Listening to the Spider: reading Hesiod’s Works and Days

 

APPENDIX                   

verse-type tables:  1  2  3  4  5

 

Syllable quantity.

            (Reference: Maas, 75-84.) Ancient Greek verse was organized in terms of quantity, rather than stress as in English and some other modern languages. One speaks of “short” and “long” syllables, although some prefer to call them “light” and “heavy,” respectively. (Of course, this is a schematic distinction which is justified by the empirical fact that it organizes the data well. The early epic poets themselves undoubtedly varied their enunciation to yield gradations in the lengths.) Specifically:

1. A syllable is short if closed and ending in, or open and followed by, a single consonant, and if its vowel is e (epsilon) or o (omicron); sometimes a (alpha), i (iota), or u (upsilon).

2. With an exception to be noted shortly, a syllable with any of the following vowels and diphthongs is long “by nature”: ē (eta), ō (omega), ai, au, ei, eu, ēu, oi, ou, ui ; sometimes a, i, u. (For the latter three, in general the quantity must be learned for each given use.) In the Greek, long vowels are sometimes printed with an iota subscript which is thought not to have been pronounced; however, this iota is transliterated here as an adscript: ēi or ōi. If the second of two adjacent vowels is to be pronounced as a separate vowel, but could be confused with part of a diphthong, it is written with the “umlaut” diacritical: aï, etc.

3. With an exception to be noted, a syllable with a short vowel is long “by position” if two consonants precede the following vowel. For this purpose z (zeta), x (xi) and ps (psi) count as double consonants. (The rough breathing, transliterated h, is not considered a consonant and has no metrical value.)

            The exception to rule (2) is the phenomenon called “correption.” Particularly in archaic hexameter, a word-ending long vowel or diphthong is almost always scanned short if the next word begins with a vowel. The exception to rule (3) is that the combination of a consonant follow­ed by a liquid, r (rho) or l (lambda), sometimes counts as a single consonant.

            Sometimes two adjacent vowels which would ordinarily form distinct syllables are scanned as a single one. The phenomenon is called “synizes­is.” The syllable is normally long, but can be shortened through correption.

            Irregularities do occur. There is “hiatus” if a word-ending long vowel or diphthong is scan­ned long after all when preceding a vowel, or if a word-ending short vowel precedes a vowel. Short vowels are sometimes scanned long to preserve the line’s cadence. Such irregularities are fairly frequent for some vowel combinations and verse positions in Homer, less so in Hesiod.

            To be sure, an apparent hiatus is often only apparent: In some cases, at least at some point in the transmitted history of the verse or formula in question, the semivowel “digamma” (transliterated w) was still pronounced as a consonant at the beginning of the next word, even though it was obsolete by the time written texts of the epics were produced and was not recorded. One cannot know when it was actually no longer pronounced; nonetheless, it is indicated in paren­theses in the present work in cases where irregular meter can be made regular with it.

 

Hexameter prosody.

            (Reference: Maas, 59-61.) The archaic Greek hexameter line in particular contains six “feet” (or “metra”). Each of the first five is either a “dactyl” (a long syllable followed by two shorts: __ vv ), or a “spondee” (two longs: __ __ ). The 6th foot has two syllables, the first of them long.

            The atypical character of the 6th foot is presumably related to its line-ending status, but the issue of how much quantity was actually enunciated in its second syllable is a controversy which will not be entered here. I simply indicate the foot metrically with the notation __ x .

            The 5th foot is most often a dactyl, to yield the characteristic closing cadence __ vv __ x .

            In almost all Homeric/Hesiodic verses there is a word end within the third foot, the “cae­sura.” It occurs after the initial long syllable or arsis (“masculine” caesura), or after the first short in some dactylic feet (“feminine”). Exceptions are treated in the Introduction above.

 

Syllable sequences.

Individual verses of the archaic epic works were first listed in accordance with their sequences of dactyls and spondees in the first five feet by J. La Roche, “Zahlenverhältnisse im homerischen Vers,” Wiener Studien, 20 (1898), 1-69, and “Untersuchungen über den Vers bei Hesiod und in den homerischen Hymnen,” ibid., 70-90. Here I give up-to-date sequence assignments for the Works and Days in particular, as the first entry in the verse listing of Table 1 (omitting vv. 93, 120, 169, 310 as spurious).

In the table, the following notes for the asterisked entries indicate disagreement with La Roche and a few other points. (Some cases result from La Roche apparently -- if without saying so -- following the Goettling-Flach text of Hesiod, influential at the end of the 19th century but with some readings no longer favored. I do not remark on cases where two vowels seem to form a diphthong although West and/or Solmsen print them as resolved, or vice versa.) I do not understand La Roche’s scansions for vv. 360 (which he gives as dsdsd), 379 (ddddd), 647 (sdddd), 683 (sdddd), 691 (sdddd), 728 (dsddd, which a typographical error in his text prints as “729”), 774 (ddsdd), and 784 (dsdsd). Apart from them: V. 167 La Roche has dsddd, possibly reading te kai ēthe’ for kai (w)ēthe’. 198 La Roche sdddd, presumably reading phareessi with short a for pharessi with it long (which is correct; see West). 292 La Roche ddddd, reading dē epeita (so also Solmsen), but for dēpeita see West (1966, 100). 369 La Roche dsddd, reading eni, but for en see West ad Th. 971. 397 I read ergazeu with Solmsen (so evidently La Roche), not ergazeo (West), which would give dssdd if scanned normally. 410 La Roche dssds; he must read t’ enn-, but for te en- see West. 430 My scansion (so also La Roche) of this difficult line follows the quantities given by West (1966, 96, 101). 436 La Roche ddsdd, presumably following Schaefer’s word order alteration, but see West. 452 La Roche dsdsd, reading bous for boas. 464 West’s widely rejected emendation would give ddsds. 485 I reject Wilamowitz’s emendation, followed by Solmsen, which would give sddsd. 492 La Roche sddsd, evidently scanning ear as ear. 521 La Roche sddsd; he and Solmsen, among others, follow van Lennep’s “correction” of erg’ eiduia to erga (w)iduia, but see West’s (ad Th. 264; 1978, 62) counterargument. 598 La Roche dsddd, I suppose following Nauck’s emendation to give the archaic form Ōariōnos for Ōriōnos here (although La Roche does not do so at 615, 619). The correction (which Edwards ad Il. 18.486 suggests is justifiable, but see West, 1978, 62-3) would make 615 ddddd instead of dddds, and 619 sdddd instead of sddds. 627 La Roche dddsd; he must read enikattheo for enkattheo, possibly rightly (the two are often variants in epic; see West ad 27). 689 La Roche dddsd, presumably reading eni n- for en n-. 693 La Roche dsdds, presumably reading ta de for kai. 699 La Roche ddddd; he must read the MS variant hina for hōs k’. 705 La Roche dsdsd, presumably reading the variant dalou kai ō- (which creates hiatus) for daloio kai ō-. 717 La Roche means to agree; his “714” listed under dddsd is a typographical error. 736a Omitted by La Roche. 756 La Roche sddsd; he may follow Triclinius’s toi for nu te.

The following statistical comparisons of frequencies for some selected verse types use La Roche’s figures for Homer. Strictly speaking, they should also be corrected, but I am not aware of publication of better figures. Fortunately, their numbers in Homer are sufficiently large that a difference of one or two for a given type would not affect the comparison appreciably, even though a similar discrepancy for our much smaller poem could do so. (Indeed, before La Roche, A. Ludwich had given totals of verse type for each Book in Homer, in Aristarchs homerische Textkritik, 2 vols. (Leipzig, 1884-85), II 301-26. His frequencies are slightly different, for example giving 18.6% rather than La Roche’s 18.9% for ddddd, but a spot check using them rather than La Roche’s in a few of the following calculations turns out to give no appreciable difference.)

The statistical comparison for each type is carried out by computing χ² from the formula

 

χ² = {|O - E| - ½}²/E + {|T - O - (T - E)| - ½}²/(T - E),

 

where O is the observed number of verses of the given type, E is the number expected from the Homeric frequency for that type, and T is the total number of verses in the sample (825 for the full poem). Standard tables for χ² with one degree of freedom then give (to be precise) the probability that χ² would be larger with a different sample, assuming that the frequencies only differ by chance. One naturally interprets this figure as the probability that the frequencies in fact differ only by chance. (This is the same procedure used in a similar context by F. Jones and F. Gray, TAPA, 103 (1972), 187-209, espec. 192-201.) The results are given in Tables 2-4. Table 2 compares the full poem with Homer. Table 3 gives the same calculation but with certain blocks of verses which contain aphorisms subtracted, in order to see if this affects the comparison. Table 4 compares these aphoristic segments with the rest of the poem rather than Homer, to test the poem’s internal consistency in this regard. In these tables the calculation is omitted in cases where the observed and expected numbers differ by less than 1.0.

Concerning the results, I take it that a probability figure of under 1% means that the two frequencies being compared are definitely inconsistent, and that factors other than chance come into play. This is still likely if the figure is under 5%, and possible if under 10%. The only exception is if the sample is very small. (For example, the 6 dssds verses of the full poem seem to yield a chance probability of only 0.6%, but if there is an error of as little as 2 verses in assigning the type, 4 dssds verses would correspond to a chance probability of as high as 21%, which would not be small enough to doubt that chance was responsible for the difference from Homer.)

If these assumptions are accepted, then as indicated in Table 2 the ddddd frequency is significantly smaller in our poem than in Homer, and the dsssd significantly larger. The dssdd, dddss, and sdssd frequencies as well are probably larger. (Jones and Gray get slightly different results, presumably because of different assumptions on what is significant and perhaps also on text readings: We agree for the cases of ddddd, and dsssd, but they find higher significance than I for sddds and sssdd.)

The differences with Homer are less striking if the aphoristic portions of the poem are removed (Table 3); nonetheless, it is still definite that this remainder deviates more from the “canonical” ddddd pattern than does Homer. On the internal consistency, the one striking result from Table 4 is that the aphoristic portions are richer in the dsdsd type. One cannot say much more.  (In principle one should do these calculations with the two sets of aphorisms, vv. 342-82 and 696-759, considered separately, since as discussed in Chapters 3 and 9, I believe they play very different roles in the poem.  However, the sample sizes would be very small.)

Finally, our poet seems to employ verses beginning with two spondees, irrespective of the number later in the verse, with greater frequency than does Homer. There are 143 such verses in the poem, or 17.3% of the total, while the figure is 14.2% for Homer. Comparison yields a χ² of 6.39, for a chance probability of 1.1%, which is significant. (It may be that Hesiod likes the solemnity of five long syllables beginning a line, as if five bells were pealing.)

 

Enjambement types.

The second entry in Table 1 is the enjambement type as defined by Higbie (1990, 29, elaborated in 28-65): “0” denotes no enjambement; “1a” inessential enjambement (as discussed in the Introduction) where the addition is not a full clause; “1b” where it is; “2a” if the added material is part of a dependent clause which begins in the first verse (a rare occurrence); “2b” if it is a dependent clause distinct from the first verse; “3” essential enjambement where the verb, object, or subject is delayed until the next verse and is really needed; and “4” if the enjambement is “violent,” i.e., either all of the clause except particles is delayed or the verse boundary separates a noun from its epithet or from a preposition. The only previously published listing known to me is Peabody’s (534-53), which employs Parry’s coarser categorization into unenjambed, inessentially enjambed, and essentially enjambed, and which follows an out-of-date text (Rzach’s).

There is some room for interpretation of Higbie’s categories, and in particular she (29) is unclear whether it is “degree of expectation” or “grammatical need” which determines whether a verse is essentially or unessentially enjambed. (For this and other relevant points, see Barnes’s review article, Bryn Mawr Classical Review, 2 (1991), 163-79.) My listing considers each case separately without formulating a general position on this issue, although I am probably biased in favor of audience expectation. Another point of contention Higbie and Barnes note is where one decides a sentence ends in relation to the punctuation imposed by editors. My position is that a clause governed only by de is considered separate for the purpose of determining enjambement type, regardless of West’s or Solmsen’s punctuation before it. Also, with Higbie (40-1), I consider the enjambement type to be 1b when I judge that kai beginning the next verse is copular; I take the same position regarding alla; and I construe 1b when the next verse is a relative clause which seems connected, granted that in semantic terms it is often difficult to distinguish the relative pronoun from a demonstrative in such cases. (In contrast, Peabody usually considers a verse simply unenjambed if it is in any of these three categories.)

Concerning the asterisked entries in the table: V. 7 Zeus is already understood, so that the subject in the next line is unnecessary. 18 Zeus could simply have “made” the good Strife, rather than “put” her/ in the roots of earth; cf. Il. 4.83. 39 I follow West in not punctuating at verse end, contra Solmsen. 55 The entire next verse is an added object of the verb (Verdenius). 70 The verb is transitive, requiring the object in the next line. 80 The verb need not specify the name it would give, as at Il. 9.515, 18.449, Od. 24.339. 113 Deilon/ gēras separates what looks like a formula, granted that it is not attested as such. 119 The sentence must end since v. 120 is to be rejected; so also 168, 309. 123 Here and at 253 the next verse is a dependent relative clause; see Higbie (39-40). 145 An attested formula (Il. 11.733) is divided. 147 Rejecting West’s and Solmsen’s treating aplastoi in the next verse as a separate clause. 174 Being “among the fifth” is sufficient, without “men”; cf. one could not tell “which (side) he was among,” Il. 5.85. 182 This and 183 could be nominal sentences, even though “will be” is added in 184 and governs them; cf. Higbie (50). 188 Assuming West’s punctuation, contra Rzach. 192 One would understand both dikē and aidōs to be “in the arm” if “will not be” did not occur in the next verse applying to the latter entity. 216 The predicate of a nominal sentence is in the next line. 248 Phazō is frequently intransitive; presumably the intensified version kataph- used here can be as well. 260 Without the next verse we would understand anthrōpoi to be the subject of apoteisēi. 268 Tade could be the subject of lēthei. 270 I still take “would be” in the next verse as unnecessary, in spite of the optative mood. However, mēt’ .../ mēt’ signals “internal” clausal enjambement; see Higbie (42) on oute. 276 I am inclined to say the audience expects specification of tonde nomon in 277, but it is not grammatically necessary, so that the enjambement could be inessential (as Peabody says). 322 The apodosis to the conditional clause of 321 does not arrive until 324 as I construe the passage, but I define enjambement so that otherwise unenjambed verses in between are unaffected in such a situation. Meanwhile, ginetai in 323 functions as would esti (see Higbie, 49-50), and “when profit deceives the mind” merely modifies ginetai. 323 “Mind” is sufficient, without “of humans.” 394 Although the chain of particles also includes the participle chatizōn, all the essential elements of the clause are in the next verse. 397 Interpreting erga ta, “those tasks” which the gods assign, in 398 to be the (unexpected) object of the verb. 405 Taking the entities to be the object of the verb in 407, not that in 404 (as West seems to imply). 414 The verb is intransitive in the parallel phrase at Il. 13.424. 415 Possibly a “4,” given the separation of the participial phrase from its noun. 416 The formula “flesh turns” lacks an object elsewhere. 426 Rejecting West’s construal of the next phrase as separate. 427 Pherō can be intransitive, yielding “1a,” but I am inclined to believe the audience expects an object. 435 Accepting West’s addition of d’ in 436. 455 I eschew the editors’ punctuation which ends the clause at verse end. 473 Taking the clause of 474 to be dependent, but that of 475 to be independent. 485 I tend to believe the audience expects the continuation. 491 The concluding verb’s object could have been “all,” in the same verse. 495 I believe the next clause is independent since it changes from third to second person. 502 I tend to believe the audience expects continuation. 512 Solmsen begins a new clause with 513 but I follow West. 517 We already know it is Boreas which blows, without being told so again in 518. 532 The continuation is unexpected, particularly given that the parallel expression “having close-constructed sties” (pukinous keuthmōnas echontes, Od. 10.283) lacks such a qualifying phrase. 533 I assume the concluding phrase is governed by the verb of 535, with 534 parenthetical. 543 Taking the clause beginning in mid-sentence as a protasis, not apodosis. 580 “Many” would be a sufficient object; cf. Il. 9.546, 12.399. 613 One could draw the “grapes” already mentioned into the jars figurative*ly. 641 To me the verse makes no sense without continuation. 683 I accept West’s punctuation, wherein harpaktos in 684 is a one-word nominal clause, not a runover. 696 Mēte signals a continuation. 722 I assume 723 begins with a separate nominal clause, as with West’s punctuation. 728 This is a conditional clause. 731 Not “1a”; what follows is a nominal sentence, not a phrase. 750 The verb kathizō is invariably intransitive in Homer; giving it an object in the continuation is a complete surprise. 753 The infinitive could be an imperative directed to the addressee. 770 An addition to this verse does occur in 772, but with the parenthetical clause 771 intervening; thus I take the latter to be what is enjambed. 771 West argues against taking a new clause to begin at 772 (also at 774). 774 I believe a men ... de correlation begins here, to conclude in 776. 786 T’ (not d’) in 787 indicates that its (nominal) clause is correlated. 798 I do not begin a new sentence in 799. 803 One expects the Furies “to tend” some object, as at Od. 20.78, and unlike there this is not yet stated. 815 Putting a yoke “on the neck” clearly refers to animals without specifying them; cf. putting a sword to the neck, Il. 16.339, 20.481. 816 Assuming Schaefer’s emendation to add a correlative te in 817.

A comparison with Higbie’s summary figures for the Iliad (she does not report on the Odyssey) is given in Table 5. The most important point of the comparison is that our poet uses type 1a enjambement, the simple runover phrase, to substantially the same degree: 22.8% as compared with 21.3%, with the χ² analysis showing the small difference to be consistent with chance. The difference in the fraction of type 1b verses I assign to our poem as compared with that for those Higbie assigns to the Iliad cannot be laid to chance, but this does not prove that the poets themselves were different in this respect. As suggested above, this is a hazy area, and it is possible that Higbie and I after all apply different detailed criteria in distinguishing a sentence from a clause (given that she does not publish an individual listing from which one could investigate the point in detail).

Finally, G. Edwards (98) feels that the amount of “unperiodic” (inessential) enjambement in vv. 201-300 of our poem is less that that in Homer. However, the difference is not statistically significant. In the first place, using his own figures, χ² for this comparison comes out to be 0.93, which corresponds to a chance probability of as high as 33%. Moreover, if one makes the comparison only on the basis of type 1a, removing clausal enjambement from the calculation, then as shown in Table 5 the frequency actually agrees with Homer.