot is syncopated (AT and TA ← <heavy syllable>). [23] On the basis of metrical evidence and evidence of the musical documents, Devine and Stephens assume that heavy syllable prolongation on the thesis mirrors any lengthening in speech, [24] including prepausal final lengthening:

Since the rhythm of Greek song was so closely tied to the rhythm of speech, one interpretation of the musical evidence is that prolongation in song reflects some degree of prolongation in speech. It could also be assumed that prolongation in song had a purely musical basis; in that case, one would still expect the prolonged syllable to be the same syllable that would undergo any independent linguistically motivated lengthening in speech.

Finally, Devine and Stephens discuss the issue of hiatus in relation to audible termination. They underline the syntactic constraints on hiatus between verses and within the verse. Their findings lead them to the conclusion that hiatus, though it is generally avoided in verse, is less unacceptable at weak word junctures in fixed phrases and appositive groups where the syllable organisation most resembled that found word internally (where hiatus is not uncommon). As elision excludes hiatus, slow rates of speech constrain elision (resulting in hiatus) in the minor phrase. In the major phrase, high rates of speech constrain hiatus, as rapid rates of speech may eliminate pause. In prose, major-phrase internal hiatus involving elidable vowels may be seen as orthographic hiatus, not necessarily as phonetic. To what extent this assumption is applicable to verse, however, is not clear. It is equally unclear whether hiatus serves as a prima facie indicator of termination: [25]

Another aspect that is unclear is whether the avoidance of hiatus in both verse and prose is simply and directly avoidance of hiatus as such, that is avoidance of difficult sequential syllable structure, or whether it is avoidance of boundary marking properties that are potentially an indirect reflex of hiatus. The former assumption is theoretically quite acceptable […]. However, if hiatus between lexical words were consistently eliminated by elision, prodelision and synizesis within a prosodic domain, then its occurrence would signal a boundary between domains, which is the hypothetical premise of the latter theory.

This is as far as Devine and Stephens go in the identification of phonetic clues for phrasing. In their view, audible pause may well result from submoraic adjustment, but submoraic adjustment is not immediately an indication for audible pause. Their examples of submoraic adjustment possibly indicating audible termination in Greek have been dealt with by other scholars. Hiatus in Homer has been discussed by many (among others Monro 1891:355–356; Parry 1971:191; Lejeune 1972:225–231; West 1982:39; Van Raalte 1986:93; Hoekstra 1989:9; Bakker 1988), whose findings are not questioned by the treatment of hiatus by Devine and Stephens [26] . Final lengthening as a marker of “completion”, however, has been discussed by Ruijgh 1987 in a way that deviates considerably from their conclusions. Another article by Ruijgh (1989) deals with a phenomenon that may be compared to heavy syllable prolongation on the thesis. Though not fully comparable to the treatment of the same issue by Devine and Stephens, Ruijgh’s discussion does contain various observations relevant for the value of final lengthening as a phonetic marker of termination in dactylic poetry.

4.3.2 “Completion”

In his 1987 article on “perfect” and “immeasurable” longa, Ruijgh identifies the phonetic reality of final lengthening as the explanation of Dionysius’s of Halicarnassus “perfect” or “complete” long element. Of course, Ruijgh’s implementation of the cross-linguistic terminology final lengthening by Devine and Stephens is based on their 1980 article summarising the general linguistic evidence for the phonetic phenomenon. Ruijgh, however, only applies the phonetic phenomenon to the word-final syllable, turning final lengthening into final-syllable lengthening (LFS).

In his article Ruijgh defends the position that the “perfect” long element, as identified by Dionysius in On Literary Composition, is in fact the long word-final syllable. Its phonetic duration, he argues, is longer than that of the non-final long syllable, taking into account that “final” here means phonetic-word final (more or less like Devine and Stephens take as a phonetic unit the appositive group centred around the lexical word). The non-final long syllable is then the “immeasurable” long of Dionysius. Ruijgh invents the “perfect” and the “immeasurable” short to match the longa: the “perfect” short is the non-final, the “immeasurable” the word-final short syllable. In case of brevia, word-final lengthening thus results in “immeasurability”. With these technical terms, Ruijgh is able to establish a correspondence between the usage of “immeasurable” by Dionysius, Aristoxenus, and Aristides Quintilianus; a correspondence sought for since Boeckh 1811 who posited a durational adjustment of both arsis and thesis to account for the “immeasurable” proportion of arsis to thesis as a result of durational adjustment of heavy syllables in anceps positions. [27] Ruijgh’s explanation of “immeasurable” still has to do with foot-internal proportion of thesis and arsis, but not primarily with the realisation of the anceps element as a heavy syllable. He rather includes the occurrence of non-final syllables into his account: if one of the two constituent elements of the foot, the arsis, is occupied by at least one “immeasurable”, that is by non-final longum or by word-final breve, the foot-internal proportion of arsis to thesis is by definition immeasurable.

Ruijgh extends his explanation to the zeugma of Porson: in accordance with the observations of others (like Snell 1962), he explains this zeugma introducing final lengthening. In his view, final lengthening is not evidenced by syllable structure: the preference for “lighter” syllable structures to account for the submoraic adjustment that is final lengthening. Ruijgh claims that final lengthening is an actual, absolute, lengthening of syllables that may disrupt the balanced podic structure of metrical rhythm. He uses the existence and persistence of metrical bridges as evidence: the occurrence of metrical bridges proves this disrupting effect of untimely final lengthening. The zeugma of Porson, for example, forbids word end after a heavy anceps as final lengthening causes too serious disruption to rhythm. As does Parker 1966, Ruijgh extends Porson’s law to the biceps of the hexameter: Homer avoids spondaic word end as final lengthening would disrupt dactylic rhythm. From this, Ruijgh concludes that, in general, the duration of a longum surpasses that of a biceps, and that the duration of a word-final longum surpasses that of a word-final double short. [28] His conclusion from especially the 17th and 20th chapter of On Literary Composition contradicts the current view (as in West 1982:20; Devine and Stephens 1984:22, 106; Wefelmeier 1994) that the dactylic double short has longer duration than the heavy arsis. Furthermore, Ruijgh concludes that any heavy element has more duration than double-short. [29] Ruijgh’s conclusions concerning the duration of the hexameter’s arsis are in accordance with the expectations concerning the durational difference between thesis and arsis based on speech rhythm, and I accept his application of final lengthening to the word-final syllable. But as much as Ruijgh’s usage of a phonetic phenomenon like final lengthening to explain “completion” (of the phonetic word) convinces me, I still find some consequences he deduces from his analysis unconvincing.

First, there is the assumption of “corresponding” light syllables. Ruijgh explains the συλλαβαὶ τελείαι of On Literary Composition 17, 12; 20, 17–20 as syllables with a λόγος-fitting duration (that is: in a specific “proportion” to the single non-final light syllable). He assumes a “perfect” light segment, the non-final light segment, as the duration that determines the λόγος. Next to the μακρὰ τελεία Dionysius mentions the μακρὰ ἄλογος, the heavy element whose duration does not suit a mathematical analysis of the λόγος between the thesis and the arsis. Ruijgh suggests that there may have been an “immeasurable” light as well: the word-final light syllable. If the immeasurable light is part of the element of the foot (or, as in the hexameter’s sixth arsis, is the element), the immeasurable light leads to an “immeasurable” foot-internal proportion as well. Like Dionysius, Ruijgh explicitly mentions the “perfect” duration of the heavy element, the μακρὰ τελεία, but defends his position with regard to the only τελείαι Dionysius actually mentions, i.e. the heavy elements.

Then there is the identification of the “perfect” long itself. According to Aujac and Lebel 1981, the μακρὰ τελεία indicates the “overlong” [30] syllable (rime: [long vowel + consonant]), the μακρὰ ἄλογος the long syllable that is long “by position” (rime: [short vowel + consonantal coda]). Ruijgh argues that the syllabic structure is not the explanation for the difference: there is no phonetic argument to suppose that there is any difference in duration between segments that are heavy due to a long vocalic nucleus (“by nature”), and those that are heavy due to a consonantal coda (“by position”). [31] Ruijgh thinks that the μακρὰ τελεία of Dionysius is the word-final heavy syllable (with LFS), and that the terminology μακρὰ ἄλογος is being used for the non-final heavy syllable. Dionysius, however, does not say that much. The verse (quoted from the Odyssey) that elicits remarks from Dionysius concerning μακρὰ τελεία and μακρὰ ἄλογος, runs:

αὖτις ἔπειτα πέδονδε κυλίνδετο λᾶας ἀναιδής

Odyssey 11.598

Then, again, the shameless boulder rolled back to the plain

Dionysius, [32] discussing the audible differences in tempo in Odyssey 11.593–598, points out that Odyssey 11.598 has a mere seven heavy syllables and that they are not τέλειοι. If the μακρὰ τελεία is the superheavy syllable, Dionysius could have pointed at verse-final -δής. If the μακρὰ ἄλογος is the heavy segment due to “position” as Aujac and Lebel defend, Dionysius would not have claimed that the line consisted of δάκτυλοι καὶ οὗτοί γε παραδεδιωγμένας ἔχοντες τὰς ἀλόγους ‘dactyls, and those with irrational syllables that are hurried along’: the line contains only two heavy segments due to “position” (-δον-, -λίν-). [33] On the other hand, if μακρὰ τελεία is the word-final heavy syllable, as Ruijgh argues, why doesn’t Dionysius take the verse-final syllable -δής into account? Ruijgh quotes Dionysius (20/93.13 U-R) πλὴν ἐπὶ τῆς τελευτῆς ‘with the exception of the positions at verse end’ and makes the quotation explains why Dionysius does not mention the verse-final syllable in this context: ‘En disant que les SL (syllabes finales) de λ 598 ne sont pas τέλειοι, Denys (20/92.19 U-R) oublie d’ajouter ‘abstraction faite, bien entendu, de la SF du vers’, restriction qu’il ajoute bien plus bas (20/93.13 U-R)’ . [34] Despite Ruijgh’s attempt to reconstruct Dionysius’s flaw of memory, the quotation from 20 (93.13 U-R) refers solely to the presence of dactyls in the line, not to the absence of τέλειοι.

The συλλαβαὶ ἄλογοι Dionysius is talking about, must be the non-final heavy theses αὖ-, -πει-, -δον-, -λίν-, λᾶ-, and -ναι-, the word-final light syllables -τα, -δε, the word-final double short -δετο, and the word-final [35] arses -τις, -ας, and -δής. Non-final light vowels ἔ-, πέ-, κυ-, and ἀν- may have been “measurable” enough themselves. The proportion (λόγος), however, existing between individual (light) syllables, has an effect on the proportion (also λόγος) between thesis and arsis: as such, only a non-final double short could have counted as “measurable”.

What, then, is a μακρὰ τελεία? I find the explanation by Aujac and Lebel misleading. Apart from the incompatibility with Dionysius’s observations, their explanation is built on the premise that there is a difference in duration between a segment with a long vowel, and a segment with a short vowel plus a consonantal coda. I agree with Ruijgh (and Devine and Stephens) that there is no phonetic reason for such a difference. Ruijgh’s identification of the μακρὰ τελεία as a word-final heavy syllable, and of the μακρὰ ἄλογος as a non-final heavy syllable, presents itself to me as much more convincing, but his analysis is not fully supported by Dionysius’s remarks.

4.3.3 Prolongation of the thesis in syncopated feet

As he announced in his 1987 article cited above, Ruijgh published a second article on final-syllable prolongation. This article, published in 1989, deals with march anapaest, and with the issue of “first” and “second” position of word end. As the numbering of the sections shows (the 1989 article starts with § 22), the 1989 article is meant as part II. The issues dealt with in the 1989 article are reminiscent of the discussion concerning heavy syllable prolongation on the thesis (AT/TA ← <heavy syllable>) in Devine and Stephens 1994; their discussion concentrated on heavy syllable prolongation due to syncopated feet in iambic and trochaic meter. The issue at hand is the maintenance of podic metrical structure. Ruijgh deals with similar issues in march anapaests: he focuses on the possibility of “equalising” durations march anapaests offer due to the rising word end on the long thesis.

In his 1989 article, Ruijgh advocates comparison of dactylic verse, like Homer’s hexameter, with the anapaestic dimeter. Dactyl and anapaest are both γένος ἴσον; their thesis and arsis have almost similar duration. The ancient ῥυθμικοί, Ruijgh points out, noted that the thesis often had a little more duration than the arsis (Aristides Quintilianus 33.26–28 in the edition of Winnington-Ingram). They labeled the dactyl ἀνάπαιστος ἀπὸ μείζονος (a maiore, cf. the ionicus), and the anapaest ἀνάπαιστος ἀπὸ ἐλάττονος (a minore). Their description makes it tempting to explain the dactylic-anapaest metarrhythmisis as a form of metrical anaclasis. It is important, however, to distinguish carefully between the rhythmical word type and the metrical phrase. Anapaests are studied as rhythmical word type by Dionysius, but as metrical cola by Ruijgh. In the anapaest-example Dionysius gives, he carefully chooses a line with a metrically “pure” appearance (holo-anapaestic), and with an anapaestic initial phonetic word:

βαρύ μοι κεφαλῆς ἐπίκρανον ἔχειν

Euripides Hippolytus 201

It is hard for me to wear a hear-net on my head

Dionysius always chooses his examples of the different ῥυθμοί according to these principles. Ruijgh [36] points out that anapaestic dimeters ought to be analysed κατὰ διποδίαν. The medial pause, a dieresis κατὰ διποδίαν, he says, facilitates the marching rhythm by equalising the durations of the two half-verses; [37] the possibility for equalising lies in the prolongation of the prepausal heavy element on the thesis. [38] In march anapaests, the exact time-value of phonetic word-final syllables depends on the amount of LFS required to maintain the rhythmical march-beat. Such application of LFS seems to be incompatible with the LFS resulting from, or rather indicating, a strong phonetic disruption. In anapaestic systems, for example in comedy, this failing relation LFS – phonetic disruption can be seen. The anapaestic system itself leads to πνῖγος “suffocation”, only to be released by a catalectic dimeter—in other words: by LFS and a phonetic disruption.

Ruijgh finds additional evidence for the realisation of non-prepausal LFS in the qualification κυκλικός “cyclic” for the anapaestic dimeter without medial pause. The κυκλικός-rhythm describes the anapaestic dimeter with caesura after the first light syllable of the second dipody. The terminology κυκλικός indicates that the realisation of the preceding heavy-syllable thesis does not facilitate phonetic disruption: Ruijgh suggests comparison with the dactylic third foot feminine caesura (following position 5½) frustrating both word end and phonetic disruption at position 5. Heavy non-final syllables on the first thesis of both dipodies of the anapaestic dimeter contribute to the κυκλικός-rhythm in the same way. [39] Thus he differentiates between a “first” word end, in order to evoke phonetic disruption, and a “second” choice for word end, not too far from the preferred location but nonetheless phonetically far less disruptive. Differences in location of word end, and of realisation of the prepausal element, between Greek and Roman poetry lead Ruijgh to the conclusion that in Greek the longum always has longer duration than the double-short.

The combination of the 1987 and 1989 articles presents Ruijgh’s analysis of the word-final lengthened syllable on the thesis. Together, the articles provide clues for the heavy word-final syllable undergoing final lengthening on a specific position within the metrical foot. The ancient ῥυθμικοί consider μακρὰ τελεία characteristic for anapaests, and μακρὰ ἄλογος for dactyls. The ῥυθμικοί are talking about the anapaestic and dactylic word-type (ῥυθμοί), not about the anapaestic and dactylic verse. The μακρὰ τελεία is hence characteristic for word-internal rhythm as in κεφαλῆς, the μακρὰ ἄλογος for a word like τεύχεσιν. The μακρὰ τελεία is therefore a word-final heavy syllable, the μακρὰ ἄλογος a non-final heavy segment. [40] In march-anapaests, every dipody ends in μακρὰ τελεία; [41] in the κυκλικός anapaest-rhythm, every dipody, like the 1st and 3rd (word-final) heavy-syllable thesis, ends in μακρὰ ἄλογος. So far, I accept Ruijgh’s observations and conclusions; first of all that phonetic final lengthening is primarily a matter of final-syllable lengthening. I argue, however, that, when combining word rhythm with verse rhythm, συμπλέκοντες τῇ μετρικῇ θεωρίᾳ τὴν περὶ ῥυθμῶν, [42] we may come to an additional conclusion.

What Ruijgh’s work calls attention to is in fact what seems to be a mismatch between the footing of speech and the footing of the continuous line. Rising word end on the thesis seems to orphan the thesis from its accompanying arsis in favour of some sort of attachment to the preceding arsis. Within the line, however, the thesis must form a foot together with the following arsis that belongs to the next word. Of the two different word-types, dactylic and anapaestic, one is at odds with the footing of the dactylic line. The anapaestic word-final heavy syllable μακρὰ τελεία is mapped on the thesis, as is the μακρὰ ἄλογος of the dactylic word type. Placing an anapaestic ῥυθμός in a dactylic line may therefore lead to a μακρὰ τελεία mapped on the thesis of the hexameter. As a result the thesis appears to orphaned from the arsis in the same foot, as if the “direction” of the line’s footing has changed. This happens as the anapaestic word-final heavy syllable coincides with word end at position 3, 5, 7, or 9.

At position 3:

τῆι δ᾿ ἑτέρηι ³ ἕθεν ἆσσον ἐρύσσατο φώνησέν τε

Odyssey 19.481

But with the other hand he pulled her closer, and spoke

At position 5:

οἳ δή μοι καμάτωι ⁵ θυμαλγέι γούνατ᾿ ἔλυσαν

Odyssey 20.118

Who make my knees weaken because of painful weariness

At position 7:

τὸν δ᾿ ἀπαμειβόμενος προσέφη ⁷ πολύμητις Ὀδυσσεύς

Odyssey 11.354

To him in reply spoke resourceful Odysseus

At position 9:

παῖδά τ᾿ ἀποκτείνεις ἐμὲ δὲ μεγάλως ⁹ ἀκαχίζεις

Odyssey 16.432

You try to kill my son, and you hurt me severely

Location has thus an effect on the metrically long syllable. Its mapping onto the various positions of the hexameter determines its suitability as a means of audible termination. In the following sections in this chapter I will show that audible termination in Homeric poetry can be described as “audible pause”, if a clear difference is maintained between phonological and phonetic demarcation. Following Ruijgh’s lead, I will distinguish two “strengths”, a stronger and a milder audible pause. Together the two pauses represent a maze of audible punctuation in the Iliad and the Odyssey. Indicators for pause are metrical-phonetic, so it is meter that offers the possibilities to pause. Additional phonetic indicators are final lengthening, mapping on thesis or arsis, and, crucial for the possibility to pause, realisation of the prepausal element not as a segment but as a syllable.

4.4 Audible Pause

Audible termination needs to be evidenced if proof is required for the intuition that the possibility for extended grammatisation in Homer stems from the reorganised alignment of metrical-cola minor phrases into major phrases. The reorganisation of this alignment is the result of audible differences in termination of minor and major phrases. What is required, in other words, is evidence for the audible pause in Homer. From my acceptance of existing views on prosodic demarcation, it follows that meter ought to be analysed for the possibilities it offers, as meter is both the phonological representation of synaphy and vowel coalescence, and the surface structure that remains intact despite phonetic disruption due to word-final lengthening or other manifestations of submoraic adjustment. Acceptance of Ruijgh’s theory of phonetic demarcation of word-final syllables implies that the possibility of word-final lengthening without damage to the surface structure of metrical footing of dactylic verse is to be found in the demarcative force of prepausal syllables; a force that depends on the realisation of prepausal locations both as a syllable, and as a position allowing for word-final lengthening. What I am looking for, in other words, is the combination of syllable and metrical position allowing for phonetic disruption, not for the combination that maintains dactylic surface structure by means of submoraic adjustment. Whereas I deviate from existing studies in looking for metrical-phonetic indicators for pause instead of purely metrical, purely phonological, or semantic indicators, the existing notions concerning terminal lengthening and maintenance of metrical surface structure remain my starting point.

4.4.1 Phonological pausa and pause

As mentioned before, the concept of “pause” in Homeric poetry has often been put on a par with the position of frequent word end in metrical colometry. The audible pause is of little importance in the extant ancient studies on meter and rhythm. [44] In chapter 3, I discussed Aristotle’s observation (Metaphysics 1093a30; the hexameter consists of 8 + 9 syllables) as concerning one specific dactylic verse-type. [45] Comments from antiquity on the “cuts” (τομή ‘cut’) of metrical colometry all refer to the positions of frequent word end. Modern studies have trouble identifying the audible pause. Korzeniewski 1968 and West 1982, following Fränkel’s colometry, do not give a reason for their identification of meaningful pauses. Both mention the τομή and its weakness in appositional groups (as do Devine & Stephens 1978), but the phonological or syntactical treatment of the verse end is not included. The value of the verse end as a meaningful pause cannot be explicated, other than by referring to the inevitability of word end. Koster [46] explains verse-final metrical indifference as elicited by the verse-end dieresis, but does not explain to what extent indifference elicits a meaningful pause. In his discussion of phrasing, Porter 1951 relied on an a priori notion of the normative effect on meaning when he discussed the rhythmical unity looking for its demarcations. Kirk took an extra step when he started looking for demarcation of unities that are “semantically coherent”: in such an approach, the identification of the unity sought for precedes the consideration of the visible points of demarcation.

Aristotle’s observation is likewise not to be seen as a permit to assume a compositional climax-principle, [47] or a persistent rhythmical division of the hexameter by its third foot. Such division of the hexameter is common though; Van Raalte [48] refers to a cognitive approach to rhythmical phrasing when explaining the division of the verse into hemistichs. Sicking 1993 and West 1997 both advocate such division. [49] Analysis of colometry, however, does not provide clues for audible termination, nor do the various ways of caesura notation. [50]

An interesting contribution to the suitability of phonological pausa as performative pause is the 1991 article by Daitz. As the title of the article (‘On reading Homer aloud: to pause or not to pause’) gives away, Daitz investigates ‘the pause at various points in the recitation and the possible effects that pause can produce on the listener’s perception of the poetic rhythm’. His ‘pause’ is audible: ‘By pause I mean a temporary interruption of phonation by the performer which is perceived by the listener as a temporary silence.’ Daitz is not very happy with the rhetorical punctuation that has permeated classical texts as we read them. Rhetorical punctuation relies mainly on 2nd century A.D. Nicanor, “the Punctuator”. [51] Daitz claims that at least 50% of printed punctuation is wrong or misleading from the viewpoint of performance. Fortunately, he claims, there are external sources to find help on the question of pause. Basically, Daitz deals with two questions: 1) is there always a pause at the end of the line? and 2) is there ever a pause within the line? As evidence for an automatic pause at verse end Daitz refers to the prosodic neutrality of the verse-final syllable (metrical indifference), hiatus at verse-end, and the clausular closure of the hexameter; testimonies from antiquity are taken from Latin authors. From Cicero (On the Orator 1.61.261) he understands that the usual practice was to recite a single verse on a single breath. Furthermore, Homer does not present hypermetric verses (against twenty in Vergil). Finally, the written practice of beginning each verse on a new line is a reflection of performance practice.

Concerning the verse-internal pause, Daitz supposes that rhythmic disruption is unwanted, but he finds evidence for lines without internal pause, again, in Cicero. He then adds phonological synaphy and vowel coalescence as additional arguments against internal pause: liaison, shortening, and elision all nullify the possibility of audible pause. Daitz allows for internal pause under specific circumstances: to avoid hiatus, to handle brevis in longo, or after certain heavy syllables, for expressive purposes. As rhetoric overcame rhythm, Daitz concludes, the scholarly procedure for semantic and grammatical analysis was unhappily transformed into a performance practice:

In the Homeric hexameter we have a form of poetry in which each verse was originally felt to be an integrated unit, centripetal in nature, knit together by the procedures of elision, correption, consonantal assimilation, and syllabic liaison. This poetry was normally read without pause from the first to the last syllable, but with a pause after the last syllable of each verse, and with sufficient flexibility of tempo and pitch to clearly convey meaning and expression without distortion of the rhythm. The overall aural effect would come closer to the rhythmic regularity and strictness of music than we are used to hearing in modern renditions of poetry. It would therefore be further removed from the rhetorical cadences of prose which we are accustomed both to hear and to see reflected in the printed punctuation of our texts, and which we unconsciously and erroneously tend to employ in out reading of ancient poetry.

Daitz analysis is in many respects relevant for the study of pause in the Homeric epic. Not because of his conclusion, as his identification of pause leans on the phonological approach (and hence on pausa), his argumentation is largely built on a specific interpretation of seemingly randomly chosen fragments from Latin authors, and his consideration for phonetics is at best “shown when wished for”. What makes it relevant is the consideration for coherent phonological phrasing as a hindrance for audible pause, the questioning of ancient written punctuation to identify audible pause, the “disappearance of recognisable meter” as the result of “rhetorical” colometry, [52] and, most important of all, raising the issue of ‘to pause or not to pause’.

4.4.2 Phonological restriction on audible pause

Daitz relies on the analysis of metrical formalism as phonology found in among others Korzeniewski 1968, Allen 1973, and West 1982. The phonological “syllabification” underlying metrical quantity is confined to the metrical unity of the verse: consonantal synaphy (συνάφεια) and vowel coalescence (συναλιφή/συναλοιφή) are proof of the phonological coherence within the verse. This coherence is broken at the metrical boundary of the verse end: shortening, liaison, and elision do not occur there. Sometimes, the coherence is broken at word end within the verse. Phonological coherence facilitates the “verse syllabification”, the division of the line into vowel-centred sounds that do not correspond either with morphological or orthographic syllables. The vowel-centred sounds are characterised prosodically, not orthographically. In the following example, as in chapter 3, I refer to this characterisation as “phonological segmentation”:

τὸν.δαὖ.τε.προς.έ.ει.πε.συ.βώ.της.ὄρ.χα.μος.αν.δρῶν (orthographical syllabification)
τὸν.δαὖ.τεπ.ρο.σέ.ει.πε.συ.βώ.τη.σὄρ.χα.μο.σαν.δρῶν. (phonological segmentation)

Odyssey 16.36

To him in turn the swineherd spoke, the men’s leader

I argued that the prosodic “cutting up” of the line can hardly be called “syllabification” as its clusters of “syllables” do not correspond with morphological lexemes, that is, with the intelligible and separable units of meaning (words and affixes). The orthographical syllables, a further subdivision of lexemes in single units of articulatory prominence, [53] do not signify metrical quantity either. The units of quantity are clusters of phonemes, groupings of sounds that have certain significance regardless of morphology. [54] When dealing with metrical text, I explained that it is better to leave the general terminology “syllable” behind and refer to the units of prosodic cutting up as “segments”. Prosodic “syllabification” is best referred to as “segmentation”.

Also to the domain of phonology belongs the identification of the prosodic unity on the level of words. On this level, the word boundary occupies a special position. As West 1982 and Ruijgh 1987 demonstrate, the unit is the phonetic word. As Devine and Stephens explain, the phonetic word, or appositive group, is the combination of the syntactically and phonologically free-standing word and its “phonological clitics”. [55] To illustrate the way phonology restricts the possibilities to “cut up” a verse, I present a few examples of verses “cut up” into phonetic words; the segments that form phonetic words are in brackets. Arrows indicate the direction in which phonological clitics “lean” on their lexical.

δῦ δὲ χιτῶν᾿ ἕλε δὲ σκῆπτρον παχύ βῆ δὲ θύραζε
(δῦ ← δὲ)(χιτῶν᾿)(ἕλε ← δὲ)(σκῆπτρον)(παχύ)(βῆ ← δὲ)(θύραζε)

Iliad 18.416

He put on a cloak, grabbed a sturdy stick, and stepped outside

ὡς δ᾿ ὅτ᾿ ἂν ἀίξηι νόος ἀνέρος ὅς τ᾿ ἐπὶ πολλὴν
(ὡς ← δ᾿)(ὅτ᾿)(ἂν → ἀίξηι)(νόος)(ἀνέρος)(ὅς ← τ᾿)(ἐπὶ → πολλὴν)

Iliad 15.80

Like when a man’s thoughts leap quickly who over a large stretch (of land has travelled)

εἰ δέ κε σίνηαι τότε τοι τεκμαίρομ᾿ ὄλεθρον
νηί τε καὶ ἑτάροις αὐτὸς δ᾿ εἴ πέρ κεν ἀλύξηις
(εἰ ← δέ)(κε → σίνηαι)(τότε ← τοι) [56] (τεκμαίρομ᾿)(ὄλεθρον)
(νηί ← τε)(καὶ → ἑτάροις)(αὐτὸς ← δ᾿)(εἴ ← πέρ)(κεν → ἀλύξηις)

Odyssey 12.139–140

But if you harm them, then I assure you, I predict total ruin | for the ship and your comrades; and as for yourself, even if you would escape

πρίν γ᾿ ὅτε δή με σὸς υἱὸς ἀπὸ μεγάροιο κάλεσσε
(πρίν ← γ᾿)(ὅτε ← δή)(με → σὸς → υἱὸς)(ἀπὸ → μεγάροιο)(κάλεσσε)

Odyssey 23.43

Before the moment when, finally, your son called me from the great hall

Consonantal synaphy and vowel coalescence (together labelled sandhi), and phonological clisis evidence the coherent phonological phrase. Absence of both sandhi and phonological clisis indicates the possible audible pause.

4.4.2.1 Resyllabification

Sandhi can only be applied within the phonological phrase: it often straddles the “metrical boundaries”. Some examples of straddling of positions of frequent word end in the thesis:

At the caesura at position 3:

καί οἱ ὑπὸ (:) ³ σκήπτρωι λιπαρὰς τελέουσι θέμιστας

Iliad 9.156

And under his sceptre they will pay him rich dues

At the caesura at position 5:

ὄψου τ᾿ ἄσαιμι (:) ⁵ προταμὼν καὶ οἶνον ἐπισχών

Iliad 9.489

I gave you enough meat after cutting it and holding wine to your lips

At the caesura at position 7:

Ἰδομενεὺς δ᾿ Ἐρύμαντα κατὰ (:) ⁷ στόμα νηλέι χαλκῶι

Iliad 16.345

Idomeneus (hit) Erumas in the mouth with the pitiless point

At the caesura at position 9:

ὧς εἰπὼν ἐς δίφρον ἑλὼν ἔναρα (:) ⁹ βροτόεντα

Iliad 17.540

Having spoken thus and after putting the bloodstained armour in the chariot

What happens at the caesura in these four examples is described by Devine and Stephens as onset to coda shift, the right to left movement of a consonant over the boundary between words: [57] ἔ.να.ρα ← β.ρο.τό.εν.τα. Such sandhi causes positions of frequent word end to be straddled by means of a consonantal coda as the closure of the segment. Thus the segment at the caesura features a rime of vocalic nucleus + segment-final coda. In segmentation the shift may also be in opposite direction: from coda to onset. [58] In that case the consonantal coda is not the closure of the segment’s rime but the onset of the next. The segment counts as light and the consonantal coda is attached to the first segment of the subsequent word as (part of the) onset. This shift may cause segmentation to straddle, for example, the trochaic word end in the hexameter’s third foot. Realisation of an audible pause following the third trochee would lead to a prepausal closed (= heavy) syllable in the third foot in the following examples:

ἠὲ φίλων ἐν χερσίν :^5½ ἐπεὶ πόλεμον τολύπευσε

Odyssey 14.368

or in the hands of friends after he had survived the war

ἔσσεται ἦμαρ ὅτ᾿ ἄν ποτ᾿ :^5½ ὀλώληι Ἴλιος ἱρή

Iliad 4.164

The day will come when finally sacred Troy must fall

Τρωιάδος τῆς εἵνεκ᾿ :^5½ ἐγὼ πάθον ἄλγεα θυμῶι

Odyssey 13.263

The Trojan (spoils) for which I endured pains in my heart

δρηστοσύνηι οὐκ ἄν μοι :^5½ ἐρίσσειε βροτὸς ἄλλος [59]

Odyssey 15.321

In handiness no other mortal would be able to compete with me

4.4.2.2 Orthographic coda on the arsis

Coda to onset shift in segmentation may be clearest seen in Homer by means of the movable (or ephelcystic) nu, whose application demonstrates the continuation of the phonological phrase over the third foot word end in one of the examples above (Odyssey 14.368). At the same time, movable nu may be used as orthographic coda, often to avoid hiatus:

τὼ καὶ ἀτασθαλίηισιν ἀεικέα πότμον ἐπέσπον

Odyssey 22.416

Precisely because of their misbehaviour they met a shameful death

In this example, however, orthography obscures what really happens at the trochaic word end. [60] The movable nu at the caesura suggests a prosodic coda (as part of the syllable’s rime), but that is metrically impossible: the syllable -σιν is heavy and impossible at this metrical position. Meter proves that the shift leads to segmentation σι.να and hence to the segment .σῐ. There is no possibility for an audible pause. Is the avoidance of hiatus so important then that the third foot caesura is being straddled? Allowing hiatus in the arsis would not have been a real problem. Within word groups that are semantically tied together, orthographic hiatus is common at the trochaic caesura and often, even phonetically, not explained as hiatus but as sandhi [61] due to the influence of a consonantal sound that has disappeared in writing:

πάντες ἐπεὶ βασιλῆα ἴδον βεβλημένον ἦτορ

Iliad 16.660

All of them, as soon as they had seen their king, mortally wounded

The combination of movable nu and an audible pause at word end (so that the nu is prepausal) can only be found when the word-final syllable occupies a position that is metrically “long”. The foot’s thesis is the preferred localisation [62] (with or without a subsequent pause):

ἀλλ᾿ ἐρέω μὲν ἐγὼν ἵνα εἰδότες ἤ κε θάνωμεν

Odyssey 12.156

But I will tell you, so that, in full understanding, we may either die (or flee)

μνησάσθω πᾶσιν γὰρ ἐπίστατο μείλιχος εἶναι

Iliad 17.671

One must remember that: for he knew to be friendly to all

Thus movable nu provides clues, and in case of metrically “short” segments, evidence for the coherence within phonological phrases, sometimes resulting in “dimming” of metrical cola. On the thesis, the movable nu can be the consonantal coda and allow for a subsequent pause. It may still be a coda without a subsequent pause, or not a coda at all. The application of movable nu at the two different third-foot caesurae (position 5 and position 5½) shows that inclusion of the orthographic coda into the rime of the segment cannot be as easily accounted for on the arsis as on the thesis.

Other phonological boundaries following the foot’s arsis where the position of the orthographic coda in the rime structure of the segment provides clues concerning the coherence within the phonological phrase (at the expense of a possible audible pause) are positions 8 and 12. When preceded by spondaic word end, the bucolic dieresis is regularly straddled by sandhi in phonological phrases that are syntactically coherent as well in the Iliad. [63] At verse end, similar straddling is found if the sixth arsis is occupied by a syllable containing a naturally long vowel or diphthong as its nucleus followed by a coda:

τὼ δὲ πνείοντε κατ᾿ ὤμων
αἰὲν ἔχ᾿ ἡνίοχος θεράπων

Iliad 13.385–386 [64]

And the servant driver always curbed | the two of them as they snorted over his shoulders

μηδέ τι δούρων
ἔστω φειδωλή

Iliad 22.243–244

And | there ought to be no thrift | of spears at all

4.4.2.3 Segment as syllable

In my account so far I have carefully distinguished between segment and syllable. Every unit in prosodic cutting up is a segment, but lack of 1) metrically demonstrable consonantal synaphy and vowel coalescence, and 2) phonological clisis results in realisation of a segment as a syllable. The combination of 1) and 2) leads to identification of the word-final syllable as the right branch demarcation of a phonetic word that is not subject to synaphy or coalescence, in other words, not subject to sandhi. The boundary of the phonological phrase, the position at which word-final lengthening may indicate audible termination and audible pause, must be realised as a word-final syllable.

4.4.3 Metrical restrictions on audible pause

Taking word-final lengthening of phonetic-word final syllables as an indication for audible pause in a metrical text, as I do following the combined lead of Ruijgh and Devine and Stephens, means that such lengthening cannot disrupt the metrical surface structure as evidenced by phonology. Lengthening cannot excessively prolong the light segment, just like moraic adjustment cannot shorten the heavy segment; at best it may prolong a short syllable to the weight of a heavy segment, or shorten a long syllable to that of light segment. As mentioned above, metrical bridges and the avoidance of spondaic word end [65] show that prolongation of both light and heavy word-final syllables on the arsis is much more restricted than on the thesis, the non-resolvable structural element of the foot. Porter and Ruijgh, amongst others, consider this as evidence supporting an intrinsically longer duration of the thesis when compared to the arsis, but scholars like Wifstrand, Irigoin, and West [66] argue that the arsis has longer duration when it is realised as a double-short. In his 1998 edition of the Iliad, West states (xxv): ‘in bicipitibus, ubi tempus paullo maius quam in thesi pedis implendum erat’.

Devine and Stephens 1994 reconstructed the rhythm of non-metrical utterances from, among other sources of evidence, metrical evidence, as they consider the basic principles of verse rhythms and speech rhythms the same, as are their basic units of organisation. As they put it: ‘poetry is like speech, only more so’. Identification of stress feet in prose utterances is explained by describing the stress foot’s structure as [thesis + arsis]. Determining the rhythmical structure of stress feet in Greek speech involves both mora count and submoraic duration: Devine and Stephens show that the phonetic realisation of syllables in speech reflects their rhythmical intensity. Often, phonetic realisation in speech, especially in other than slow speech, mirrors significant submoraic adjustment. Still, Devine and Stephens note that the submoraic adjustment of prima facie mora count is much more restricted in poetry than it is in prose. The higher level of regulation in verse is not so much due to the non-metronomic character of syllable durations, but rather to ‘verse rhythms being much more regular than speech rhythms in the durational patterns of their performance’. Still, the study of what is allowed in verse made it possible to learn things about the nature of rhythm (thesis and arsis) in speech, and about the qualities of different syllable structures in different word-internal positions. Especially the study of how allowances in verse change over time as the severity of restriction changes over time is very instructive. In their 1994 study, Devine and Stephens assume that metrical composition remains sensitive to the rhythmical properties that words have in normal speech. Ruijgh’s conclusion that the foot’s metrical thesis is longer than the metrical arsis, even when thesis and arsis have the same weight, corresponds to the durational ratio of the rhythmical thesis and arsis in speech. As word-rhythm shows a resistance to putting heavy final syllables in metrical arsis positions, so final heavy syllables in verse are rhythmical theses. In 4.3.3 I explained my reasons for accepting Ruijgh’s argumentation in favour of accepting the phonetic disruption of final lengthening as final-syllable lengthening. If I combine the arguments in favour of a syllable (instead of a segment) as phrase-demarcating with Ruijgh’s conclusions with regard to the proportion [TA] (conclusion: T ≥ A), I find reason to repeat my additional conclusion (additional to the observations by Ruijgh) concerning the possibility for audible pause, this time strengthened by metrical restrictions.

As Ruijgh starts from the notion of foot-internal durational ratio, the proportion T ≥ A is indicative for the specific type of meter, as it indicates the “direction” of the “rhythmisis”. As metrical patterns with comparable durational proportions may easily be mixed up or put on a par with one another, in non-stichic metrical strophes iambs are easily analysed as trochees, just like, in a similar way, anapaests and dactyls seem to be interchangeable. But Ruijgh’s view on “direction” shows that they are not. The way in which the duration of the thesis influences that of the arsis (as in the dactyl) or, alternatively, adapts its duration to that of the arsis (as in the anapaest) becomes clear from the opportunities that segmentation offers for both phonological coherence and phonetic demarcation. As I follow Ruijgh and, implicitly, the structural analysis of Devine and Stephens, I accept that the durational proportion of the metrical foot assigns longer duration to the segments that are mapped on the foot’s thesis. In a quantifying language like ancient Greek, the demarcative value of segments is thus restricted in three ways: (a) in order to demarcate a phonological domain through word-final lengthening, segments must have certain duration; (b) segments must be mapped onto the foot in such a way that their mapping allows for final-syllable lengthening, and (c) segments must be realised as syllables, that is not subject to sandhi or resyllabification, if they are to be phrase-final.

4.5 Audible Pause in Homeric Prosody

I argue that it follows from these three restrictions together that the likeliest position for audible pause is a heavy syllable (not just a segment) that is mapped on the thesis. Localisation of metrical word types, and the avoidance of spondaic word end show that word end in a heavy syllable means word end in a marked heavy syllable—on the thesis, the dactyl’s first element that cannot be resolved into two light segments. Such a syllable indicates termination of the audibly demarcated phrase. In the following example word end in a syllable on the thesis is underlined:

τῶν ἕν᾿ ἀειραμένη Ἑκάβη φέρε δῶρον Ἀθήνηι

Iliad 6.293

Having lifted one of these Hecabe carried it as a gift for Athene

As mentioned earlier, word end in a marked heavy syllable brings out a different metrical pattern: that of the anapaest. Word end in a marked heavy syllable gives the thesis the chance to “equalise” its duration to the preceding arsis (ἕ̆ν᾿ ἀ̆εῑρᾰμέ̆νη̄ ἑ̆κά̆βη̄). At the same time, any additional word-final lengthening exaggerates the separation of the orphaned monosyllabic thesis (T[A]) from the following arsis that belongs to the same metrical foot (TA). On the level of the word, and of the phrase, the orphaned thesis is heard as the thesis attached to the preceding arsis in the rhythmical synaphy of the dactylic line in a rhythmical word that by itself seems to be rising. The metrical shift is a prosodic reality, that of metarrhythmisis (τὰ μέτρα μεταρρυθμίζεσθαι): the shift from one metrical pattern (dactyl) to another (anapaest). It is, I argued, in fact a change of “direction”: the orphaned thesis becomes attached to the preceding arsis, thus suggesting a change in foot-structure. ‘P’ stands for the possibility of phonetic demarcation:

τῶν ἕν᾿ ἀειραμένη Ἑκάβη φέρε δῶρον Ἀθήνηι
[T A ][T A][T/P A][T/P A] [T A ][T A] “dactylic rhythm”
[T ][A T][A T/P][A T/P][A T] [ A T][T/P] “metarrhythmisised”

Word end in a light syllable will also carry “extra duration” (διάστασις). Extra duration as submoriac adjustment of light word-final syllables is mapped onto the arsis and hence more severely restricted. Ruijgh [67] suggests that LFS of a light syllable may result in approximately half the additional lengthening compared to LFS of a heavy syllable. Restricted demarcative final lengthening signals the branching of “smaller phrases”; it depends on the rate of speech whether these “smaller phrases” are processed as separate phrases, or as parts of larger-scale phonological phrases. [68]

There are therefore two types of audible pause depending on the additional lengthening the prepausal syllable may carry. The variance in pause is reminiscent of Ruijgh’s ‘coupe primaire’ and ‘coupe secondaire’, but his labels refer primarily to the frequency of phonological boundaries. Audible pauses ought to be differentiated by the phonetic disruption they cause, in as far as phonetic disruption is allowed by metrical position.

As phonetic word end on the thesis allows for considerable phonetic adjustment/additional lengthening of the prepausal syllable, I will refer to it as the primary pause. The primary pause is realised in various positions in the verse (indicated as ‘:’):

At position 3 (“trithemimeres”):

Ἀργεῖοι :³ καὶ δ᾿ αὖτε μεθίετε Ἕκτορι νίκην

Iliad 14.364

Greeks, do you grant Hector victory again

At position 5 (“penthemimeres”):

ὧδε δὲ μυθέομαι :⁵ Ζεὺς δ᾿ ἄμμ᾿ ἐπὶ μάρτυρος ἔστω

Iliad 7.76

That is what I think, and Zeus must be here for us, present as witness

At position 7 (“hephthemimeres”):

νῦν δ᾿ ἅμα τ᾿ αὐτίκα πολλὰ διδοῖ :⁷ τὰ δ᾿ ὄπισθεν ὑπέστη

Iliad 9.519

But here and now he gives many, and others he promised in due course

At position 9 (“ennehemimeres”):

ῥίμφ᾿ ἔφερον θοὸν ἅρμα μετὰ Τρῶας :⁹ καὶ Ἀχαιούς

Iliad 17.458

quickly brought the swift chariot in the midst of Trojans and Greeks

Non-restricted phonetic disruption on the thesis stems from the phonetic word-final syllables containing a long vowel or a diphthong, or the word-final superheavy syllable (‘:’ indicates the pause): [69]

ἦν δέ τις Εὐχήνωρ :⁵ Πολυίδου μάντιος υἱός

Iliad 13.663

There was a certain Euchenor, son of the seer Polyidus

σχέσθε φίλοι :³ καί μ᾿ οἶον ἐάσατε κηδόμενοί περ

Iliad 22.416

Don’t, my friends, and please allow me alone, concerned though you are,

οὔ τι κατακτείνει :⁵ πλάζει δ᾿ ἀπὸ πατρίδος αἴης

Odyssey 1.75

He does not kill him, but makes him wander far from his native land

τίς :¹ πόθεν εἲς ἀνδρῶν :⁵ πόθι τοι πόλις ἠδὲ τοκῆες

Odyssey 19.105

Who are you? From where among men are you? Where are your city and parents?

It may also stem from the word-final closed syllable containing a short vowel, [70] when mapped on the thesis: [71]

ὁ ξεῖνος ἐμέθεν (: ?) ⁵ ἐθέλω δέ μιν ἐξερέεσθαι

Odyssey 19.99 [72]

(that) the stranger (may listen) to me; I wish to interrogate him

οἴκαδε νισσόμενον (: ?) ⁵ ὃ δ᾿ ἔβη μετὰ πατρὸς ἀκουήν

Odyssey 4.701

On his return home; he went searching for word of his father

ἐλθὼν ἐξ ὄρεος (: ?) ⁵ ὅθι οἱ γενεή τε τόκος τε

Odyssey 15.175

Having come from the mountain, where his origin is and his offspring

I will refer to the word-final lengthening whose “demarcative strength” in prosody depends on the rate of speech, as the secondary pause. As said, it may or it may not be exploited to suggest a “smaller phrase” within the larger-scale phonological phrase. To the ear, the difference between the primary and the secondary pause may also have been the difference between the breathing and the nonbreathing pause, as this difference would be influenced by rate of speech: at higher rates of speech, the rhythmical disruption caused by pauses decreases at the expense of the number of breathing pauses. I cannot say whether the secondary pause is only a phonetic disruption due to the submoriac adjustment of final lengthening or a breathing pause as well. For that reason I will not make any distinction between the secondary pause as phonetic disruption and as audible pause: in both cases I will treat the phonetic disruption as submoraic adjustment of the word-final syllable, not as a measurable “pause” (or “silence”) as a separate audible phenomenon. [73]

Phonetic word end in a light syllable is a necessary condition for secondary pause. As light syllables are restricted to the foot’s arsis, so is the phonetic adjustment that is secondary pause. A metrical position often used for secondary pause is the feminine third foot caesura (5½):

ῥώοντ᾿ ἄσβεστος δὲ :^5½ βοὴ γένετ᾿ ἠῶθι πρό

Iliad 11.50

They swarmed; endless shouting arose before dusk

Realisation of the prepausal segment as word-final syllable means that the shortened word-final syllable (^ss) is excluded, just like the elided syllable (^es), and the resyllabified segment (^rs):

καί οἱ (^ss) ἐγὼ τόδ᾿ (^es) ἄλεισον (^rs) ἐμὸν περικαλλὲς (^rs) ὀπάσσω

Odyssey 8.430

I will even give him this cup of mine, most beautiful

Mapping of a syllable on the arsis, as part of a double-light or as a single heavy syllable, only facilitates a secondary pause. The disruption may be audible, but with only minimal word-final lengthening (and not enough additional lengthening to take a breath) at normal rate of speech. Elision frustrates a rhythmical pause on its own right branch; elision of an enclitic, with its prosodic ties to both the preceding and the subsequent constituent, frustrates a rhythmical pause altogether.

It will be my working hypothesis that as primary and secondary pause make phrases audible in the continuous flow of dactyls of the Iliad and the Odyssey, they will suggest patterns that remind us of the hierarchical constituency of spoken language: [74] the primary pauses recall the boundaries of the major phonological phrase, the secondary pauses that of the minor phonological phrase. In the phonetic analysis of spoken language, the major phrase, or utterance, encompasses the larger scale syntactical unit, like the “clause” or the “sentence”. The minor phrase consists of smaller scale units like constituents and word groups. The minor phrase also includes the extra-clausal constituents. Syntactically, the minor phrase occupies a hierarchical position within the major phrase. Attractive as this resemblance between prosodic and syntactical units may seem, there are still quite a few problems to solve when applying something like “phonological phrasing” to the way the Iliad and the Odyssey are presented to a listening audience. The value of phonological boundaries as “audible punctuation” needs to be accounted for—as does the demarcative value of the verse end. Issues concerning Homeric composition must be compared to the comprehensibility of what reaches the ear of Homer’s audience. What demands are there on syntactical phrasing if an audience of listeners is the performer’s main target group? And is the larger scale syntactical unit in Homer shaped as a major phonological phrase? Or will audible punctuation bring to light other syntactical patterns? I will first deal with the distribution of primary and secondary pause as evidenced by phonology.

4.6 Audible Punctuation in Homer

The location of word end in the verse is subject to a rule: the end of a phonetic word / prosodic unit is granted on the condition that disturbances to the flow of dactyls are avoided. This rule underlies both the localisation of the limited number of “metrical word-types”, [75] and the various metrical bridges. Phonetic word end is regularly put on a par with metrically demarcative positions: [76] 1st foot arsis (position 2), 2nd foot thesis (position 3: trithemimeral word end), 3rd foot thesis (position 5: penthemimeral word end), 3rd foot trochee (position 5½: trochaic caesura), 4th foot thesis (position 7: hephthemimeral word end), 4th foot arsis (position 8: bucolic dieresis), [77] 5th foot thesis (position 9: ennehemimeral word end), and 6th foot arsis (position 12: verse end). Some of these positions (3, 5, 7, 9) immediately follow the irresolvable [78] long element on the thesis; others, the diereses (positions 2, 8, 12), occur in the transition from the arsis to the thesis, a position whose duration is most severely restricted by the maintenance of dactylic rhythm. Within the arsis (when realised as a double short) there may be word end following the first light segment. Phonetic word end is not allowed after the fourth trochee (position 7½; Hermann’s bridge). At every position of frequent word end, an audible pause may occur. The realisation of audible pause, however, does not depend on the frequency of the metrical boundary. Not even in case of the verse end.

The (now) spondaic ῥυθμός could have been placed elsewhere, with word end mapped on the thesis. The same applies to the spondaic ῥυθμός followed by hiatus:

κούρη Ἰκαρίοιο περίφρων Πηνελόπεια

Odyssey 20.388

Daughter of Icarius, clever Penelope

νωμᾶι ἔνθα καὶ ἔνθα κακῶν ἔμπαιος ἀλήτης.

Odyssey 21.400

He checks it from every angle, master of mischief, beggar

Spondaic word end in Odyssey 16.350, Odyssey 20.388, and Odyssey 21.400 is a deliberate avoidance of a μακρὰ τελεία, the “perfect” heavy element on the thesis that indicates a strong audible pause. In other feet of the hexameter, especially in the fourth, the same may occur: the deliberate creation of spondaic word end as the result of “unnecessary” localisation of a rhythmical word type that may as well be localised so as to avoid spondaic word end. It is possible, after all, to avoid it. Examples from the Iliad’s first book are:

κλῦθί μευ ἀργυρότοξ᾿ ὃς Χρύσην ἀμφιβέβηκας

Iliad 1.37 = Iliad 1.451

Hear me please, Silverbow, who stand protectively over Chryse

ἔκλαγξαν δ᾿ ἄρ᾿ ὀιστοὶ ἐπ᾿ ὤμων χωομένοιο

Iliad 1.46

The arrows rattled on the shoulders of the enraged

οἴωι φαινομένη τῶν δ᾿ ἄλλων οὔ τις ὁρᾶτο

Iliad 1.198

Appearing to him alone; of the others no one saw her

οὔτέ ποτ᾿ ἐς πόλεμον ἅμα λαῶι θωρηχθῆναι,

Iliad 1.226

And not ever to arm yourself for battle together with the soldiers

ἀλλ᾿ ὅδ᾿ ἀνὴρ ἐθέλει περὶ πάντων ἔμμεναι ἄλλων,

Iliad 1.287

But this man here wanted to be high above all others

καί σφωιν δὸς ἄγειν τὼ δ᾿ αὐτω μάρτυροι ἔστων,

Iliad 1.338

And please give her to them to take; let the two of them be wittnesses themselves

δίπτυχα ποιήσαντες ἐπ᾿ αὐτῶν ὠμοθέτησαν,

Iliad 1.461

Having made it into a fold they put the pieces of raw meat on top of it

νεικείηισι πατήρ σὺν δ᾿ ἡμῖν δαῖτα ταράξηι,

Iliad 1.579

Lest father resorts to scolding and disturbs the meal for us with it

ῥῖψε ποδὸς τεταγὼν ἀπὸ βηλοῦ θεσπεσίοιο [88]

Iliad 1.591

He grabbed me by the leg and threw me from the divine threshold

The localisation of bacchic word shapes, or words ending in bacchic meter (⏑ | – – |) [89] similarly leads to word end in a μακρὰ ἄλογος, mapped on the foot’s arsis. Under Wernicke’s law such word end may be possible before the bucolic dieresis, a rather frequently occuring localisation pattern. Consider the following instances in the Iliad’s first book:

Iliad 1.2         οὐλομένην ἣ μυρί᾿ Ἀχαιοῖς ἄλγε᾿ ἔθηκε
Iliad 1.44       βῆ δὲ κατ᾿ Οὐλύμποιο καρήνων χωόμενος κῆρ
Iliad 1.71       καὶ νήεσσ᾿ ἡγήσατ᾿ Ἀχαιῶν Ἴλιον εἴσω
Iliad 1.89       σοὶ κοίληις παρὰ νηυσὶ βαρείας χεῖρας ἐποίσει
Iliad 1.91        ὃς νῦν πολλὸν ἄριστος Ἀχαιῶν εὔχεται εἶναι
Iliad 1.121     τὸν δ᾿ ἠμείβετ᾿ ἔπειτα ποδάρκης δῖος Ἀχιλλεύς
Iliad 1.159     τιμὴν ἀρνύμενοι Μενελάωι σοί τε κυνῶπα
Iliad 1.244     χωόμενος ὅ τ᾿ ἄριστον Ἀχαιῶν οὐδὲν ἔτισας
Iliad 1.278     ἀντιβίην ἐπεὶ οὔ ποθ᾿ ὁμοίης ἔμμορε τιμῆς
Iliad 1.371     ἦλθε θοὰς ἐπὶ νῆας Ἀχαιῶν χαλκοχιτώνων
Iliad 1.384     πάντηι ἀνὰ στρατὸν εὐρὺν Ἀχαιῶν ἄμμι δὲ μάντις
Iliad 1.412     ἣν ἄτην ὅ τ᾿ ἄριστον Ἀχαιῶν οὐδὲν ἔτισε
Iliad 1.430     τήν ῥα βίηι ἀέκοντος ἀπηύρων αὐτὰρ Ὀδυσσεύς
Iliad 1.551 (cf. 568) τὸν δ᾿ ἠμείβετ᾿ ἔπειτα βοῶπις πότνια Ἥρη
Iliad 1.567     ἆσσον ἰόνθ᾿ ὅτε κέν τοι ἀάπτους χεῖρας ἐφείω
Iliad 1.571     τοῖσιν δ᾿ Ἥφαιστος κλυτοτέχνης ἦρχ᾿ ἀγορεύειν

Of these examples, one is not in accordance with the special feature that is the most important condition for word end under Wernicke’s law: the length of the vocalic nucleus. Generally, the heavy segment mapped on the hexameter’s fourth arsis contains a long vowel or diphthong so that the word-final syllables take bacchic shape before the word juncture. The formula βοῶπις πότνια Ἥρη (14 times in the Iliad, never in the Odyssey) contains a spondaic word end due to a consonantal coda (“position”) instead of the natural length of vowels. That the only exception is found in a noun-epithet formula bears some significance: I think it points at the possibility that spondaic word end under the conditions of Wernicke’s law is itself a licence with regard to an older allowance concerning spondaic word end due “to position”. Of course, one formulaic example is insufficient proof.

It is more significant that hardly any of the examples, both of rhythmical-spondaic word types and (final-)bacchic word-shapes coinciding with metrical-spondaic word end, is phonologically indicative of a strong audible boundary at its completion. Overwhelmingly most of the examples feature a word-final segment that is subjected to sandhi. Devine and Stephens suggest that much when they point out that mapping on the arsis is much more likely for a segment ending in a long vocalic nucleus (possibly as the result of coda-to-onset shift) than for a segment that is heavy due to position. [90] This strikes me as peculiar as spondaic word end under Wernicke’s law is interpreted as an allowance under specific conditions: the rhythmical closure of the verse-final spondaic word end with LFS is then being duplicated at another, verse-internal position. The resulting disturbance of hexametric rhythm is taken for granted. [91] Any such disturbance, however, would originate from a μακρὰ τελεία on the arsis, possible under the conditions of Wernicke’s law. The law would allow for word end that disrupts dactylic rhythm due to non-observance of submoraic adjustment. The underlying premise is then that the metrical boundary following position 8 inevitably doubles as a phonetic disruption. I do not think it does.

Wernicke’s law does not automatically result in a μακρὰ τελεία at the metrical boundary. As Parry already observes, [92] “overlong syllables” are gradually being avoided towards verse end; following Devine and Stephens, we may take this observation as an indication of submoraic adjustment to maintain podic rhythm despite final lengthening. In the (final-)bacchic examples from the Iliad’s first book, few of them allow the metrical boundary at position 8 to double as phonetic disruption. Most of the instances of spondaic word end at position 8 are subject to sandhi in their segmentation. The word-final element, in other words, is realised primarily as a segment rather than a syllable. Metrically, the segment is often heavy instead of superheavy (as a syllable); phonologically, the element is nonprepausal instead of prepausal. [93]

Of the (final-)bacchic examples from the Iliad’s first book, three (44, 121, 371) contain a superheavy syllable in the fourth foot arsis: sandhi cannot turn the syllable-final consonant into a segment-initial onset. Of these three lines, two (121, 371) allow the use of a superheavy syllable in order to apply a formula (ποδάρκης δῖος Ἀχιλλεύς, Ἀχαιῶν χαλκοχιτώνων). My conclusion tends to be the same as the one concerning βοῶπις πότνια Ἥρη in Iliad 1.551: the more restrictive allowance of spondaic word end under Wernicke’s law postdates the localisation of certain formulas. It is noteworthy that, with few exceptions, the use of these formulas is restricted to the Iliad.

The remaining superheavy syllable at position 8, καρήνων in Iliad 1.44, results in a marked διάβασις, the audible phonetic closure of the phrasal domain. The complete realisation of LFS before word end makes word-final -νων resemble the monosyllabic foot (T[A]/[A]T), a thesis that has been orphaned from its arsis in the preceding or in the following word. The final syllable lengthening of -νων exaggerates the separation from the following thesis in the fifth foot. At the same time, it makes the preceding thesis within the fourth foot look like a syncopated thesis. The lengthened “thesis” on position 8 looks like the second in a series of three theses: T(A)⁷(A)T⁸T⁹A¹⁰. The marked διάβασις suggests an audible pause at position 8: in this case, a pause separating two ῥήματα, two finite verb forms. [94] Any change of “direction” in the rhythm indicating metarrhythmisis to anapaests here, is no longer metarrhythmisis on the level of the rhyhtmical word type; it is metarrhythmisis on the level of the phrase, as termination is not merely in an isolated thesis, but in catalexis. [95]

Finally, the first book presents its audience with at least two molossus-shaped words [96] ending at position 8: ἀλλ᾿ ἄγε διογενὲς Πατρόκλεις ἔξαγε κούρην in Iliad 1.337, and οἶσθα τί ἦ τοι ταῦτ᾿ εἰδυίηι πάντ᾿ ἀγορεύω in Iliad 1.365. Both molossus-shaped words can be used in other positions within the hexameter. The phonetic effect of their localisation is the same as for spondaic and (final-)bacchic word shapes ending at position 8: their word-final μακρὰ ἄλογος renders, at best, a mild pause.

What spondaic word end allowed under Wernicke’s law and various instances of hiatus, both in the thesis and in the arsis, have in common is their frequent word end in a μακρὰ ἄλογος. The unmarked διάβασις and διάστασις resemble the non-disruptive duration and synaphy of the non-final μακρὰ ἄλογος. With only few exceptions, submoraic adjustment maintains the podic structure of dactylic rhythm. Synchronically, Wernicke’s law appears to be an allowance for word end in order to make a choice concerning verse-end-resembling rhythmical patterns within the hexameter. Diachronically, however, the non-disruptive word end in a μακρὰ ἄλογος in the hexameter’s arsis developed together with the συλλαβὴ ἄλογος in the arsis of the sixth foot. It is therefore tempting to assume the possibility of non-disruptive διάβασις at verse end for at least a number of the Homeric hexameters.

4.6.4 Options at verse end

The γένος ἴσον restriction on metarrhythmisis (“dactyls cannot be metarrhythmisised to trochees or spondees”) has consequences for the verse end as well. Word end at the hexameter’s verse end can only be prepausal if the verse-final element is a marked heavy syllable due to metarrhythmisis to anapaests. As in the case of spondaic word end at position 8, metarrhythmisis at position 12 means metarrhythmisis on the level of the phrase instead of on the level of the rhythmical word. For the verse-final element to be marked as a heavy syllable, the verse-final metrical colon needs to end in catalexis (–´–´ ||) so that the metrical colon starting from position 5 is actually a catalectic anapaestic dimeter. Starting from another position, metarrhythmisis to anapaests is more gradual, within a metrical phrase. The following two examples feature both types of phrasal metarrhythmisis to anapaests. In both examples the metrical colon from position 5 is an anapaestic dimeter, whereas the metarrhythmisis is more gradual in the first metrical colon of the line, as it is on the level of the rhythmical word:

ὧς ἔφατ᾿ εὐχομένη : ἀνένευε δὲ Παλλὰς Ἀθήνη
[T A ][ T A][T]:[A T][A T][A T][T]

Iliad 6.311

Thus she spoke praying, but Pallas Athena shook her head in refusal

τὴν δ᾿ ἀπαμειβόμενος : προσέφη πολύμητις Ὀδυσσεύς
[T] [ A T ][A T ]:[ A T][ A T][A T][T]

Odyssey 19.499

Answering her clever Odysseus spoke

A catalectic verse-final dactyl, as if the verse would end in a trochee, [97] is as improbable as phonetic disruption at position 5½:

χάλκεον αὐτὰρ ἔπειτα (: ?) σάκος μέγα τε στιβαρόν τε (|| ?)

Iliad 16.136

Of bronze, and then the big and solid shield

καί τέ με νεικείεσκον (: ?) ἐγὼ δ᾿ οὐκ αἴτιός εἰμι (|| ?)

Iliad 19.86

And they keep casting it in my teeth, but I am not responsible

Prepausal mapping of a heavy syllable on the fourth arsis under Wernicke’s law equally requires phrasal metarrhythmisis to anapaests.

Rhythmically, the verse end marks the transition to descending, dactylic rhythm. The verse end as a transition is then clearly distinct from word end at position 5: the latter facilitates the transition from ascending to ascending rhythm; the former from ascending to descending—that is, if the verse-final syllable is a marked heavy syllable on the thesis. There is, of course, the possibility that the verse-final syllable is not marked, and that the verse end is not the closure of catalectic anapaestic rhythm but the spondaic closure, due to metrical indifference on the sixth arsis, of dactylic rhythm. In that case, the dieresis at verse end separates descending rhythm from descending rhythm, as the dieresis would within the line. [98] Mild “demarcative strength” is not unusual at the dieresis, for example at position 8: [99]

μή μ᾿ ἐθέλοντ᾿ ἰέναι κατερύκανε :⁸ μηδέ μοι αὐτή
ὄρνις ἐν μεγάροισι κακὸς πέλε᾿ οὐδέ με πείσεις

Iliad 24.218–219

Do not try to stay me who am minded to go, and do not for me yourself | prove a bird of ill omen in my halls; you will not convince me

ἥμισυ τῶι ἐνάρων ἀποδάσσομαι :⁸ ἥμισυ δ᾿ αὐτός
ἕξω ἐγώ τὸ δέ οἱ κλέος ἔσσεται :⁸ ὅσσον ἐμοί περ

Iliad 17.231–232

The half of the booty I will share with him, the half I myself | will keep; the glory for him will be the same as for me at least

Dactylic prepausal word end does allow for LFS without serious disruption to the flow of dactyls, if the additional final lengthening would be minimal. Qualification as secondary pause gives rise, however, to an important consequence for the analysis of the verse-final foot. In order to mark a primary pause, the verse-final metrical colon (whatever the position it starts from) should undergo phrasal metarrhythmisis to anapaests. Brevis in longo, applied, as it commonly is, to indicate metrical indifference at verse end, [100] will not do to grant the verse end demarcative value as a primary pause. As metrical anceps element, the verse-final arsis may be indicative of a strong metrical boundary, [101] but as it offers choice in realisation, it also offers choice when it comes to its demarcative value. Without metarrhythmisis, its realisation is not so much prosodically neutral, but reminiscent of what Devine and Stephens call the “subordinated” segment: a segment whose rhythmical prominence is kept in accordance with its position within the proportion of the metrical foot. Devine and Stephens [102] use the terminology “subordination” to describe adjustments to heavy syllables to retain the unidirectional structure for stress feet in a phrase. Contiguous stresses, suggesting a sequence of theses, are replaced by a thesis-arsis stress foot structure by defooting a stress foot that is preceded and followed by a thesis. The defooted heavy syllable is consequently refooted and mapped onto arsis ([T][T][T] → [TA][T] or [T][AT]). Devine and Stephens note that

‘implicit in the hypothesis of subordination is the assumption that, when a heavy syllable is mapped onto arsis, it is pronounced with less duration than a heavy syllable mapped onto thesis.’

Having accepted that the durational ratio of arsis and thesis resembles A ≤ T, I find their terminology “subordination” very useful, outside the reconstruction of “stress feet”, to describe the restriction on prepausal mapping of heavy segments onto arsis to retain the unidirectional podic structure. [103] Like the heavy segment on position 8, the verse-final anceps element offers the choice comparable to the heavy κυκλικός-element in both the dactyl and the anapaest. [104] Like phrasal metarrhythmisis, realisation of the verse-final element as a subordinated syllable is an example of phonetic adjustment: this time frustrating the realisation of a prepausal syllable. As non-prepausal, the verse-final metrical anceps requires no more than maintenance of the dactylic γένος ἴσον foot with its foot-internal proportion and realisation of the verse-final element as a metrically indifferent arsis. The thesis, as the first element of the sixth foot, must be the only marked element: only then can be avoided that the verse-final element is mapped as a prepausal syllable. Evidence pointing at the correctness of this observation can be found in the general avoidance of lexical monosyllables on the verse-final element of the hexameter. As the sixth foot anceps allows for a choice among options, there is no reason to assume, as Ruijgh does, [105] that the sixth foot ends in a μακρὰ τελεία. Attractive as his assumption may seem from the metrical point of view, phrasal metarrhythmisis (to anapaests) cannot automatically be taken for granted. Within the line, phrasal metarrhythmisis is evidenced by “spondaic word” end in a superheavy syllable, that is, by an “overlong” segment without the possibility of synaphy or ambisyllabism, with the exception of [long vowel + ς]. The parallelism between the allowance for such word end under Wernicke’s law and the law’s raison d’être, the verse end, proves that caution is equally required when considering the possible realisation of position 12 as a μακρὰ τελεία. It may very well be possible (phrasal metarrhythmisis to anapaests; orphaned verse-final thesis [T]), but the metrical and phonological considerations presented so far are not sufficient to consider the verse-final syllable as other than a συλλαβὴ ἄλογος.

As heavy anceps, the arsis of the sixth foot must have mora count of at least two: the mora count of the verse-final anceps is more or less equal to that of the non-final heavy segment. The metrical indifference of the final syllable, including the light syllable, of the verse may result in a verse-final foot that might be labelled σπονδαιοείδης ‘resembling a spondee’. It looks like the verse-final foot inevitably resembles, or even takes, the shape of a spondee. Dionysius of Halicarnassus does not describe the verse-final foot as a spondee: if he did, it would imply that a spondaic word type is regularly found in the sixth foot, and that word end in the fifth foot would prepare for a spondaic ending of the line. At first sight, this happens in the Homeric text, but only seldom in the Iliad [106] (type: Iliad 18.379 κόπτε δὲ δεσμούς || ‘and he was hammering the fastenings |’). Dionysius’s example of the spondaic ῥυθμός (ποίαν δῆθ᾿ ὁρμάσω ταύταν || ἢ κείναν ἢ ταύταν) easily allows spondaic word type. His example may, however, be a holospondaic realisation of anapaestic verse. [107]

The automatism that presents position 12 as a metrically “long” element is based on the indifference that characterises metrical finality. A verse-final foot scanned |– –|| is spondaic because it features a monosyllabic arsis: the final syllable does not represent double light. [108] As the monosyllabic arsis is in no way the alternative for a resolvable element, its realisation would not allow for a primary pause (= ||). If position 12 would be considered a metrically “long” element due to brevis in longo, the brevis in longo itself, due to the identification of verse end as a pause, [109] invariably marks the transition from pendant to descendant rhythm. Its potential as an indicator of pause should therefore be compared to that of the verse-internal dieresis, or to the trochaic caesura. Both are secondary pauses. As the mark of pause, any LFS of the verse-final element is only about half the LFS of the μακρὰ τελεία. Before the trochaic caesura, the word-final syllable [short vowel + consonant] shows that the phonological constrains are stronger than the phonetic disruption. In such cases word-final -ν must be considered as subject to resyllabification:

μάψ ἀτὰρ οὐ κατὰ κόσμον ἐμοι δ᾿ ἄχος οἱ δὲ ἕκηλοι

Iliad 5.759

Recklessly and without restriction; for me it means sorow, but the others calmly

It functions as a means of segmental synaphy; not as a coda (resulting in a prepausal heavy syllable), but as a mere onset (expanding the phonological phrase forward to ἐμοι-). The word-final light syllable preceding the secondary pause may be additionally lengthened into a βραχεῖα ἄλογος, but prepausal word-final lengthening cannot turn it into a συλλαβὴ μακρά. The appearance of movable nu at the trochaic caesura points in the same direction:

τοῖσι δὲ βοῦν ἱέρευσεν ἄναξ ἀνδρῶν Ἀγαμέμνων

Iliad 7.314

And for them the lord of men Agamemnon immolated a bull

Movable nu is metrically harmless, phonologically useful, and phonetically non-disturbing. In epigraphical evidence, [110] the trochaic caesura figures as a secondary pause despite punctuation:

ἁλ(λ)όμενος νίκησεν : Ἐπαίνετος οὕνεκα τοῦδε: (from Eleusis)
ϟου[φαγόρας μ᾿ ἀνέθη]κεν : Διὸς γλαυϟώπιδι ϟούρηι (from Athens)

At word end on the thesis, brevis in longo may occur, and the licence is extended to word end on a thesis that is not prepausal. Brevis in longo does not appear in the arsis preceding verse-internal dieresis. If I combine the observations concerning syllable duration and weight before various positions of word end in, or following, the arsis, I can reasonably postulate that brevis in longo is restricted to the completion of blunt rhythm, and hence not, I argue, a feature of the rhythm at verse end, unless metarrhythmisised on the level of the phrase. Only when metarrhythmisised may a verse-final heavy syllable function as a metrical thesis and be equal to the μακρὰ τελεία. Usually, however, it is ἄλογος. Dionysius’s comment on the hexameter-example Odyssey 11.598, however puzzling, at least endorses this conclusion:

αὖτις ἔπειτα πέδονδε κυλίνδετο λᾶας ἀναιδής

Odyssey 11.598

Then, again, the shameless boulder rolled back to the plain

The verse-final long syllable -δής in the example indeed counts as a μακρὰ ἄλογος: Dionysius does not “forget” to count it among the “perfect” long elements. He rightly does not consider it a “perfect” long element, as, in my terms, this verse end qualifies as a secondary pause.

The metrical anceps on the sixth-foot arsis thus offers a choice between three options, three types of segmental realisation. As (I.) a heavy, or rather superheavy segment, the superheavy syllable may equal the “perfect” long. As (II.) a light syllable, the prosodic realisation of the verse-final segment results in a heavy verse-final element (resembling brevis in longo) due to metrical indifference. Both realisations, however, take a strong audible pause at verse end for granted. And both require some form of phrasal metarrhythmisis to allow for that pause. As the verse end cannot automatically be seen as such, I would add that in many instances ‘dimming of the rhythm of the end of the hexameter’ [111] through realisation of an anceps element as (III.) subordinated syllable at position 12 is most likely. [112] Realisation as a subordinated syllable is an example of preparatory prosodic organisation: the metrical repetition, and the maintenance of the metrical “direction” provided by the verse-final foot reduce the verse-final syllable to non-prepausal, or, at most (depending on rates of speech), prepausal to a secondary pause. The resulting “dimming of the rhythm” at verse end creates the possibility of phonological continuation over the verse end.

4.6.5 Repetition

Mention has been made of the resemblance of allowance of spondaic word end under Wernicke’s law and at verse end. On the one hand, under Wernicke’s law there is a possibility for the mapping of a specific type of word end that may otherwise only apply to verse end; on the other hand, word end under Wernicke’s law depends on specific conditions as the verse audibly terminates too early. Both explanations, of the allowance and of its restrictive use, are based on the assumption that similar phonetic disruption is repeated within one single hexameter. Repetition of similar phonetic disruption is not a persistent phenomenon in Homeric poetry though, not even in verses featuring word end allowed under Wernicke’s law. In the sample in the Appendix, 14% of the lines of the Iliad and 13% of the lines of the Odyssey show such repetition. Repetition is significantly clustered in and around (through the introductory formula) direct speech: 37% of the verses featuring phonetic repetition in the Iliad, 69.2% of the verses featuring phonetic repetition in the Odyssey are found in direct speech. The combination of word end under Wernicke’s law, an audible pause at position 8, and a primary rhythmical pause at verse end hardly ever occurs in Homer: in the Iliad’s first book, only 1.44 and 1.371 qualify as examples . [113] Avoidance of repetitive rhythmical patterns can be made audible: verses with word end at the bucolic dieresis, even in a secondary pause, do not feature prepausal postpositives both before the bucolic dieresis and the verse end. Consider, for example, the distribution of postpositives (underlined) in Iliad 4.524–6:

θυμὸν ἀποπνείων ὁ δ᾿ ἐπέδραμεν ὅς ῥ᾿ ἔβαλέν περ
Πείρως οὖτα δὲ δουρὶ παρ᾿ ὀμφαλόν ἐκ δ᾿ ἄρα πᾶσαι
χύντο χαμαὶ χολάδες τὸν δὲ σκότος ὄσσε κάλυψε

Iliad 4.524–526

With his life on his last breath; but he came running, the one who had hit him, | Peirous, and he stabbed him with the spear near his navel; out all | his intestines came pouring on the ground; as for him, darkness covered both his eyes

Several postpositives are elided. Most are on the arsis. Some of them are both elided and on the arsis. Their demarcating effect as phonetic-word final is limited, or in case of elision, absent: due to the elision, they may even no longer be appositive. [114] The only postpositives localised before a possibly stronger pause are περ and δὲ (in Iliad 4.526). Most of the time, the “demarcative character” of the postpositives is nearly or completely frustrated. And even in the case of περ and δὲ (in Iliad 4.526) there are more arguments in favour of additional lengthening than there are to come to a conclusion about a rhythmical pause: both particles do not make a “perfect” heavy syllable.

The sample also shows that there are hardly any verses that combine word-final syllable [short vowel + consonant] before a syntactically meaningful caesura at position 5½, with word end in a heavy syllable elsewhere within the same line, except for verse end, word end under Wernicke’s law (as in Iliad 1.370), and due to linguistic adaptations (as in Odyssey 12.7). In the sample from the Iliad and the Odyssey the combination occurs in 2% of the lines (Iliad 1.9; Iliad 1.99; Odyssey 1.37; Odyssey 1.70).

4.7 Audibly Patterning Homeric Syntax

In chapter 3, the connection between the shape of the hexameter as a metrical unity and that of the grammatical clause as a phonological phrase has turned out to be quite loose. The great variety in ways and positions from which to start a grammatical clause shows that the metrically repeated pattern of the stichic hexameter does not automatically lead to a “stichic” clausal-grammatical pattern. Often, the grammatical clause runs over the verse end into the subsequent verse without a rupture in the clause-internal metrical repetition. The metrical feet serve the continuation within the grammatical clause, not primarily the division of the hexameter into smaller metrical phrases. This metrical continuation within the grammatical clause redirects attention to the pauses that audibly demarcate the phrases in the continuous flow of dactyls. In this section I will focus on the realisation of audible pauses vis-à-vis the grammatical clause. This section aims to provide a fundament for the intuition that grammatisation develops due to audible punctuation.

4.7.1 Audible variation

As shown, the start of the hexameter is itself not a preferred position for clause start, unless the preceding verse end doubles as παραγραφή, the end of a syntactical unity. In fact, the start of a grammatical clause is frequently formed by a light segment (⏑), a tendency that is further strengthened by the avoidance of word end in a spondaic foot. The preference for clause start by means of a light segment results in a majority of ascending clause starts, in other words, clause starts by means of words that are themselves iambic (⏑ –) or anapaestic (⏑ ⏑ –), or whose first two or three segments fit a iamb or an anapaest. Rhythmically, ascending starts are predominant as well. [115] It is therefore not surprising that the start of the hexameter, being descending (beginning with a marked heavy element), is only sometimes the start of a grammatical clause. As the Homeric grammatical clause is commonly preceded by extra-clausal transitional constituents, some additional lengthening at the preceding verse end is required. As a phonological minor phrase, however, the extra-clausal constituent is demarcated by restricted additional lengthening. In an example like Odyssey 5.315, lengthening of the verse-final syllable, as the final syllable is formed by an enclitic, will appear as artificial as the lengthening “by position” of δέ at positions 5 and 7:

τῆλε δ᾿ ἀπὸ σχεδίης αὐτὸς πέσε πηδάλιον δέ (:)

Odyssey 5.315

Far away from the raft he himself fell, and the rudder

Apparently, verse end as rhythmical disturbance is not the ideal point of transition from a minor-phrase pragmatic constituent to the start of the grammatical clause. Not surprisingly, the bucolic dieresis proves equally unfit. If the start of the new grammatical clause is postponed until the start of the new hexameter (for example for formulaic reasons), the verse end rather facilitates the extension of a minor-phrase transitional constituent into αὐτὰρ ἔπειτα or ἔνθα ἔπειτα, despite the prosodic artificiality.

The predominance of ascending clause start appears to be on strained terms with the division of the hexameter by the third foot caesura into ἡμιστίχια. The current consensus, however, grants this division a theoretical background: on the level of verbal meter, the two ἡμιστίχια of the hexameter are said to be deliberately dissimilar. The second hemistich is longer than the first (the so-called “principle of climax” or “Gesetz der wachsenden Glieder”), and the two belong to different “types”, with the first being descending and the second ascending. Subsidiary pauses [116] (positions 2, 3, 7, 8, and 9) are introduced to further divide the ἡμιστίχια. They create dissimilarity between the resulting word groups (= metrical cola) not unlike the effect of the third foot caesura for the hexameter as a whole. Rhetorical improbability, or the impossibility of third foot caesura (approximately 1 in every 150 verses in Homer), turns position 7 into the main caesura of the verse. In such cases, position 3 is generally accepted as a subsidiary pause (the so-called “tripartite hexameter” [117] ) in order to maintain the principle of climax: further up the hexameter the word groups are longer. According to this widely accepted concept of verbal meter, the looked-for rhythmic variation [118] is limited within the hexameter: as a rule, descending and ascending rhythmical ἡμιστίχια alternate regularly. The only variation lies in the completion of the descendant one: before position 5, completion is blunt; before 5½, it is pendant. Only the absence of a third foot caesura can give subsidiary pauses a unique rhythmical value and importance.

In this concept, I argue, rhythmical variation is not just limited, it is practically lacking. This “lack” of variety is caused by four persisting concepts, all of them proven as wrong by my own approach:

I. the verse end is an unavoidable and strong pause;

II. the third foot caesura is almost unavoidably a strong pause;

III. subsidiary pauses are neither unavoidable nor strong;

IV. the stichic repetition of paired ἡμιστίχια is maintained with disregard for the non-stichic series of grammatical clauses and the transitional pragmatic constituents.

This is not a random order. As I showed above, the verse end does not automatically claim “prosodic neutrality” despite the metrical indifference. The verse-final syllable is not unavoidably prepausal-heavy, [119] but spondaic word end at verse end is only under conditions similar to those of Wernicke’s law at position 8. Dimming of the metrical irregularity at verse end is optional despite the unavoidable word end. Word end, and phonetic word end, is nearly always found in the third foot as well. Due to this strong tendency, a main third foot caesura is in general only denied to verses containing a word that bridges at least the entire third foot:

Τηλέπυλον ³ Λαιστρυγονίην ⁷ ὅθι ποιμένα ποιμήν

Odyssey 10.82

Laestrigonian Telepylus, where herdsman (calls to) herdsman

In my opinion, [120] however, it is illogical to deny the possibility of rhythmical “dimming” to verses with elision at the caesura within the third foot:

ὣς ἄρα φωνήσασ᾿ ἡγήσατο Παλλὰς Ἀθήνη

Odyssey 7.37

Having spoken thus, Pallas Athena led the way

The punctuation of many modern texts of the Iliad and the Odyssey bears witness to the strength with which grammatically structuring text-elements suggest an alternative, and much more varying pattern. [121] In my view, position 3 is not a subsidiary pause in Iliad 11.289:

Ζεὺς Κρονίδης :³ ἀλλ᾿ ἰθὺς ἐλαύνετε μώνυχας ἵππους

Iliad 11.289

Zeus, the son of Cronus; come on, drive your hoofed horses straight forward

The line features a third foot caesura, but does not feature a third foot pause. Consider the realisation of, respectively, position 7, 8, and 9 in the following examples:

σχέτλιος ἀλλ᾿ ἀκέσασθε φίλοι :⁷ δύναμις γὰρ ἐν ὑμῖν

Odyssey 10.69

So miserable; come, make us recover, please, friends; for you have that power

κείνου ὅπως δὴ δηρὸν ἀποίχεται :⁸ οὐδέ τι ἴδμεν

Odyssey 4.109

Because of him; remember how long he has been away by now; we do not even have a clue

ἀνέρες ἠρήσαντο παρεστάμεναι :⁹ δύναται γάρ

Odyssey 4.827

Men have begged her to stand by them; for she is able to do it

What these examples (and many more can be added) have in common is that the phonetic realisation of the verse, as does the syntactical structure, suggests a pattern that disregards the prosodic guidance of the two ἡμιστίχια of the hexameter. The force with which phonological boundaries, other than the third foot caesura and verse end, determine the shape and internal organisation of the phrases (at the cost of the importance of the third caesura and verse end) is mirrored by the grammatical structure of the hexameter or parts of the hexameter. The patterning of the grammatical clause and the various transitional and preparatory extra-clausal constituents results in a non-stichic variety of phrases.

4.7.2 Patterning the grammatical clause

[ὣς φάτο] καί [σφιν νῶτα βοὸς παρὰ πίονα θῆκεν
ὄπτ᾿ ἐν χερσὶν ἑλών] τά ῥά οἱ [γέρα πάρθεσαν αὐτῶι]
[οἱ δ᾿ ἐπ᾿ ὀνείαθ᾿ ἑτοῖμα προκείμενα χεῖρας ἴαλλον]
αὐτὰρ ἐπεὶ [πόσιος καὶ ἐδητύος ἐξ ἔρον ἕντο]
δὴ τότε Τηλέμαχος [προσεφώνεε Νέστορος υἱόν
ἄγχι σχὼν κεφαλήν] ἵνα μὴ [πευθοίαθ᾿ οἱ ἄλλοι]

Odyssey 4.65–70

Thus he spoke, and at their side he placed for them fat pieces of a cow’s back that had been | roasted, taking in his hands the pieces that had they had placed next to him as gifts of honour for him; | and they stretched their hands to the prepared dishes lying before them; | but when they had done away with their longing for drink and food, | at that moment Telemachus addressed Nestor’s son, | and he brought his head close to him, lest those others might notice

The descending start of the first grammatical clause of this passage is due to discourse shift (the completion of direct speech: ‘thus he spoke’). Note the straddling of verse end between lines 65–66, and the repeated disregard for the third foot caesura as a strong pause. Thirdly, it is remarkable that the two clauses with blunt ending are both built around a participle as their verbal nucleus (ρῆμα): ἐν χερσὶν ἑλών ‘taking in his hands’ and ἄγχι σχὼν κεφαλήν ‘and he brought his head close to him’. If we count the participle as the nucleus of a separate clause, the wide rhythmical variation in this passage results in only two grammatical “clauses” that correspond with a ἡμιστίχιον on the level of meter: προσεφώνεε Νέστορος υἱόν ‘addressed Nestor’s son’ and ἄγχι σχὼν κεφαλήν ‘and he brought his head close to him’. Only in these two half-verses, 69b and 70a, the grammatical “clause” coincides with the metrical colon, and hence with the widely accepted “rhythmical” colon. In the remaining verses and parts of verses, we find larger and shorter phrases that are invariably started from, and completed at, positions of frequent word end. The phrases formed by grammatical clauses are themselves built from a nuclear predication and satellites that are regularly confined to the metrical configurations available between positions of frequent word end. In other words, the phonological major phrases are built from metrically demarcated minor phrases. Thus, in oral composition, the metrical building blocks of the clause retain their long-existing minor-phrase shape, without necessarily resulting in repetitive audible boundaries. It is exactly this “mismatch between prosodic domain and corresponding metrical domain” [123] that proves the phonetic determination of the grammatical clause—and the great prosodic freedom with which it can be internally organised and externally demarcated.

4.7.3 Clause, colon, phrase

According to these principles, Iliad 1.1–16 can be divided into grammatical clauses onto phonological phrases. The grammatical clauses consist of words and word groups that fit recurring metrical shapes. They are being strung together by means of extra-clausal constituents that also fit recurring metrical shapes. In the presentation below, the grammatical clauses are in brackets. The transitional constituents are underlined. The recurring metrical shapes are separated by means of a comma:

1        [μῆνιν ἄειδε] θεά , [Πηληϊάδεω , Ἀχιλῆος ,
          οὐλομένην] , [ἣ μυρί᾿ , Ἀχαιοῖς , ἄλγε᾿ ἔθηκεν] ,
          [πολλὰς δ᾿ , ἰφθίμους , ψυχὰς , Ἄϊδι , προΐαψεν ,
          ήρώων] , αὐτοὺς δὲ , [έλώρια , τεῦχε κύνεσσιν ,
5        οἰωνοῖσι τε πᾶσι] , [Διὸς δ᾿ , ἐτελείετο βουλή] ,
           ἐξ οὗ δὴ , [τὰ πρῶτα , διαστήτην , ἐρίσαντε] ,
           Ἀτρείδης τε ἄναξ , ἀνδρῶν , καὶ δῖος Ἀχιλλεύς ,
          [τίς τάρ σφωε θεῶν , ἔριδι , ξυνέηκε μάχεσθαι] ,
          [Λητοῦς καὶ Διὸς υἱός] , ὁ γὰρ , [βασιλῆϊ χολωθεὶς , or (] , [)
10      νοῦσον ἀνὰ στρατὸν ὦρσε , κακὴν] , [ὀλέκοντο δὲ] λαοί (or: λαοί]?)
           οὔνεκα , [τὸν Χρύσην , ἠτίμασεν , ἀρητῆρα] ,
           Ἀτρείδης , ὁ γὰρ [ἦλθε , θοὰς , ἐπὶ νῆας Ἀχαιῶν , or (] , [)
          λυσόμενός τε θύγατρα , or (] , [) φέρων τ᾿ , ἀπερείσι᾿ ἄποινα , or (] , [)
          στέμμα τ᾿ ἔχων , ἐν χερσὶν , ἑκηβόλου , Ἀπόλλωνος ,
15      χρυσέωι ἀνὰ σκήπτρωι] , καὶ [λίσσετο , πάντας Ἀχαιούς] ,
           Ἀτρείδα δὲ [μάλιστα , δύω , κοσμήτορε λαῶν] ,

Metrical-phonetic analysis show the predominance of ascendant words and word groups between positions with frequent word end (indicated as ,), as well as the wide variety of phrase-shapes between the strong (indicated as ˚) and mild (indicated as ˙) phonetic disruptions in Iliad 1.1–16:

Annotation to the analysis:

Line 1: The vocative is considered an extra-clausal constituent, but phonologically it is closely tied to the preceding word group. [124] Word end at position 3½ does not fall at a subsidiary pause. Word end at position 5 does. Whilst breaking the grammatical clause in two, the vocative creates a possibility for one convenient primary pause in a longer grammatical clause. Similar pauses are facilitated by, for example, γέρον in Iliad 1.26, and θεά in Odyssey 1.10. [125]

Line 2: The elision at position 5½ dims a third foot caesura.

Line 3: The elision at position 2 dims an audible dieresis after the first foot.

Line 5: The elision makes the clause straddle position 7.

Line 7: As a connector, prepositive καί contributes to prosodic bridging, rather than emphasising (as postpositive δέ often does), a possible audible pause.

Line 9: If the predicative participle is considered as equally fit as a finite verb to form the centre of a grammatical clause, the end of line 9 is both a grammatical and a phonetic boundary.

Line 10: λαοί can be a tail if the audible pause at position 10 is strong enough. Regularly it only is if the verse end is neither a grammatical nor a phonetic boundary.

Line 12: Left-dislocation of ὁ by means of postpositive γάρ requires an unusual pause following position 4.

4.7.4 Patterning extra-clausal constituents

The phonological shape of the extra-clausal constituents requires special attention. Their appearance signals the completion of a grammatical clause; their completion the start of the next grammatical clause. Extra-clausal constituents carefully avoid taking a shape that resembles the major phrase: of the two phonetic demarcations (on the left and on the right) not more than one is realised as a primary pause. In other words: either the constituent may be preceded by a metrical position that allows for considerable additional lengthening, or its final syllable is mapped at such a position. A secondary pause on both the left and the right side is of course possible:

ὧς φάτο ² τὸν δὲ ^3½ ἄνακτα χόλος λάβεν οἷον ἄκουσε

Iliad 6.166

Thus she spoke; as for him, anger seized the king at what a dreadful thing he heard

Iliad 1.13 ὁ γάρ is an example of an extra-clausal constituent starting from a primary and ending in a secondary pause. Alternatively, the sequence of pauses is reversed (Iliad 1.9 ὁ γάρ). When starting from a primary pause, demarcation on the right may be further obscured due to a postpositive ending of the phonetic word at an uncommon metrical position (Iliad 1.13 ὁ γάρ⁴ [dieresis after the second foot]). Vocatives regularly start from an unmarked position, and end in a position that allows considerable additional lengthening. [126] The phonological shape of the extra-clausal constituent is that of the minor phonological phrase, itself part of a larger scale major phrase. Its participation in the syntax of the major phrase may as well be concluded from the frequent rhythmical reorganisation due to the elision of the constituent-final syllable of δέ or ἄρα:

[τὸν δ᾿ εὗρ᾿ ἀμφ᾿ ὤμοισι τιθήμενον ἔντεα καλὰ
νηὶ πάρα πρυμνῆι] [τῶι δ᾿ ἀσπάσιος γένετ᾿ ἐλθῶν

Iliad 10.34–35

And him he found busy putting the beautiful armour around his shoulders | near the ship’s deck; for him his arrival was most welcome

Audible variation is the result of the combination of grammatical clauses and extra-clausal constituents. Together they create a sequence of phrases in which hardly any metrical or phonological shape is being repeated instantaneously. The grammatical clause in the dactylic hexameter surprises the ear of the audience with its scope and size despite the hexametric repetition.

4.7.5 Audible enjambment

Enjambment has been discussed in two different ways in the preceding chapters; it is either the mere acknowledgement of mismatch between metrical and syntactical phrasing, or it is some sort of affective prosody (emphasis) as well. With the possibility of prose-like clause formation on the one hand, and the natural chunking of spoken language on the other, it is theoretically unsound to grant emphasis to words based on their metrical location. Enjambment as affective prosody must stem from the mismatch of grammatical / syntactical phrasing and the phonetic disturbance caused by primary pauses. Dionysius [127] comments on such disturbance: the audible pause (ἀναβολὴ χρόνων ‘delaying of durations’) is a ‘clashing of syllables’ (ἀνακοπὴ συλλαβῶν) like ‘blockades formed by articulate sounds’ (ἀντιστηριγμοὶ γραμμάτων). I think that, if such an ἀντιστηριγμὸς γραμμάτων evokes an audible pause within a grammatically coherent clause, poetically meaningful enjambment may be assumed.

This is not the place to identify all instances of such “real” enjambment in Homer, [128] but maybe two examples may be presented to conclude this chapter on audible punctuation. The first deals with possible enjambment at the verse end:

ἀφείλετο νίκην
ῥηιδίως

Iliad 17.177–178

He took away victory | easily

The second features similar enjambment [129] verse internally:

ὅσσ᾿ ἐμοὶ ἐκ πασέων Κρονίδης Ζεὺς ἄλγε᾿ ἔδωκεν

Iliad 18.431

As many sorrows as to me, out of all, Cronus’s son Zeus has given

Again, I would like to stress that enjambment is possible considering the phonetic realisation, not that it is unavoidable. Nor do I claim that possible enjambment at such positions lends emphasis to either the word preceding, or the word following the phonological realisation. I do not know what exactly the poetic effect of such enjambment might have been, I merely analyse the possibility for an audible pause that disturbs the grammatically coherent unit. There is, nonetheless, one specific aspect that I do want to point out: both at verse end, and verse internally, instances of “true” enjambment possibly disturbing clausal grammar are rare, especially in the Iliad. [130]

4.8 Conclusion

In this chapter, I have discussed some recent theories on visible and audible demarcation in Greek poetry, in order to find proof for the intuition that larger-scale grammatisation in Homer is facilitated by prosody. I have shown that metrical boundaries demarcate syntactical unities only when combined with another type of realisation. This other type of realisation is evidenced in phonetics, the clues taken from submoraic adjustment and word-final lengthening for audible termination. Devine and Stephens deal with phonetic disruption to the continuous flow of metrical feet as an indicator of termination. One of the cross-linguistic phenomena they identify in ancient Greek on empirical evidence is final lengthening. Dealing with dactyls and anapaests in relation to the “perfect” and the “immeasurable” heavy syllables in Greek, Ruijgh shows that word-final lengthening is actually final-syllable lengthening. Combined with observations by Dionysius of Halicarnassus and Ruijgh’s own conviction (equal to an observation from the analysis of Greek speech rhythm by Devine and Stephens) that the foot-internal proportion thesis-arsis is A ≤ T, Ruijgh defends the position that the “perfect” long syllable is the word-final heavy syllable as the result of final-syllable lengthening, the “immeasurable” the non-final heavy syllable.

From these existing theories I started looking for the audible pause in Homer. Whereas I deviate from existing studies in looking for metrical-phonological indicators for pause instead of purely metrical, purely phonological, or semantic indicators, I accept the notion of terminal lengthening and the foot-internal proportion as defended by Ruijgh. Phonologically identified pauses, like the pauses Daitz assumes, do contribute to the identification of phonetic disruption: audible pause can only stem from realisation of the phonological segment as a syllable. Taking final-syllable lengthening as the indication of termination, I also accepted that this has consequences for the possibilities meter offers to pause: the thesis allows for considerably more lengthening than the arsis. Therefore I have identified two types of audible pause depending on the amount of additional lengthening the word-final syllable allows: the primary pause on the thesis or the metarrhythmisised arsis (A  T due to phrasal metarrhythmisis), and the secondary pause on the arsis. The variety in pauses and their location functions as audible punctuation, creating a patchwork of phrases. This patchwork shows remarkably little repetitive patterning since pause is a choice in Homeric prosody. Both within the verse, and at the verse end, the phonetic realisation of syllables reflects the choices made. As an arsis, the verse end occupies a special position: it can be realised as a stronger pause, a milder pause, or as no pause at all. The rhythmical possibilities of the metrical anceps element at verse end are applicable, though only partially, to a few other positions in the line. Audible realisation of the secondary pause depends on rates of speech.

Applying the identification of audible pauses to the grammatical clause in Homer, I have shown that developing grammatisation in the Iliad and the Odyssey profits from the phonological phrasing of grammatical clauses and extra-clausal constituents: as audible punctuation shows, minor phrases are reorganised into major phonological phrases. The metrical phrases from phonological analysis still play a part in the phrasal patchwork, but the different combinations of segmental weight, metrical position and word-internal position lead to a type of phrasing, even a kind of rhythmical repetition, that audibly differs from metrical colometry.

One could assume that sensible coherence is at times determined by prosody. On a larger scale, this holds true, I think—for the performance of Homeric poetry as well as for its composition. In the next chapter, my aim will be to show the reflection of the prosodic process on the sensibility of composition: phonological phrasing as evidenced by audible punctuation also determines the prose-like composition of Homeric discourse.

Footnotes

[ back ] 1. Golston and Riad 2000 show that the conscious disregard of phonetic adjustment results in a lack of rhythmical periodicity for most meters, except the anapaestic dimeter.

[ back ] 2. Hermann’s bridge and the avoidance of spondaic word end in the dactylic hexameter. Outside the dactylic hexameter: Wilamowitz’ bridge and Knox’ bridge in the iambographs; Porson’s bridge after the penthemimeres in the iambic trimeter; Havet and Porson’s bridge for the 1st and 3rd anceps of the trochaic tetrameter. Cf. Snell 1982:11; Ruijgh 1987:325n27.

[ back ] 3. Devine & Stephens 1994:100–101.

[ back ] 4. Devine & Stephens 1994:47–48.

[ back ] 5. Devine & Stephens 1994:50.

[ back ] 6. Lidov 1989 describes it as the alternation between more stable and more changeable positions.

[ back ] 7. Devine & Stephens 1994:215 underline the “dynamic” aspect of stress. They also point at the terminological confusion concerning “stress” in the typological literature: ‘As such, ‘stress’ is not a very suitable term for nonaccentual durational prominence. Nevertheless, there are signs that it is beginning to be used in this latter sense in the typological literature, and should this sense of the term become current, there would be little reason not to speak of ‘stress’ in Greek.’

[ back ] 8. As already done by, among others Allen 1973; Sommerstein 1973.

[ back ] 9. Devine & Stephens 1994:195–204; 214.

[ back ] 10. Devine & Stephens 1994:215–216.

[ back ] 11. Devine & Stephens 1994:215.

[ back ] 12. Devine & Stephens 1994:100–101.

[ back ] 13. Devine & Stephens 1994:49–52.

[ back ] 14. Devine & Stephens 1994:156.

[ back ] 15. Devine & Stephens 1994:79–84.

[ back ] 16. Devine & Stephens 1994:146.

[ back ] 17. Devine & Stephens 1994:147.

[ back ] 18. Devine & Stephens 1994:148.

[ back ] 19. Devine & Stephens 1994:151–152.

[ back ] 20. Devine & Stephens 1994:272.

[ back ] 21. Devine & Stephens 1994:401–402.

[ back ] 22. Devine & Stephens 1994:432–433.

[ back ] 23. Devine & Stephens 1994:129.

[ back ] 24. Devine & Stephens 1994:135.

[ back ] 25. Devine & Stephens 1994:255.

[ back ] 26. In addition to their view that hiatus and elision are mutually exclusive depending on rates of speech, and not automatically an indication for pause, I present the following data from the sample in the Appendix:

– Hiatus on the thesis cannot automatically be seen as phonetically disruptive; in Iliad 1.1–100 and Odyssey 1.1–100 there are in total 12 instances (4x Iliad, 8x Odyssey) of “hiatus” with a glide on the thesis;

– In the same sample hiatus at the trochaic caesura only occurs twice, both instances in the Odyssey (1.46, 1.64). More often, hiatus is being prevented: Iliad 1.82, Odyssey 1.4, 1.7, 1.9, 1.27, 1.34, 1.54, 1.69;

– Long vowels and diphthongs followed by hiatus while mapped on the arsis are relatively rare; within the line, the sample yields one instance for the Odyssey (following position 6), and one for the Iliad (following position 2). Hiatus following the heavy sixth-foot arsis occurs in 14.5% of the verses.

[ back ] 27. Such is the explanation of “immeasurable” proportion of arsis to thesis according to Westphal 1867, after Hermann had pointed out that there was no ancient evidence for the explanation by Boeckh. Devine & Stephens 1994:117 do not include Ruijgh’s explanation.

[ back ] 28. In their reconstruction of the rhythm of Greek speech, Devine & Stephens 1994:124 state as a rule that the final syllables of a word ending in two heavy syllables can be mapped TT of AT, but not TA.

[ back ] 29. Contra Wifstrand 1933:26–34, who considers the double short longer than any single heavy element, especially in dactylic verse.

[ back ] 30. Dionysius of Halicarnassus On Literary Composition 22/106.2 U-R: τὸ δίκαιον ὑπεραίρουσαι μέτρον. Devine & Stephens 1994:76–79 prefer the terminology “overlong segment” which is theoretically more precise, as the segment is still subject to phonological sandhi. If realised as prepausal, Devine & Stephens speak of the “superheavy syllable”.

[ back ] 31. Contra Wefelmeier 1994 who follows Herodian II,709.

[ back ] 32. Dionysius of Halicarnassus On Literary Composition 20/92.2–93.19 U-R.

[ back ] 33. Ruijgh 1987:324, 352.

[ back ] 34. Ruijgh 1987:325 n25.

[ back ] 35. Devine and Stephens 1984:4–5,11 state that the non-lexical character of words like αὖτις makes their phonological boundary rather weak: especially in the first half of the verse, the final syllable undergoes synaphy.

[ back ] 36. Ruijgh 1989:310–311.

[ back ] 37. Ruijgh 1989:313 ‘c’était donc le diérèse entre deux dipodies successives qui permettait de synchroniser le rythme du débit avec celui de la marche: en manipulant la durée supplémentaire de la syllabe longue finale de chaque dipodie, on pouvait égaliser les durées des dipodies’. Cf. Korzeniewski 1968:88.

[ back ] 38. Ruijgh 1989:314 claims that, at the same time, the analysis κατὰ διποδίαν allows for non-repetitive realisations of the various arses and theses, under the influence of LFS, within the διποδία. He demonstrates this by referring to the higher frequency of spondees in march-anapaests compared to dactylic verse.

[ back ] 39. Ruijgh 1989:320; Korzeniewski 1968:43; Goodell 1901:181; Rossi 1963:45, 90.

[ back ] 40. Ruijgh 1987:324–326.

[ back ] 41. Ruijgh 1987:325.

[ back ] 42. ‘interweaving the observations concerning rhythmical word types with those on meter’, Aristides Quintilianus (p. 38.15 in the edition of Winnington-Ingram). Cf. Goodell 1901:9; Korzeniewski 1968:42.

[ back ] 43. There is, of course, always the possibility that the word-final syllable is not τελεία but ἄλογος due to κυκλικός-rhythm (that is: non-prepausal realisation). Still, the anapaestic verse rhythm shows that, in order to be τελεία, the word-final συλλαβὴ μακρά has to be mapped on the thesis. I argue that that is the reason why Dionysius claims that Odyssey 11.598 contains 7 heavy syllables, but that they are not τέλειοι. It is not because he excludes the verse end from his analysis (20,21 πλὴν ἐπὶ τῆς τελευτῆς) Dionysius, in fact, counts 7 heavy syllables (20,18), that is, including verse-final -δής. Verse-final -δής is heavy, but it is a μακρὰ ἄλογος since it is on the arsis of the final foot.

[ back ] 44. In Dionysius of Halicarnassus On Literary Composition 16/61.20–68.6 U-R, Dionysius describes the ἀναβολὴ χρόνων as a ἀνακοπὴ συλλαβῶν or ἀντιστηριγμοὶ γραμμάτων.

[ back ] 45. See Dee 2004 for the frequency of the remaining 31 hexameter patterns in Homer.

[ back ] 46. Koster 1953:69–70.

[ back ] 47. Known as the ‘Gesetz der wachsenden Glieder’.

[ back ] 48. Van Raalte 1986:18.

[ back ] 49. They differ in their concept of third-foot caesura though: where Sicking 1993:54 extends caesura to “rhythmically relevant word end”, West 1982 considered caesura “regular word end”.

[ back ] 50. The notation most used is that of O’Neill 1942; I use it throughout this study. Alternative notations, not always that different from O’Neill’s, are presented in Porter 1951, Beekes 1972, Kirk 1966, Sicking/Van Raalte 1993. Hagel 1994–1995 presents a notation based on mora count (¹[longum]2[breve]3[breve]4 ²[longum]6[breve]7[breve]8 etc.). Janse 1998:138 and 2003 offers a notation ¹[longum]1a[breve]1b[breve]1c ²[longum]2a[breve]2b[breve]2c etc.

[ back ] 51. Various scholia on the Iliad mention issues concerning punctuation (Erbse 1983:482; 1988:135ff.), but their main concern is the syntactical phrasing. Devine & Stephens 1994:420–422 show that any “prosodic element” stems from a reconstruction of a hierarchy of pause durations based on Nicanor’s system of punctuation in descending rank order relying on mora count and only applicable in the purely metrical approach.

[ back ] 52. Like others before him, Daitz refers to the Danae-passage in On Literary Composition 26/140.18–142.13 U-R, and to roman authors contradicting his quotations from Cicero (Daitz 1991:159n18): ‘W.S. Allen has suggested in private correspondence and also implicitly in his book, Accent and Rhythm (Cambridge 1973) 335–42, that Latin poetry, and particularly the hexameter, was read aloud as prose. This would accord with Quintilian’s suggestions for reciting Vergil with pauses appropriate to prose, but would be at variance with Cicero’s observation.’ I find the recent approach to these various issues in relation to Simonides in Lidov 2010:34–38 very stimulating.

[ back ] 53. Devine & Stephens 1994:24–31.

[ back ] 54. Henderson 1973.

[ back ] 55. Devine & Stephens 1994:303–375, and see chapter 2.

[ back ] 56. Or (τοι  τεκμαίρομ᾿).

[ back ] 57. Devine & Stephens 1994:243–248.

[ back ] 58. Elision looks like a coda to onset shift, but is in fact reduction of a syllable to an onset (see 3.1.1). Elision creates a segment-initial consonant; Devine & Stephens 1994:256–266. West 1973:226–229 considers elision as disappearance of the vowel without a trace. He points at light syllables as result of elision where crasis results in heavy syllables. In his 1982 publication, however, he notices that correption is almost always followed by a short word-initial vowel (1982:11). In my view, these observations must lead to the conclusion that ἔκθλιψις only affects the syllable-initial realisation of consonants (‘l’attaque’, cf. Ruijgh 1987:351n87; Devine & Stephens 1994:255–256) and not that of syllable-initial vowels, without any effect on the mora count of the affected syllable. Elision itself is always the disappearance of a single mora in mora count. It is also important to note that elision makes it possible for a penultimate syllable ending in a short vowel to maintain its phonetic realisation: type Iliad 1.1 μυρί᾿ Ἀχαιοῖς ἄλγε᾿ ἔθηκεν.

[ back ] 59. The shift is by means of a glide: μο.ιε.ρίσ.σει.ε (Devine & Stephens 1994:255–256).

[ back ] 60. Cf. the reasoning in West 1998:xxv–xxvi.

[ back ] 61. If, on the other hand, hiatus were to indicate the demarcation of some sort of coherence, one “expects” hiatus, on syntactical grounds, in a verse like Iliad 15.402: σπεύσομαι εἰς Ἀχιλῆα ἵν᾿ ὀτρύνω πολεμίζειν ‘I will hurry to Achilles, that I may urge him to join the battle’.

[ back ] 62. With the exception of the verse end, movable nu is used on the heavy arsis once in the first 100 lines of the Iliad (Iliad 1.66) versus five times on the thesis. For the first 100 lines of the Odyssey the ratio is 1 (Odyssey 1.71) :1.

[ back ] 63. See 3.4.4.2 and the Appendix for Iliad 1.1–100 and Odyssey 1.1–100 as a sample. Seemingly deviating examples feature a word-final spirans (Iliad 1.121; Iliad 1.89) in a formulaic expression. In Iliad 1.44 and Iliad 1.46, the word-final coda before the bucolic dieresis separates verb forms. In the Odyssey, however, things do not seem to work like that anymore: syntactical coherence is no longer supported by prosodic coherence over the bucolic dieresis. This observation should make us cautious: the way segmentation suggest metrical coherence may not be the only aspect of prosody in which the Odyssey differs from the Iliad.

[ back ] 64. Janko 1992:96–97 points at the remarkable modification of formulaic ἡνίοχος θεράπων, which is otherwise found in accusative case at verse end. The modification is interesting as it allows the formulaic expression to occupy the final position in a syntactically coherent phrase that has phonologically straddled the verse end.

[ back ] 65. Parker 1966 explains the avoidance of spondaic word end in the hexameter end as an extension of Porson’s law.

[ back ] 66. Porter 1951:20; Ruijgh 1987:335–339; Wifstrand 1933:26–34; Irigoin 1965:224–231; West 1982:20, 36.

[ back ] 67. Cf. Ruijgh 1987:346 ‘syllabe longue finale, dont la durée supplémentaire était probablement le double de celle d’une syllable brève finale’.

[ back ] 68. Cf. Devine & Stephens 1994:148,272.

[ back ] 69. Devine & Stephens 1994:70–84.

[ back ] 70. Dionysius of Halicarnassus On Literary Composition 15/57.9–61.18 U-R; cf. Devine & Stephens 1994:50–51.

[ back ] 71. My suggestion is that the frequency of metrical lengthening in the thesis (especially the thesis of the third foot) points at a licence due to rhythmical lengthening (submoraic adjustment) in prepausal mapping.

[ back ] 72. Cf. the remarks on the metrical lengthening in Russo, Fernandez-Galiano & Heubeck 1992:79.

[ back ] 73. Devine & Stephens 1994:421 mention Nicanor’s (reconstructed) punctuation system as it has been used for the measurement of pause in rhythmic units, but they point out that the reconstruction only deals with phonological categorisation.

[ back ] 74. Devine & Stephens 1994:376–378, 384–385, 412–420.

[ back ] 75. O’Neill 1942.

[ back ] 76. Terminology as used by Korzeniewski 1968.

[ back ] 77. For exceptions to this tendency, see below.

[ back ] 78. Descriptions of alternating rhythm speak of the “more regular expression”, cf. Lidov 1989:68.

[ back ] 79. Fränkel 1926; Visser 1987:80–82; Bakker 2005:1–21.

[ back ] 80. The development of the hexameter actually leads to a strong preference for third foot word division at the trochaic caesura in Nonnus, thus avoiding metarrhythmisis in the third foot.

[ back ] 81. Van Raalte 1986:93; Bakker 1988:2n3.

[ back ] 82. Leaf 1900–1902:app.D C2; Steinrück 2005.

[ back ] 83. Steinrück 2005:495 analyses the hexameter as an asynartetic verse: the third foot caesura separates a dactylic basis (hemiepes) from a choriambic expansion. Koster 1953:26 suggests that the trochaic sixth foot of the hexameter might be considered as a μεταρρύθμισις of the dactyl (cf. the dactyl as μεταρρύθμισις in the trochaic tetrameter). This analysis presents the sixth foot as a logaoedic (λογαοιδός; dactyl + trochee) expansion to a dactylic pentapodic basis. It does not exclude that the sixth-foot thesis of the στίχος μείουρος is an anceps (examples in Leaf 1900–1902,ap.D,C3); in this verse-type, the final thesis may be occupied by a light syllable, just like the verse-initial thesis in στίχος ἀκέφαλος, as in e.g. Iliad 22.379 ἐπεὶ δὴ τόνδ᾿ ἄνδρα θεοὶ δαμάσασθαι ἔδωκαν ‘as the gods have finally granted me to restrain this man’, and Iliad 23.2 ἐπεὶ δὴ νῆάς τε καὶ Ἑλλήσποντον ἵκοντο ‘as they had reached the ships and the Hellespont’ (further examples in Leaf 1900–1902:ap.D,C1). Both the verse-final and the verse-initial foot of the hexameter give rise to the assumption of an earlier stage in which both feet only had to comply with Aeolic isosyllaby.

[ back ] 84. Contra the identification of the punctuation following initial vocatives as “4 morae” (in the summary of Nicanor’s system quoted by Devine & Stephens 1994:421).

[ back ] 85. Judging from the sample (see Appendix) spondaic word end is found on position 2 in 7% of the verses of the Iliad, and 12% of the Odyssey. For positions 4 and 8, the percentages are 1% (both Iliad and Odyssey), and 5% (Iliad) / 1% (Odyssey) respectively. The occurrence of metrically spondaic word end at position 12 (verse end) shows significance difference: 41% of the lines of the Iliad features verse-final spondaic word end, versus 78% of the Odyssey.

[ back ] 86. West 1982:39.

[ back ] 87. The verses in which the third foot caesura is not being observed do not feature spondaic word end on the third foot.

[ back ] 88. If the word juncture following the prepositive after position 6 is hardly felt in lines 46, 287, 461, and 591, the localisation of the prepositive + lexical may better be compared to the metrically restricted bacchic word type: in that case, the possibilities for localisation are as restricted as those under Wernicke’s law.

[ back ] 89. Metrical word shape as opposed to rhythmical word type, as the final heavy element of the bacchic counts as a thesis and cannot be resolved.

[ back ] 90. Devine & Stephens 1994: 131. They speak of “subordination” when discussing the mapping of heavy syllables on the arsis; see further 4.6.4 below.

[ back ] 91. Devine & Stephens 1976. Cf. Van Raalte 1986:94. Basset 1938:144 credits the bucolic pause the possibility to ‘restore the dactylic rhythm’.

[ back ] 92. Parry 1971:93–94.

[ back ] 93. Its duration is comparable to that of any other μακρὰ ἄλογος, both on the thesis, and on the arsis. The examples quoted show that its duration as μακρὰ ἄλογος serves a κυκλικός–rhythm: it avoids rhythmical disturbance, in a way similar to ῥυθμὸς κυκλικός of the non-final thesis in anapaestic meter.

[ back ] 94. Ruijgh 1987:342n62, following Dionysius’s terminology concerning the participle.

[ back ] 95. Description of rhythm as alternating (cf. Lidov 1989:69) does not allow phrasal metarrhythmisis within a period, but ‘between periods (that is, at a “pause”) a break in the alternation should not be construed as an abnormality of special interest.’

[ back ] 96. Metrical word shape, as opposed to rhythmical word type. In the rhythmical word-type, the word-final syllable is mapped on the thesis, though some of the instances of the non-contracted prototype, Ionicus a minore, treat the word-final syllable as an arsis.

[ back ] 97. Snell 1962:7; Van Raalte 1986:29, 92–108 mentions, among other features, the avoidance of spondaic word end as a characteristic of acatalectic verse.

[ back ] 98. Basset 1938:150.

[ back ] 99. Cf. the bucolic dieresis as the most common spot for printed punctuation in the hexameter.

[ back ] 100. Strictly speaking, brevis in longo applies only to the metrical thesis.

[ back ] 101. Korzeniewski 1968:9.

[ back ] 102. Devine & Stephens 1994:131–132.

[ back ] 103. From Porter’s notation of alternating rhythm (in Loomis 1972, cf. Lidov 1989:72n17) it becomes clear that the metrical anceps is the regular alternative between instances of double short (as the realisation of the “more changeable expression”, the arsis) in dactylo-epitrite.

[ back ] 104. Rossi 1963:44–49.

[ back ] 105. Ruijgh 1987:349 ‘Denys oublie de signaler que la syllabe finale en fin de vers est bien une μακρὰ τελεία’.

[ back ] 106. In 9% of the verses of the Iliad (21,9% of all verses ending in spondaic word end), but 23% (29,5% of the verses ending in spondaic word end) of the Odyssey, according to the sample in the Appendix.

[ back ] 107. Ruijgh 1987:317; West 1982:121.

[ back ] 108. Van Raalte 1986:17 compares a double light ending for the hexameter with the final heavy element of the iambic trimeter as a licence for brevis in longo. Apparently, she does not presume a correspondence between this licence and the rhythmical closure (blunt versus pendant) of the metrical colon.

[ back ] 109. Van Raalte 1986:17 explains such brevis in longo by saying ‘that the actual quantity of the verse-final syllable (i.e. the rhythmic realisation of the final metric element) is irrelevant in virtue of its prosodic neutrality, the rhythmic series […] being interrupted as soon as the metric series has been rhythmically completed.’

[ back ] 110. Examples from Usener 1887:38–40.

[ back ] 111. Parry 1971:263.

[ back ] 112. From the analysis of Iliad 11–100 and Odyssey 1–100 in the Appendix the following statistics emerge:

Percentage of lines ending in:

	Iliad	Odyssey
Perfect heavy after metarrhythmisis	16%	16%
“Brevis in longo”*	21%	20%
Subordinate syllable	63%	64%

Table 6: various rhythmical line endings.
* The alternative would be to describe this possibility as “perfect heavy < light syllable after metarrhythmisis”, taking metarrhythmisis for granted. Metrically, there is no distinction between the two ways to end the line. It is exactly the metrical resemblance that obscures the choice of pause: the anceps allows for secondary pause, metarrhythmisis for primary pause. In case of “brevis in longo” there is no a priori reason to assume metarrhythmisis, apart from the notion that brevis in longo, strictly speaking, is limited to the thesis.

[ back ] 113. Punctuation may suggest otherwise as in, for example Iliad 4.364: ὧς εἰπὼν τοὺς μὲν λίπεν αὐτοῦ βῆ δὲ μετ᾿ ἄλλους ‘having spoken thus he left them there, and went to the others’. There is no reason, however, to consider the verse-final syllable a “perfect” heavy element, as Iliad 4.365 starts with a vowel: εὗρε δὲ Τυδέος υἱὸν ὑπέρθυμον Διομήδεα ‘and he found the son of Tydeus, brave Diomedes’.

[ back ] 114. Devine and Stephens 1994:324–326.

[ back ] 115. Van Raalte 1986:73–83.

[ back ] 116. Labeled A1 (position 2), A2 (position 3), C1 (position 7), and C2 (position 8) in the overview of Clark 2004:120–121. The third foot caesura is labeled B1 (position 5) or B2 (position 5½). Position 9 is not mentioned in this overview.

[ back ] 117. Of which the “rising threefolder” (Kirk 1985:18–24) is the analysis for verses without third foot word end.

[ back ] 118. Devine & Stephens 1994:89–90, 400.

[ back ] 119. In Latin hexametric poetry, the spondaic verse-final foot is often deliberately created, cf. Hardie 1920:5.

[ back ] 120. As in the opinion of many others. Arguments against “prosodic disturbance” at elision are summarised in Devine & Stephens 1994:256–266.

[ back ] 121. The use by modern editors of punctuation like colon, semicolon, and comma, if to be accepted, must be judged on the basis of their observance of audible pause, for which reason Daitz 1991 considers 50% of all printed punctuation wrong or misleading. In addition to his general criticism, I would like to point out that one common application of printed punctuation in particular is questionable, the use of the comma following the second foot trochee before a vocative as in Iliad 1.1 μῆνιν ἄειδε θεά ‘sing of the wrath, goddess’, and Iliad 11.287 ἀνέρες ἔστε φίλοι ‘prove yourself men, comrades’. It seems that Meyer’s law for the Callimachean hexameter is observed in Homer as well: in the examples quoted as elsewhere the segmental synaphy of the developing phonological phrase frustrates the possibility of phonetic word end at the verse’s second trochee.

[ back ] 122. In the sample in the Appendix the ratio [Ascending start : Descending start] for grammatical clauses is [65.2% : 34.8%] for the Iliad, and [53.8% : 46.2%] for the Odyssey.

[ back ] 123. Devine & Stephens 1994:400.

[ back ] 124. Devine & Stephens 1994:417–418.

[ back ] 125. Though refusing in principle the possibility for a verse-internal pause in Homer’s hexameter, Daitz 1991:156 grants Odyssey 1.10 an ‘optional pause’ at position 5 ‘for expressive purposes’.

[ back ] 126. Devine & Stephens 1994:418 speak of “parentheticals” as “right dislocated afterthoughts”.

[ back ] 127. Dionysius of Halicarnassus On Literary Composition 16/61.19–68.6 U-R.

[ back ] 128. The sample in the Appendix provides some interesting data though. “Real” enjambment occurs 22 times, of which 17 times in direct speech (where direct speech is 52% of the sample). In the Odyssey, there are 34 instances, 24 of which in direct speech (where direct speech is 59% of the sample). The 22 instances of enjambment in the Iliad all break up the clause, and all but one are verse-internal. A remarkable example is Iliad 1.82, where the enjambment is the result of phrasal metarrhythmisis on position 4. Of the 34 instances of enjambment in the first 100 lines of the Odyssey, 33 break up the grammatical clause, and 30 are verse-internal. In Odyssey 1.72, I see no enjambment, but a syntactical structure in which the apposition in nominative case is not tied to the preceding, but to the subsequent genitive case apposition. The only examples of “real” enjambment in accordance with Higbie’s system are Iliad 1.67, Odyssey 1.52 (integral enjambment), 1.18, 1.61 (adding external enjambment), 1.39 (clausal internal enjambment), 1.23, 1.59. 1.83, (clausal external enjambment), though Odyssey 1.23, 1.39 and 1.59 are verse-internal.

[ back ] 129. Phonetic word end in [long vowel + ς] does not necessarily lead to a superheavy syllable, cf. the syllabification of ς in consonant-clusters in Devine & Stephens 1994:43.

[ back ] 130. The sample in the Appendix provides the following statistics:

	Iliad			Odyssey
	Clause end*	“Enjambment”	Other**	Clause end*	“Enjambment”	Other**
Pos. 3	8.10%	9.30%	8.1%	7.50%	12.50%	13.75%
Pos. 4	0.00%	0.00%	0.00%	0.00%	1.25%	0.00%
Pos. 5	17.44%	5.81%	3.49%	6.25%	7.50%	3.75%
Pos. 7	11.63%	5.81%	3.49%	5.00%	10.00%	2.50%
Pos. 8	2.32%	0.00%	0.00%	0.00%	1.25%	2.50%
Pos. 9	2.32%	2.32%	1.20%	0.00%	3.75%	1.25%
Pos. 11	0.00%	0.00%	0.00%	0.00%	1.25%	0.00%
Pos. 12	12.79%	3.48%	2.32%	7.50%	5.00%	7.50%

Table 6: Demarcation due to the primary rhythmical pause as percentage of total number of primary pauses.
* Including the right branch demarcation of the predicatively used participle.
** The right branch demarcation of a transitional constituent or a cluster of transitional constituents.

About

Fellowships

The Library

Programs