3. Metrics and Phonology of Grammatical Clauses

The real compositional principle behind the verse-to-verse structure of both the Iliad and the Odyssey seems to be a progressive tendency that strings together a large and possibly infinite number of single-clause informational units. The progression results from continuation, itself due to transitional constituents with clear pragmatic value, but little semantic function. The grammatical clauses that this progressive tendency produces are not guided by the grammatical requirements of the sentence. There is little predicational hierarchy, and hence no complex predication. Coordination and subordination stem from discourse markers that are themselves characterised prosodically, that is as metrical chunks. This prosodic characterisation isolates discourse markers from grammatical clauses in a way that differs [1] from written classical Greek. In Homer, their prosodic isolation in the phonetic word reduces, or even frustrates, usage as sentential particle. With the exception of the particle-introduced metrical-colon clause, the syntactical function of conjunctions and particles in larger scale grammatisation is severely hampered by the prosodic characterisation, the isolation, of the phonetic-word chunk. The pragmatic transitional constituents are both syntactically, and prosodically, extra-clausal. [2]

The majority of the Homeric transitional constituents appears to fill a metrical colon. In chapter 2, however, it was shown that elision, turning prosodic isolation impossible, in a way “reorganises” the juxtaposition of transitional constituents and grammatical clauses: the transitional constituent was no longer prosodically isolated, but incorporated into a single phrasal domain together with (part of) the subsequent clause. It is tempting to discuss “reorganisation” in terms of a compositional process that takes an earlier state and a new state, both of which are identifiable in the Homeric epic. The earlier state and the new state are then thought of as mental stages in the process that created the change in the surface structure from [transitional constituent]+[clause] as two prosodically characterised domains, into [transitional constituent + clause] as a single phrasal domain. Elision turned out to be a means to leap from one mental stage to the other; it may not be the only means to do so. There may be other phonological phenomena that reduce or frustrate prosodic isolation “on the right branch boundary” (at the completion) of the transitional constituent, thus reducing the isolating effect of metrical colometry. Seen from that angle, “reorganisation” of phrasal domains is no more that a development from an earlier mental stage in which colometry is understood as the reflection of surface structure, to a new stage in which surface structure at times mutes the positions of frequent word end. As both “mental stages” are still recognisable in the Homeric epic, phonology must provide the conditions under which the leap from one state to another could be undertaken. In order to determine under which conditions the transitional phonetic word may constitute one phrasal domain together with what follows the phonetic-word boundary, we need to look at that boundary. If the transitional phonetic word is syntactically functional within a larger scale grammatical unit, I would expect it to be phonologically appositive [3] as well. Phonological clisis is evidenced by resyllabification and phonetic reorganisation. Meter does not change in the process, but phonology will mirror the rationalisation of phonetic reorganisation. The phonological realisation of the (transitional) phonetic word and the larger scale grammatical unit may resemble that of the minor and major phonological phrase. If so, the level of syntactical arrangement and grammatical governance will appear to be higher than the persistence of nonconfigurationality and apposition in Homer. Phrasal variability stems from, but does not reflect, the metrical repetition.

I aim to show that the metrical and phonological shape of the grammatical clause (and of transitional constituents) will show the result of reorganisation of phonetic words in a prosodic structure that mirrors that of the minor and major phonological phrase. The positioning of the minor phrase within the major phrase is itself a type of reorganisation of phrasal domains. First, however, I will explain how the phrasal domain is organised internally through coherence, and how it is demarcated at its beginning and end. In a metrical text like the Homeric epic there is a role to play here both for metrics and phonology.

The surface structure of metrical text shows that within the metrical unit of the verse, there is no orthographic syllabification, but “prosodic syllabification”: The syllable structures are such that they correspond to the weight of the metrical positions they occupy. Prosodic syllabification is confined to the metrical unit of the verse: prosodic syllabification phenomena like shortening, liaison, and elision demonstrate the συνάφεια “juncture” within the verse. This “juncture” is broken at the metrical boundary of the verse-end: shortening, liaison, and elision do not occur there. Sometimes, the ‘juncture’ is broken at word-end within the verse. The συνάφεια indicates a type of coherence within the metrical unit of the individual verse. It is this coherence that facilitates the division of the verse into vowel-centred sounds that do not correspond with morphological or orthographic syllabification: the vowel-centred sounds are characterised prosodically, not orthographically. The prosodic “cutting up” of the line, however, can hardly be called “syllabification” as its clusters of “syllables” do not correspond with morphological lexemes, that is, the understandable and separable units of meaning (words and affixes). The orthographical syllables, a further subdivision of lexemes in single units of articulatory prominence, [4] do not signify quantity. [5] The units of quantity are phonemes, sounds that are formed regardless of morphology. It is better to leave the general terminology “syllable” behind and refer to the units of prosodic cutting up as “segments”. As phonology determines the exact shape of the segments, prosodic “syllabification” is best referred to as “phonological segmentation”. The segmentation itself is coherent within the verse, but cannot transgress the boundaries of the single line. As mentioned above, the phonemes of phonological segmentation are best referred to as segments. The terminology “syllable” is then only applicable to segments that are free from the influence of sandhi: liaison of consonants (or συνάφεια), and vowel coalescence (or συναλοιφή) through elision and shortening.

Within the line, phonological segmentation joins individual words together into a string of inseparable word groups by means of sandhi, much like the musical legato. Not every string is a whole verse, nor is every verse one whole string. The segmental approach of meter, with regard for the phrasal domain as the result of sandhi and for demarcation as the result of syllable-segments, suggests coherence and demarcation on a scale that is at times smaller or larger than the single verse. Coherence often disregards or nullifies the ‘ideal’ metrical colometry: phonological segmentation shows that positions of frequent word end may not automatically be considered as boundaries of the phrasal domain: sandhi often straddles these metrical boundaries. [6] Where phonological segmentation runs counter to orthographical syllabification, Devine and Stephens speak of resyllabification: either as onset to coda shift, the right to left movement of a consonant [7] (ἔ.να.ρα.βρο.τό.εν.τα --> ἔ.να.ραβ.ρο.τό.εν.τα), or as coda to onset resyllabification. Coda to onset resyllabification may be clearest seen in Homer by means of the movable nu, whose application regularly demonstrates the continuity of the phrasal domain over, for example, the third foot word division: As was shown in the modification to the analysis of Iliad 1.1–16 in 2.2.3, elision looks like the coda to onset type of resyllabification, but is in fact reduction of a syllable to onset. In other words, elision creates a segment-initial consonant [8] . The elided vocalic sound is still present, though severely reduced: elision affects the pitch pattern in case of an elided accented syllable, and in case of elision of clitics with secondary accentuation: In case of an elided monosyllable, the remaining consonant turns into an orthographic onset, that is, the syllable-initial consonant(s): Hiatus, on the other hand, is a means of demarcating the phrasal domain. Following a heavy syllable containing a long word-final vowel or diphthong, hiatus can be avoided in the arsis by shortening: [9] Apart from hiatus, demarcation of the phrasal domain is indicated by the simultaneous completion of both phonological segment and orthographical syllable. Demarcation of the domain is hence found at the appearance of phonetic word-final syllables, including not-shortened heavy phonetic word-final syllables containing a long vowel or a diphthong: [10] Next to word-final light syllables on the arsis, another indication for demarcation is the syllable containing a short vowel on the foot’s first element: The coherence and demarcation of the phrasal domain show that there is a second type of “prosodic unit”, next to the phonetic word: there is a phonological phrase that is larger than the metrical-colon appositive group, but not necessarily equal in size to the single hexametric verse. At the same time, this larger phrase seems to incorporate the, smaller size, phonetic words. The larger phrase resembles the major phonological phrase, the phonetic word the minor phonological phrase. If this comparison makes any sense for Homeric Greek, how do these two types of phonological phrases create one pattern of phrasal domains?

Phonological phrases have been identified in Greek, though not specifically in Homer, by Devine and Stephens (1994). Their description of the minor phrase shares most of the characteristics of Homer’s intonation unit as identified by Bakker: Devine and Stephens identify the minor phrase as a prosodic unit that forms the domain for grammatical development because of its characterisation as a domain. In Greek, the combination of clisis and metrical phrasing is evidence of the coherence of the phonological domain, and hence, as Devine and Stephens point out, of some coherence within a syntactical domain. They point at the positions of frequent word end, the occurrence of sandhi, and the minor phrase boost of the accent (as evidenced in the musical settings of the Delphic hymns) as proof of the prosodic coherence within the minor phonological phrase. In most respects, the minor phrase in Greek works, I would say, like the intonation-unit metrical phrase in Homer as presented as the unit of Homeric discourse by Bakker. In chapter 2, I showed that both prosodic and syntactical clisis turns enclitics into “bidirectionals” as intonation-unit transitional constituents turn into sentential prepositive phonetic words. That way, the minor-phrase intonation unit is drawn into a larger whole (Devine and Stephens speak of ‘prosodic unit’), reflecting a higher level of grammatisation, and demarcated by audibly stronger disruptions at normal rates of speech. These larger scale units resemble what Devine and Stephens describe as major phonological phrases: The demarcation of minor and major phonological phrases is hence a matter of both meter and apposition (for the minor phrase), and phonetics (for the audible phenomena demarcating the major phrase). In the remainder of this chapter, I will first analyse the metrical pattern of the grammatical clause and its intonation units. It is easy to identify and the easiest way to provide evidence for the prosodic determination of phrasal-domain reorganisation. The phonological pattern is a different story entirely; it provides the rationalisation of the possibilities and impossibilities of phrase termination, but phrase termination can not change the metrical surface structure. Phonology evidences the internal coherence of the phonological phrase and its demarcation. Phonetic clues for phrase termination are hence identifiable. Their identification requires a new approach of the audible pause that is partly my own contribution in this study. The description of Homeric syntax as a progressive tendency featuring grammatical clauses will finds its reflection in the patterning of audible pauses, discussed in detail in chapter 4. Before that, following the description of the grammatical clause’s metrical and phonological pattern, I will assess the clues from antiquity for the perception of Homeric discourse as a progressive tendency due to a series of grammatical clauses.

In this section I will examine the way grammatical clauses are located onto the hexameter or parts thereof. The notion, rejected in chapter 1, that grammatical requirements of the clause are preferably met within the boundaries of the individual hexameter suggests that the possibilities for location are restricted to the phrasal domain of the single line. I will show that this is not the case: the single line is not the phrasal domain of the grammatical clause. In other words, the compositional impulse of the metrical hexametric unity is limited. The remarks made so far concerning the Homeric grammatical clause provide the examination with a serviceable basis. It is possible to start the identification of the metrical pattern or patterns of the clause, as the grammatical clause shows a clearly marked beginning. The beginning of the grammatical clause is of course a much more convenient point to start than its completion. The extra-clausal intonation-unit constituent branches the boundary of the clause on the left, that is, its start. The combination of syntactical and prosodic framing (clisis) is a clear indication for clause-start. Branching on the right is not so easily determined in Homer’s adding and appositional style: on the right branch, there is only prosodic framing. Syntactical framing, if any, is on the left branch of the subsequent clause or transitional constituent. That makes the Homeric grammatical clause rather open-ended.

Metrical clause-patterns are to some extent comparable to metrical word-types. Certain patterns fit the hexameter, others would not. The latter category is either excluded from the Iliad and the Odyssey, or has been modified. [14] There is, however, apart from the difference in size, another, more essential, difference between the metrical clause-pattern and the metrical word-type. Whereas the word type necessarily fits the hexameter, the clause-pattern may have the stichic [15] hexametric series as its phrasal domain. As mentioned above, the completion of a grammatical clause is primarily signalled by the syntactical left branch framing of the subsequent clause or transitional constituent. Transition to a subsequent predicate frame is a clear means of syntactical left branch framing, as is a pragmatic transitional constituent. Theme or tail constituents are not always syntactically conclusive. Sometimes a tail constituent doubles as preparatory theme constituent, as in Iliad 1.9 Λητοῦς καὶ Διὸς υἱός. Prosodic features like clisis must then be taken into account to determine the exact pragmatic value: is the transitional constituent pre- or postpositive? The realisation of metrical boundaries is the decisive factor: is the metrical boundary realised as an audible pause, or is it straddled by any means?

The metrical boundary most commonly considered as an audible pause is the verse end. Grammatically, the verse end is straddled easily. The verse end does not necessarily indicate the completion of a grammatical clause, though it coincides with the transition from one grammatical clause to the next many times. The verse end is often not itself the point of transition, but it is preceded or followed by extra-clausal constituents. A clause can easily be distributed over (parts of) more than one hexameter without being—grammatically—hampered at all.

Grammatical clauses can be found to start after numerous positions within the verse and at the beginning of a verse. Listed below are examples of clauses starting after different metrical positions. I have chosen to make the clauses start after the different metrical positions: that makes it easier to account for the various ways a clause may start depending on the metrical realisation of the foot’s second element. The difference between clause start in –, ⏑, or ⏑ ⏑ is now presented more clearly. I have tried to present a complete list of the metrical possibilities for clause start, so extra categories have been added to account for the various metrical ways in which a clause can start. Clauses are cited till verse end, but may extend into the next line. This list will show the freedom and variety to both create an open-ended clause, and to start a new one, almost regardless of metrical position. [16] It also shows the metrical adaptations and licenses required (and used) to attain such freedom and variety. Many more examples can of course be added. The start of a grammatical clause after position 1 by means of a heavy segment at position 2 is not common. When used, it is often in combination with an elided particle at position 1, in a formula, or with an elided particle in a formula, like τὸν δ᾿ αὖτε προσέειπε ‘and to him, in turn, said’. In this formula, the sandhi due to elision makes the grammatical clause start at the beginning of the verse: another indication that clause start with a heavy segment at position 2 is avoided. The example Iliad 10.344 is clear enough, but looks like a modification of a prototype starting from 1 [+ ⏑ ⏑]. Leaf [17] suggests original ἀλλ᾿ ἐάωμεν πρῶτα, pointing at the irregular position of μιν, but retains the elision. It is difficult to determine whether or not reorganisation through elision postdates the modification of formulaic prototypes: both have the major-phrase hemistich as their domain. [18] It would be theoretically sound, I think, to correct to ἀλλὰ ἐῶμεν πρῶτα (as do Brandreth and Von Christ). The start of a clause after position 2 can of course be preceded by a heavy syllable as in Iliad 10.203 ἀγχοῦ δ᾿ ἱστάμενος ἔπεα πρερόεντα προσηύδα ‘standing close to him he spoke the winged words’. Similar to clauses starting after 1 [+ heavy segment], clauses starting after 3 [+ heavy segment] may show signs of modification. Hilberg’s law is being observed. The phonetic-word boundary following enclitic τοι, is reversed by the prepositive character of οὔ τοι: The particle τοι becomes bidirectional [19] (phonetic word [οὔ τοι πάντες]). Contrary to what one may expect, the number of grammatical clauses starting with double short after position 5 (after the masculine third foot caesura) is small. Clause start after position 6 is rare. Not surprisingly, as it falls between two positions of frequent word end, not being a position of frequent phonetic-word end itself. In Odyssey 10.79 there is no phonetic-word end due to shortening. [20] Phonetic word end at position 6 is not impossible, though. Since clause start with a heavy segment is regularly preceded by a postpositive, light syllable, clause start after position 6 is unlikely when position 6 is itself occupied by a heavy syllable. A heavy syllable on position 6 may be followed by elision of the particle, as in Odyssey 8.505: The heavy syllable + elided particle is rare at this position throughout the Iliad and the Odyssey. With double-light at position 6, the same remarkably often applies: the phonetic word between positions 5 and 7 is kept intact. Reorganisation is clearly visible in Odyssey 7.110: original πέρι γὰρ σφίσι δῶκεν Ἀθηνη is likely. Odyssey 5.146 shows reorganisation in the elision of δ᾿.

Contrary to position 5, clause start after position 7 is fairly common, both in the Iliad and in the Odyssey. When preceded by a short particle or a pronoun, the transitional particle is phonetically lengthened at position 7: This phenomenon is common to the Iliad and the Odyssey. Phonologically, position 7 is straddled due to (geminate -μμ-) onset-to-coda resyllabification: δὲμ.μέ.γα. [21] Elision of the particle in Odyssey 5.146 is a clever echo of Iliad 1.5: Διὸς δ᾿ ἐτελείετο βουλή (including reference to a μῆνις). The elision is equally common: Clause start after an elided particle is common at this position, in this example allowing for a pronoun at position 7½. A combination like ὁ δ᾿ maintains the word end at position 7 and Hermann’s bridge: there is no room for another word end in the fourth foot, demarcating the extra-clausal constituent. When the transitional particle is given in full (not elided) at position 7½, a feminine fourth foot caesura is avoided by adding another postpositive particle or a pronoun: The bucolic dieresis can easily be used to start a new clause. There is a clear preference for starting the grammatical clause after position 8 over starting after 7½. It is noteworthy that clause starts after position 8 are not as often preceded by preparatory pragmatic constituents (like theme) as starts after other positions. The clause start can be preceded by the transitional constituent καί, as in Odyssey 14.404: Alternatively, the clause after position 8 starts with a verb form (including the participle), or with a proclitic connector: The well-known continuation after position 8 αὐτὰρ ἔπειτα ‘but then’ is itself a transitional pragmatic constituent, to be compared with other pragmatic constituents that are intonation units fitting the metrical pattern after the bucolic dieresis. These observations strongly suggest that the bucolic dieresis can be realised as both a minor-phrase and a major-phrase boundary. Prepausal location of the word-final biceps on position 8 is suggested by punctuation [22] at the bucolic dieresis, but is not markedly realised in phonetics: In Odyssey 15.82, the realisation of phonetic-word end is frustrated by shortening, [23] in 15.88 by elision. Realisation of phonetic-word end is equally frustrated by sandhi in formulaic expressions: The realisation of sandhi over the metrical boundary frustrates phonetic word end. As such, the presence or absence of sandhi is a feature of phrase boundary. The bucolic dieresis must be demarcated by the minor-phrase boundary in order to be realised as a prosodic boundary. If there is no phonetic word end at the bucolic dieresis, there is no minor-phrase boundary: Though frequent as a position of word end, position 9 is not often used to demarcate the phonological phrase. It is phonologically straddled by enclitics (ἡ :⁹ μὲν ἔπειτα / οἱ :⁹ δὲ δὴ ἄλλοι / παῖδας :⁹ τ᾿ ἄλοχόν τε), or proclitics (αἴ κ᾿ :⁹ ἐλεήσηι / ὡς :⁹ ἐτέτυκτο / ὣς :⁹ τὸ πάρος περ). Odyssey 13.389 is not an example of clause start with a heavy segment after position 9; it is an example of position 9 as the demarcation of a phonological phrase. Clause start with a heavy segment after position 9 requires a spondaic verse with word end in the fifth foot. Split fifth-foot spondees are rare in Homer, let alone clause start with a heavy segment after position 9. If one reads formulaic ὄφρ᾿ ἐὺ εἰδῶ as ὄφρ᾿ εὖ εἰδῶ (Odyssey 1.174), one gets close, but the elision shows that in the dactylic hexameter the split spondee is reorganised to a metrical bridge in the fifth foot. Just as after positions 7½ and (in some cases) 8, the start of a new clause after position 9½ is often preceded by a pragmatic constituent filling the metrical pattern of the minor phrase from the nearest phonetic-word end (in this case, at the bucolic dieresis φίλε :⁸ σοὶ). Compare the use of ὄφρα, or αὐτάρ in, for example, Odyssey 1.215: Clause start after position 10 is infrequent thought not impossible. The actual clause start may be after 9½ with position 10 being filled with a clitic pronoun or a particle like καί, τοι or δή. Proper clause start after position 10 can be found in, for example, Iliad 18.358–359 with σεῖο at the close of the hexameter: Another example is Odyssey 7.69: The prepositive character of καὶ is, as so often, further strengthened by shortening. The resulting single verb form ἔστιν remains elliptic and awkward: in his commentary, Hainsworth [24] suggests Plato Symposium 195b μετὰ δὲ νέων ἀεὶ ξύνεστί τε καὶ ἐστι as a parallel. Extension through particles may, however, postpone the true clause start until the start of the subsequent hexameter: On the other hand, clause start after position 10 may even result in completion of that clause at position 12 in the same line: Clause start after position 11 only seldom occurs in Homer, and the example quoted stands in brackets: it can only be accepted if the participle is seen as equivalent to a finite verb. Verse end may be in enclitic particles like τε, γε, περ, or in monosyllabic words like δῶ, κῆρ, φώς, βοῦς, χρῆ or Ζεύς, but verse end in a combination [transitional extra-clausals + monosyllabic] (for example: *ὁ δὲ Ζεύς || ) is not found. Iliad 1.128 τριπλῆι τετραπλῆι τ᾿ ἀποτίσομεν αἴ κέ ποθι Ζεύς ‘threefold, fourfold we will compensate you, if Zeus ever’ comes close, but of course the minor phrase contains at least [αἴ κέ ποθι Ζεύς]. The prosodic bridge that hampers monosyllabic ending of the hexameter renders phonetic-word end facilitating clause start from position 11 nearly impossible. Still, just as at position 9 there appears to be a possibility for a nearly straddled minor-phrase boundary: Clause-start after position 12 means: the start of a grammatical clause coincides with the start of a hexameter. As after other positions, clause start after position 12 may be anticipated by extra-clausal constituents in the preceding line: Both examples show bucolic anticipation, one of the “hooks” identified by Clark. [26] The anticipation is not in terms of the clause, as suggested by Clark 1997: the grammatical clause is yet to start with the next verse. The anticipation is syntactical only from the point of view of the major phrase. Phonologically, the anticipation itself is a minor phrase, with a minor-phrase demarcation at verse end.

When identifying the completion of a grammatical clause, secure identification is the result of identification of the start of the subsequent clause, or of a pragmatic extra-clausal constituent. As the pragmatic constituent may be ‘reorganised’ within the major phrase, its identification as indicator of completion is not as relevant for the analysis of Homeric clause formation as its identification as clause start. [27] Regardless of the relevance of identification of completion of a grammatical clause, the identification itself is not identification of the grammatically complete clause. [28] Though clauses are grammatically complete at the end of the line, they may continue well into the next: The verb form ἀγέσθω has a complete predicate frame in 146, but the frame is nonetheless extended into 147 by means of the satellite πρὸς οἶκον Πηλῆος. The true completion of the predicate frame is marked by an extra-clausal constituent with the pragmatic function theme: ἐγὼ δέ in 147. The description of πρὸς οἶκον Πηλῆος varies according to the theoretical point of view. If the verse end is acknowledged as a must-be observed metrical boundary, πρὸς οἶκον Πηλῆος is an example of run-over. [29] Taking a start from the whole-verse utterance, πρὸς οἶκον Πηλῆος stands in adding internal [30] enjambment: the addition is not a grammatical requirement, and ‘belongs to the same clause as the leading line’. [31] As an addition, it occupies a metrical colon, and forms a minor phonological phrase. Standing between the verse end and the start of a new syntactical unit, πρὸς οἶκον Πηλῆος may be described as a phrase. In performance, πρὸς οἶκον Πηλῆος may well be a single intonation unit. Both from the syntactical and the prosodic point of view, πρὸς οἶκον Πηλῆος demonstrates the open-ended character of the grammatical clause.

Lines 146–148 contain another, similar example. The clause ἐπὶ μείλια δώσω is grammatically complete at the end of line 147, but continues into the next line until completed by the appearance of, and hence, at ὅσσα in line 148. The grammatical clause is ἐπὶ μείλια δώσω πολλὰ μάλα.

The metrical shape of the Homeric grammatical clause is highly variable and rather free: the alternation and combination of grammatical clauses and extra-clausal constituents (with their specific shape) creates a metrical patchwork, despite the repetition of the dactylic hexameters. Grammatical clauses vary in size. Transitional constituents (both within the clause and extra-clausal) occupy constantly changing metrical cola between positions of frequent word end.

Phonology rationalises the perceptible metrical surface structure, and will hence not run counter to the requirements of this surface structure. Regardless of the metrical surface structure—or rather, without being hampered by it—the clitic character of the transitional constituents as intonation-unit minor phrases illustrates their “relation” to the adjacent grammatical clauses. As minor phrases, the transitional constituents contribute to syntactical organisation and to prosodic variety. It is not to be automatically assumed, however, that the grammatical clause coincides with a major phonological phrase, or merely with a part thereof. It may well be possible that the grammatical clause is one major phrase, but given the abundance of minor-phrase extra-clausal constituents, the grammatical and phonological unity of the grammatical clause will often have only mild audible disruptions on its left or right branch. It may be reasonably expected that the major phonological phrase encompasses several smaller intonation-unit minor phrases, some of them clauses or parts thereof, some extra-clausal constituents. The major phrase may hence encompass several intonation units that are themselves phonologically realised as minor-phrase unities. In order to find the much stronger audible disruptions that demarcate the boundaries of the major phonological phrase, we have to look for phonetic clues in the phonological and metrical patterning of grammatical clauses. Devine and Stephens have suggested such clues: intonation, additional duration (final lengthening), pause, intensity (stress) and pitch (tone). Section 3.4.1 deals with such concrete phoneticism. Then, in section 3.4.2, I will investigate the phonological rationalisation of metrical surface structure within the grammatical clause to identify the phonetic clues, however scarce, that we find in observations from antiquity.

The distinction between phonology and phonetics is likened to the distinction between metrics and rhythm by Devine and Stephens. [32] In their attempt to reconstruct the prosody of Ancient Greek speech they draw attention to the fact that twentieth century durational approaches to Greek meter have developed some awareness for the durational differences between segmentally longer and shorter types of heavy and light syllables. The introduction to their analysis of the duration and weight of the syllable (1994:43–53) takes into account that segmentation of the metrical utterance phonologically rationalises the metrical surface structure. Mora count of the segments does not run counter to the metrical pattern of light and heavy syllables, despite durational differences between the various heavy and the various light segments. Such differences are referred to as differences in submoraic duration. Devine and Stephens point out, however, that in the philological tradition there had been a tendency to assume that differences in submoraic duration arising from differences in segmental structure (subcategorical syllable duration) are relevant to the analysis of meter. They list a number of metrical phenomena to illustrate the metrical relevance to durational difference in the philological tradition. Differences in subcategorical and intrinsic syllable duration have been referred to, as Devine and Stephens summarise, to explain the possibilities of, and restrictions on, resolution, and to explain the occurrence of specific segmental structures at specific metrical positions. With Devine and Stephens I do not try to find metrical relevance to differences in subcategorical and intrinsic syllable duration. But although these differences are not relevant to metrical analysis, I do not dismiss them altogether: I will approach differences in submoraic duration as relevant to the analysis of phonological phrase. The approaches summarised in the overview by Devine and Stephens turn out to be relevant to phonological phrase structure: they all derive concrete phoneticism from their observations concerning varying segmental structures. Such concrete phoneticism is called for when we investigate the termination of both the clause and the phonological phrase. The phonetics of termination, however, cannot harm the metrical surface structure, and can only be derived from the phonological analysis of segmental structure.

Ancient sources throw light on the disparity between the phonological and syntactical phrases on the one hand, and the metrical phrases on the other. For the point I want to make in chapter 4—clauses and phrases differ from each other in the way they terminate—the observations from antiquity, however scarce, may serve as our point of departure to retrieve the phonetics of the prosodic unit.

Terms from antiquity referring to prosodic phrasing and prosodic units are difficult to interpret. A serious problem is the application of the same or similar technical terms to both prosodic phrasing and syntactical unities. Terms like κῶλον, κόμμα, κομμάτιον, and περίοδος have been used as indicative of grammatical and syntactical unities for the past century of scholarship. A closer look at hexametric poetry in general shows that the equation of κῶλον to both the metrical colon and the grammatical clause, and that of περίοδος to both the verse and the sentence, is roughly correct with regard to Hellenistic and Roman poetry. Though grammatical clauses may reflect a certain correspondence with metrical phrasing, there is a severe disparity between the grammatical phrasing and the metrical cola of “inner metrics”. The correspondence between the two that characterises later hexametric development is not a persistent feature of the Iliad and the Odyssey.

Prosodic terminology from antiquity is commonly understood as referring to metrical phrases and cola, the metrical unities demarcated by positions of frequent word end. Today, [33] the positions of frequent word end are identified as rhythmical pauses in Homer. Such identification, however, has no basis in sources from antiquity. A remark by Aristotle sheds light on the syllabic division of specific hexametric verse variants. In Metaphysics 1093a30 Aristotle subdivides the hexameter into two hemistichs of 8 and 9 syllables respectively. This subdivision, however, is related to musical phenomena concerning scale, and cannot be applied to any metrical subdivision of the hexameter in general. Aristotle’s subdivision equals the metrical phrasing created by the caesura at position 5½ (in one specific verse type), but has nothing to do with the metrical boundary at position 5½, or with caesura in general.

The first mention of caesura comes from Hermogenes (Περὶ ἰδεῶν, 2nd century AD), who may have followed Heliodorus (1st century AD). Hermogenes used caesura (τομή) as the term for the metrical boundary within the line, the demarcation of smaller metrical phrases, or cola. Phrase demarcation at word end is never a topic in the technical manuals of antiquity. The only points of demarcation ancient metricians discuss are the metrical boundaries separating metrical cola. Hardie and Basset point out that the identification of the metrical cola of the stichic hexameter is in accordance with their likeness to cola known from non-stichic lyric poetry. [34] Basset argues that the subsequent labelling of the metrical boundary as “caesura” refers to its persistent, or at least frequent, realisation as word juncture, but not to a possible realisation as audible pause. Such realisation would stem from the combination of metrical and phonological phenomena; it depends on audible effects, on phonetics. In the treatment of prosody in manuals from antiquity, phonetics was not completely ignored: the effect of the coda on the mora-count of the metrically light word-final syllable [→ short vowel + consonant], of the movable nu, and of the syllable-initial consonant(s) are all treated by, among others, Dionysius of Halicarnassus (in his On Literary Composition).

In later hexametric poetry the disparity between phonology and phonetics looses its relevance. [35] In post-classical epic poetry the metrical colon serves, more and more, as the syntactical unit as well; the repetitive stichic metrical unit of the verse serves, more and more, as repetitive syntactical period. Not surprisingly, the sense-pause tends to fall, more and more, at the positions of metrical demarcation. According to these observations Basset explains the Byzantine development of the caesura into a rhetorical (sense-)pause. [36]

To gauge the Homeric phonological phrase, and the audible pause (as opposed to the pausa, the metrical boundary) with it, there are some helpful comments of the ancient ῥυθμικοί, as opposed to those of the μετρικοί. The latter group of scholars from antiquity, the μετρικοί, study what is now called outer metric. To them, the realisation of metrical rhythm is purely a regularised pattern of light and heavy segments. Every single segment is either heavy or light: both are absolute durations in accordance with the mora of the metronome. There is nothing in between. For the μετρικοί, the positions of frequent word end frame the metrical colon as a repeated sequence. Grouped together, the repeated metrical sequences form metrical periods, or metrical strophes. The theories of the μετρικοί are presented by Hephaestion. [37]

On the other hand, there are the ῥυθμικοί, whose ideas can be traced back to Aristoxenus. These scholars approach the issue of meter and rhythm with the aid of phonological tools that are very similar to those used in modern phonological research. Aristides Quintilianus, for example, says that the prosodic realisation of segments and syllables does not only depend on their metrical position (light/heavy), but on their position within the word as well: word-final position results in additional lengthening of both light and heavy word-final syllables. [38] As is the view of modern scholars like, among others, Devine and Stephens, the interpretation of duration is not as absolute but as relative, depending on position of the segment within the word.

The possibility of additional lengthening and the absence of phonological phenomena to avoid such lengthening (like elision or shortening) is described as a phenomenon occurring in the presence of pause by Devine and Stephens. [39] Their approach may thus provide tools to deduce the phonetics of termination from the combination of metrics and phonology.

Manuals from antiquity mainly deal with rhythmical word types, the ῥυθμοί. In his list of ῥυθμοί, Dionysius of Halicarnassus [40] cites exemplary κῶλα that often, but not necessarily, [41] equal metrical cola or periods, that is, verses. Dionysius carefully chooses examples that start with the rhythmical word type that gives its name to the metrical phrase as a whole. In specific cases, he quotes an example that completely consists of one single rhythmical word type. [42] In his example of the dactyl, however, Dionysius quotes a Homeric line without a dactylic word type: Not that he had to: Basset suggests that Dionysius might have cited Iliad 1.214 [43] :

The ancient ῥυθμικοί described the phonetic phenomenon now known as lengthening of final syllables (LFS). The additional lengthening of the word-final syllable (τονή) [44] is partly due to assimilation of the pause to the word-final syllable. Lengthening affects the final syllables of rhythmical word types. At word end, lengthening is described as διάστασις “separation” and κενὸς χρόνος “empty unit of time”. Other terminology indicating the same phenomenon is μεταξὺ χρόνος “in-between unit of time”, χρόνος μέσος “middle unit of time”, and σιωπή “silence” . [45] None of our sources explicitly says so, but it is tempting to assume that such terminology may also indicate a true pause of measurable duration. [46]

In ῥυθμοί, rhythmical word types, Dionysius of Halicarnassus (On Literary Composition 17) describes realisations containing more heavy syllables in more favourable terms than those with more light syllables. Still, Dionysius considers the dactylic word type “noble”, despite the preponderance of light syllables in the line he cites to illustrate [47] dactylic ῥυθμός: Furthermore, Dionysius distinguishes between word end that is masculine (following the foot’s thesis) and feminine (following the foot’s trochee). No specific terminology is being used for word end at the metron dieresis. As heavy syllables are preferred over light ones, so word end in a heavy syllable is better than word end in a light syllable. I argue that from these observations from antiquity another observation, metarrhythmisis, can be deduced.

As the localisation of metrical word types, and the avoidance of spondaic word end show, word end in a heavy syllable means word end in a heavy syllable on the thesis, the dactyl’s first element that cannot be resolved into two light segments. In the following examples word end on the thesis is underlined: Often, word end in a heavy syllable on the thesis brings out a different metrical pattern: that of the anapaest. The metrical shift is a prosodic reality on the level of the rhythmical word, that of metarrhythmisis (τὰ μέτρα μεταρρυθμίζεσθαι): the shift from one metrical pattern (dactyl) for phonetic words to another (anapaest). [48] One of the ancient ῥυθμικοί, Aristides Quintilianus, notes that the thesis often had a little extra duration compared to the arsis. He labels the dactyl ἀνάπαιστος ἀπὸ μείζονος (a maiore, cf. the ionicus), and the anapaest ἀνάπαιστος ἀπὸ ἐλάττονος (a minore). [49] This description makes it tempting to explain the dactylic-anapaest metarrhythmisis as a form of metrical anaclasis on word level.

Anapaests may be studied as rhythmical word type too. 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: In fact, Dionysius always chooses his examples of the different ῥυθμοί according to these principles.

Dionysius’s choice for Odyssey 9.39 may well have been deliberate. The line of his choice brings out clearly the word-final long thesis due to the process of τὰ μέτρα μεταρρυθμίζεσθαι. One look at Homeric poetry shows that the word end that results from dactylic word type is not the preferred localisation for metrical boundary; not even at the verse end where dactylic word type never occurs. Metarrhythmisis creates word juncture at positions other than metron dieresis: in the example from the Odyssey, position 5 (με φέρων :), and 7 (ἄνεμος :) are phonetic-word-final.

Can these few clues to the phonetics of termination shed any light on the relation between the major phonological phrase and Homer’s grammatical clause? Phrase-internal phonological cohesion stems from sandhi (vowel coalescence and liaison) as evidenced by metrical segmentation. Analysis of metrical segmentation shows that the Homeric grammatical clause is not automatically to be put on a par with the major phonological phrase. Phonological phrases do not always terminate where clauses terminate, just as repetitive metrical units and clauses do not automatically terminate at the same position. Sometimes they do, and sometimes they do not. The resulting patchwork-like pattern of metrical units, phonological phrases, and grammatical clauses will bring out possibilities for extended grammatisation, as I will show later in this study. At this point, our identification of the grammatical clause may benefit from analysis of the way major phrases terminate. Caution is needed: grammatical clauses can not be identified just looking at the pattern of major phrases. Both have to be studied in their own right, only to draw conclusions from the comparison of the separately identified patchworks of syntactical units and phonological phrases.

In order to provide an example of an ambiguous grammatical situation where phonology may come to our aid, I will turn to the request of Patroclus’s ghost to Achilles to wake up and bury his dead body: Metrical structure (and my translation) suggests two grammatical clauses in asyndeton. The metrical boundary dividing the two grammatical clauses is the trochaic third foot caesura (τάχιστα :^5½ πύλας). The metrical boundary dividing the two grammatical clauses is not a strong phonetic boundary: the phonological phrase develops well despite the third foot caesura. But then there is another (minor-phrase) phonetic boundary that is potentially much more disruptive: the dieresis after the first foot (θάπτέ με :² ὅττι), realised after the postpositive of a metrical-colon phonetic word. As an audible disruption, this boundary softens both the asyndeton and the hiatus: θάπτέ με·² ὅττι τάχιστα πύλας Ἀίδαο περήσω ‘bury me; that I may soonest pass the gates of Hades’s house’. Van Leeuwen, in his 1895 edition of the Iliad, already suggested this rhetorical punctuation; I think phonetics proves him right. Even in a line where grammatical structure seems to follow metrical colometry, phonological phrasing may point at another grammatical structure.

Written punctuation, treated as rhetorical punctuation, needlessly obscures phonological phrasing. Consider the lines in Homer that are made up of more than two complete grammatical clauses. This leaves hardly any room for extra-clausal constituents: In the edition of Leaf, there is a full stop after κεῦθε, but a comma after ἐξαύδα. The use of the comma after ἐξαύδα connects the imperative in asyndeton to another imperative. This usage of punctuation is in line with Nicanor’s system of punctuation, [50] but it is not consistently applied. A full stop after an imperative as the first word of the line is not uncommon in Leaf’s edition (for example Iliad 4.29 (= 16.443, 22.181), Iliad 21.128, Iliad 22.365 [51] ), but this edition consistently avoids two full stops within one line, that is, without counting a full stop at the verse end. As a result, the treatment of similar syntactical phenomena is not always equal: at times, it appears to be rather inconsistent.

Phonological phrasing may be applied to point at another rhetorical structure. The word-final syllable of ἐξαύδα is a heavy syllable on the thesis. A heavy syllable on the thesis may indicate phrase boundary. As a phrase boundary, position 3 resembles the full stop in rhetorical punctuation. [52] A full stop after κεῦθε is not supported by phonetics. The phrase that started with μὴ κεῦθε continues over the metrical third foot boundary. [53] The verse-internal phonetic boundaries seem to be a lot stronger in Iliad 1.363, the line that served as a model for Iliad 18.74: Here, both imperative phrases are right-branch demarcated by a heavy element on the thesis. Line Iliad 1.363 consists of three complete grammatical clauses, as does Iliad 18.74. And, equally, Iliad 1.363 cannot be analysed as a phonological “threefolder”: [54] the heavy syllable on position 7 is subject to gliding, [55] the phonological liaison bridging the word division, hence weakening the demarcative effect of the word-final heavy syllable on the thesis. Contrary to what is suggested by written punctuation, it is difficult to find an example of repetitive phonological phrasing within a Homeric line. Can it be found in Iliad 22.450? Here we have a directive, with a second one following in asyndeton. The first directive is not an imperative, but an adverb. Mostly this adverb is used with an imperative or a subjunctive in one clause (Iliad 7.350, Iliad 14.128). Here, however, combining the adverb and the imperative does not make sense, so Leaf and West, I think rightly, separate them into two separate clauses: the adverb is then used as an imperative [56] without an accompanying verb in the same clause (as in Iliad 13.481, Odyssey 8.307). What applies to a verb acting as an imperative, applies to an adverb acting as an imperative as well: if such an imperative forms a clause on its own and is followed by another clause in asyndeton, there is reason to consider the possibility of phonological phrases corresponding to each imperative. [57] Again, there is no repetitive pattern. Read δεῦτε·^1½ δύω μοι ἕπεσθον·^5½ ἴδωμ᾿ ὅτιν᾿ ἔργα τέτυκται and both imperatives are right-branch demarcated by word end that does not justify such harsh rhetorical punctuation. Alternatively, δεῦτε δύω·³ μοι ἕπεσθον ἴδωμ᾿ ὅτιν᾿ ἔργα τέτυκται creates an imperative phrase shaped as a metrical colon, but its phonological shape is not repeated within the line.

A number of open-ended grammatical clauses do not end with verse end. [58] The clear metrical break at verse end may or may not result in a break in the domain of the phonological phrase. Therefore the phonological realisation of the verse end of the hexameter calls for attention. There is a possibility that the phonological realisation of verse end allows for continuity of the phrasal domain: the metrical structure of the hexameter is based on repetition, but the sequence of grammatical clauses shaped as (parts of) phonological phrases aims at variation. There is, of course, room for coincidence of metrics (stichos-end) and syntax at verse end. Consider verses where the start of the hexameter is the start of a new grammatical clause as well. If the new clause has not been prepared for by means of a pragmatic constituent at the close of the preceding line, the verse end may be expected to double as a phonetic disruption. An example is Iliad 11.591–592 where the verse end of 591 demarcates the boundary of the phonological phrase: The phonetic realisation of the metrical value of the preceding verse-final syllable -όν may have been to some extent metrically indifferent (“anceps”)—thought the deceleration of speech tempo towards verse end would probably have made its duration equal to that of a heavy syllable. [59] When introduced by a pragmatic constituent in the preceding line (for example αὐτὰρ ἔπειτα), clause start from the start of the hexameter still leaves room for a metrically light element ending the preparatory constituent, and ending the hexameter. The resulting metrical pattern – ⏑ ⏑ – ⏑ would be similar to the pattern preceding clause start after position 3½, and resemble clause start after the trochaic third foot caesura (5½). Still, of course, the clause about to start after position 3½ or 5½ starts as an ascending clause, with a single light segment. Descending clause start following a single light element would be a rhythmical impossibility [60] within the hexametric major phrase sequence. Starting a descending clause with a marked heavy element within the line often requires, as I showed in chapter 2, phonological sandhi at the right side of a preceding constituent-final segment. The descending clause is always preceded by either double-light or a heavy element, never by a single light element. Major phrases straddling the verse end obey to the same rules.

Then again, clause start from the start of the hexameter with preparation at the closure of the preceding line is not predominant in Homer. Other forms of continuation straddling the verse end within the clause are much more frequent, and make use of the verse-final metrical anceps. A clause may have the metrical pattern ⏑ – ⏑ ⏑, – ⏑, ⏑ – ⏑, – – – ⏑, ⏑, – ⏑ ἑλώρια τεῦχε κύνεσσιν οἰωνοῖσί τε πᾶσι, when its rhythmical pattern is ⏑ – ⏑ ⏑, – ⏑, ⏑ – X, – – – ⏑, ⏑, – ⏑. Many other examples can be added featuring verse end in a single, metrically light, syllable. [61]

This suggests that verse-final syllables all serve the same phonetic function: the syllable at stichos end satisfies the expectation of a metrically heavy syllable. It can do so because of its metrical indifference: the metrical anceps is itself indicative of “termination” of the verse. The expectation of a heavy syllable in turn suggests that the verse-final syllable itself is not only metrically anceps (as it is either metrically light or heavy—the verse-final position is anceps “indifferent”, in that it allows for both metrical realisations), but that it may be rhythmically “indeterminate” due to ongoing grammatisation over the verse end. [62] Metrically, indifferent refers to the metrical position of a syllable within the verse. With rhythmically indeterminate, I refer to the realisation of a syllable that satisfies the expectation of a metrically heavy syllable.

Rhythmical “indeterminability” leaves room for continuity of the phrasal domain over the verse end, as the particular rhythmic character of the segment might be determined by sandhi. It also implies that phonological phenomena at verse end can be compared to those within the phrasal domain.

Let us take as an example the single light syllable on arsis at position 12. Within the phrasal domain, sandhi determines its rhythmic character, just as it does for the single light syllable on any other arsis. The word-final syllable [→ short vowel + consonant] may fill both positions of the double short, as the word-final syllable was pronounced as ambisyllabic. [63] Ancient grammarians heard a difference between ἔστι Νάξιος and ἔστιν ἄξιος, more or less like the audible difference between English ‘an aim’ and ‘a name’, given that the tempo of speech is relatively slow [64] and that the speaker pronounces in a clear and polished manner. If sandhi in ancient Greek implied that the final consonant of a word, before a vowel, was pronounced as an ambisyllabic consonant, a word-final syllable [→ short vowel + ambisyllabic consonant] remains metrically light within the hexameter in any position: And as a word-final syllable [→ short vowel + consonant] before a consonant within the hexameter counts as heavy, so a verse-final syllable [→ short vowel + consonant] may count as heavy as well: τεῦχε κύνεσσιν = – ⏑ ⏑ – –. In such cases there is a true heavy syllable to satisfy the expectations of the verse-final metrical position. The strongest visible evidence of ambisyllabism in this occurrence lies in the movable nu at verse end. Moveable nu reflects the rhythmical indeterminability, rather than the avoidance of hiatus. [65]

Allowing (ambisyllabic) sandhi to cross verse end would result in many more phonological realisations of the indeterminate syllable to account for the metrical indifference of the sixth-foot arsis, and that includes the avoidance of epic shortening at position 12. In total, there would be eighteen phonologically different realisations of the indeterminate verse-final heavy element. [66] Some types of phonological verse-end realisation are practically absent. Phrase-internal verse-final [→ long vowel/diphthong + movable nu] is very rare: I would argue that it is unwanted within the phrasal domain, and is with equal ease avoided. Phrase-internal verse end in [→ long vowel/diphthong + consonant] is regularly in -ς, and in –ν, the two consonants that are used very often as coda, and both highly ambisyllabic. The rarity of (phrase-internal) verse end in [→ diphthong with short vowel + consonant] finds additional explanation in the linguistic “falsity” of the diphthongs ου and ει. The examples quoted in footnote 66 give rise to the notion of “prosodic neutrality” of the verse-final syllable: they do not prove that verse end may be a continuation of sandhi both since hiatus is possible, and since liaison, elision and epic shortening are either non-existent or seemingly impossible to demonstrate.

It is, however, possible to question the irrelevance of the metrical indifference of the verse-final syllable, and hence its “prosodic neutrality”. Examples of grammatical clauses that run over the verse end into the subsequent hexameter all feature verse-final syllables are not only the completion of a stichic metrical shape (the hexameter), but also take their position within the metrical pattern of the grammatical clause. [67] The listening audience expects continuation of the metrical repetition within a clause. [68] The verse-final foot must be realised as a spondee to maintain the metrical repetition within the grammatical clause. In other words, any anticipatory metrical organisation at verse end requires a spondaic verse-final foot.

Metrical indifference that accepts a verse-final trochee as a verse-final spondee does not only give rise to the concept of phonological indeterminability, it also creates phonological difficulties. With the hexameter ending in a spondaic verse-final foot, as well as with word end, spondaic word end is the greatest phonological threat to clause continuation over the verse end. Within the hexameter, Hilberg’s law (no spondaic word end at position 4), the third foot word end, and Naeke’s law (no spondaic word end at position 8) together sufficiently guarantee the avoidance of spondaic word end. Homer, however, sometimes features spondaic word end within the line at positions 4 and 8. Wernicke’s law states that disregard of the spondaic bridge at positions 4 and 8 in the hexameter is allowed under certain circumstances: spondaic word end at position 8 features a word-final syllable [naturā-long vowel]. Wernicke’s law creates the possibility to use words before the bucolic dieresis that may otherwise only be used at verse end. In that respect, the verse end, with its nearly unavoidable spondaic word end, seems to enable, if not force, the composer to allow for spondaic word end within a grammatical clause that runs over into the next line. I find it odd that within the phrasal domain word end is easily allowed at one position, whereas at another, similar, position it can only be allowed under specific conditions. Word end at position 8 under Wernicke’s law seems highly restricted, but word end at position 12 appears to be possible in anything.

There seems to be, however, a restriction in the usage of spondaic word end at position 8 under Wernicke’s law, that does also appears at position 12. At position 8, word end under Wernicke’s law hardly ever results in a superheavy word-final syllable [naturā-long vowel + consonant] within the grammatical clause. In the Iliad, only few exceptions can be found, which occur especially in formulaic expressions: Save for the verse-final syllable [naturā-long vowel + ς] at position 12, examples of superheavy word-end syllables within the grammatical clause are equally hard to find (compare the types 14 and 17 in footnote 66): In fact, the licence that Wernicke’s law represents, the condition for spondaic word end, is similar to the condition for spondaic word end at verse end within a run-over grammatical clause. In other words, the metrical repetition within a grammatical clause is maintained despite the apparent “prosodic neutrality” of the verse-final syllable: it is not “prosodically neutral”, but rhythmically indeterminate.

In chapter 1, I listed the different types of verse-end enjambment identified by Higbie. Hers is the most elaborate classification of verse-end enjambment. Her classification categorises enjambment looking at the grammatical break caused by the verse end. As enjambment rises from the concept of verse-internal grammatical completeness, the acknowledgement of grammatical incompleteness at verse end determines the strength of the enjambment. Above, I have argued that grammatical completeness is preferably not sought after within metrical boundaries. The demarcation of the phonetic word, the minor phrase, and the major phrase tells us more about the level of syntactical development than do the recurring metrical boundaries. In chapter 2, I argued in favour of a new approach to clausal enjambment based on phonological phrasing. Now, with a complete set of different phrasal domains, Higbie’s classification of enjambment types can be reorganised in accordance with the phonetic realisation of the metrical verse end: Higbie’s classification disregards that a) the realisation of phonetic-word boundary is not automatically perceptible as a pause, b) the realisation of minor phrase boundaries is in accordance with grammatical organisation, [69] and c) within the phonological phrase there is no “affective prosody”, [70] that is, no emphasis due to localisation. The first constituent of the Homeric hexameter following enjambment does not automatically receive emphasis, not even if it is, grammatically, a mot-en-rejet. Looking at Higbie’s classification, and from her point of view, I still find it advisable to change the label “violent”: it is the only label that does not refer to what the word(s) following the enjambment contribute to the grammatical (in)completeness of the line ending in enjambment at verse end.

Phonetically, violent enjambment resembles necessary, clausal internal, and adding internal enjambment. The only difference is that violent enjambment cannot coincide with a minor-phrase boundary. To the ear, necessary, clausal internal, and adding internal enjambment not coinciding with a minor-phrase boundary are exactly the same as violent enjambment.

Grammatical clauses may thus be recognisable from their phonological shape as well as from the ongoing development of the predication. Grammatical expectations do not necessarily depend on repetitive metrical phrasing. The expectations of an audience, relying on its ears, depend on what is audible. The grammatical completeness of the “sentence” or “clause” does not have to be audibly suggested or frustrated by the verse end. Proper assessment of a feature like intonation has not even been given due attention, though it is a part of the prosodic repertoire from antiquity (μέλος [71] ). It is not illogical to put the domain for intonation on a par with the phonological phrase, rather than with the metrical colon as an “intonation unit”. Even when put on a par with the metrical colon, it may be safer to dismiss the term intonation unit altogether, as it is not automatically the equivalent of the phonetic word or the (minor) phonological phrase. As intonation can not be seen as something that is unrelated to the structure of language, I will refrain from using the term intonation unit in chapters 4 and 5. For now, I will return to the phonological clues for larger scale grammatisation.

The phonological phrase may easily straddle the verse end, or the third foot word juncture. At the same time, the grammatical clause in Homer may extend over these metrical boundaries with equal ease: the grammatical clause is rather open-ended, regularly at the expense of the verse end as a meaningful grammatical or syntactical boundary. Is it useful to describe Iliad 7.394–397 as a sequence of “skewed clauses”, as does Clark? [72] Metrically, αἴ κ᾿ ἐθέλητε παύσασθαι πολέμοιο δυσηχέος, εἰς ὅ κε νεκροὺς κείομεν, and ὕστερον αὖτε μαχησόμεθ᾿ εἰς ὅ κε δαίμων ἄμμε διακρίνηι are “skewed”, but syntactically they are not. The syntactically coherent clauses may be demarcated as phonological major phrases: the verse end does not impede such analysis. The phonetic realisation of the verse-final element is not necessarily a phonological clue for audible termination. Some “subordination” of the metrical boundaries and subsequent “yielding” of metrical to phonological phrasing are the result of the mismatch of boundaries and cola with phonetic boundaries. I realise that this seems to be a circular argument, but it is not: it would be if, as Parry and his followers take for granted, the Homeric utterance is equal to the hexametric metrical phrase. Later hexametric Greek poetry shows that the equation of metrical and sense-phrase gradually developed into a poetic norm. Latin hexametric poetry accepted the norm, but realised that there were exceptions: in Latin prosody, however, such exceptions could not be imitated by means of a verse-final metrical foot, as there is no possibility for realisation of the verse-final syllable as rhythmically indeterminate. Furthermore, verse-final syllable [→ short vowel + consonant] in Latin is not metrically heavy. [73] Straddling the hexameter’s verse end with a phonological phrase in Latin requires resyllabification: in other words, elision at the end of the hypermetric verse. [74]

I argue that the thought whose coherence is evident from grammatical organisation is prosodically characterised as (part of) a phonological phrase. The phonological phrase is hence the domain for the grammatical clause. [75] The relatively “variable clause-length” is an original feature of Homeric poetry. Homeric clausal grammar is the grammar of the phonological phrase, not of the metrical phrase. This means that there is quite a distance between the “expectation” and the “grammatical need” for what follows metrical boundaries, including the verse end.

So far, the concept of the whole-clause formula has been under attack, but not banned as something that is irrelevant. It clearly is relevant, for example in a formulaic line like Odyssey 6.186: Here, however, the elision points at phonological reorganisation. A better example would be Odyssey 10.(214–)215: Here the extra-clausal constituent is separated from the phonologically coherent grammatical clause by the verse end as phonological-phrase boundary. But examples of transitional constituents (within the clause and extra-clausal) with the verse end as their right branch demarcation (type: αὐτὰρ ἔπειτα ||) are relatively rare. [76] It is far more common that the single hexameter contains (parts of) both a grammatical clause and one or more (extra-clausal) transitional constituents. As shown above, the (extra-clausal) transitional constituents are themselves sentential-prepositive minor phrases. As such, they are part of the larger scale unit of the major phrase, while retaining their prosodic characterisation of isolated chunks. When elided on their final syllable, the transitional constituents are reorganised into the major phrase.

Extra-clausals are commonly quite short, though there are instances of combined noun-epithet formulas occupying a hexameter (Iliad 1.7 Ἀτρείδης τε ἄναξ ἀνδρῶν καὶ δῖος Ἀχιλλεύς). The shorter extra-clausals appear occupying almost any metrical phrase between positions of frequent word end. Their appearance reflects, like a negative, the highly variable sizing of the intervening grammatical clauses, as the analysis of Iliad 1.1–16 in chapter 2 shows. The liberty to create grammatical clauses of any desired size is a characteristic of prose. Not just because of the variable size, but mainly due to the resulting obfuscation of incidental metrical phrases.

Homer, of course, is not prose: it is poetry. But it is unwise, as I have shown, to put hexametric meter on a par with hexametric phonological phrasing. The former is clearly visible. The well-known metrical bridges all show the importance of retaining the podic structure of the hexameter. Phonological phrasing, however, is not the product of the metrical hexameter. Not every verse contains two phonetically demarcated hemistichs, or four cola, nor is the verse automatically a phonological major phrase. The grammatically coherent phonological phrase is phonologically defined regardless of metrical phrasing. If the metrical phrasing corresponds with the phonological phrasing, the metrical cola that characterise poetry appear as phonological phrases. If the two do not correspond, meter has to give way to phonology. The result is a temporary subordination of meter. That is a true first step towards prosaic movement. Or should we say, considering the increasing rigidity in Greek hexametric composition, that what we see in Homer is actually its last trace?

Is Homer prosaic then? And is there no such thing as “meaningful verse end enjambment”, only phonetic demarcation? Yes and no. The Iliad and the Odyssey seem to represent a transitional stage in the development of stichic hexametric poetry.

An interesting parallel for the developments in Homer can be found in hexametric inscriptions and the punctuation used in these inscriptions. Some of these inscriptions [77] are “run-over”-type hexameters. An example like IG V 1.720 looks like a dactylic line, but not like a hexameter: It rather looks like a heptameter. What makes it appear as a hexameter is the punctuation following υἱὺς. To Friedländer, the punctuation suggested that ‘παι[δί] seems to fall outside the verse’. Comparison with other metrical inscriptions suggests that his conclusion may have been precipitate.

In metrical inscriptions, punctuation is based on a variety of principles. [78] Sometimes punctuation is found at the end of the metrical period: in our case, that would suggest a hexameter. Then again, punctuation is sometimes linguistically based. For this inscription (IG V 1.720), it is hard to decide which explanation is to be preferred. What is not so hard, is to see that the grammatical clause extends beyond the hexameter, and phonologically phrased by means of word end at position 7 and 12: the metrical phrase στῆσ᾿ Ἀνθίδα υἱὺς is followed by the run-over παιδί. Whatever punctuation marks, it is not an audible pause. Other examples, featuring visible verse end: The inscriptions vary ‘among hexameter, prose, and hybrid forms’ [79] . In all examples cited the verse end may function as a linguistically defendable minor phrase boundary. Such a minor phrase boundary, however, is impossible in IG I² 689: This inscription contains a formula familiar from prose dedications. Here, the run-over word ἀπαρχήν belongs to the predicate frame of ἀνέθηκε. Is there any reason to assume affective prosody due to verse-end enjambment in this inscription? Compare an inscription on the Chest of Cypselus: Further consider the formation of grammatical clauses in IG I² 975: The first line is not a hexameter: Γνάθωνος τόδε σημα is the start of a hexameter, and θέτο δ᾿ αὐτὸν ἀδελφὴ the completion, but in between there is an iamb missing. The metrical phrase between positions 5½ and 7 is not there. In Homer, this specific metrical phrase may be occupied by an extra-clausal constituent [80] . If a Homeric example like Iliad 10.164 is compared with IG I² 975, the inscription seems to disregard the need for maintaining podic structure [81] within the metrical period: The Homeric example, by contrast, shows the way in which maintaining podic structure can be used to single out the phonetic demarcation between the extra-clausal constituent and the subsequent grammatical clause. In Homer, localisation of extra-clausals proves to be an antidote to stichic metrical repetition. In the inscription, such appliance of an antidote is not required. At the same time, the metrical period is disregarded: θέτο δ᾿ αὐτὸν ἀδελφὴ fills the metrical phrase between positions 7 and 12, and ἡλίθιον νοσηλεύσασα cannot be analysed as a dactylic metarrhythmisis [82] . The combination of prose-elements and dactylic verse-patterns appears to be old and persistent. Compare the “enjambed” dactylic (or anapaestic?) clausula to an otherwise prose inscription: As in some of the inscriptions, the freedom in clause-formation in Homer may disregard the metrical boundary that is the verse end. The resulting phonological phrasing turns the period-final metrical colon into a minor phrase. The rhythm is still dactylic, but the metrical phrasing of the period, the verse, yields to the phonological phrasing: This slightly resembles prose rhythm in that the phrase structure is more important than the metrical structure in the experience of the verse. Of course, such prose rhythms are not predominant: there remain many verses that have a less subordinated verse end, either as a minor or major phrase boundary. With the possibility of prose-like rhythms 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 localisation. Affective prosody (as emphasis) must stem from the mismatch of grammatical / syntactical phrasing and the phonetic realisation of positions of frequent word end.

In this chapter, I have discussed the way developing grammatisation in Homer is determined not so much by metrical phrasing, but by phonological phrasing. The prosodic demarcation of grammatisation, as expressed by phonetics, is not as easily found as the syntactical demarcation. Indicators for phonetic demarcation have been suggested by modern scholars (hiatus, additional lengthening), but our ancient sources comment much more on phonological phrasing than on phonetic demarcation. Comparison with the punctuation of inscriptions serves as proof that metrical phrasing underlies variable phonological phrasing, a provisional conclusion already drawn on the basis of the free formation of grammatical clauses in Homeric poetry.

Chapter 4 will deal with phonetic demarcation as it has been discussed in recent scholarship. To these observations I will add a few of my own. I will argue that the combination of a (super)heavy syllable and the impossibility of coda to onset resyllabification is the phonological marker for deliberate phonetic disturbance. Dionysius [83] comments on such disturbance: the audible pause (ἀναβολὴ χρόνων ‘reset [84] of durations’) is caused by a ‘clashing of syllables’ (ἀνακοπὴ συλλαβῶν) that is the result of ‘blockades formed by articulate sounds’ (ἀντιστηριγμοὶ γραμμάτων). If such an ἀντιστηριγμὸς γραμμάτων evokes an audible pause within a grammatically coherent clause, poetically meaningful enjambment may be assumed.

In this chapter, I have argued that the Homeric epics benefit from a clausal grammar that is in accordance with phonological phrasing. In the process, the usefulness of the commonly used types of verse-end “enjambment” has been reduced.

So far, this study did not question the metrical phrasing of Homeric poetry, nor the fact that it is phrased according to phonological and syntactical units. What it did question is the identification of metrical units as either phonological or syntactical units. Both identifications are not automatically correct. On the one hand, there is the problem of syntactical units. I have argued that the syntactical unit is not necessarily equal to either the hemistich, or the dactylic hexameter. The syntactical unit can be constructed with the aid of smaller scale metrical phrases, with licenses that are reminiscent of clause composition in prose. On the other hand, there is the phonological phrase. This phrase is to be identified by its phonetically realised demarcation, a type of prosodic realisation that is not so readily identified. Of course, there is an identifiable stichic hexametric pattern underlying the Iliad and the Odyssey, but to the ear the stichic repetition may yield to phonological phrasing. Within the grammatically coherent clause, the verse end may be subordinated: the podic rhythm is maintained despite the period-final sixth foot. Intonation and maintenance of the podic structure contribute to clause constitution regardless of stichic metrical phrasing. The questions remains: how to identify phonetic disruption—audible pause?

The extent of grammatisation strengthens the supposition that Homeric epic represents a transitional stage in hexametric poetry: both prose composition (rhythm which subordinates meter) and poetic composition (metrical rhythm) are present: they exist next to one another, without any proof of either being the “original situation”. Prosody and syntax strengthen each other: grammar and syntax are furthered by means of phonological reorganisation; they do not commonly create any “mismatch”. The clausal grammar of the Homeric epic resembles, in part, the free, that is less metrical, clause formation of rhythmical prose. Metrical colometry remains identifiable by and large, but there is definitely no ‘beat’. [85]

The likeness to prose-composition stems from the way prosodically variable clauses and extra-clausals team up to form larger scale units of varying size. Within the grammatically coherent whole, the minor phrases of the extra-clausals and those of the clauses together build larger phonologically characterised “utterances”. The “utterances” themselves start from variable metrical positions, thus creating a patchwork of phonological phrases that reflects the sense of the perceptible units. I have shown the way the Homeric clause receives its sensible coherence from prosody. One could say, in fact, that sensible coherence is at times determined by prosody, both by metrical and phonological phrasing. 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 identify the corresponding phonetic disruption, the “audible punctuation”. Sources from antiquity comment more on phonology than on phonetics, so I have to turn to modern linguistic theory for help. Starting from the observations by modern scholars, I will identify phonetic disruption as a combination of the phonetic realisation and metrical position of the segment as a syllable.