1. The Unit of Homeric Discourse: Enjambment, Special Speech and Metrically Defined Grammatisation

For the modern reader, the punctuation of a text furthers comprehension of its contents. In spoken language, the person speaking may use differences in intonation, tempo and articulatory prominence to further the audience’s comprehension of what is being said. Punctuation in printed texts mimics to a certain extent the purpose of such audible differences in spoken language. It may, however, not cover all the nuances these differences add in spoken language.

Understanding spoken language demands an ability to interpret what is being transmitted that is different from the ability to understand a printed text. Experienced readers are trained in self-structuring the soundless mental process of reading through recognition and processing of various means of punctuation. The punctuation of a printed text makes the text accessible to a reader. To a certain extent, however, the accessibility created through punctuation is in accordance with the editor’s interpretation of the text. This applies to current editions of the Iliad and the Odyssey as well. Choosing a different edition of the Homeric text might mean, on the level of wording and syntax, choosing and reading a different text. An editor may consider critical mass in the various emendations and lectiones in manuscripts sufficient reason for publishing a new text. Still, he will be forced either to follow or to make choices on the text’s division in visible units. In editions of the Iliad and the Odyssey, [1] the use of the full stop is prominent, in alternation with other punctuation devices like comma and colon (raised stop).

The punctuation of the Homeric text does not receive much attention from commentators. The ancient scholia, especially those related to Nicanor, comment on rhetorical punctuation in a limited number of lines. [2] Comparison of the recent commentaries on the Iliad and the Odyssey shows that few if any critical remarks are made on the differences in punctuation between, and even within, various text editions. [3] At the same time, the punctuation which is applied to the text causes terminological obscurities. Modern studies on the prosodic, especially the metrical, shape of the Iliad and the Odyssey and their syntactical structure, show the problematic use of everyday words like “sentence” and “clause”. [4] It is my aim, as well as that of such modern studies, to illuminate the alleged disparity between the metrical and the syntactical structure. [5] It is equally a problem that an everyday word like “sentence” may not indicate for me what it indicates for others. Is a “sentence” syntactically demarcated? If so, where does it start? Where does it stop? Can “markers” separating syntactical units be identified and valued? And is the result then a visible division of the text into “units of sense” of some sort? Does the metrical structure influence the “sentence structure” of the Iliad and the Odyssey? Without a proper assessment of the units of Homeric discourse there is hardly a chance to even start illuminating any disparity.

What is the unit of Homeric discourse and what is its relation to the metrical structure? In the following sections, I will examine existing descriptions of Homeric syntax. Attempts to describe Homeric syntax have always paid much attention to grammatisation over the verse end. Due to the formulaic, oral nature of Homeric composition, grammatisation over the verse end inevitably leads to remarks on verse-end enjambment, a phenomenon that has drawn much more attention in the dactylic hexameter than in other metrical shapes that have long lines. Therefore I will start by recapitulating important studies in this field by Parry, Kirk, Higbie and Clark. I will also present criticism on the shortcomings of the approach to enjambment. Clark, however, seems to draw conclusions that seem to contradict the premise of the Parryan approach, which sees the single hexameter as the basic unit of composition. The descriptions of Homeric syntax by Chantraine and Bakker equally deal with the relation between the structural norms of hexametric poetry and the compositional norms as evidenced by semantic phrasing. Falling in with Chantraine and Bakker, I will then discuss the concept of grammatical completeness beyond their descriptions. Important differences in descriptions of Homeric syntax depend on the importance granted to the metrical circumstances of the text and the structuring impulse from meter.

The approach of enjambment as a special phenomenon in the Homeric epic [6] mirrors the importance that is granted to the single hexameter as the unit of Homeric discourse. Parry’s idea was that ‘the easiest formula for the oral poet to handle is that which is both a whole sentence and a whole verse’. [7] That such an approach stems from an implicit concept of syntax becomes clear from his remark that ‘the art of the oral poet is largely that of grouping together whole fixed verses’. [8] As Clark puts it: ‘the methods of composing oral poetry can be expected to produce a preponderance of lines consisting of one complete clause apiece’. [9] The Parryan focus on whole-line sentences has led to some commonly accepted conclusions. Parry writes: The whole-line sentence is seen as the basis of oral composition and as the traditional, compositional unit that provides the oral singer with his necessary formulas. More recent enjambment studies, like Higbie’s and Clark’s, agree with this analysis and approach. In their own subsequent approach to enjambment in Homer, they follow Parry in considering enjambment at the verse end to be the result of the whole-line formula expanding ‘beyond the limits of a single hexameter’. When discussing enjambed lines with unenjambed doubles, Higbie says: Clark agrees: Their acceptance of verse end enjambment in Homer is based on two premises. First: that, ever since the first lines of Homeric narrative were composed, the verse end has always been the strongest prosodic boundary in the hexameter used. Second: that at a certain, early, stage in the development of the Iliad and the Odyssey as oral narratives, the verse end was the syntactical boundary of choice. It was seldom, or perhaps never, crossed to continue the syntactically coherent unit into the next verse. At that stage, most, if not all, lines were end-stopped and consisted of one single sentence.

Trading the “sentence” for the “clause”, as Parry suggested and scholars working on his legacy adopt, widened the scope. Understanding “whole-line formula” to mean both whole-line “sentence” and whole-line “clause”, increases the total number of whole-line formulas considerably. Parry already uses whole-line “clauses” as examples [13] of whole-line formulas: [14] This exact line is used seven times in the Iliad, and fourteen times in the Odyssey. It indicates the end of the first part of the meal and signals the start of the second part, the conversation and entertainment. The frequency of the line in the Homeric narrative causes Parry to use it as an example of a formulaic expression, consisting of one “clause”, taking up exactly one verse. Higbie and Clark do the same when they refer to a ‘succession of eight whole lines’ as a ‘series of unenjambed verses’: Parry already suggests that the whole-line clause is more or less equal to the whole-line formula. This equation caused problems for later scholars. One of these problems is the “couplet”, verses that are grammatically or syntactically tied together. An example is the couplet of correlative formulas: [15] The couplet occurs again at Iliad 11.84–85. The lines from the couplet make their appearance separately at Odyssey 9.56 and Iliad 16.778. Clark [16] expects his readers to consider such correlative couplet formulas an addition to the list of whole-line formulas (both whole-line clauses and whole-line sentences or whole-sentence lines). He continues with presenting his readers with examples of longer passages consisting of “whole lines” or “unenjambed lines” (like Odyssey 21.269–276, cited above, and Iliad 8.143–152).

Higbie’s and Clark’s classifications of enjambment-types show to what extent the various types of enjambment are distinguished looking at the grammatical completeness at the enjambed verse end. In Higbie, we read: This approach, and the resulting classification, has proven to be a useful tool for the study of Homeric verse making and the appreciation of Homeric poetics from a reader’s point of view. The notion of “expectation” counts as an equivalent of “grammatical need”. Hence Iliad 7.394–397 can be described as a sequence of “skewed clauses”: [18] To date, Higbie’s classification of all the types of enjambment is the most elaborate; Clark uses it as his starting point. The development of Higbie’s system out of those by Kirk and Parry is best presented in a table: When Higbie uses the word violent to label a specific type of enjambment, her labelling makes clear that different types of enjambment ought to be classified based on what it is exactly that they separate. Violent enjambment (enjambment indicated as ^e in the Greek text) separates words that belong to a single word group: Necessary enjambment separates words that need each other to form a grammatically complete clause; [19] in other words, they belong necessarily to a single grammatical clause. The constituents of a coherent grammatical clause can be separated easily within the line. Such constituents consist of single words or word groups: Clausal internal enjambment [20] separates words that need each other to form a grammatically complete clause, just like necessary enjambment. The only difference is that clausal internal enjambment separates constituents that are tied together in a single grammatical clause by correlative adverbs: Clausal external enjambment separates grammatical clauses, the first of which is accompanied by a constituent suggesting hierarchical syntactical organisation in combination with the subsequent line. This means that the first clause is a subordinate clause, or that the main clause in the enjambed line prepares for the subordinate clause, or another main clause, by means of a correlative adverb. Two clauses thus tied together need not immediately follow one another: there may be more subordinate clauses, parentheses and independent main clauses in between: Adding internal enjambment separates words belonging to a single grammatical clause, though the clause met all its grammatical requirements before the verse end. What follows the verse end can be left out without affecting the grammatical completeness of the preceding line:

Finally, adding external enjambment separates grammatical clauses that turn out to be hierarchically organised only after crossing the verse end: the “sentence” met all its grammatical requirements before the verse end of what turns out to be an enjambed clause. As in adding internal enjambment, what follows the verse end can be left out without affecting the grammatical completeness of the preceding line: Higbie’s classification is based on a) observance of the verse end as an inevitable boundary, and b) the identification of the grammatically coherent unit with the metrical phrase.

Clark follows Higbie in not automatically allowing for emphasis on the first constituent of the Homeric hexameter following enjambment, not even if it is, grammatically, a mot-en-rejet. [21] In his 1997 publication, Clark uses the terminology enjambment as the acknowledgement that the verse end does not double as a syntactical boundary. His focus is not on grammatical completeness at verse end. He points at the frequency and ease with which Homeric composition is extended beyond the hexameter. In fact this extension takes place so often that a model can be reconstructed in accordance with which specific run-over words and verse-final “anticipations” provide the composer with formulaic material to continue his clause or sentence with. A runover or an anticipation (for example the verse-final constituent following position 8, the bucolic dieresis) is part of a formulaic verse: using the runover-word or the anticipation always extends the formulas used backwards or forwards towards the nearest verse end. A few examples from Clark 1997: Just like the formulaic material that can be used to compose whole-line verses or parts thereof, the runovers and anticipations of Clark’s model provide formulaic material to compose beyond the boundaries—the metrical, but from a diachronical point of view the syntactical boundaries as well [22] —of the individual hexameter. Runovers and anticipations function as semantic “hooks” to link formulas over the verse end.

Enjambment at verse and is a way out of the disparity between metrical structure and semantic phrasing in a system of whole-line formulas. Grammatical completeness at verse end defines the “strength” of the resulting enjambment. Grammar decides whether we hear violent enjambment due to the breaking up of a preposition group in hyperbaton, or necessary enjambment since ἐν serves as an adverb: Especially interesting are instances of violent enjambment that feature an attributive adjective at verse end. The undisputed examples are: The Iliad, however, provides several other, similar examples from the point of view of Higbie’s system: There are many more examples of enjambment resembling the few quoted above, which are not counted as instances of violent enjambment by Parry’s followers. In such cases the adjective is numerical or partitive/ablatival, like πᾶς, πολύς, ἄμφω, ἄλλος, ἄκρος, and θαμύς. The adjective may be used as a noun, judging solely from the line ending in enjambment. The adjective πᾶς for example, is used attributively in Iliad 1.78 but acts as a noun in Iliad 1.122: Parry uses Iliad 1.78 when he explains the resulting enjambment. In his terminology, the transition from 78 to 79 is an example of “necessary” enjambment: Parry’s estimate of this phenomenon’s frequency in Homer was based on a limited number of lines (Iliad 1.1–100, 5.1–100, 9.1–100, 13.1–100, 17.1–100, 21.1–100, Odyssey 1.1–100, 5.1–100, 9.1–100, 13.1–100, 17.1–100, 21.1–100). His focus on the whole-line formula led to the conclusion that an enjambed line ending in an attributively used adjective, is in fact an expansion of an original whole-line formula ending in an adjective used as a noun. That explains the preponderance of adjectives like πᾶς, πολύς and ἄλλος. Surely, Parry would not have objected to Higbie’s classification “violent” for Iliad 22.410–411: The adjective ἅπασα has no “meaning” until it is joined with the noun Ἴλιος. Modern commentators struggle with the appropriate place of these examples in the enjambment-system. [28] Rightly so, as the approach of these examples makes clear that, if Higbie’s classification of enjambment is used, lines with similar features receive different labels. A transition like Iliad 9.74–75 should be labelled “violent enjambment” in Higbie’s system: The adjective ἀρίστην does not have a “meaning” until it is joined to its noun βουλὴν in the next line. What is more, the adjective has no understandable semantics at all unless it is joined to βουλὴν. The explanation for the origin of this type of enjambment still goes back to Parry’s. [29] His suggestion to take a closer look at instances where the adjective was a numerical or partitive/ablatival one that might be used as a substantive, proved to be an incentive for his followers to try and limit the frequency of violent enjambment. It looks as if they try to go to great length to make sure that the compositional building block of the Homeric epic remains limited in size to the single hexameter: every type of run-over runs the risk of being labelled as some sort of aberration. It can be easily shown, however, that this approach is incorrect.

If Parry’s view were right, Homer would not only present his audience, as he does, with many lines ending in a form of πᾶς, πολύς, ἄμφω, ἄλλος, ἄκρος, or θαμύς. If such lines were the compositional units of his poetry, the natural way to expand on them would be by means of a subsequent, verse-initial noun that agrees with the adjective in “adding” enjambment. But such “adding” hardly ever happens, save for a rare example: Lines ending in a form of πᾶς, πολύς, ἄμφω, ἄλλος, ἄκρος, or θαμύς, and with a clause that is grammatically complete at the end of the line, are very rare. Normally, if a line ends in a form of πᾶς, πολύς, ἄμφω, ἄλλος, ἄκρος, or θαμύς, and the subsequent line features a noun which agrees with the adjective, [30] the verse-end enjambment is more than merely adding. In most instances, the enjambment is at least necessary: The appearance of the noun as the verse-initial word is the result of out-of-line grammatisation. The difference between Iliad 16.657–8 κέκλετο δ᾿ ἄλλους || Τρῶας ‘and he called out to the other | Trojans’ and Iliad 19.262–3 οὔτε τευ ἄλλου || ἀλλ᾿ ‘nor of something else, | but’ is a difference in grammar not a coincidence due to appositional alignment of verses. Remarkable are the ease and the frequency with which the clause that enjambed the verse end continues with hyperbaton of the adjective and its awaited noun: Several more examples, below, will make clear that Parry’s attempt to reduce the frequency of “unusually harsh” necessary enjambment (Kirk 1966 and Higbie 1990: violent enjambment) misleads his followers. I consider it very unlikely that these instances are ‘due to a chance interplay of formulas’ [31] in Homer.

An extraordinary example of alleged violent enjambment is Iliad 16.104–105: the clause contains two adjectives (δεινήν and φαεινή) both separated from their nouns (καναχήν and πήληξ) by the verse end: This line has a feature in common with necessary enjambment: one of the essential elements of the clause, the verb, appears only after the enjambment. With only one exception, this is always the case when scholars quote examples of violent enjambment. The only exception is Iliad 13.611–612, where an essential element, other than the verb, is postponed till after the enjambment: The clause in line 611 has all the elements required for grammatical completeness except for the object. The object’s appearance—and only the object’s—is prepared for by the adjective: the audience will expect it to immediately follow καλὴν, even more so as grammar at the verse end of 611 does not require any other elements or words (in other words, as the enjambment is not merely “necessary”). If verse-end enjambment were to be taken for granted as the acknowledgement of run-over, a better name for this type of enjambment would be “noun-preparatory necessary enjambment”—both to distinguish it from the broader definition of violent enjambment as used by Higbie and others, and to bring it in agreement with the basic feature of enjambment again: the expectations of the audience.

For all the other examples of “unusually harsh” enjambment, the choice between labelling “noun-preparatory necessary enjambment”, “violent enjambment” and simply “necessary enjambment” would be a tough one. More new technical terms, solely expressing the audience’s expectations, would be desirable. At the same time, sufficient reason remains to question the distinction between various types of verse-end enjambment.

The metrical position that most obviously suffers from somewhat lessened importance as a grammatical boundary is the verse end: any poetic effect resulting from verse-end enjambment is not to be assumed automatically in Homer. “Noun-preparatory necessary enjambment” is actually the easiest way to do away with the automatism of poetically effective verse-end enjambment. Some examples, again with lines ending in πᾶς, πολύς, ἄμφω, ἄλλος, ἄκρος, and θαμύς, will illustrate this. Iliad 6.498 features what Kirk considers an “awkward enjambment”: [32] Kirk points out that other verses ending in a form of πολύς are smoother, causing Higbie to label this type “violent”. Kirk wonders why an adverb like ἔνδοθι ‘inside’ has priority over the object-noun: the clause as it stands might have been grammatically completed within the metrical phrase. Verse 498 is grammatically complete (‘and inside she found many’) and might have been complete in meaning if Andromache was expecting only one type of women inside the house at this time. Was she? The question of course is not what Andromache might expect to find, but what the listening audience was expecting. There may have been a role for intonation here, but any role of intonation must be seen in the correct perspective.

Intonation belongs to the performer, but its use is the performer’s way to guide the audience’s expectations. Intonation, however, cannot be seen as unrelated to the structure of language. Any expectation that the performer creates using intonation is limited by the pattern of intonation within the units of Homeric discourse. Any remark with regard to the units of discourse is in a way a remark that takes some sort of independent intonation pattern for granted. The pattern of intonation is not necessarily the key to phrasing, but it surely is a side effect of phrasing. What expectations may the performer’s use of intonation have evoked in Iliad 6498–499? What does it learn us about the strength of enjambment? Intonation might have made the audience expect continuation of the clause, possibly a noun agreeing with ‘many’, but the actual noun could hardly have been a surprise. Here, “noun-preparatory enjambment” is not “strong”, nor “emphatic”: [33] there might have been noun-preparation, but there is no poetic effect to the enjambment. What is “awkward” about πολλὰς || ἀμφιπόλους (to use Kirk’s qualification) is that it resembles the continuation of a clause regardless of the metrical phrase: the way it is found in prose. There appears to be nothing special going on with the words preceding and following the verse end: Homer simply continues the clause into the next line. Trying to explain the reader’s disturbance in terms of enjambment here results in failure to explain what is “awkward” about it. Something similar happens whenever the choice of the verse-initial noun will not have been a surprise: In terms of giving emphasis or building up tension, such enjambment is no less than disappointing. The clause simply continues into the next line, despite the grammatical completeness at the straddled verse end. Another example is Iliad 13.797–799, where the Trojans are compared to a storm: Line 798 supplies only one requirement for grammatical completeness, the subject. The predicate is missing, so ἐν is to be understood as ἔνεστι. [35] The audience is waiting for the subject and expects the first word of line 798 to be it. Whether or not intonation suggested more to come, the first word, the subject κύματα, is not likely to be a surprise for the audience. Expectations will probably shift from the verse-initial word to the remainder of the line as the audience is eager to hear what will follow this rather dull location of the subject. Three of the four words in line 798 appear four times in epic together as a word group, not interrupted by other words. The fourth word, παφλάζοντα, is separating the word group this time, and happens to be a hapax legomenon in Homer. [36] “Enjambment” does not seem to disrupt the structure of the clause as a grammatical unit: the clause simply continues into the next line. Enjambment does not create special expectations involving the first word of line 798. There is nothing special about its first word. If intonation strengthens the phrasal structure of the developing clause, the audience’s attention will focus on the remainder of the line, though grammar only prepared them for the awaited noun. Is it a coincidence that the second word in this line is a hapax legomenon? Either way, there is no poetic device realised as a result of what Higbie labelled “violent enjambment”. Consider the following examples: The adjective θαμειάς remains meaningless until joined with the noun in the next line. The choice of the noun αἰχμάς is not a complete surprise for the audience. The addition, however, of the superfluous ἐκ χειρῶν in Iliad 12.44 would weaken any “emphatic” use of αἰχμάς. What the addition shows and underlines is that apparent “noun-preparatory necessary enjambment” is not a mot-en-rejet.

The missing noun may be accompanied by a relative clause: Grammatically, line 591 λίσσετ᾿ ὀδυρομένη καί οἱ κατέλεξεν ἅπαντα is complete, but 592 continues with a verse-initial noun agreeing with ἅπαντα. The remainder of line 592 is expanding on the “sorrows”. Line 592 is an example of what Clark identifies as a “hook”, resulting in out-of line composition. The composition, however, is not just about continuing the developing clause: the continuation develops into a line that is not grammatically complete by itself, and depends grammatically on the preceding verse. Was it clear to the audience that there would be more to the clause than just the noun “sorrows” as soon as they heard [37] it follow the ἅπαντα of line 591? Was the intended expansion already audible at the end of line 591? Looking at the examples cited we may safely assume that intonation at verse end in Iliad 14.422–423 was slightly different from that in Iliad 12.44–45. In the latter example, the addition ἐκ χειρῶν has little to do with the audience’s expectations: the audience is being prepared for the verse-initial noun (αἰχμάς), but they hear that the clause does not end with it. At the same time they come to know that the “addition” is not a grammatically required element. The same holds for Iliad 9.591. Could the audience hear that what would follow the verse-initial noun was merely an “addition” in Iliad 9.592?

A completely different (but, I hope, clarifying) situation is found in Iliad 12.51–52, a couplet of lines that is generally understood as being divided by violent enjambment: [38] Opinions about this line, however, may vary. In Iliad 12.52, the actual meaning of the verse-initial noun is in accordance with the audience’s expectations: they already heard that Hector encourages his comrades to try and cross the τάφρος in line 50, but now his own horses do not dare to leap the gap. If this enjambment is to be considered “violent”, the translation should be ‘but they whinnied loudly, standing on the outermost, the outermost of the rim’. Parry could have used these lines as an example of “non-periodic enjambment” (Kirk 1966: “progressive”; Higbie 1990: “adding”) following a form of ἄκρος used as a substantive. The clause is grammatically complete at the end of line 51: ‘but they whinnied loudly on the outermost (, standing on the rim)’. [39] But the clause is in fact not brought to completion. The word group ἐπ᾿ ἄκρωι χείλει ἐφεσταότες may well constitute a syntactical unit, predicate-centred around the verbal form ἐφεσταότες. Any effect of the use of the participle ἐφεσταότες on intonation [40] would already have been noticeable and audible from ἐπ᾿ ἄκρωι, that is, before the verse end of line 51. It is hard to think of verse-end crossing intonation as a coincidental result of appositional alignment of verses: alignment as we find it here is not appositional but rather due to the pre-existing possibility of verse-end crossing units. Thus the audience not only expects a noun to go with ἐπ᾿ ἄκρωι but something else as well. Grammatically, Iliad 12.44 and 12.51 both raise the same expectations concerning the first word of the subsequent line, but contrary to Iliad 12.44, Iliad 12.51 raises very different expectations melodiously. If enjambment prepares the audience for more to follow than just a verse-initial noun to agree with the adjective, it is inappropriate to try to classify it as “noun-preparatory necessary enjambment”. Intonation can be taken into account when determining “types” of enjambment when the types are judged in relation to phrasal structure. The content of the subsequent line must be judged by standards other than “grammatical requirement”.

In cases of what Higbie describes as “violent enjambment”, the enjambment is usually “necessary” as well, as shown above. Often the finite verb or another required verbal form is also postponed until after the enjambment. If at “violently enjambed” verse end the line ends with an adjective, the accompanying noun, as a rule, is verse-initial. Exceptions exist, especially in cases involving the adjectives πᾶς, πολύς, ἄμφω, ἄλλος, ἄκρος, and θαμύς. Thus, lines like Iliad 24.163–164 show what appears to be standard practise: The adjective πολλή has no meaning until it is joined to the noun in the next line. Then the finite verb follows with κεφαλῆι τε καὶ αὐχένι τοῖο γέροντος added, though not necessary for grammatical completeness. Of course, this addition is an important piece of information: without it, the grammatically complete clause (‘all around there was a lot of dung’) would be pointless. Enjambment brings κόπρος in verse-initial position, but the attention of the audience is immediately drawn to what follows κόπρος: the finite verb. Something similar happens in other verses despite the fact that the first word following the enjambment is all but spectacular: The choice of the noun (‘Greeks’) annuls, I think, any poetic effect of violent necessary enjambment. The effect of the non-violent necessary enjambment is smoothened by the choice of the finite verb (‘I ordered’): it immediately incites the audience to focus on what will follow the finite verb itself. In Iliad 15.80–81 the finite verb is postponed until the end of line 81: The prepositional phrase ἐπὶ πολλὴν awaits its noun; intonation prepares the audience for the noun to be the next word. The audience does not know what noun to expect in this simile, but as soon as it hears γαῖαν its attention will again be drawn further. The audience’s attention focuses on a word—awaiting the finite verb—that will link the imagery of this clause with the central imagery of the simile proper: the νόος ἀνέρος. But the finite verb is postponed. First by the participle that concludes the phrase ὅς τ᾿ ἐπὶ πολλὴν γαῖαν ἐληλουθὼς. Then by the word group φρεσὶ πευκαλίμηισι that ends a line twice in the Iliad, but is ‘moved from the verse end’ [41] here, as in Iliad 20.35. When the finite verb νοήσηι finally appears, the need for grammatical completeness might have prepared the audience for something following the “necessary” [42] enjambment. What follows is a unique quotation in Homer: To the ear, the demarcation in intonation resembles the audible pause of the colon (:) read aloud. There seems to be a connection between two successive cases of “enjambment” in two successive verses; in this instance an accumulation of the audience’s expectations resulting in a surprising continuation after the second “enjambed” verse end. The mere suggestion of such a connection is not plausible, and almost unexplainable, in a model for Homeric verse-making that is based on the assumption of the whole-line formula as the compositional unit, and the subsequent explanation of enjambment as ‘expanding beyond the hexameter’. It also runs counter to the idea ‘that the length of the enjambment can vary from a single syllable to a complete line’. [45] In this case it develops into a second “enjambed” verse end and hence another subsequent line.

The verb the audience is waiting for is not always the finite verb. Consider Iliad 24.113–114, repeated with a few modifications at Iliad 24.134–135: At first sight these lines seem to illustrate Parry’s idea that everything following the enjambment is “added”: line 113 is a complete clause (or rather, two complete clauses: ‘tell him that the gods are angry with him and, most of all, that I am’, understanding ἐμέ as the second accusative with the infinitive σκύζεσθαι). In Higbie’s system, this can hardly qualify as violent or necessary enjambment: the meaning of πάντων is clear thanks to θεούς in the same line, and the clause ἐμὲ δ᾿ ἔξοχα πάντων is readily understood as an equivalent of σκύζεσθαι οἱ εἰπὲ ἐμὲ ἔξοχα πάντων. But this description might be nowhere close to what the audience heard and expected based on line 113. If intonation prepared them for a noun agreeing with πάντων to be the first word of the next line and for something else to follow this noun, they might have been waiting for an infinitive to which ἐμέ is the subject-accusative. In other words, any preparatory effect of intonation makes the line end in “violent necessary enjambment”. As in previous examples, such “enjambment” is not likely to emphasise the first word of the next line: ἀθανάτων. The noun is by no means surprising or even meaningful. If enjambment is capable of protracting the tension felt by the audience, the word to receive emphasis is the infinitive κεχολῶσθαι, not the noun ἀθανάτων. Is it? The infinitive perfect κεχολῶσθαι does underline the appropriateness of ἔξοχα πάντων compared to the other gods’ σκύζεσθαι. But how much emphasis remains if Iliad 24.113–114 is compared to Iliad 16.657–658? Here, the clause κέκλετο δ᾿ ἄλλους is grammatically complete; the adjective at the end of the line does not need a noun; the verse-initial noun is not surprising, nor meaningful; intonation may have prepared the audience for a verse-initial noun and for something extra. But compared to Iliad 24.113–114, the infinitive in Iliad 16.658 is as superfluous and meaningless as the noun following the enjambment. Nonetheless, the infinitive and the noun are probably both required melodious elements: as intonation is related to phrasing, the crossing of the verse end by the phrase can hardly have been without some sort of verse-end crossing intonation pattern. Seen as an audible phenomenon that is related to the structure of language, intonation, rather than the grammatical completeness of Iliad 16.657, would be a reason to label any enjambment “violent necessary enjambment”. While connecting Iliad 16.657–658, “enjambment” does not contribute to the tension in the clause. It does not emphasise one of the elements of the clause following the verse end of 657. In fact, allowing a verse end “pause” between lines 657 and 658 while reciting would only unnecessarily disturb the audience’s perception and enjoyment of the clause, and practically annul the ‘running a word group over from one verse into another and thus dimming the rhythm of the end of the hexameter’. [46] Similarly, a “pause” following line Iliad 24.113 is only effective in that it creates an awkward, disturbing break between two words and between two parts of a grammatically coherent clause.

As illustrated above, intonation is best considered a feature of spoken language that is not unrelated to the units of discourse. Intonation is thus a factor in the judgement that grammatical completeness of the line ending in enjambed verse end is not by itself decisive for identification of enjambment types. This becomes particularly clear in verses where the grammatical structure of the straddling clause at verse end leaves room for several options for continuation. In Iliad 14.35–36 emphasis (if any) in the description lies on the large number of ships and the hardly sufficient space on the beach to draw them all up: The verb ἔρυσσαν indicates the soldiers’ action of drawing the ships onto dry land. Subsequently, the verb πλῆσαν indicates a soldiers’ action as well. [47] Based on the grammatical completeness of line 35, the audience might understand the soldiers to have filled the ships with something “complete”. [48] I do not believe that the poet deliberately misleads his audience into thinking that the soldiers had filled the ships with something, as readers, taking one verse at a time, might understand at first. Instead, intonation must have prepared the listeners for more to follow in the next line than just the noun agreeing with ἀπάσης. It turns out that mentioning ἡιόνος only further postponed an object. Despite this unexpected twist of grammar, the object does not render a very special or unexpected idea; even less so, since the actual object (στόμα) has some semantic overlap with the noun in the verse-end straddling word group (ἀπάσης ἡιόνος), [49] which is not very special or unexpected either, in spite of “violent enjambment”.

I argue that interruption of a word group by the verse end only proves to be another way of connecting two verses, so that the clause may continue with as little interruption as possible. The verse end is not used as a poetic device here to obtain a certain effect that can not be achieved with the metrical or rhythmical opportunities available within the verse. The verse end is being ignored or muted to allow the clause to “run over” (rather: “continue”) into the next line. In Iliad 14.35–36 the breaking up of an adjective-noun word group is not an intentional creation of enjambment for a certain poetic effect. Nor is it a deliberate use of the verse-end enjambment to highlight the adjective or the noun. Rather it is the result of using the reduced emphasis on verse end as a metrical break to create a clause of the length desired by the poet: unhindered by the metrical boundary that is the verse end.

Other examples lead to the same conclusion. Consider Iliad 9.97–98: The adjective is the last word before the straddled verse end and the agreeing noun the first of the next line. The finite verb of the clause (ἐσσί) is in the line following the enjambment too. In this case, another noun, used predicatively, is also in the second line. The adjective πολλῶν could have been used on its own, without a noun. Here, there is a noun following the adjective in “violent necessary enjambment”. The audience hears intonation prepare for a noun to immediately follow the adjective, but the meaning of the noun is in no way a surprise. In fact, after hearing the noun λαῶν the audience waits for the finite verb that has been postponed, seemingly without urgent need, through the use of the noun. Having heard ἐσσί the following ἄναξ finally gives the audience the information it has been waiting for: Nestor addresses Agamemnon as the king, only to remind him that ‘kingship has its duties.’ [50]

In the same way, the adjective φίλτατος explains to the audience why Hector takes Apollo’s advice when he is disguised as Phaenops: Apollo carefully impersonates a very close friend of Hector to make sure that his encouragement will have the desired effect: the audience waits for the poet’s explanation of Apollo’s choice. That Phainops is actually a ξεῖνος is not the most important piece of information this clause has to offer. The fulfilment of the audience’s expectation for the verse-initial word cannot have been a surprise. As before, verse content must be taken into account. Compare Iliad 5.1–3, where Pallas Athena supplies Diomedes with extra strength: True, Diomedes will enjoy the consequences of this extra strength throughout the book that is about to start, but the audience is not being prepared for what will follow, neither will they remember the enjambment in these lines as particularly significant once they have heard all Diomedes’s exploits in Iliad book 5. It is noteworthy, however, that the line ending in enjambment at verse end contains the non-formulaic μένος καὶ θάρσος and the Homeric hapax legomenon ἔκδηλος. Again, the “enjambment” seems to exert only a superficial influence on the words following it, but appears in combination with noteworthy phenomena preceding it in the same line. This is a fine example of the licence to compose “free-size” clauses overruling the “predominance of whole-sentence lines” thanks to metrical boundaries, an issue I will discuss in more detail in the final section of this chapter.

In cases of “violent necessary enjambment”, the noun agreeing with the verse-final adjective is not always verse-initial, thus resulting in hyperbaton. The verse-initial word may be the finite verb: Alternatively, the verse-initial word is one of the other remaining requirements of the grammatically complete clause: The verse-initial word of 34 is another required element in the clause, in this case a very special one: αἰγιαλός stands in hyperbaton (εὐρύς περ ἐὼν … αἰγιαλὸς) due to the verse-end straddling word group (πάσας … νῆας) that stands in hyperbaton itself. There are only two other examples in Homer of verse end allowing a double hyperbaton. The first one is the already cited Iliad 16.104–105, but at least there one of the divided word groups was divided only by the verse end (φαεινὴ || πήληξ). The second one is Iliad 1.283–284: It is tempting to understand μέγα as an adverb (‘highly’) here, as it is used elsewhere: On several, similar occasions the adverb is more closely attached to the verb. [51] If Iliad 1.78–79 is seen as a fortuitous combination of formulas, the translation may be more like ‘who, high above all, rules over the Greeks, and whom the Greeks obey’. Such a translation is in accordance with a line-to-line approach as well. Still, neither Iliad 1.283 nor Iliad 1.78 will have left any doubt for the audience as to how μέγα was to be understood. For the poet and his audience, there must have been a significant difference in intonation at the end of the two verses. The latter prepares for the next line to continue the grammatical clause unhindered. The former includes the metrical break at verse end in the application of the hyperbaton. In other words, not only the audience was likely well aware of the grammatical function of μέγα as an adjective in Iliad 1.283 but the poet as well.

So, does the poet at least try to keep the words of a word group that is divided by verse end as closely together as possible? Does he only make an exception for the verse-initial position in order to keep it safe for another grammatical requirement? Does the poet aim to keep the words needed to express the grammatical requirements of the clause as closely together as possible when the clause straddles the verse end? Obviously not: the ease with which hyperbaton is used in the following examples of “violent necessary enjambment” demonstrates, I argue, Homer’s ability and freedom to compose a grammatically complete clause despite the verse end: Iliad 14.98–199, 24.83–84 and 24.98–99 have all the characteristics of “necessary” enjambment up until and including the first word following the enjambment. The rest of the line (‘immortals and mortal men alike’, ‘sea-goddesses … in an assembly’, ‘ever-living, blessed gods … in an assembly’), looks like an addition, from the point of view of grammatical completeness. Semantics and intonation surely run counter to such analysis. With the exception of Iliad 15.121–122, all examples allow hyperbaton in favour of another grammatical requirement (finite verb) to occupy verse-initial position. The hyperbaton allows other, grammatically non-necessary elements to be inserted in between as well (εὖ Iliad 15.412; χαμαί Iliad 21.181; ἐπισσώτρων Iliad 23.505; ὁμηγερέες ἅλιαι Iliad 24.84; ὁμηγερέες μάκαρες Iliad 24.99).

Hyperbaton involving “violent necessary enjambment” also arises from the position of the adjective in the line ending in enjambment at verse end. Iliad lines 1.283–284, 14.33–34 and 16.104–105 have been quoted before as examples: In Iliad 15.84–85 the adjective ὁμηγερέεσσι stands in hyperbaton with the agreeing noun, but is not the last word of the line, nor is the noun the first of the next: In his commentary on Iliad 13–16, Janko points out that the word group ἀθανάτοισι θεοῖσι is ‘transposed from the verse-end’, [52] where it can be found 38 times. An explanation for this transposition will also shed light on the nature of the enjambment at the end of line 84. For now, it has to suffice that I point out that hyperbaton allows Homer to create a grammatically complete clause of the desired length without the verse end being a hindrance that forces the poet to complete the syntax of the clause within the boundaries of the single line. As noted before, the syntax of the line does not even have to be completed as soon as possible after the enjambment. Again, the strength of the verse end as a boundary—syntactical boundary this time—does not differ very much from boundaries that appear frequently within the line. What is more: Homer’s compositional technique is not simply ‘adding lines, one after another’. It may require preparation in the line ending in enjambed verse end before any “adding” takes place. Homer’s technique is rather one that allows for liberally using (combinations of) successive recurring metrical patterns to express a well-prepared and coherent thought. He does not have to fit his thoughts in with successive recurring metrical patterns like the single hexameter.

The Homeric technique presupposes that verse makers originally used a wide variety of “compositional units”, only one of which was the whole line. Descriptions of Homeric syntax so far took little notice of composition over the boundaries of the single hexameter. Not that they deny it exists. Of course it exists, but it is treated as non-standard. [53] In this section I have provided arguments for, and examples of, grammatisation despite the verse end. Some of the conclusions that are drawn in Clark 1997 contradict the premise of the Parryan approach, but his treatment of out-of-line composition is an attempt to provide a model for certain types of composition beyond the hexameter. Clark himself stresses that “enjambment remains the basis of the study,” [54] and hence the concept of the whole-line formula. My examples in this section, however, justify the search for discourse units other than the single verse for at least a large number of Homeric lines. The description of Homeric syntax by Chantraine and Bakker provides several other units of Homeric discourse.

Within the verse, Homeric discourse seems to feature various units. The variety of discourse units is the result of varying levels of syntactical organisation. In many verses, the Homeric epic features a syntax that seems to reflect a high level of grammatical segmentation. Chantraine [55] describes Homeric syntax as characterised by apposition and autonomy. Apposition, because Homer’s utterance seems to take shape through constant adding: adding of subjects, attributes, epithets, and whole clauses. In his analysis of Homeric syntax, Chantraine uses the description “construction appositionelle”: This appositional character of Homeric syntax is evident from the paratactic organisation of the discourse. Devine and Stephens consider the ‘rather flat, as opposed to hierarchical, phrase structure’ [56] of the Homeric epic as illustrative for the “nonconfigurationality” of ancient Greek, the fact that there is hardly any hierarchical phrase structure. Characteristics of nonconfigurationality [57] can indeed be found everywhere in the Iliad and the Odyssey. I will list eight of these characteristics, and illustrate each with one or more examples from the Homeric epic:

1. (grammatically) free word order:

2. pronouns that are implied in the verb: [58]

3. dislocation of constituents and their replacement by clitic pronouns:

4. failure of agreement: [59]

5. restricted use of prepositional phrases. In Homer, the “adposition” is regularly better analysed as an adverb, [60] or as a morphological prefix to verb or adjective:

6. lack of definite article. In Homer, the definite article has “embryonic status”: [61] when applied, it rather functions as demonstrative. [62]

7. comitative adposition in stead of noun phrase coordination. Nouns can be coordinated to be the subject of a verbal form, together: ‘A and B walked to school’: In Homer, such noun phrase coordination has an alternative in comitative adposition: ‘A walked to school, and so did B’ / ‘A walked to school, and B with him’ / ‘A walked to school, ánd B’:

8. parataxis. In Homer, parataxis is most remarkable in instances of grammatical prolepsis, when a constituent is put first, separated from its clause by an intervening clause. Thus the constituent is “externalised”, isolated from the clause in which it had an important, but unmarked semantic function: The syntax of the Homeric epic is quite different from the syntax found in post-Homeric literary works. As opposed to the syntax in poetic writings of, for example, Hellenistic poetry, or that of prose writings of, for example, Plato, Homeric syntax seems to be rather flat and paratactic. [63] It consists of highly autonomous words, word groups, and clauses. The autonomy on these various levels is represented by the description of Homeric style as adding style: adding more word, word groups, and clauses is almost always possible, but the additions themselves are hardly ever necessary for proper understanding. Single words define their own role in the discourse, [64] and grammatical governance is not the standard.

The analysis of the syntactical structure of the Iliad and the Odyssey is based on the implicit assumption that the syntactical structure of classical Greek prose may serve as the starting point. In recent years, more objections have been made against this comparison of the Homeric epic with the syntax of classical Greek. Starting from Chantraine’s work in the field of Homeric syntax, Bakker, in a series of publications (Bakker 1990, 1997a, 1997b, 1999, 2005), discusses a large number of syntactical and stylistic issues in Homer, like enjambment, discourse markers and the sentence.

For Bakker, the starting-point of the study of Homeric syntax is the study of Homeric discourse as speech, or spoken language. With “spoken language” Bakker refers to the medium aspect of language, as distinct from the conceptional aspect: language is about linguistics (the conceptional aspect) and communication (the medium aspect). He agrees with linguists like Halliday & Hassan 1976, and Chafe 1994, who, in the words of Oesterreicher: The conceptional aspects of orality can be illustrated by marking the differences with the conventions of written discourse. Doing so for Homer’s Iliad and Odyssey would suggest that both epics were composed in an environment in which there were written texts as well as oral compositions. Since it is not probable that this is the case, Homer’s “oral style” is not the counterpart of written literature. [66]

Parry, whose work [67] made the concept of Homer as oral poetry widely acceptable, focused on the compositional aspects of the formula as oral poetry’s building block. [68] His observations have nourished the persisting notion of “primitive parataxis” [69] as a description of Homeric syntax, but Parry already showed that the formula is not the product of naivety. [70] Still, for Parry the conventions of written literature were the touchstone to judge the syntax of the Homeric epics. Bakker, on the other hand, stresses the receptional aspects of the language of the Iliad and the Odyssey. As he analyses Homeric discourse as spoken language (speech), he applies the findings of general discourse analysis, as presented by, among others Chafe 1994, to the Iliad and the Odyssey.

In the analysis of Chafe, spoken language is the verbalisation of consciousness. The cognitive terminology “consciousness” is defined as focus on a little piece of what the brain knows in relation to the egocentric model of the environment. [71] This focus changes constantly, and rapidly. In time, consciousness is a mental process consisting of a flow of foci of consciousness. Spoken language reflects the speaker’s mental processes, and, inevitably, the constraints of consciousness. Most important of these constraints is the one new idea constraint. [72] This cognitive constraint limits the size and the span of verbalisation to the single focus of consciousness. The resulting “one-idea-unit” is ‘most typically characterized prosodically’ according to Chafe. [73] Based on physically realised properties of the unit like rhythm and intonation, Chafe coins it intonation unit.

In Chafe’s view, intonation units consist of four to five words, [74] and their beginning and end are marked by intonation boundaries. The intonation boundary may or may not coincide with exhalation and inhalation. In verbalisations of foci of consciousness in spoken language, the intonation units appear as the spurts that, as shown by the study of the phenomenon in modern spoken languages, characterise unplanned speech. [75] Spoken discourse is a process with a very communicative purpose. Characteristic for unplanned speech on the conceptional level, however, is that its syntax not so much organises, but ‘reflects the speaker’s mental processes’. [76] As these processes equal shifts in focus of consciousness, spoken language is “chunked” both from the receptional (audible “spurts”) and the conceptional point of view (lack of larger scale syntactical organisation). The syntactical structure of spoken language often resembles, at least partly, a “checklist”. [77]

In applying the findings from linguistic research as Chafe’s to the Iliad and the Odyssey, Bakker shows that the Homeric epics have much in common with unplanned spoken language chunked into intonation units by discourse analysis. As an example, he presents Iliad 1.1–7 as chunked text (Bakker 1997b:291–292): In Bakker’s analysis, the text of the first 7 lines of the Iliad is chunked ‘with the units into which the passage easily divides’. [78] The resulting units correspond to Chafe’s intonation units according to the one new idea constraint. The syntax of Iliad 1.1–7 corresponds to the audible structure of the spurts: it resembles a checklist, in this case a “preview”. Bakker calls Iliad 1.1–7 ‘a series of island-like ideas’, [79] with “orientation” as main purpose. The checklist-structure can be found in many verses in the Iliad and the Odyssey, often in the shape [A – verbal form – B || addition to A || addition to B]. The apparent checklist-structure is further strengthened by the autonomous usage of words and word groups: words define their own role in discourse independent of “governance” by other words; intonation units can often be considered elliptic clauses. [80] Autonomy of words and word groups results in appositional style—as can be expected in unplanned spoken language. Bakker asserts, however, that the discourse of the Iliad and the Odyssey cannot be fully put on a par with spoken language in general. The use of formulas and the restrictions due to meter turn the spoken language of the Iliad and the Odyssey into “special speech” or “marked speech”. [81] The intonation units that Bakker identifies are identical to the metrical cola that Fränkel [82] identified as the structural units of the hexameter. Bakker thus puts the structural unit of meter on a par with the compositional unit of discourse. The metrical shape of the Homeric epic is to be considered as a presentation of intonation units, be it a stylised presentation. [83]

Bakker’s description of Iliad 1.1–7 as ‘a series of island-like ideas’ underlines his doubts concerning the usefulness of the terminology “sentence” for the description of Homeric syntax. In an earlier article [84] he states: The concept of “sentence” seems hardly fit to describe the way in which the Homeric discourse is organised internally. Comparative studies have shown that the principles guiding the adding-style verse-making in Homer are similar to those underlying oral poetry from other centuries and cultures. Like these other products of poetry, the Iliad and the Odyssey present their narrative to the listening audience in a chain-of-thought-style. This style, that resembles a checklist, is the result of the chunking of information, as described above.

Bakker points out [85] that the localisation of the chunks of information fits the metrical patterns of cola between positions of frequently occurring word end. [86] In his view, there is a correspondence between the cognitive restraints on the intonation unit, and the size of phrases. In poetry, stylisation regularises the phrases, the metrical cola of the single hexameter: That is why the units that Bakker mentions are “stylized into metrical properties”. [88] The syntactical “segmentation” of the Homeric discourse can hence be clearly felt. The units that are strung together to represent Homeric syntax are still recognisable as individual intonation units; their shape responds to frequently used metrical patterns. [89] The stylisation of intonation units into metrical properties visualises Homeric discourse as the equivalent of metrical colometry:

The requirements of grammar are secondary to the existence of the linguistic units: grammar “emerges as a response to a recurring task”. [90] What is this recurring task?

Bakker argues that metrical unities consist of a nucleus and one or more fillers. Fillers are (1) context neutral, (2) metrical variable, and (3) interchangeable. [91] Fillers facilitate the nucleus of a metrical unity to extend towards a metrical boundary, a position of frequent word end like the caesurae at position 3, 5, 7, 9, or the dieresis at position 8 or 12. The result is some level of grammatical governance and hence “grammatisation”. Between metrical boundaries within the verse or even the hemistich, metrical fillers allow the semantically more important constituents, often the verb, to occupy the position their metrical shape restricts them to. Bakker [92] demonstrates this compositional principle by means of the well-known smaller scale narrative pattern ‘A kills B’ in the Iliad. Such a pattern requires a killer, a victim, and a verb expressing the action of the killing (fillers are underlined): All verses in the cluster Iliad 6.29–33 contain metrical fillers like the weapon, an epithet, or a patronymic/geonymic. The filler expressing the weapon itself contains another filler: the epithet χαλκείωι / φαεινῶι.

Discussing particles and discourse markers, [93] Bakker shows that on a larger scale Homeric syntax is a movement, that demands the audience’s attention for what is to follow, with only little regard for what has already been said. The periodic sentence of written discourse regularly requires the addressee to interpret, or even re-interpret, the subordinate clause upon engaging the subsequent main clause. In spoken discourse, on the other hand, intonation units reflect the speaker’s mental processes. The emerging grammar does not necessarily organise the units according to a logical or hierarchical principle. To illustrate “movement” of discourse in Homer, Bakker [94] presents two types of discourse-relations. On the one hand, there are additions (genitive attributives, participle used as adjectives, nominatives as appositions), on the other, continuations (connective δέ, additives). The various discourse-relations keep the narrative going: they facilitate continuation. Bakker consistently translates the Homeric text while maintaining his division in intonation units. [95]

Increasingly, it becomes clear that the analysis of the alleged sentence-structure of the Homeric poems does not contribute to the identification of anything like “sentences”. A description of Homeric syntax must start from a different angle, as the work of Bakker clearly shows. Analysis of Homeric discourse supports the notion of syntactical movement and continuation. It would be more appropriate, I argue, to speak of Homeric syntax in terms of “continuation” within larger discourse units: in the Iliad and the Odyssey, there are many more devices to facilitate continuation within discourse units than there are clearly marked pauses in sense. [96] In planned written discourse, the sentence can be identified as a unit of discourse-internal organisation. In spoken discourse, what seem to be sentence-structuring features, both in syntax and in meter, can be deceiving. Starting from criticism on the concept of “grammatical completeness” I will present the grammatical clause as the syntactical unity of Homeric discourse.

In the vast scholarship on Homeric sentence-structure and verse-making, it has been argued at an early stage that the Homeric dactylic hexameter often seems to coincide with a grammatically complete whole. As Parry [97] wrote when comparing Greek and South Slavic heroic song: It did not take long for Parry to realise that usage of the word “sentence” is the cause of problems: It is remarkable that Parry keeps trying to fit the sentence-structure into the pattern of whole hexametric lines. He allows for a grouping together of several sentence-verses into a stanza—a possibility for which evidence is considered insufficient nowadays. [99] Parry notices that our sentence is not a very useful instrument when analysing the structure of the Iliad and the Odyssey. He looks for a different type of ending or completion for semantic phrases. Terminology like “sense” [100] may suggest that ending or completion. But a sense pause does not necessarily coincide with verse end, nor is every verse end a sense pause. [101] “Varying portions of time” may in fact straddle the verse end. Still, in Parry’s view, the chain of thoughts presented in the Homeric epic is cut up into “portions” that preferably occupy whole hexameters or parts thereof. For Parry and his followers, the sense of completeness at verse end is so strong in Homer’s lines that ‘the reader of the Iliad and the Odyssey soon comes to mark in them, as a part of a larger movement of the thought, the way in which the sense passes from verse to verse.’ [102] Parry 1929, Kirk 1966, Higbie 1990, and Clark 1997 judge the syntactical construction within the verses of the Iliad and the Odyssey by its alignment to the metrically repeating pattern built by these stichic verses. In other words, the force of the verse end as a prosodic boundary is seen as confirmation of its force as syntactical or sense boundary. Parry understands whole-line expressions as whole-line formulas, due to their frequency in the Homeric epic. Famous examples are the speech introductions, the vocative phrase cum interrogative clause, and combined clauses: Indeed sometimes these “whole-liners” construct larger passages (Iliad 8.145–152, Odyssey 21.269–276 [103] ).

The precise definition of a “formula” (and its usefulness) is not the issue here: its relevance in relation to enjambment must be discussed elsewhere. The point to be made here is that, in Parry’s view, singers like Homer had a clear preference for “expressions” or “combinations of expressions”, that take up exactly one line and fill one hexameter. This assumption—though supported, at best, by less than half the verses [104] —has had far-reaching influence. Many scholars of oral poetry, both working on enjambment and other phenomena, tend to head for the same direction. [105] This means that a line like Iliad 1.73 should be considered an appropriate unit to facilitate composing for an oral poet: As does a line featuring a vocative cum interrogative clause: Another very useful, though not strictly formulaic, line might then be: In Iliad 1.202, the vocative gives the line a tripartite structure; Iliad 2.555 contains two complete clauses. [106] Some of these lines are repeated within the Iliad, like Iliad 1.73, others (Iliad 2.555) are not. If a line contains two complete grammatical clauses, often at least one of the two is either a direct command or a question. It is not surprising, therefore, that such verses appear frequently in direct speech. Some of these verses in direct speech are repeated within the Iliad, like Iliad 1.362 = Iliad 18.73, others are not: they are even unique in Homer. Some of the unique lines containing either one or two complete clauses suit their specific context very well; tentatively describing them as “formulaic” lacks all evidence.

The idea of grammatical completeness is thus closely related to the metrical shape of the Homeric narrative. The grammatical need and expectation felt at the end of the line are felt because the line, the verse, has an end. But the status of “grammatical completeness” is still questionable considering that the grammar and syntax of the Iliad and the Odyssey seem to be in an early stage of development. Chantraine [107] has clearly shown that the level of word autonomy is high in the Homeric epic. Many words define their own role in discourse, without any hierarchical word group organisation through grammatical governance. Similarly, larger units, centred round a predicate, are juxtaposed in parataxis: a form of autonomy on the larger scale level. There are, of course, instances of hierarchically organised word groups and predicates. Some development of grammar and syntax is hence found.

Within smaller scale units, a distinction between semantically nuclear and peripheral elements is made by Bakker. In the process of larger scale composition, modifications of grammatical categories and word endings is caused and required by the frequent application of the smaller scale unit. Examples are the alternation τὸν δ᾿ ἀπαμειβόμενος / τὸν δ᾿ ἀπαμειβομένη, and τὸν δ᾿ ἀπαμειβόμενος / τὴν δ᾿ ἀπαμειβόμενος. Bakker defines grammar as governance within the metrical unit of the colon. If the unit within which the process of “grammatisation” takes place is the metrical colon, the unit is minor scale. Compared to the whole of the epic framework, or even the book or the passage, the unit of such grammatical segmentation is not remarkably striking: There is the possibility that all grammatical segments are formulas. Lack of primary sources, however, deprives us of proof for that hypothesis. [108] As grammatical segments, formulas are shaped metrically. Bakker explains that their formation finds its origin in the extension of a nucleus (translated in capital letters) into a metrical phrase (underlined): The principle of formulaic economy suggests that the formation of formulas was defined by metrical requirements, rather than by semantics. Noun-epithet formulas, for example, are specific extensions to occupy specific metrical positions. I agree that maintenance of metrical boundaries as positions of word end contributes to the sense of grammatical segmentation. After all, it has not proven difficult to identify recurrent metrical colometry based on positions of frequent word division (indicated by the number of the metrical position in both the Greek example and its translation): [109]

into two hemistichs: that can both be divided into two cola: into four cola: into a three-colon verse: into a “rising threefolder”: [110] Analysis of single hexameters as metrical units is, of course, the foundation of such colometry. The hexameters as the metrical units, and the metrical colon as the metrical phrase, seem to underlie the composition of the Homeric narrative. As such, they precede Homer’s semantic phrasing. There are clear examples of verses that consist of one complete “sentence”, [111] or “clause”; [112] and of verses that naturally divide into metrical cola. [113] Fränkel’s colometry is based on its usefulness for internal articulation: the hexameter features three “zones” into which the positions of frequent word end can be found. “Displacement” of the caesura demarcating the phrase does not question the division of the verse into four units. Fränkel [114] points out that the metrical cola are regularly “units of meaning”, and that the word breaks which limit them are “sense-pauses”. Porter [115] shows, however, that this is not always the case. He argues that the persistence of metrical cola is due to prosodic reasons (‘an expected sequence of syllables produced by a brief rhythmical impulse’ [116] ), and that cola only tend to comprise organic word groups. He states that the main explanation for their persistence lies in sentence-articulation: colon-structure is concerned with aural effect. [117] Porter argues for a modification of Fränkel’s system. He looks for “balance” between the phrases that together constitute the hexameter, and accepts that the metrical colon is not invariably a semantic unit. His determination to balance the metrical cola, and to identify the colon as a word group rather than as a word, leads him to ignore the statistical evidence for the frequency of word end at position 7. As Porter’s colon is a rhythmical unit that exercises a normative effect on the meaning, Ingalls 1970 rejects Porter’s criticism of Fränkel when he concludes that Porter works from a priori notions (Porter seems to reject the idea of an iamb-shaped colon between positions 5½ and 7). Kirk meets with similar criticism. His addition to the basic colometry of the hexameter into two or four cola is offered in his 1966 publication. Kirk starts from verses where metrical boundary in the third foot is impossible. [118] As the frequency of word end determines the third foot caesura as the second most important metrical boundary in hexametric verse, it is straddled rarely in metrics. Identification of other metrical boundaries in the verse where the third-foot word end is impossible, leads him to the tripartite colometry of the “rising threefolder”, soon extended to incorporate “semantic rising threefolder” verses.

The extension incorporating semantic threefolder verses shows how readily metrical boundaries are seen as prosodic indicators of sense-structure or, alternatively, of cognitive structure. These are the a priori notions Ingalls 1970 and 1972 criticises Porter and Kirk for. On the level of composition, however, there is a correspondence between metrical cola and formulas. [119] My examples in this section do not contradict that metrical boundaries play a part in the demarcation of unities of grammatisation.

The study of Homeric discourse as spoken language [120] gives new life to the generally accepted concept of metrical colometry. The appositional juxtaposition of metrical cola resembles the natural “chunking” of speech. Bakker’s analysis of the Homeric discourse [121] reveals that the metrical cola share many of the characteristics of the spurts of spoken language: [122] metrical cola are limited in size, characterised prosodically, [123] and usually express one “focus”, one single new idea. [124] Not all observations from the general studies of spoken language can be applied to the Iliad and the Odyssey. [125] The Homeric epic cannot be described as “unplanned” speech. As much as the metrical cola resemble the spurts of unplanned speech, certain stylised features, in addition to meter, turn the discourse of the Iliad and the Odyssey into special speech. Special speech is represented in the chunks (or spurts) of spoken language, but attains a different formal level, as it uses metrical formulas, not the units of intonation, as its chunks. [126] In metrical special speech, the common size of a chunk is the hemistich, but smaller chunks may be identified. Bakker chunks ‘with the units into which the passage easily divides’. [127] Bakker values both semantic and metrical considerations. As our consciousness is chunked, so is our language when we speak—and so is Homeric discourse.

I accept that the stylisation is indeed an adaptation of single words or word groups to metrical circumstances, and I am in favour of a division into ‘the units into which the passage easily divides’. [128] As Bakker explains, the chain of thoughts is physically divided into utterances that each fit a spurt of vocalisation. One single spurt or intonation unit contains just one idea. The highly autonomous character [129] of single words and word groups in Homer is in accordance with the checklist-structure of the syntax. Words and word groups are sufficient indicators of their own role in discourse: their semantic function remains clear enough despite the appositional rendez-vous of the intonation units. In Homer, semantic functions do not stem from grammatical governance.

At the same time, the checklist-structure facilitates grammar’s emergence as the consequence of the juxtaposition of self-defining parts of speech. Grammar emerges from the combination of intonation units or chunks into larger wholes. [130] Together, a series of chunks may result in a clause that meets certain grammatical requirement, for example the use of a verb. The chunks themselves, especially those without a verbal form, are either preparatory or additive to preceding and subsequent clauses. As grammar emerges from the alignment of intonation units, so the appositional alignment itself emerges from the movement of Homeric discourse.

Existing studies are reluctant to allow, as the result of the appositional alignment, for a level of syntactical organisation that exceeds the grammatical governance of the syntactical unit that is the metrical phrase. A description of Homeric style as “adding” is in accordance with the concept of grammatical completeness at fixed and recurring metrical boundaries. The model introduced by Clark, in which verse-initial and verse-final constituents (runovers and anticipations) function as semantics “hooks” to link formulas at verse end, seems to take out-of-line composition for granted under certain circumstances. I agree with Clark on that point, but his approach lacks the question whether or not such composition allows for, or leads to, a higher level of syntactical organisation. Examples of this lack are Clark’s treatment of the correlative pair in clausal-internal enjambment, and of the well-known phrasing of Iliad 1.2 οὐλομένην ἣ, which is regularly cited as an example of the frequently recurring model of an ‘emphatic runover-word cumulation developed in what follows’. [131] Syntactically, the “development” seems to contribute to continuation in a way similar to the adjective. It has, however, the advantage of a larger and more adaptable appearance. Both the correlative pair and the “development” are regularly labelled “subordinate”. It is not evident, however, that the syntactical qualification “subordinate” can be applied.

Chantraine reconstructs a development in Homeric discourse towards subordination: adverbs, originally correlative, develop into conjunctions, due to the disappearance of the first limb of the correlative pair. Syntactical units introduced by conjunctions show further development suggesting subordination: the position of the conjunction and the use of modes other than the indicative. From the point of view of grammatical necessity, as taken in current enjambment studies of for example Higbie, [132] verses introduced by a subordinating conjunction are never a requirement for understanding the preceding verse. In adding style, they cause verse-end clausal external enjambment (Higbie’s terminology). An example from the first lines of the Iliad is Iliad 1.10–11: On the other hand, we find in the Homeric epics verses introduced by subordinating conjunctions that precede what may be considered their main clause. The first example in the Iliad is Iliad 1.39–41: Followed by a main clause, (ταύρων ἠδ᾿ αἰγῶν) τόδε μοι κρήηνον ἐέλδωρ, the verse end of Iliad 1.40 constitutes clausal internal enjambment (Higbie’s terminology). [133] This type of verse-end enjambment is considered more effective than the clausal external type: Iliad 1.39–40 raise grammatical expectations. There, identification as subordinate units requires continuation by means of a main clause. Iliad 1.39–41, however, do not show any preparations in the grammar of the subordinate unit to prepare for either another subordinate unit (Iliad 1.40), or for a main clause (Iliad 1.41b). Not surprisingly, in Homer the main clause can easily be deferred for several lines, or not appear at all. [134] Chantraine [135] argues that the uncertain status of forms of ὅς is proof of the “paratactic character” of the “Homeric phrase”. Bakker and Clark seem to agree with Chantraine. The alleged paratactic character, however, has a blurring effect [136] on the syntactical division into main and subordinate “clauses”. Chantraine, among others, advocates the analysis that the Iliad and the Odyssey mark a stage of the development from parataxis to a broader use of hypotaxis.

Inclined as I am to subscribe to Chantraine’s point of view, there is still something missing in the analysis. The approach by Bakker of metrical phrases as appositional intonation units makes the feeling that something is missing even stronger. Describing the Homeric narrative as a stage in the development to hypotaxis makes it look like the development of a syntactical feature like hypotaxis is more or less coincidental, the by-product of the random apposition of internally coherent grammatical unities. Clark’s model of the semantic “hooks” to link formulas reduces, correctly in my view, the importance of grammatical completeness at the metrical boundary that is the verse end, [137] but there are so many more examples of grammatically more complex composition due to the straddling of various metrical boundaries—not to the chance meeting of appositional chunks at the metrical boundary. Examples of grammatisation over the metrical boundaries of the accepted metrical phrases do not always rely on “hooks”, and that includes crossing the verse end: An example crossing the third foot caesura is Iliad 1.48: And, I argue, there is crossing in all instances of elision at the metrical boundary: Other positions of frequent word end require similar attention. Grammatical coherent phrases start at various positions in the hexameter, and completion of grammar may disregard the available (and often used) metrical boundaries. Both in the Iliad and in the Odyssey, whole passages consist of such diverging grammatical and metrical patterns. I will cite four larger scale examples. The first two are examples of character speech: they illustrate what Bakker calls the “fugal effect” of more emotional passages. Numbers 3 and 4 do not have a similar excuse: they show the ordinariness of the non-alignment of metrical and grammatical boundaries. The grammatical boundaries are not always those of Bakker’s chunks, that is, with strict observance of metrical colometry. In the Greek text, I indicate the positions of frequent word end that are “straddled” by ongoing grammatisation. In all four passages grammatically coherent phrases are not equivalent to metrical cola. The way grammatically coherent phrases are being located onto the metrical pattern in fact disrupts the repetition that characterises the sequence of metrical cola of the hexameter. The juxtaposition of grammatically coherent phrases is much smoother than metrical cola suggest.

Expanding grammatisation does not disregard metrical boundaries. The identifiable tools of pre-enjambment preparation and post-enjambment postponement are combined with the use of “metrical fillers” [138] to facilitate larger scale grammatical development. Such development makes use of the metrical boundaries, but not all of them. Some boundaries are straddled, only to allow others to demarcate units of discourse. In this section I will show that as grammatisation extends the nucleus towards the nearest metrical boundary, on a larger scale it extends the clause towards a metrical boundary. Semantically innocuous fillers are used to make sure that grammatically required constituents can occupy a fitting metrical position without jeopardising the syntactical continuation.

Even without necessary enjambment, metrical fillers serve to make the clause extend towards specific metrical boundaries. In many instances of verse end in enjambment, “grammatical necessity” for enjambment gives way to “semantic superfluity”: Especially the last two lines, Iliad 4.295–296, seem to be superfluous, and, worse, incompatible with information concerning Pylian leaders elsewhere in the Iliad. [139] Taking Iliad 4.295–296 out of the text would not significantly damage or alter the passage. Why are these verses here? Semantic superfluity is not enough reason to take them out. Ancient scholars like Aristarchus left the verses where they found them, noting the difficulties they caused, but not removing them.

Additional verses may also lead to interesting digressions, as in Iliad 6.288–292: The addition over the verse end is a verse, or a cluster of verses, and may completely consist of metrical fillers, as in Odyssey 15.404–406: In other verses, additions consist of metrical fillers occupying a part of the line. The metrical filler then bridges the half-verse gap before another addition that has more semantic value: The usefulness of such metrical fillers seems to be determined by the location of semantically more valuable additions. The Odyssean formula ἐν σπέσσι γλαφυροῖσι occurs seven times, always occupying the first half of the hexameter. In Odyssey 1.15 λιλαιομένη πόσιν εἶναι occupies the second half of the hexameter, as does θεμιστεύει δὲ ἕκαστος in Odyssey 9.114. The formula ἐν σπέσσι γλαφυροῖσι is thus bound to an invariable metrical position, but so is its role as a bridge towards an equally restricted addition with higher semantic value: λιλαιομένη πόσιν εἶναι is metrically restricted as well.

On a smaller scale, that is, between metrical boundaries within the verse or even the hemistich, metrical fillers have a similar function: they allow the syntactically more important constituents to occupy the position their metrical shape restricts them to. As mentioned in 1.5, Bakker demonstrates this compositional principle by means of the narrative pattern ‘A kills B’ in the Iliad (fillers are bold): Similar to the larger scale units, the smaller fillers enable semantically more important constituents to occupy the metrical position they are restricted to without jeopardising the comprehensibility of their syntactical function: the epithet ἄναξ ἀνδρῶν in Iliad 6.33 makes it easier, not more difficult, to recognise Agamemnon as the subject of the clause in which Elatus is the object.

In the cluster Iliad 6.29–33 cited above, the usage of metrical fillers clearly shows the compositional principle of extension towards the nearest metrical boundary, as explained by Bakker. [140] Extension meant that a phrase containing one or more “nuclei” [141] may be extended towards the nearest metrical boundary by means of fillers. The fillers proved to be (1) context neutral, (2) metrical variable, and (3) interchangeable. [142] Variable fillers like ἔγχει / ἔγχει χαλκείωι and δουρὶ / δουρὶ φαεινῶι show both their metrical usefulness and their semantic superfluity. At the same time, the fillers redirect our attention towards semantically more important constituents.

A filler like ἐν σπέσσι γλαφυροῖσι in Odyssey 1.15 answers to the same rule: its metrical shape is different from ἔγχει χαλκείωι in Iliad 6.31, but it equally occupies the first hemistich. Another similarity is the fact that both fillers, ἐν σπέσσι γλαφυροῖσι and ἔγχει χαλκείωι, are “linked” to both the preceding and the subsequent clause: semantically, both constituents belong to the preceding clause; metrically, both seem to be forward extensions of a semantically more valuable clause in the same hexameter towards the nearest metrical boundary, the beginning of the verse.

Does a filler like ἐν σπέσσι γλαφυροῖσι have a structuring semantic value as well as metrical functionality? In other words, can it be more than a free-floating addition that happens to fill a certain metrical position? A metrical position that was in danger of remaining unoccupied when the poet first arranged the semantic nuclei of the passage in his mind?

True, the metrical fillers appear to be applied in order to extend nuclei like the verb towards metrical boundaries. The usage of metrical fillers often demonstrates the restrictions on the location of other smaller scale units. These restrictions are very compelling in case of the location of certain structuring discourse markers like the start and completion of direct speech, and the resumption of an extended simile by means of ὡς. In both these cases, direct speech and the extended simile, the transition over the verse end is the only position from which the poet starts. That makes it necessary to fill out the line before the start of direct speech: When the verb ‘to speak’ is confined to verse-final position, [143] fillers may bridge the gap to the verse end: When grammatical completeness requires another constituent that must be postponed to the following line, the subsequent line is still extended towards the verse end: Extensions over the verse end due to the requirements of grammatical completeness may even result in the addition of another verb not denoting the actual act of speaking before the verse end: What the various extensions have in common, regardless of their size and exact metrical shape, is their usefulness in extending an utterance towards the verse end. The verse end, however, is not always the nearest metrical boundary.

Consider the resumption of the extended simile, which seems to be equally restricted to the beginning of the hexameter: [144] What the extended simile in particular shows, is the usefulness of additional constituents up until the verse end before resumption: constituents and even whole lines may be freely added, as long as the addition as a whole, together with possibly required constituents, [146] ends with verse end: The start, or completion, of direct speech, like the resumption of an extended simile, is a special instance of syntactical organisation in Homer. Syntax is still rather flat and paratactic [147] in these cases, but the compositional process is not that of mere adding anymore. There may still be some semantically more or less important additions following the verb denoting ‘to speak’ or the resumption (ὡς) of a simile. The verb ‘to speak’, however, prepares for a specific type of continuation, that is direct speech. Similarly, the start of an extended simile [148] prepares for resumption by means of ὡς or τοῖον. The start of the constituents, whose appearance is prepared for, is confined to the start of the hexameter. Above, I argued that the fixed localisation of constituents due to their metrical shape does not harm the comprehensibility of their syntactical function. Semantically, the metrical fillers contribute to this comprehensibility. In the case of direct speech, as in that of the resumed extended simile, the syntactical function of constituents is confined to one specific metrical location, the start of the line. Even if the verb denoting ‘to speak’ were not in the preceding line, there are various ways to mark the start of the first line of direct discourse as a shift in discourse type. I will give examples of the various markers:

Vocative:

Imperative:

Prohibitive:

Emotional outcry:

Shift to first or second person:

Asyndeton: In many instances, as in the examples cited, various markers of discourse shift are combined. The discourse shift itself seems confined to verse-initial position. But some doubt might arise in the absence of verse-initial semantic markers of the shift: Or: Both analyses are equally possible.

The completion of direct speech always coincides with verse-end [149] . The constituent marking the shift back to the narrative is normally in verse-initial position:

Resumption:

Emphatic use of the personal pronoun:

Emphatic use of a narrative-structuring adverb: Metrical fillers show that certain constituents may be used despite their invariable position, and that the verse end facilitates grammatical completion by means of extension of the discourse unit that ends with it: The invariable verse-initial position of certain discourse markers strengthens the idea that the verse end of Homer’s dactylic hexameter is the first position of choice for the completion of one syntactical unit and the start of the next. Apart from direct discourse and extended similes, many other verses feature verse-initial start, or verse-end completion of syntactical units.

A number of verses feature both. Some verses have another syntactical “break” within the line, at the third foot caesura: The third foot caesura is not the only position for an additional, verse-internal “break” in syntax: And then there is a second type of phrasing, as not every line in the Iliad and the Odyssey has the start of a syntactical unit at its beginning, or the completion of one at its end. Verse-internal metrical boundaries do not always feature a syntactical “break” either. Metrical boundaries like the verse end and the third foot caesura are regularly straddled by syntactical units. The principle of using metrical fillers to facilitate the start of a syntactical unit from a metrical boundary is equally used to continue the syntactical unit over the metrical boundary (fillers are in bold): As in previous examples, the application of metrical fillers results in metrically variable and context neutral (innocuous) constituents. As such, the constituents are peripheral material, as opposed to the nuclei, the semantically more important constituents. The metrical fillers do not jeopardise the comprehensibility of the semantic function of nuclei. When demarcated by a “break” in syntax (indicated by square brackets) on either side, the filler is semantically (and syntactically) tied to either the preceding, or the subsequent syntactical unit: Often, however, metrical fillers are not the start or closure of a syntactical unit. Many metrical fillers have their place within the syntactical unit: The hyperbaton, so to say, of the nuclei (Μηριόνης … ηὔδα) belonging to the syntactically coherent unit, does not hinder the syntactical coherence. The fillers themselves are an integral part of the syntactical unit. As syntactical coherence is maintained despite “disturbances” like epithets, so it is maintained despite more structural embedded constituents, like the “embedded predication”: [151] The examples so far show that the nuclei, the semantically more important constituents, may well keep their metrically restricted position. The gap between the nuclei, bridged by means of peripheral material, regularly extends the nuclei forwards or backwards to a position of frequent word end (metrical boundary). The “adding” that characterises Homeric style presents itself as adding of metrical phrases: nuclei (printed in capitals in the following translations) are extended to positions of frequent word end in order to constitute metrical phrases like cola and hemistichs before being added: In these examples, the juxtaposition of metrical hemistichs results in juxtaposition of smaller scale syntactical units. Such adding of metrical phrases, I would argue, is paratactic in more than one sense: it is paratactic both metrically and syntactically. Metrically, positions of frequent word end (in this case: the third foot caesura) are observed; syntactically, the organisation does not imply any further hierarchical organisation.

Metrical phrases like hemistichs, created around nuclei, may constitute larger scale syntactical units (indicated by square brackets), often coinciding with verses: Passages may be built using verses that are syntactical units: If a passage consists of verses that are syntactical wholes, the verse end as position of demarcation is observed, and paratactic syntactical organisation aptly describes the juxtaposition of verses. In that case, however, metrical boundaries within the line no longer demarcate syntactical units that may be applied as syntactically paratactic additions. The single syntactical unit is then no longer equal in size to the shorter metrical colon or hemistich, but to the single hexameter. Extended nuclei constituting metrical phrases are themselves part of larger scale syntactical units. The larger unit still consists of metrical phrases: the continuing additions take the repetitive shape of dactylic hexameters.

The extension of the nucleus to the nearest metrical boundary, an adaptation in fact due to metrical circumstances, requires a certain level of grammatical organisation within the metrical unit. Metrical fillers cannot be put in as free-floating elements, without any semantic or syntactical tie to surrounding constituents. In order to create a larger scale syntactical unit or a verse, one or more nuclei team up with peripheral material that underlines the semantic and syntactical cohesion: As Bakker put it, grammar “emerges” from such adaptation. [153] When put side by side, the single words loose some of their autonomy: not all of it, as they still define their own role in the discourse, [154] and grammatical governance is not the standard.

But there is reason to take one more step beyond the level of grammatisation that Bakker has shown. Some more autonomy is lost, as the syntactical unit grows bigger: the “hyperbaton” of semantically coherent nuclei (printed in bold face in the following two examples) in a verse is more demanding for the composer, and for the audience, than the “apposition” of nuclei in, for example, the hemistich: The more metrical boundaries are being straddled within a syntactical unit, the stronger a sense of higher level syntactical organisation becomes. Straddling a metrical boundary within a syntactical unit means that the second part of the unit is not a paratactic addition: it may contain a semantically relevant constituent of the developing syntactical unit. As the extension of the nucleus to the metrical boundary led to a certain level of “grammatisation”, so the straddling of metrical boundaries leads to a larger scale grammatisation. On a larger scale, grammatisation may be equally due to straddling the verse end, the third foot caesura, or both: Larger scale grammatisation is the process of extending the expression of thought, not by means of addition, but by means of semantic cohesion. Certain points of demarcation appear: not the metrical demarcation that easily allows for endless continuation through repetition, but semantic-syntactical contours that create a sense of completion, or expectation. I argue that there is reason to assume that the syntactical contours of the Homeric epic are not so much limited by metrical boundaries, but rather brought into being by a degree of selection of metrical boundaries.

My assumption reduces the importance of nonconfigurationality for Homeric discourse, but it was pointed out in 1.4 that the occurrence of phenomena of nonconfigurationality does not make Homeric discourse nonconfigurational: the characteristics of nonconfigurationality can be found from time to time, but they are not pervasive. Some of these characteristics are characteristic of Greek in general, and can even be found in the highly organised syntax of written Classical Greek. Others appear to be the rule-confirming exceptions when compared to more “figurational” Homeric discourse. Both types of discourse can be found, side by side. Homeric syntax does, however, show certain extended grammatical development. Grammatical governance can be identified; syntactical subordination does not seem to be fully absent in Homer, but it may be too much to say that it is in a stage of development. [155] The chunks of spoken language, the metrical phrases of the Iliad and the Odyssey, are themselves either formulas, [156] or grammatical adaptations of verb forms: the adaptation extends the verb to the nearest metrical boundary, for example by means of fillers. [157] Alternatively, Homer may have used not so much formulas but single words as compositional units. [158] Either way, the combination of (some) nonconfigurationality and (some) extended grammatisation in Homer seems to me to confirm and deny the importance of metrical boundaries at the same time: the words in the line can be separated into metrical phrases, but the metrical phrases are often not semantically coherent units. The metrical phrases show internal grammatical organisation, but contribute to larger scale grammatical organisation at the same time. Recognition of this contribution over the verse end is labelled enjambment. Verse-internally, however, contribution to grammatical development over metrical boundaries is even more common. [159]

As pointed out earlier, metrical boundaries may join together metrical phrases comprising clauses that are in some sort of semantic hierarchy to one another, like the main and subordinate clause. Such joining at the verse end creates expectations for the verse to follow when the subordinate clause comes first, or when the main clause contains a constituent suggesting semantic limitation to follow (like correlatives and certain particles): Beginning with the subordinate clause and using a correlative can of course be combined. The resumption by means of the main clause is then more accurately marked: Resumption, however, may be equally marked within the line: I have rejected any “strength” of enjambement, and the possibility for emphasis due to metrical position, but I think it is safe to state that the grammatical organisation at the metrical boundary is not paratactic in these examples. The continuation over the metrical boundary is not a mere addition in the sense that ‘it could have been anything’, any new focus of consciousness. The continuation is prepared for, and it stands in a semantically hierarchichal relation to what preceded it. Even a continuation by means of another main clause, as in Iliad 21.177 τὸ δὲ τέτρατον, has a semantic function prepared for by a constituent in the preceding main clause: [160] without it, the usage of τρίς remains without proper meaning.

Metrical boundaries fulfil more roles in organising the syntax of Homeric discourse—apart from providing a break between clauses, and resulting in “enjambment”. Semantic hierarchy makes it possible to separate discourse markers and sentential particles from their clause at the caesura and the verse end (compare the many instances of αὐτὰρ ἔπειτα at verse end). The metrical boundary then frames a constituent that facilitates the transition from one clause to the next: This phenomenon may of course be compared to the so-called dislocation of constituents, a characteristic of nonconfigurational language discussed above in 1.4. Many instances of “enjambment” can then be removed from the various categories, as they reflect a cognitive process: [161] an implicit constituent is merely made explicit through left- or right-dislocation. Prosodically “isolated” nouns may thus function as “heralding” of the clause to come, or as an “afterthought” to the clause that has just been completed. From a semantic point of view, such labeling is not always in order: the prosodically isolated noun is often essential for the comprehension of the narrative. For example, as the isolated noun signals the shift to another grammatical subject: It is not impossible, as it would not be in spoken discourse, to consider the isolated noun a heralding (or theme) of the clause to come, but the semantic shift has already been indicated by a conjunction (αὐτὰρ) whose domain stretches beyond the single noun [162] Ἀχιλλεὺς. The verse end is hence a grammatical transition within a larger scale syntactical whole. Punctuation in the translation is helpful to show the syntactical coherence between the conjunction and the clause, despite the “dislocated” noun: As free-floating as the dislocated constituent seems to be, it is syntactically incorporated into a developing clause. The same is true for the “dislocated” discourse marker or sentential particle: The discourse markers and sentential particles are separated from their clauses by a metrical boundary, but the development of a syntactically coherent whole is not hindered by the verse end. Within the hexameter, metrical boundaries do not hinder syntactically cohesive extension of the clause either: Nor do verse-internal metrical boundaries necessarily create “dislocation” of other constituents: Still, it is remarkable that so many discourse markers and sentential particles, often together with other constituents, are metrically “isolated”, as they constitute metrical phrases. As such, they do not comprise independent “units of meaning”. The metrical phrase contains constituents that can not be left out without rendering the clause incomprehensible.

What is left of the metrical “isolation” then? If grammatical completeness is extended over several metrical phrases, why is there still what looks like a persistent “natural division” of parts at the metrical boundary in so many clauses? Focusing on the discourse markers and sentential particles (as examples of “transitional” constituents) for now, their metrical phrasing often shows the use of fillers. Their metrical isolation, possibly together with another nucleus, is obvious: In this last example, Odyssey 16.263–265, the leap from an isolated metrical phrase that can be juxtaposed paratactically, to a semantically and syntactically integrated part is especially perceptible: just take out the underlined metrical phrases. Semantically, the integration is partly the result of the way the constituent is extended towards the nearest metrical boundary: the phrase comprises a sentential particle and another semantic nucleus, or a discourse marker and another semantic nucleus, or a constituent suggesting syntactical hierarchy and another semantic nucleus. In other words, the transitional constituent is tied to something more to follow or to be implicated: the chunk that is the metrical phrase is not in itself markedly autonomous or syntactically paratactic. The chunk is characterised as transitional: it both continues from what was already said (sometimes by contrast), and it needs further elaboration in subsequent syntactical development. The chunking of “transitional” constituents has an equivalent in linguistics: in the pragmatic approach, these constituents are not always considered as part of the clause proper, but as extra-clausal, if they hold functions like theme and tail. Their syntactical relation to the clause itself becomes apparent through prosodic characterisation. [163] On a larger scale, semantic hierarchy also nullifies the equation of “metrical parataxis” and syntactical parataxis.

The choice between the use of “clause” and “grammatical units” is still as close to Scylla as it is to Charybdis. As Bakker shows, the requirements from the descriptive analysis of written language turn out to be ill-fitted for the reconstruction of Homeric syntax. The assumption that “inner metrics”, or metrical cola, structure Homeric syntax is equally misguided. Metrical phrases are not always the grammatical units of discourse. In the analysis of Homeric discourse, there is an abundance of discourse-unit structuring elements, [164] of which the sentence appears to be least suited as a tool of study. Clark is right in stating that the grammatically complete clause proves to be the most useful instrument, but then we must accept, contra Higbie, that grammatical completeness is not per se determined by recurrent metrical boundaries like the verse end. Terminology like “enjambment” accounts for the mere acknowledgement of a visible disparity between metrical phrases and syntax: it cannot teach us anything about what the audience actually perceives—though the temptation to explain enjambment as a poetic effect is never far away, as recent commentaries on the Iliad and the Odyssey show.

I have shown that Homeric syntax is not so much concerned with the due completion of “sense” in accordance with metrical phrasing. The frequent divergence of syntactical and prosodic patterns serves continuation despite metrical repetition. Such divergence is at least as frequent as the audible structuring of the discourse into metrically well-balanced grammatically complete wholes. I do not deny that syntactical phrasing or completion of “sense” is regularly enclosed within metrical phrases like cola, hemistichs, and verses: I have cited examples above, and others have given many more. I merely argue that there is a second type of syntactical or rather “sense” phrasing: a type whose units do not seem to be restricted in size by the recurring metrical phrases. At the same time, this second type of phrasing inevitably leads to a higher level of grammatical organisation than witnessed within the boundaries of the recurring metrical phrase. Descriptions of Homeric syntax like Chantraine’s and Bakker’s, who both stress the compositional principle of appositional alignment of words and metrical phrases, do not account for this emergence of large scale grammatisation.

The emergence of larger scale grammatisation is reflected in the progressive movement that characterises Homer’s style: at first sight the “nonconfigurationality”, so reminiscent of Chantraine’s approach, of the merely paratactic structure seems to reflect the sense in Homeric narrative, but looks naive when compared to the syntax of the complex periodic sentence of written language. But Homeric syntax is not naïve: its aim is not to progressively categorise the various units to form a larger scale well-ordered syntactical hierarchy—a hierarchy, though, whose sense often may only be understood from re-reading. On the contrary, Homeric syntax aims to retain the attention of the listening audience, without them having to rearrange previous chunks along the way. In Homer, syntax develops from a progressive tendency. This progressive compositional technique demands parataxis, non-disturbing metrical boundaries, audibly framed chunks and a creative poet. Its also requires the possibility of perceptible reorganisation of grammatical wholes despite metrical repetition, without loosing sight of sense in the narrative. Listening to Homer, the audience must rely on their own ability to construct a mental picture of what is told from the order in which it is told. The narrative of the Iliad and the Odyssey presents its audience with certain specific discourse structuring elements, both in syntax and in semantics. These elements not only structure the discourse, but are also supportive in constructing the mental picture in the listener’s mind. The listener’s attention can be channelled into focusing on the continuing chain of thoughts. Metrical cola may be neglected as sense-units in the process, syntactical coherence can not. Still, the listener’s attention cannot be channelled without any aid from the side of prosody. As Bakker assumes, prosodic phenomena characterise the units into which spoken language naturally divides. For Bakker, these prosodic phenomena are the recurrence and fixed shape of metrical phrases. The resulting metrical-phrase chunks were labelled “intonation units”, but unfortunately meter does not provide us with clues for a prosodic phenomenon like intonation: intonation is related to the structure of language and syntactical phrasing, not to colometry. We are looking for further prosodic clues that are not limited to just meter, but take some other form of audible coherence of units into account. First, however, a theoretical model is needed for the analysis of the units of Homeric discourse.

Then what are these units? How large are they? Are their boundaries signalled somehow, if not by the verse end? How do the units relate to the grammatically complete clause, if the size of a clause may vary? Must a unit contain a verb? [165] In anticipation of chapter 2 I say it often will, as grammar emerges from the extension of, especially, verbs. Clauses must contain at least a unit featuring a verb. Units without a verb can only be preparatory, or an addition, to units with a verb, but often several units are required to answer to the requirements of developing syntax. The progressive movement of Homeric discourse seems to facilitate rather open-ended continuation. In chapter 2 I will show that the open-ended character of Homeric clause formation is still subject to grammatical completeness. I will present the model of the grammatical clause as it is defined in functional grammar as the suitable tool to analyse the syntax of Homeric discourse.