3. On the Origins of Dactylic Hexameter
I propose that the archetype of the dactylic hexameter is the pattern pher3d, as it is still used by Alkaios:
1
| U | U | - | u | u | - | u | u | - | u | u | - | u | u | - | U |
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 |
This proposal entails that the original hexameter must have operated on an inherited principle of isosyllabism (as still maintained by Alkaios). Since a pher3d consists of 16 syllables, this conjectured prototype of Greek epic verse matches the syllable-count of the s/loka, the basic unit of Indic epic versification. (By origin, the s/loka is composed of two 8-syllable AnuSTubh verses.)
In order to derive the dactylic hexameter from pher3d, I postulate the following developments in pher3d: (1) the optional replacement of -uu by --; (2) the replacement of UU by --, which is optionally replaced by -uu. Let us first examine how the equation of two shorts (in -uu) with one long (in --) may have come about. Possibly, it is the reflex of a dramatic change in the Greek language, namely, the contraction of vowels after the disappearance of intervocalic *s and *y. Imagine that, where V = 'vowel', the original patterns *...[Vbreve]1s[Vbreve]2... and *...[Vbreve]1y[Vbreve]2... of certain words in certain formulas become ...[Vmacr]3... in the natural language. In that case, the formulas that contained these words set the precedent for two shorts equaling one long within their metrical framework. Then the equation 'two shorts equal one long' may become extended beyond the original confines of the formulas that generated it, so that new formulas with spondee (--) instead of dactyl (-uu) are admitted to epic hexameter. In another work,
2 I have elaborated on such metrical phenomena to the extent of drawing up an axiom. We might say that traditional meter and traditional phraseology are formally correlate in origin. If a formal change occurs within the phraseology, it may trigger a corresponding change in the meter. Thus formulas may set new precedents for meter, which may then accommodate new formulas.
Whatever the linguistic origins of substituting spondees for dactyls, it is possible to demonstrate from the internal evidence of epic hexameter that in certain respects the spondee is alien to the 2nd, 3rd, 4th, 5th feet but not alien to the 1st. Unlike dactyls, spondees generally cannot be followed by a word-break in any foot except the 1st (and of course the 6th). Such a constraint against spondees implies that they may not be basic to the 2nd, 3rd, 4th, 5th feet; conversely, the absence of this constraint at the 1st foot implies that the spondee may be basic there. Thus the current dynamics of epic hexameter allow for at least the possibility of a pher3d prototype:
| U U | - u u | - u u | - u u | - u u | - U |
| 1 2 | 3 4 5 | 6 7 8 | 9 10 11 | 12 13 14 | 15 16 |
| 1st | 2nd | 3rd | 4th | 5th | 6th feet |
We are left with serious problems, however. Why, for example, should UU (1 2) be gradually replaced by -- (and -uu, 1 2 2½)? Also, what was the internal cause for any constraint against word-breaks after spondees? For answers we may find it useful to examine Allen's theory of ictus. Even before we begin such an examination, however, I should note that Allen's theory is not indispensable for my own theory of pher3d origins for the hexameter. It is, however, at the very least, a useful point of departure.
According to Allen, ictus in Greek verse is a direct reflex of stress in the Greek language.
3 The Greek stress-system was originally independent of the Greek intonation-system, the patterns of which are well known from the Attic and Ionic evidence. (For example, we learn that proparoxytone becomes paroxytone when the ultimate vowel is long.) By contrast with the manifold rules for intonation, two will suffice for stress. Allen's formulation is as follows: '(a) [monosyllabic words generally do not count, since they afford no opportunity for contrast of stress within the word]; (b) '
4 (The dots underneath syllables will henceforth indicate stress.) For example,
ἄνδρα μοι ἔννεπε Μοῦσα πολύτροπον ὃς μάλα πολλά
ὦ κοινὸν αὐτάδελφον )Ιςμήνης καρᾶ.
The testing of this formula on vast numbers of dactylic hexameters and iambic trimeters led Allen to his theory that
5 Clearly, the dictum applies: the dynamics of poetic language, whether or not they are still dependent on the natural language, nevertheless originate from the actual grammatical rules of the latter.
6We must take note of an elusive but important implication in these two stress-rules: a given word such as one ending in
-ος will have two different stress-patterns, depending on whether the next word begins with a vowel or consonant. The pattern
-ος V- lacks ictus, vs. the pattern
-?ς C-, with ictus. When a pause follows the final syllable of a word, stress is indifferent and subject to determination by the preceding rhythm.
7From the synchronic point of view, the dactylic hexameter displays the following patterns of ictus:
-/M-/M-/|M-/M-/M-/U (with penthemimeral caesura)
-/M-/M-/u|u-/M-/M-/U (with trochaic caesura)
-/M-/M-/M-/|M-/M-/U (with hephthemimeral caesura)
(The verse-final arrangement is -/U, not -U/, in accordance with the preceding pattern of descending rhythm.)
8 All major types of word-break have been indicated here, except for one:
-/M-/M-/M-/uu|-/M-/U (with bucolic diaeresis)
At this point, we must add a modification based on Allen's theory: whereas the verse-final pattern -/uu|-/M-/U conforms to the ictus-rhythm, any corresponding spondaic type --/|-/M-U upsets ictus-rhythm. The ictus in this instance is forced into the second part of the foot. Accordingly, of plurisyllabic Homeric words with heavysyllable final, only about 7% end with the 4th foot, as opposed to over 50% of plurisyllabic words ending in a dactyl.
9 As for the 5th foot, the apparent innovation of substituting a long for two shorts has had the least effect--a sign of archaism that we might expect from the ending of the verse. But even in the 5th foot, the spondaic incursion is significant enough to leave one spondee for every twenty dactyls. By contrast, a spondaic-final word that ends with the 5th foot is practically nonexistent, with only three certain instances in the whole Homeric corpus.
10 In the 2nd foot as well, the frequency of spondaic word-ending is extremely low: of plurisyllabic words with heavy final, only about 2.6% end with the 2nd foot--this despite the fact that spondees are even more common in the 2nd foot than in the 4th.
11 In sum, spondees may occur in every foot of the hexameter, but spondaic word-endings cannot normally coincide with foot-endings. According to Allen's theory, this constraint helps avoid disruption of the ictus-rhythm in the verse.
The one exception is the 1st foot. Although spondees are only very slightly more common here than in the 4th foot, spondaic word-endings are three times more common; in fact, there is a spondaic word-ending in the 1st foot once in every 10 lines.
12 These facts suggest that the 1st foot of the epic hexameter is not by origin a dactyl (-uu) but a spondee (--). Here, in terms of my theory, the equation of -uu with -- was operative in the reverse order: from spondee to dactyl, not dactyl to spondee as in the 2nd, 3rd, 4th, 5th feet. The spondee of the 1st foot can in turn be traced back to the Aeolic base (UU), which internal lyric evidence shows being gradually restricted to -- and -u in preference to u- and uu.
13
I return to my reconstruction of the epic hexameter as the archetypal pattern pher3d still preserved by Alkaios:
14
| U | U | - | u | u | - | u | u | - | u | u | - | u | u | - | U |
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 |
If we allow for the replacement of UU in (1 2) by -- (1 2) or -uu (1 2 2½) and the optional replacement of -uu by -- anywhere from slots 3 through 14, then our pher3d schema is an adequate means of describing the epic hexameter. For the present, of course, it remains just that and nothing more.
We may test the pher3d hypothesis by applying it to the problem of the major word-breaks in epic hexameter. In terms of our metrical schema, they are found in the following positions:
| U | U | - | u | u | - | u | u | - | u | u | - | u | u | - | U |
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 |
| | * | | | * | * | | * | | * | | | | | |
- ...3|, trithemimeral caesura
- ...6|, penthemimeral caesura
- ...7|, trochaic caesura
- ...9|, hephthemimeral caesura
- ...11|, bucolic diaeresis
Of course, some of these word-breaks are more important than others, and the occurrences of each vary from hexameter to hexameter in any given epic text. Only the penthemimeral and trochaic caesuras approach the status of invariables, with one or the other occurring in about 99% of Homeric hexameters. Beyond these two, the figures drop sharply. The next most common type of word-break, the bucolic diaeresis, occurs in about 60% of Homeric hexameters.
15 The dynamics of these various patterns have been the subject of much fruitful study leading to a refined appreciation of epic meter and its relation to diction,
16 but there has been no successful application to 'a problem like that of the aetiology of a complicated metrical form.'
17 Perhaps the word-breaks do tell us something about the origins of epic hexameter, but it is difficult to understand such regular variations as the one between the penthemimeral and trochaic caesuras. Are we to suppose that the hexameter was partially built from smaller metrical units shaped either UU-uu- or UU-uu-u? Even if it were so, how would we explain the metrical sequences that follow these shapes in hexameter? Presumably, we would have to divide uu-uu-uu-U and u-uu-uu-U into smaller units with shapes depending on whether the next word-break is a hephthemimeral caesura or a bucolic diaeresis. And what about those instances in hexameter where both occur? In aword, the problems are endless if we assume that the five major word-breaks of hexameter mark the junctures of smaller meters. As an alternative, I will apply Parry's theory that these word-breaks mark the junctures of formulas.
In his thèse complémentaire,
18 Parry has shown that two of the most common metrical irregularities in Homeric hexameter occur regularly at the five major word-breaks:
(a) - short in place of expected long before
- trithemimeral caesura -Mu|M-M-M-M-U
- penthemimeral caesura -M-Mu|M-M-M-U
- hephthemimeral caesura -M-M-Mu|M-M-U
(b) - short vowel in hiatus before
- trochaic caesura -M-M-u|u-M-M-U
- bucolic diaeresis -M-M-M-uu|-M-U
These metrical irregularities, Parry noticed, result from formulaic juxtapositions. As a random selection of respective examples, consider the following sets of Homeric hexameters. In the first member of each set, there is a metrical irregularity caused by juxtaposition of a formula. In the second member, we see the same formula juxtaposed in a metrically regular context.
(a)
“μεσσηγὺς | Ἰθάκης τε Σάμοιό τε παιπαλοέσσης
”
vs. metrically regular
“μεσσηγὺς | Τενέδοιο καὶ )/Ιμβρου παιπαλοέσσης
”
θώρηκος γύαλον | ἀπὸ δ᾽ ἔπτατο πικρὸς ὀιστός
vs. metrically regular
“θώρηκος γύαλον | διὰ δ᾽ ἔπτατο πικρὸς ὀιστός
”
ἔνθεν δὲ προτέρω πλέομεν | ἀκαχήμενοι ἦτορ
vs. metrically regular
“ἑξῆμαρ μὲν ὁμῶς πλέομεν | νύκτας τε καὶ ἦμαρ
”
(b)
“λαῶν οἵ οἱ ἕποντο | ἀπ᾽ Αἰσήποιο ῥοάων
”
vs. metrically regular
“λαῶν οἵ οἱ ἕποντο | Τρίκης ἐξ ἱπποβότοιο
”
ἀλλ᾽ ὅ γε μερμήριζε κατὰ φρένα | ὡς Ἀχιλῆα
vs. metrically regular
“τρὶς μὲν μερμήριξε κατὰ φρένα | καὶ κατὰ θυμόν
”
I must stress that the process of formulaic juxtaposition is a basic feature of Greek epic versification,
19 not some sort of deviation from normal techniques. More often than not, this process may be expected to maintain instead of disturb metrical regularity. Where lapses do occur, we are dealing simply with minor faults within the larger framework of a metrically workable system. Accordingly, the second members of the sets listed above should not be viewed as models upon which the first members were imperfectly patterned. Both members of each set are formulaically regular. Each first member is cognate with the second, not derived from it. Here we have the essence of an important yet elusive aspect of Parry's theory, which I would rephrase as follows:
20 Occasionally, formulaic regularity is achieved at the expense of metrical regularity, and the first member of each set above is simply meant to serve as illustration of this hierarchy. The point remains that the poet does not create metrical licenses: he merely inherits them by virtue of inheriting the formulas with which he composes.
21 When we observe the 'striking contrast between the care lavished on the inner metrics of the formulae and the indulgence extended to metrical faults at their junctures with other words or formulae,'
22 we must keep in mind that this 'indulgence' is an archaism, not an innovation.
Let me restate the metrical irregularities at issue: (a) short in place of expected long before trithemimeral, penthemimeral, or hephthemimeral caesura; (b) short vowel in hiatus before trochaic caesura or bucolic diaeresis. Notice that both these features, short for long and short in hiatus, would cease to be metrical irregularities if they occurred in verse-final position rather than verse-medial position before caesura or diaeresis. Furthermore, we actually know that these metrical irregularities occur for the most part in formula-final position. Consequently, I propose that the formulas that end in such metrical irregularities had originated from verses shorter than pher3d. In other words, the pher3d pattern of the archetypal epic hexameter had accommodated the formulas of shorter verses, and it is these formulas that eventually led to the attested caesura- and diaeresis-system of the larger verse.
Pursuing my main hypothesis that the epic hexameter originated from pher3d, I will first assume that the most obvious type of shorter verse capable of transferring its formulas to pher3d is the simple pher itself, UU-uu-U. Let us examine how formulas shaped pher might fit into pher3d:
| B | | | A |
| U | U | - | u | u | - | u | u | - | u | u | - | u | u | - | U |
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 |
| | | B' | | | | | A' |
Such original placements, I will argue, would have led to the generalization of word-break patterns after slot 7 (trochaic caesura), after slot 9 (hephthemimeral caesura), and after slot 11 (bucolic diaeresis).
Complete pher formulas could fit either segments A or B, though those that have any long syllable in the Aeolic base would have had to be screened out of A. Also, B would tend to exclude those complete pher formulas that have uu, u-, or -u at the Aeolic base--in decreasing order of unsuitability. These restrictions on the Aeolic base would have encouraged the incursion of partial pher formulas with the pattern -uu-U, which I symbolize as A' B' vs. A B in the schema above. That is, pher formulas shaped UU|-uu-U, with word-break after the Aeolic base, could fit A' or B' with the actual omission of the segment shaped UU. Since internal expansion by dactyls had originally occurred between UU and -uu-U,
23 the segmentation UU|-uu-U should be commonplace in terms of the simple pher itself. As for the epic hexameter, the presence of complete/partial pher formulas in segments A/A'
24 seems borne out by statistical evidence. A word-break after slot 9 occurs in 45.4% of Iliadic verses and in 51.7% of Odyssean verses.
25 Furthermore, 19.8% and 25.4% of Iliadic and Odyssean verses respectively have word-break after both slots 9 and 11, so that approximately one-fourth of Homeric verses have at least a formula-segment A when there is no formula-segment A'.
26 For an illustration of interacting A'/A formulas, consider the following doublet:
A' “
...λῦσε δὲ γυῖα #”
（Δ 469, etc.), -uu-U = pher minus Aeolic base
vs. A “
...ὑπέλυσε δὲ γυῖα #”
（Ο 581, etc.), uu-uu-U = pher
Such formulas are functional variants in Homeric diction, as we may see from the formal mirroring of the above expressions in the mediopassive voice:
27
A' “
...λύντο δὲ γυῖα #”
（Η 16, etc.), -uu-U = pher minus Aeolic base
vs. A “
...ὑπελύντο δὲ γυῖα #”
（Π 341), uu-uu-U = pher
Because of the rhythmical differences between UU (1 2) and uu (10 11), where our current hypothesis has UU (1 2) evolving into -- or -uu, we should not expect to find formulas shared by segments A and B. On the other hand, formulas shaped -uu-U may fit both segments A' and B'. The shape -uu-U is familiar from Lyric, where it is called the Adonic after the famous Sapphic refrain
“ὦ τὸν Ἄδωνιν
”
(168LP)
-uu-U
Among Adonic formulas shared by segments A' and B' of epic hexameter, there is one important restriction: B' accommodates formulas shaped -uu-u but screens out those shaped -uu--. In other words, formulas shared by A'/B' have an overt short at slot 7 (u) and a latent short at slot 16 (U). From Hainsworth's tables,
28 I cite the following list of Homeric formulas which occur in both segments A' and B'. At the right of each formula is the number of Homeric occurrences in the A' and B' positions respectively:
Among these formulas shaped -uu-u, the proportion of occurrences in segments A' vs. B' suggests that the primary locus of diffusion was in the verse-ending, A'. In some instances, we may even find grammatical proof of A'-provenience. For example, the archaic neuter dual adjective
φαεινώ (u--) in the verse-final formula
ὄσσε φαεινώ# (N 3 etc.) is switched to an innovated neuter plural
φαεινά (u-u) in the corresponding formula which appears before the trochaic caesura,
ὄσσε φαεινά| (N 435)-- this despite the fact that the substantive
ὄσσε remains a neuter dual.
29Adonic formulas may also occur at two other positions within the hexameter. One of these is slots
; for example,
νηυσὶ θοῇσι occurs in this position (
δ 173) as well as in slots
(= A') and
(= B'). We see here an innovation in terms of our hypothesis, since such a placement would be possible only after the Aeolic base
was replaced by a dactylic
. The second other position possible for Adonic formulas is
; consider the 15 attestations of
πατρίδα γαῖαν in these slots, vs. the 62 instances of the same formula in slots
(= A') and the solitary instance in
(= B'). But this second too is an innovation, as we will see below.
30Of course, I do not intend to suggest that all B' formulas stem from A' counterparts. There are, in fact, several Homeric formulas shaped -uu-u which regularly occur in section B', never in A'. Again using Hainsworth's tables,
31 I cite the following examples, with the number of Homeric occurrences at the right:
On the basis of such evidence, I am reluctant to rule out at least the possibility that some complete or partial pher formulas were originally restricted to the opening of the pher3d meter. Of course, there are also a vast number of formulas shaped -uu-u that are restricted to the ending of the hexameter. Consider the following list of Homeric formulas that occur in A' but never in B':
32
Besides formulas shaped -uu-U, there is also another readily available configuration of partial pher formulas, namely uu-U (from UU-|uu-U). Parallel to the A'/B' alternation of formulas shaped -uu-u, we find an a/b alternation of formulas shaped uu-u:
| | B' | | A' |
| U | U | - | u | u | - | u | u | - | u | u | - | u | u | - | U |
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 |
| | b | | a |
What follows is a list of examples,
33 with the number of Homeric a vs. b occurrences at the right:
Again, we also find formulas shaped uu-u which occur in section b to the exclusion of a, such as
34
or in section a to the exclusion of b, such as
35
So far, I have tried to motivate the epic word-breaks after slot 7 (trochaic caesura), slot 9 (hephthemimeral caesura), and slot 11 (bucolic diaeresis) by assuming the accommodation of complete or partial pher formulas within a pher3d meter:
| | B' | | A' |
| | b | | a |
| U | U | - | u | u | - | u | u | - | u | u | - | u | u | - | U |
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 |
| B | | A |
There remains the task of motivating the functional variant of the trochaic caesura in epic hexameter, namely the penthemimeral caesura (word-break after slot 6). I propose that this pattern arose from the accommodation of formulas shaped pherd, specifically uu-uu-uu-U = bracket C:
| | C |
| U | U | - | u | u | - | u | u | - | u | u | - | u | u | - | U |
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 |
| | A |
Consider the following sets of A vs. C formulas lodged in the endings of Homeric hexameters:
A “
...δέπας ἀμφικύπελλον #”
（Α 584, etc.), uu-uu-U = pher
vs. C “
...δέπας οἴσεται ἀμφικύπελλον”
（Ψ 663, etc.), uu-uu-uu-U = pherd
A “
...νέας ἀμφιέλισσας #”
（Ρ 612, etc.), uu-uu-U = pher
vs. C “
...νέες ἤλυθον ἀμφιέλισσαι #”
（Ν 174, etc.), uu-uu-uu-U = pherd
A “
...βοὸς ἀγραύλοιο #”
（Κ 155, etc.), uu-M-U = pher
vs.C “
...βοὸς ἔρχεται ἀγραύλοιο #”
（Χ 403), uu-uu-M-U = pherd
A “
...χθονὶ πουλυβοτείρῃ #”
（Γ 89, etc.), uu-uu-U = pher
vs. C “
...χθονὶ πίλνατο πουλυβοτείρῃ #”
（Ψ 368), uu-uu-uu-U = pherd
A “
...χροὸς ἀνδρομέοιο #”
（Ρ 571, etc.) uu-uu-U = pher
vs. C “
...χροὸς ἄμεναι ἀνδρομέοιο #”
（Φ 70) uu-uu-uu-U = pherd
Such formulaic intrusions of
οἴσεται, ἤλυθον, ἔρχεται, πίλνατο, ἄμεναι in Epic correspond to the metrical intrusions of the dactyl in Lyric, a process that I have called internal expansion.
36In this connection, I now cite another factor in lyric meter for which we find an analogue in epic formula:
37
- pher = UU -uu -U (or UU + d' + -U）
- pherd= UU -uu -uu -U (UU + d d' + uU）
- pher2d= UU -uu -uu -uu -U (UU + d d d' + uU）
- pher3d= UU -uu -uu -uu -uu-U (UU + d d d d' + uU）
Although d' was the model of d from the standpoint of my reconstruction, it is bound to lose any formal distinction from the plain dactyl. From the mechanical standpoint of poetic composition, the d' of an expanded pher is perceived simply as one of a series of dactyls, with the series framed by UU on one side and -U on the other. It follows from this inherent symmetry that the dynamics of prosody allow for dactylic expansion of a formula at either of two junctures rather than at only one:
either UU + -uu + -uu-U
or UU-uu + -uu + -U
= UU-uu-uu-U, pherd
As we have just seen, alternations between formulas shaped pher and pherd in epic hexameter reveal expansions of the type
...uu + -uu + -uu-U
In such instances, the formulaic expansions of Epic match the metrical expansion of Lyric. Accordingly, I should also expect epic hexameter to reveal formulaic expansion of the type
...uu-uu + -uu + -U
For verification, consider the following sets of A vs. C formulas in the endings of Homeric hexameter:
A “
...μεγαλήτορι θυμῷ #”
（Ι 109), uu-uu-U = pher
vs. C “
...μεγαλήτορι ἥνδανε θυμῷ #”
（Ο 674), uu-uu-uu-U = pherd
A “
...περικαλλέα δίφρον #”
（υ 387), uu-uu-U = pher
vs. C “
...περικαλλέα βήσετο δίφρον #”
（Γ 262), uu-uu-uu-U = pherd
A “
...κορυθαίολος (/Εκτωρ #” (passim), uu-uu-U = pher
vs. C “
...κορυθαίολος ἠγάγεθ᾽ (/Εκτωρ #”
（Χ 471), uu-uu-uu-U = pherd
A “
...Τελαμώνιος Αἴας #” (passim), uu-uu-U = pher
vs. C “
...Τελαμώνιος ἄλκιμος Αἴας #”
（Μ 349), uu-uu-uu-U = pherd
A “
...φύγαδ᾽ ἔτραπεν ἵππους #”
（Θ 257), uu-uu-U = pher
vs. C “
...φύγαδ᾽ ἔτραπε μώνυχας ἵππους #”
（Θ 157), uu-uu-uu-U = pherd
Again, the formulaic expansions with
ἥνδανε, βήσετο, ἠγάγεθ᾽, ἄλκιμος, μώνυχας in Epic correspond to metrical expansions with the dactyl in Lyric.
At this point, a word of caution is in order. I am not claiming that dactylic expansion of formulas at the end of epic hexameters should be viewed as some sort of operative device at the time of Homeric composition. Rather, the formulaic variants shaped ...uu|-uu-U vs....uu|-uu|-uu-U and ...uu-uu|-U vs. ...uu-uu|-uu|-U are parallel inheritances. As such, I treat these formulaic patterns as survivals of archaic precedents that at various points of their evolution may or may not have triggered other precedents in epic formula and meter. My task now is to evaluate some of these other precedents, with the purpose of linking them to the phenomenon of dactylic expansion.
In the patterns ...uu -uu -uu-U and ...uu-uu -uu -U,
we find an important application to epic meter as well as formula. Let us begin with the obvious. The word-breaking pattern
...uu-uu-uu|-U
is a common phenomenon of epic hexameter. Among Homeric words with the shape --, 41% occur in the 6th foot.
38 By contrast, the word-breaking pattern
...uu-uu-uu-|U
is rare. In Books
ΑΒΓΔΕΖΩαβγω, it occurs in only 1% of the hexameters.
39 Notice too that the word-breaking pattern
...uu-u|u-uu-U
is even more rare, occurring in only 0.1% of Homeric hexameters. In this instance, the constraint against word-breaking is so noticeable that it goes by a name, Hermann's Bridge.
40 Let us now contrast these rare patterns
...uu-uu-uu-|U
...uu-u|u-uu-U
with the common patterns
...uu-uu-uu|-U
...uu-uu|-uu-U
I submit that we see here a reflex inherited by hexameter from pher and pherd formulas. A pher formula can be expected to shun the patterns
UU-uu-|U
U|U-uu-U
for the simple reason that dactylic expansion could then result in pherd formulas shaped
| UU-uu -uu -|U | = | UU-uu|-uu|-|U |
| U|U -uu -uu-U | = | U|U|-uu|-uu-U |
besides
| UU -uu -uu-|U | = | UU|-uu|-uu-|U |
| U|U-uu -uu -U | = | U|U-uu|-uu|-U |
The unwieldy succession of two monosyllabic words in the first two of these four types rules them out as viable patterns. So far, then, the decisive factor seems to be the structure of a pherd, not that of a pher.
As a corollary to Hermann's Bridge, we see a reflex of dactylic expansion in the marked avoidance of spondees in the 4th foot, slots 9-10-11 (-uu), if a word-break follows. This word-break is the bucolic diaeresis:
| U | U | - | u | u | - | u | u | - | u | u | | - | u | u | - | U |
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | | 12 | 13 | 14 | 15 | 16 |
I have already discussed this constraint against spondaic word-endings in general terms of ictus. To be more specific now, I note that the innovation of substituting one long for two shorts has significantly pervaded the 4th foot. Spondees occur there with noticeable frequency. Here are statistics from a sample of 12,866 Homeric verses, the first twenty books of the Iliad:
- in 1st foot, 5,070 spondees
- in 2nd foot, 5,171 spondees
- in 3rd foot, 1,966 spondees
- in 4th foot, 3,849 spondees (!)
- in 5th foot, 788 spondees41
Thus the proportion of dactyl to spondee in the 4th foot, 2.3 to 1, does not differ appreciably from the overall average for the first four feet of hexameter, which is, 2.6 dactyls to 1 spondee.
42 By contrast, when a word-break occurs after slot 11, dactyls (-uu) outnumber spondees (--, 9 10=11) 8 to 1; when there is a clause-break after 11, the figure is even higher, 20 to 1.
43 Where a word shaped ...-u might have ended at 9 10, we find instead a variant shaped ...-uu covering 9 10 11. I choose my examples from Meister's collection:
44
- πτολιπόρθιον at 7 8 9 10 11 (ι 504, 530)
vs. πτολίπορθος, etc. elsewhere (passim) - ἀέθλια at 8 9 10 11 (θ 108, φ 62)
vs. ἄεθλον, ἄεθλα, etc. elsewhere (ca. 30x) - Αἰτώλιος at 7=8 9 10 11 (Δ 399)
vs. Αίτωλός elsewhere (Δ 527, etc.) - δαφοινεόν at 8 9 10 11 (Σ 538）
vs. δαφοινός, etc. elsewhere (Β 308, etc.) - μητέρος at 9 10 11 (Τ 422, Ω 466, ξ 140, ο 432, ς 267）45
vs. μητρός elsewhere (passim) - ὤρετο at 9 10 11 (Μ 279, Ξ 397, Χ 102）46
vs. ὦρτο elsewhere (ca. 50x)
From such formulaic evidence, I infer that ictus, Hermann's Bridge, bucolic diaeresis, and the dactyl preceding it are all interlocking phenomena, reflecting an original pherd formulaic structure with dactylic expansion between uu and -uu-U. The expansion results in the word-breaking pattern ...uu-uu|-uu-U of epic hexameter.
It remains to ask why the expansion does not leave an equally distinct imprint on the other side of the dactylic shape. To put it another way, the question is why we should not find the pattern
...|uu|-uu|-uu-U
as consistently as the pattern
...|uu-uu|-uu-U
I propose that the answer lies in the alternative pattern of dactylic expansion, between uu-uu and -U. The pherd formulas with this shape should yield the following pattern within the meter of pher3d:
...|uu-uu|-uu|-U
The result is a strong precedent for formulas before the bucolic diaeresis which are shaped uu-uu without necessarily having word-break between uu and -uu: hence the type
πτολιπόρθιον.
We are still left with the major problem of accounting for formulaic elements shaped u-uu before the bucolic diaeresis, such as
ἀέθλια, δαφοινεόν, etc. Notice that the word-break preceding such shapes is the trochaic caesura:
| X | y |
| U | U | - | u | u | - | u | | u | - | u | u | | - | u | u | - | U |
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | | | 8 | 9 | 10 | 11 | | | 12 | 13 | 14 | 15 | 16 |
I have already proposed that the trochaic caesura is a reflex of pher formulas placed at the opening of a pher3d meter, resulting in what I will now call segment X. This explanation of the trochaic caesura will have to be modified in the succeeding discussion, and we may start by asking the obvious question: what is segment y? I submit that this y is by nature a pherd formula minus the first syllable. From the metrical point of view, we could call it an acephalic pherd (= ^pherd). Of course, this hypothesis will have to be reconciled with the functional variant of the trochaic caesura, the penthemimeral caesura:
| x | Y |
| U | U | - | u | u | - | | u | u | - | u | u | - | u | u | - | U |
| 1 | 2 | 3 | 4 | 5 | 6 | | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 |
(To justify my designation of functional variant, let me reiterate: in 99% of the Homeric hexameters, we find either a trochaic caesura or a penthemimeral caesura.) Did Y/y originate from formulas shaped pherd/^pherd?
Before going any further, I must first refine my definition of sector Y. We expect it to contain formulas shaped pherd, but its formulas need not consistently bear the overt marks of dactylic expansion. We have already seen that the word-break patterns
- (1) ...uu|-uu|-uu-U (L'|R) and
- (2) ...uu-uu|-uu|-U (L|R'),
which preserve both frames of internal dactylic expansion, are not nearly as frequent as the pattern containing only the bucolic diaeresis,
...uu-uu|-uu-U,
where only the right frame (R) of pattern 1 and the left frame (L) of pattern 2 survive. The statistical predominance of the bucolic diaeresis, which occurs in about 60% of Homeric hexameters, is due partially to the convergence of R and L. On the other hand, the divergence of R' and L', which mark the two divergent ways of expanding pher into pherd, weakens the precedent for observing word-break at either of those two places. In short, the word-break patterns in sector Y of epic hexameter, uu-uu-uu-U, serve to show that it is by origin the locus of pherd formulas, but we must now shift from the standpoint of reconstruction to that of current dynamics in the Homeric composition of verses.
In the verses of the Iliad and Odyssey, the formulaic input of sector Y is not confined to pherd formulas. Granted, I am arguing that pherd formulas had provided the initial precedent for the shape of sector Y, as I infer from the statistics of word-break and word-bridge patterns (especially the bucolic diaeresis and Hermann's Bridge) and from the shapes of formulas like
“
...| δέπας | οἴσεται | ἀμφικύπελλον #”
（Ψ 663, etc.)
“
...| περικαλλέα | βήσετο | δίφρον #”
（Γ 262)
Nevertheless, within the doubtless lengthy prehistory of epic hexameter, sector Y must have evolved a formulaic framework that is more versatile than that of any original pherd formulas. Within the format of Epic, pherd formulas could not exist in their own right, but rather, must have coexisted with partial or complete pher formulas encased at the end of a pher3d meter. There is, however, an observable hierarchy between such pherd and pher formulas, in that the second are subordinate to the first. For example, partial pher formulas shaped -uu-u may shift positions within the framework of a pherd or ^pherd pattern (slots 7-16 or 8-16):
| | A'' | |
| U | U | - | u | u | - | u | u | - | u | u | - | u | u | - | U |
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 |
| | A |
A partial pher formula like
πατρίδα γαῖαν, which occurs 62 times in sector A of Homeric hexameter, also occurs 15 times in sector A''. From Hainsworth's tables,
47 I cite the following additional examples, with the respective number of Homeric A vs. A'' occurrences at the right:
The subordination of these partial pher formulas to a larger pherd framework is also made evident by an interesting constraint. If a formula shaped -uu-u is to be switched from sector A to sector A'', it must not consist of a two-syllable word followed by a three-syllable word. Notice that such a combination would violate the pattern of dactylic expansion in a pherd. In terms of a pherd formula within a pher3d meter, dactylic hexameter requires a word-break after slot 11 rather than 10. To rephrase from the standpoint of epic hexameter as it is attested in the Homeric phase, we may say that such a combination would violate Hermann's Bridge.
Besides partial pher formulas shaped -uu-u, those shaped uu-u are also found in shifting positions within the framework of a pherd or ^pherd pattern (slots 7-16 or 8-16):
| | a'' | |
| U | U | - | u | u | - | u | u | - | u | u | - | u | u | - | U |
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 |
| | a |
From Hainsworth's tables,
48 I cite the following examples, with the respective number of Homeric a vs. a'' occurrences at the right:
Here too we find a constraint that illustrates the subordination of these partial pher formulas to a larger pherd framework. Formulas shaped uu|-u may be shifted from slots 13 14 15 16 (= a) into slots 10 11 12 13 (= a''), but not into slots 7 8 9 10. Such a shift would violate the pattern of dactylic expansion in a pherd. In terms of a pherd formula within a pher3d meter, dactylic expansion requires a word-break after slot 11 rather than 10. Again, I have expressed this constraint from the standpoint of my reconstruction. From the standpoint of the hexameter as it is attested in the Homeric phase, we may again say instead that such a shift would violate Hermann's Bridge.
To sum up: sector Y(/y) of epic hexameter is a composite that evolved out of pherd (/^pherd) and pher formulaic elements. We see the traces of these elements in the following word-break patterns:
- ...(u) u|-uu|-uu-U, (^)pherd formula
- ...(u) u-uu|-uu|-U, (^)pherd formula
- ...(u) u-|uu-uu-U, pher formula within (^)pherd
- ... (u) u-uu|-uu-U, partial pher formula A within (^)pherd
- ... (u) u|-uu|-u|u-U, partial pher formula A'' within (^)pherd
- ... (u) u-uu-|uu-U, partial pher formula a within (^)pherd
- ... (u) u-|uu-u|u-U, partial pher formula a'' within (^)pherd
It is, then, an oversimplification to say that sector Y consists of pherd formulas. Rather, it consists of formulas that evolved within the framework of a pherd formulaic structure.
Having recorded this important qualification, I return to the problem of the two fundamental caesuras in epic hexameter, the penthemimeral and the trochaic:
| x | Y |
| U | U | - | u | u | - | u | u | - | u | u | - | u | u | - | U |
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 |
| X | y |
| U | U | - | u | u | - | u | u | - | u | u | - | u | u | - | U |
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 |
Notice that Homeric formulas shaped Y have functional variants shaped y:
49
The Y/y variations are especially noticeable when they are composed entirely of epithet + name combinations, as in the genitive singular:
Turning to corresponding epithet + name combinations in the nominative singular, however, we find a highly interesting constraint in Homeric diction. Despite numerous instances of epithet + name shaped y, there is a marked absence of variants shaped Y applying to the same person. For example, consider the lack of Y formulas corresponding to the following designations with y formulas:
As Parry has noticed,
50 there is a basic difference in combinatory behavior between such nominative formulas and any corresponding genitive formulas. The first are normally subjects of preceding verbs, while the second serve as dependents of preceding nouns far more often than as objects of preceding verbs.
51 Parry infers that it may well be the shape of preceding verbs that results in the preponderance of nominatives shaped y over nominatives shaped Y. Let us examine the verbs immediately preceding a formula like
πολύτλας δῖος )Οδυσσεύς:
μερμήριξε, ἦσθε (2x),
ἄκουσε, ἐσάκουσε, καθεῦδε (2x),
ἠρᾶτο, γήθησε (2x),
ἐνόησε, ἀνάειρε (2x),
θηεῖτο, ἄρχε (2x),
ἐβόησε, προσέειπε (8x). Parry links such formal evidence with a basic functional feature of Greek Epic. Consisting primarily of narrative, Homeric diction furthers the preponderance of past tense and third person, for which the basic Greek form is
-ε.52 If, the reasoning goes, the verb is to be placed immediately before the nominative epithet + noun combination, then our formula must be shaped u-M-M-U (= y) to make room for the last syllable of the verb. Presumably, any variant shaped uu-M-M-U (= Y) would not be of much use.
It is possible to go beyond Parry's analysis of the situation by reexamining it from the standpoint of X and x, the initial sectors of hexameter. The shape UU-uu- (= x) differs from the shape UU-uu-u (= X) in that it should not be expected to end with a verb. So much for a reformulation; the problem is, it simply does not work. I can think of several Homeric verbs that could theoretically be placed in exactly this position, at the end of a sequence UU-uu-:
ἀπέβη, μίγη/
ἐμίγη, μετέφη, προσέφη, etc., not to mention any 3rd singular
-εν followed by a consonant starting the next word. And yet, strangely, there is no such regular pattern in Homeric diction. For example, let us consider the common Homeric formula for introducing a speech:
X #
τοῖσι δὲ καὶ μετέειπε |...,
followed by such nominative epithet + name combinations as
etc.
Contrast the total absence of Homeric formulas shaped
*x #
τοῖσι δὲ καὶ μετέφη |...
In fact, Homeric diction never features
μετέφη before the penthemimeral caesura:
*x #-uu-
μετέφη|...
And yet, there is no purely metrical factor that should prevent the occurrence of this verb
μετέφη in the sequence uu- before the penthemimeral caesura. Consider its regular placement in the sequence uu- before the hephthemimeral caesura:
“
τοῖσιν δ᾽ )Αντίνοος μετέφη | Εὐπείθεος υἱός”
（δ 660, etc.)
“
τοῖς δὲ δολοφρονέων μετέφη | πολύμητις )Οδυσσεύς”
（ς 51),
φ 274 etc.
Even on those rare occasions when a nominative epithet + name combination shaped Y is available, such as
...
|ἱερη )ὶς Τηλεμάχοιο #, uu-M-uu-U,
the preceding formula (introducing a speech) eschews
μετέφη:
“
τοῖσι δὲ καὶ μετέειφ᾽|ἱερὴ )ὶς Τηλεμάχοιο”
（β 409, etc.),
instead of
*
τοῖσι δὲ καὶ μετέφη|ἱερὴ )ὶς Τηλεμάχοιο
Nor does it matter that
μετέφη is here followed by a vowel. Consider again
“
τοῖσιν δ᾽ )Αντίνοος μετέφη|Εὐπείθεος υἱός ”
（δ 660, etc.),/bibl>53 Why, then, does sector x fail to end with
μετέφη? Since meter does not seem to be a direct factor, perhaps we should look to formulaic behavior for an answer.
Let us pursue the issue by taking up another example. We have a host of Homeric attestations for the formula
X #
τὸν δ᾽ αὖτε προσέειπε|...,
but there is not a single instance of
*x #
τὸν δ᾽ αὖτε προσέφη|...
In fact, out of the 213 instances of
προσέφη in the Iliad and Odyssey, the word occurs 211 times before the hephthemimeral caesura, but only twice before the penthemimeral:
“
ὥς πού σε προσέφη|σοὶ δὲ φρένας ἄφρονι πεῖθε”
（Π 842)
“
ἔνθα χ᾽ ὅμως προσέφη|κεχολωμένος, ἤ κεν ἐγὼ τόν”
（λ 565)
The anomaly of these verses is noticeable not only from the position but also from the usage of
προσέφη. Unlike the vast majority of Homeric verses containing
προσέφη, these two neither introduce a quotation nor give the nominative epithet + noun for the person who is being quoted. Contrast the following familiar patterns:
The verb
προσέφη can be followed by a subject like
|“
μένος )Αλκινόοιο” # (
η 178, etc.)
but not by
...
|ἱερὸν μένος )Αλκινόοιο #
The latter formula, shaped Y, is available as the subject, and yet we never find a Homeric verse shaped
*
τὸν δ᾽ αὖτε προσέφη | ἱερὸν μένος )Αλκινόοιο
The first part of such a verse, shaped x, is wanting. Furthermore, whenever it happens in the Odyssey that
ἱερὸν μένος )Αλκινόοιο serves as subject of the clause, any verb immediately preceding it will originate not from some x formula, but from a modified X formula:
In such instances of ?|Y, I infer that there is no inherited formula x which would be available to take the place of X when a subject shaped Y takes the place of y. Furthermore, from our data on the verbs
μετέφη and
προσέφη, I infer that inherited x formulas resist verbs in final position. Note again the contrast with the X formulas, which regularly accommodate verbs in final position.
For another striking example of X/x contrast, I cite the formulaic behavior of verbs in 3rd singular
-ε plus movable
ν before single consonant (C-) starting the next word. In Homeric diction, the pattern
-εν|C- is extremely rare at the penthemimeral caesura, while it is relatively common at the hephthemimeral. Consider the following statistics:
54
I do not mean to suggest, however, that these verbs cannot end on syllable 6 of our schema for the hexameter:
| U | U | - | u | u | - | ↓ | u | u | - | u | u | - | u | u | - | U |
| 1 | 2 | 3 | 4 | 5 | 6 | | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 |
Such verbs as
δῶκεν can and do regularly end on syllable 6, but only if they are followed by a monosyllabic particle:
αὐτὰρ ἐπεὶ δῶκέν τε|καὶ ἔκπιον, αὐτίκ᾽ ἔπειτα
Technically, the initial part of this verse ends at the trochaic caesura, not the penthemimeral. The
τε is syntactically dependent on and inseparable from the
δῶκεν, so that the formula ends as
...δῶκέν τε. In other words,
δῶκεν functions here as part of an X rather than an x formula. What, then, would happen with such a combination at the hephthemimeral caesura? Let us compare the behavior of a verb like
λοῦσεν when it ends at slot 6 as opposed to when it ends at slot 9:
...λοῦσέν τε|καὶ ἔχρισεν λίπ᾽ ἐλαίῳ #
,
κ 364,
450
...λοῦσεν|καὶ χρῖσεν ἐλαίῳ #
,
ω 366
The combination
λοῦσέν τε καί would have been metrically impossible in the latter instance. If a verb ending at slot 9 is followed by a monosyllabic enclitic, this enclitic in turn must be followed by another enclitic:
...πόρεν δέ ἑ|Φαίδιμος ἥρως #
,
ο 117
...πόρεν δέ οἱ|ἀγλαὸν υἱόν #
...πόρεν δ᾽ ὅ γε|σήματα λυγρά #
...δῶκεν δέ μοι|εὖχος )Αθήνη #
55
In other words, Hermann's Bridge is not violated. The central issue, however, remains this: when it ends at slot 6, a verb like
λοῦσεν functions as part of an X rather than x formula.
Consider also the common X formula
#
ὣς φάτο μείδησεν δὲ |...,
followed by such epithet + name combinations as
Other such formulas are
In these contexts, we even see two instances where the nominative epithet + name combination is shaped Y rather than y:
ὣς φάτο μείδησεν δ᾽ | ἱερὴ ἲς Τηλεμάχοιο
ὣς φάτο γήθησεν δ᾽ | ἱερὸν μένος )Αλκινόοιο
To accommodate Y, the original X is simply modified, apparently for lack of any inherited formulaic variant of X shaped x. Likewise, among the rare Homeric attestations of
-εν|C- at the penthemimeral caesura, the x formula involved may be the result of modification instead of direct inheritance. For example, a line like
ὣς ἄρ᾽ ἐφώνησεν | τῇ δ᾽ ἄπτερος ἔπλετο μῦθος
seems to be part of a subset which includes the common types
Accordingly, the pattern
ἐφώνησεν|C- at the penthemimeral caesura cannot serve as counter-example to my argument that verbs in
-εν are wanting at the end of inherited x formulas.
Let us return to the original problem as viewed from Parry's standpoint. He observed that there are very few occurrences of nominative epithet + noun formulas shaped
uu-M-M-U,
as opposed to a multitude of functional equivalents shaped
u-M-M-U.
What I am proposing here is that this Y/y distinction in formulaic behavior is but a symptom of a more basic x/X distinction. To rephrase the problem, we must ask why Parry's solution for the Y/y distinction was that the past 3rd singular verbs that immediately precede the main caesura have the shape that requires a trochaic rather than penthemimeral caesura. It becomes obvious from our survey, however, that there are several past 3rd singular verbs that could ideally fit before the penthemimeral caesura and yet shun this very position. I infer that we see here an aspect of the formulaic heritage: formulas shaped x behave differently from those shaped X because they are not inherited variants. Consider such X formulas as
The verbs of these formulas recur in verse-final position as well:
As I have already argued,
56 such interchange in positions before caesura and before verse-end is a characteristic of partial pher formulas:
| | B' | | A' |
| U | U | - | u | u | - | u | | u | - | u | u | - | u | u | - | U |
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | | | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 |
| | b | | a |
What I am now proposing is that X is a repository of partial pher formulas shaped B' and b,
57 and that it is this capacity that makes such a segment X distinct in formulaic behavior from x. This distinction, I should repeatedly emphasize, implies that X and x were not formulaic variants by origin.
Going even further, we can say that UU-uu-u (= X) and UU-uu- (= x) cannot be interpreted as metrical variants either, since there is to my knowledge no such thing as an Aeolic meter shaped
*UU-uuU = pher^ (catalectic pher),
derived from
UU-uu-U = pher
By contrast, y and Y are functional variants both from the standpoint of epic formula, as in
and from the standpoint of lyric meter:
U-uu-uu-U = ^pherd (acephalic pherd),
derived from
UU-uu-uu-U = pherd
The two types ^pherd and pherd are metrical variants not only in structure but also in usage. Consider the lines of Sappho 111LP:
58
ἴψοι δὴ τὸ μέλαθρον: ---uu-U = pher
ὐμήναον:
ἀέρρετε τέκτονες ἄνδρες: u-uu-uu-U = ^pherd
ὐμήναον.
γάμβρος ἔρχεται ἴσος )/Αρευι: ---uu-uu-U = pherd
ἄνδρος μεγάλω πόλυ μέζων. --uu-uu-U = ^pherd
Compare also the Rhodian folk-song known as the Chelidonismos or 'Swallow Song' (848P), where ^pher and pher are clearly variants in the first nine lines:
ἦλθ᾽ ἦλθε χελιδὼν --uu-U ^pher
καλὰς ὥρας ἄγουσα, -u-uu-U pher
καλοὺς ἐνιαυτούς, 59--uu-U ^pher
ἐπὶ γαστέρα λευκά, uu-uu-U pher
ἐπὶ νῶτα μέλαινα. uu-uu-U pher
παλάθαν σὺ προκύκλει uu-uu-U pher
ἐκ πίονος οἴκου --uu-U ^pher
οἴνου τε δέπαστρον --uu-U ^pher
τυροῦ τε κάνυστρον. --uu-U ^pher
The type pherd is used line-by-line (=
κατὰ στίχον) in Sappho 110LP, as also ^pherd + -u-u-U in Archilochos (168-171W).
60 In sum, the lyric metrical variation of verses shaped pherd and ^pherd corresponds to the epic formulaic variation of segments shaped Y and y.
Once the innovation of substituting one long for two shorts takes effect in the evolution of the epic hexameter, any original ^pherd formulas shaped
--uu-uu-U (which we will designate Y')
may then be inserted in place of pherd formulas shaped
uu-uu-uu-U (= Y).
From the metrical standpoint, the shape Y' is a variant of Y. From the formulaic standpoint, however, Y' may also be a variant of y:
As Parry has noticed,
61 the epithets in the last example are found in complementary distribution. Either epithet can apply to any hero, depending on the space available before the bucolic diaeresis. After a trochaic caesura, use
δι?φιλος; after a penthemimeral,
δουρίκλυτος. We may wish to reinterpret as follows:
δι?φιλος and
δουρίκλυτος are reflexes of epithets that were originally meant to start ^pherd formulas shaped
u-uu|-uu-U (= y)
and
--uu|-uu-U (= Y')
respectively. Also, the word-break pattern within these shapes is simply an indirect reflex of dactylic expansion. Epithets were not inherited to fill gaps between caesura and diaeresis. They were inherited to fill out formulas. In other words, the bucolic diaeresis is motivated by the length of epithets like
δι?φιλος and
δουρίκλυτος, not vice versa.
Within the metrical framework of a pher3d, both Y' and Y formulas require a preceding x formula, while y requires X. As I have already argued, the X sector is distinct from x by virtue of its capacity to accommodate partial (or complete) pher formulas. The only portion of the pher3d meter that cannot accommodate some kind of Pherecratic formula is sector x. Accordingly, I propose that the least traditional component of epic hexameter is to be found in the x sector. By the same token, this sector must afford the maximum opportunity for phraseological improvisation. Having just outlined such a hierarchy of XxYY'y, I should now attempt to speculate on the mechanics of composing an epic hexameter. I envisage the following set of original alternatives:
- 1. Begin with any group of words shaped UU-uu-; continue with a formula shaped pherd, featuring initial uu instead of u-, -u, --
- 2. Begin with a pher formula, or at least with a group of words that end with a pher formula; continue with a formula shaped ^pherd, featuring initial u instead of -
At a later stage of evolution, another alternative develops:
- 3. Begin as in no. 1; continue with a formula shaped ^pherd, featuring initial - instead of u
After the Aeolic base (UU) was replaced by the dactyl/spondee (-M) in epic hexameter, the initial sequence x becomes identical to the Hemiepes, -uu-uu-.
62 The traditional name Hemiepes begs the question, of course, in view of the fact that dactylic hexameter is the Epos par excellence. Still, it remains valid to claim that the elegiac pentameter,
-M-M-|-uu-uuU,
did indeed originate from an acephalic/hypersyllabic Prosodiakon in distich form.
63 The strict maintenance of isosyllabism in the second part of the distich is a sure sign of archaism. Constraints here against the substitution of one long for two shorts indicate that a verse-ending is involved, not some hypothetically reapplied verse-opening. A less direct but still significant sign of archaism is the intentional practice of rhyming the last syllable before the caesura with the last syllable of the verse.
64 For example, in Theognis 174, 176, 178, 180, 182, there is a whole series of this pattern in the pentameters between the hexameters:
καὶ γήρως πολιοῦ|Κύρνε καὶ ἠπιάλου
ῥιπτεῖν καὶ πετρέων|Κύρνε κατ᾽ ἠλιβάτων
οὔθ᾽ ἕρξαι δύναται|γλῶσσα δέ οἱ δέδεται
δίζησθαι χαλεπῆς|Κύρνε λύσιν πενίης
ἢ ζώειν χαλεπῇ|τειρόμενον πενίῃ
In fact, such rhyme-patterns exist in one out of every seven Theognidean pentameters. The effect is to accentuate the rhythmical symmetry of the elegiac pentameter vs. the asymmetry of the epic hexameter.
65 Of course, the actual coexistence of epic hexameter with elegiac pentameter in the framework of an elegiac couplet favors the incursion of formulas with the shape of a Hemiepes at the start of hexameter verse. To put it another way, the frequency of penthemimeral caesura in hexameter may be due partially to the influence of the obligatory caesura in pentameter.
So far, I have accounted for the following types of word-break simply by arguing for the placement of pherd and ^pherd formulas at the end of a pher3d meter:
| bucolic diaeresis | |-uu-U |
| penthemimeral caesura | |uu-uu-uu-U |
| trochaic caesura | |u-uu-uu-U |
The bucolic diaeresis could also be caused by the placement of complete or partial pher formulas within the framework of a pherd or ^pherd. As for the hephthemimeral caesura, it could be caused by the placement of complete pher formulas, again within the framework of a pherd or ^pherd:
|uu-uu-U
I have yet to reckon with the trithemimeral caesura, the least significant among the major types of word-break:
|uu-uu-uu-uu-U
Even here, it is possible to cite formulaic behavior for precedent. Note the existence of formulas shaped pher2d:
...θάνατόν τε κακὸν καὶ κῆρα μέλαιναν#
(
Φ 66, χ 14) vs. formulas shaped pherd:
...θάνατον καὶ κῆρα μέλαιναν#
(
β 283, etc.) Admittedly, the shape of such a formula is pher2d from the standpoint of meter only. As we apply it to hexameter, the 2d of the symbol pher2d cannot uniformly designate two words shaped -uu any more than the d of pherd automatically designates one word shaped -uu. It seems fitting to close this chapter by stressing this point once more. The principle of dactylic insertion in Pherecratics can be described as a phraseological phenomenon only on the diachronic level. So much for origins. On the synchronic level, as we observe the operative mechanics of attested Greek epic versification, dactylic insertion is a metrical phenomenon which is merely accommodated by the inherited phraseology. A formula like
...θάνατόν τε κακὸν καὶ κῆρα μέλαιναν#
represents simply the internal metrical lengthening of a formula like
...θάνατον καὶ κῆρα μέλαιναν#
