Chapter 2: Homer and Textual Criticism

Having examined and discussed the various principles used in editing classical and biblical texts, I turn my attention to the specific case of Homer. [1] My argument will be that since Homer is different from other classical texts, we cannot simply apply the canons of textual criticism to the surviving manuscripts of the Iliad and Odyssey. I recall my introductory remarks, where I stressed the need to consider not only the evidence but also the theory; [2] many scholars pay lip-service to the theory of oral composition and transmission for the purposes of literary analysis and criticism, but then when they look at the Homeric textual evidence they tend to treat it in much the same way as they would that of Apollonius Rhodius or some other poet who wrote. [3]

There are even those who envision Homer as a sort of literary collector and combiner of myths, traditions, and geographical details, who has already made choices between variant readings for us. All we need to do, it is supposed, is to find out exactly what he wrote down, and then we will have the “real Homer,” in the same way as we might have the “real Virgil.” [4] I compare the way in which Elias Lönnrot, in compiling what is known as the Kalevala, a collection of epic material from various parts of Finland, “merged variants of songs from different regions”; and further “had access to manuscript collections containing variants of songs from various regions ... and he chose elements from those variants.” [5] Here we see a literate scholar surveying the widely varying material and himself making the decisions as to what, on the one hand, achieves canonicity or authenticity, and what is rejected—just like the modern editor of a written work.

It is the purpose of this chapter to show that the variation in our surviving manuscripts of Homer (and other sources) is inconsistent with a single archetype, but rather points back to a multiplicity of archetypes, a situation which arises from the oral nature of the transmission of Homeric epic. By attempting to follow the traditional text-critical techniques illustrated in the previous chapter, it will become apparent that such a methodology breaks down in the case of Homer. And unlike the case with Virgil’s Aeneid, it will not be contamination alone which prevents the creation of a neat stemma—contamination merely being (as stated above) the horizontal influence between manuscripts; nor is it merely the quantity of manuscripts of Homer which leads to this conclusion, but also their quality—i.e. the degree to which they diverge from any kind of “norm,” whether it be the “vulgate” or other traditional standard; these divergences occur in ways which can be characterized neither as horizontal nor as vertical influence.

I began the previous chapter with this statement: “The primary goal of textual criticism has traditionally been to establish the actual text that the author wrote, so far as this is possible.” [6] In working with Homer, one immediately runs into problems with the formulation “the actual text which the author wrote,” in particular with the very concept of an “author” and the idea that he “wrote” anything. [7] The fact that there are difficulties even in expressing the purported goal of textual criticism (except for those few scholars who believe Homer actually did sit down and write the Iliad and Odyssey in much the same way that Virgil wrote the Aeneid) should make us cautious as we think about applying this goal and its methods to the extant manuscripts of Homer.

An associated benefit of textual criticism discussed above is that it enables us to learn more about Homer’s style, language, and poetic (including metrical) usage. [8] In the case of Homer, the term “style” includes the concept of the “formula”; and we shall see that in the context of “formulaic” material the very concept of “variant” needs to be redefined. [9]

If we have learned anything from the work of Milman Parry [10] and Albert Lord, [11] it is at least this: that epic poetry in the Troy tradition was performed orally long before any part of what we now know as the Iliad and Odyssey was ever written down, and that no one of these oral performances was identical to any other; hence no one performance—or any written record of it—could lay claim to possessing ultimate and unique authority as being the “original version” of the Iliad or Odyssey (it seems that only with the appearance of the “vulgate” did any such authoritativeness become attached to a written text of Homer). Each time an ἀοιδός (‘performer of Homeric epic poetry’) sang his song, it was a unique and “original” performance, yet one firmly rooted in tradition, inasmuch as it was expressed by means of traditional language and themes. I offer this typical quotation from Parry: Lord too would have us get away from the idea of an “original” text: “From one point of view each performance is an original”; “each performance is ... a re-creation”; “each performance is ‘an’ original, if not ‘the’ original”; and finally “the author of an oral epic ... is the performer.” [13]

Thus if more than one of these performances were to be recorded in writing, we would expect to find texts of Homer that differed significantly from one another, more so than surviving texts of originally written works such as Virgil’s Aeneid. And unlike the situation with Virgil, one particular Homeric manuscript would therefore not necessarily be derived from another through the process of copying and the inevitable errors associated with scribal activities. Rather, a manuscript of Homer could be derived from an oral performance which was more or less different from any other performance, thus giving rise to “variants” which would be inexplicable if one were to depend solely on the canons of textual criticism as applied to written works.

I am not completely ruling out the existence of scribal errors in our texts of Homer—indeed such errors do occur both in Homeric papyri and in the medieval manuscripts of Homer (although the type of analysis I am here arguing for makes it more difficult to simply write off any given variant as an “error”). Rather I am proposing to treat in a special way variant readings of the type which “differ markedly from the traditional text in a way which cannot be explained by the processes of merely mechanical corruption.” [14] I plan to do this by giving such variants the benefit of the doubt, as it were, and not automatically assuming that one reading is right and all others are wrong, as would be reasonable in the case of the text of, say, Virgil’s Aeneid. [15]

As far as manuscript evidence is concerned, what survives in the case of Homer is firstly, as mentioned briefly in the previous chapter, the large number of (mostly fragmentary) papyri from the Ptolemaic, Roman, and Byzantine periods, [16] and secondly, the sizeable collection of medieval manuscripts—fewer in number than the papyri but generally containing a far greater quantity of text, and in much better physical condition. We also have the indirect evidence of the scholia, as well as quotations from ancient authors such as Aeschines and Plato. Finally there is the evidence of early inscriptions and vase paintings. [17] When it comes to agreement with the “vulgate,” [18] it is the earliest papyri—those from the Ptolemaic period—which show the most significant divergences, while the later medieval manuscripts differ much less from our “modern” text (e.g. the OCT). The other types of evidence (inscriptions and iconographic objects) are also more likely, the older they are, to contain significant deviations from the “received text” (an equivalent term for “vulgate”), often preserving readings (and versions of episodes, if we include the evidence of vase painting) as “wild” as those of the earliest Ptolemaic papyri. [19] As time passes, the frequency and the extent of textual and other variation diminish; in fact there is a distinct terminus at around 150 BCE. This date is presumed roughly to coincide with the end of the editorial work of the greatest of the Alexandrian textual scholars, Aristarchus of Samothrace (ca. 216–144 BCE), [20] thanks to whom, apparently, the “eccentric” variations were largely eliminated from subsequent texts of Homer. [21] I observe at this point that the way in which the Alexandrian scholars carried out their textual work (in particular the reasons behind some of their editorial decisions) is somewhat unclear to us; modern scholars can be quick to label a reading of, say, Zenodotus as a conjecture, [22] whereas a more careful examination of the evidence illustrates that these earliest “editors” of the Homeric text were often more scholarly than we give them credit for. [23] The problem is that almost none of their textual readings found their way into the later manuscripts, an issue which will be discussed below.

In the case of a written text, one tends to find the opposite situation from what exists for Homer. For example, for the text of Virgil’s Aeneid there are seven manuscripts dating from the fourth to the sixth centuries, and a mass of medieval manuscripts starting in the late eighth and early ninth century. The older witnesses are considered more trustworthy and authentic, and are “the editor’s mainstay,” [24] whereas it is the later ones which contain more significant and complicated deviations from the “standard” text. The conventional theory behind such a state of affairs is that over the passage of time, errors gradually creep into a text as it is copied and recopied; the details of this process were discussed in the previous chapter. Such errors can arise inadvertently, or they may result from a deliberate decision to alter the current text. An example of the latter in the Aeneid is book II, lines 567–588; these lines are not in any of the best and oldest manuscripts. Either they did originally belong in the text, and were deleted as reflecting badly on Aeneas, or they were added later, in order to fill a perceived lacuna in the narrative. There is presently no scholarly consensus on the question of their genuineness; from the point of view of establishing the “original text” they are either genuine or they are not. [25]

Typically of most classical works, the earliest manuscripts of the Aeneid are the most reliable in establishing the “true text.” The later manuscripts tend to be so corrupted with errors, many of which arise from contamination as different readings are compared and collated, that their usefulness is significantly reduced. In the case of the text of the Aeneid, the complex medieval manuscript tradition makes it hard to see the wood for the trees. [26] With Homer, as noted above, the reverse is the case: it is the earliest witnesses to the text which contain the most divergent readings, while the later manuscripts tend to converge toward the “received text.”

In order to properly comprehend this state of affairs, we need to return to Parry and Lord’s fundamental findings about oral traditional poetry. As Parry himself said in 1932, Apparently this prospect is unsettling to some scholars. M. van der Valk feels that if the Homeric text was transmitted originally orally, and was thus exposed to various “vicissitudes and alterations,” then one must accept that “[i]t was so to speak in a state of continuous evolution and metamorphosis. It is obvious that the acceptance of this theory has far-reaching consequences. The whole basis of our Homeric text becomes uncertain.” [28] It is as if we lose “our” Iliad and Odyssey, and even “our Homer”; in exchange we get something that lacks the kind of stability to which we are accustomed; instead of a poet whom we can admire we get a tradition consisting of many anonymous poets. And yet, a proper understanding of oral-traditional poetry should lead us to feel that we have gained, rather than lost, as we look again at the Homeric text, freeing ourselves from the shackles of anachronistic assumptions regarding written transmission and all that follows from them. [29]

Other scholars, though more objective in their approach than van der Valk, still see no necessary connection between the unusual variants of the “eccentric” papyri and the nature of oral tradition. [30] It comes back to viewing the evidence in the light of a clear understanding of oral poetry and how it is composed, performed, and transmitted.

I cite two biblical analogies, although neither is an exact parallel, inasmuch as neither involves oral transmission in anything similar to the Homeric situation. The Dead Sea Scrolls, first discovered in 1947, gave scholars manuscripts of the Hebrew Bible (among other works) which pre-dated the earliest existing texts by at least a thousand years. A few conservative writers (theologically speaking) were quick to point out that the small differences between, say, the Isaiah scroll and the Massoretic text of Isaiah were so small as to inspire confidence in the belief that the Hebrew text had been transmitted “essentially unchanged” since its original putting down in writing. Others, however, looked at these same “small differences” and were able to show that there must at that time have existed differing texts of the book of Isaiah, and furthermore that each of these texts could be assigned a different geographical location, the so-called theory of “local texts.” [31] Similarly, and with an appreciation of the differing situations, I suggest that we need to look at all Homeric textual variations, whether they appear “significant” or “insignificant,” and understand them in the light of the nature of the poetry with which we are dealing.

The second parallel is in the field of New Testament textual criticism. E. J. Epp, in reflecting on the future of the discipline, writes that “we need to face the complex and perhaps unsettling notion of multivalence in the term original text. In other words, the issue is more difficult, has wider implications, and is also richer and potentially more rewarding than we might have imagined.” [32] He continues “there is a real sense in which every intentional, meaningful scribal alteration to a text ... creates a new Textform, a new original.” [33] If instead of the words “scribal alteration to a text” we read “performance variation,” we can almost hear the voice of Albert Lord. [34] Epp concludes by repeating his phrase “multivalence of the term ‘original’ text” and by adding the further suggestive words “dimensions of originality.” [35]

As discussed above and in the previous chapter, when one is dealing with a text which was written down as it was composed, whenever a variant occurs it is usually (but not always) possible to choose one reading as original, and the other as a corruption of it (for example by the theory of the lectio difficilior). One is thus justified in distinguishing between the “genuine” and the “spurious” readings. With Homer, I shall be considering variant readings, each of which appears as “genuine” and “authentic” as the others; in these cases by comparing and weighing both internal and external evidence, [36] I hope to show that neither variant can be shown to be the correct reading—rather both can be considered “correct.” Thus I will use terms such as “authentic,” “original,” “genuine,” etc., to characterize readings which appear to be Homeric in both nature and lineage, but not in order to rule out other readings as “spurious” or “inauthentic.” For the same reasons I shall avoid using such terms as “superior” and “inferior.” [37]

As an example of what I suggest is a more appropriate way of dealing with variant readings in Homer, I consider the very beginning of the Iliad (I leave out for the moment the fact that according to the A scholia Zenodotus athetised these two lines). Zenodotus (according to Athenaeus 12F) read: I start by looking at various types of internal evidence in support of each reading. One immediately notices that Zenodotus’ text presents a neater parallel between lines 4 and 5: a chiastic arrangement involving accusative–dative–dative–accusative: ‘prey for dogs and for birds food’. It also gives line 5 a great deal more alliterative effect—note the preponderance of dental stops (including dental nasal ‘n’, as also in line 4): Much of this effect is lost by the removal of δαῖτα in the “vulgate” rendering. Now a conventional textual critic might be inclined to reject Zenodotus’ reading precisely because of the “nicer” parallel it offers between lines 4 and 5, as well as the alliteration in line 5; the reasoning would be that if δαῖτα were “original,” no scribe would deliberately remove it and replace it with πᾶσι, a reading which destroys the parallel chiastic construction and undermines the alliterative force of line 5. In other words, the vulgate’s lectio difficilior must be original, since one can easily understand how it was changed into Zenodotus’ lectio facilior, while a change in the opposite direction is much harder to explain.

On the other hand, the “usual” reading πᾶσι has its own internal support. It can be shown to be formulaic [38] in this position [39] (as indeed can δαῖτα—cf. Iliad I 424); also it provides a less obvious (and hence more subtle) balance between the two phrases—i.e. not the simple chiastic structure of the reading with δαῖτα. In addition, Nagy adduces the concept of “hyperbolic allness,” found in such passages as Iliad V 52 and Odyssey xviii 85, as support for the “vulgate” reading. [40]

So much for internal evidence. In terms of external evidence (manuscript or other testimonial support) δαῖτα, while it lacks quantity of manuscript support, nevertheless makes up for it in age: when we consider the evidence of Aeschylus in the Suppliants, lines 800–801, [41] it becomes likely that he had access to a version of Iliad book one which had δαῖτα rather than πᾶσι. I note further, following Nagy, [42] that the term “version” here would refer to the performance tradition rather than the manuscript tradition, in other words that Aeschylus will have heard a performance of this part of the Iliad which included δαῖτα. Thus this reading can be dated back into the first half of the fifth century. The “vulgate” reading πᾶσι has (by definition) the overwhelming support of the manuscript tradition.

So far we have seen that both readings have both internal and external support. Therefore I would claim that both can be considered to be “genuine.” Both fit well in context, and each has an ancient tradition behind it. However, traditional text-critical thinking, divorced from the realities of oral performance and transmission, forces scholars to make a choice. We can compare various opinions about the two readings: Pfeiffer [43] thinks that Zenodotus’ δαῖτα is the genuine original reading, which Aristarchus replaced with his own πᾶσι. Kirk, [44] on the other hand, sees πᾶσι as authentic, and, without giving any reasons, characterizes Zenodotus’ δαῖτα as “a fussy change of the vulgate.” Similarly van der Valk [45] describes the reading of Zenodotus as “a subjective conjecture.” However van der Valk also explicitly states as “absurd” the possibility that Aeschylus and Sophocles had access to (written) texts of Homer, which contained respectively the readings inherited by Zenodotus and the vulgate. [46] Since he cannot accept the consequences such a view would entail, he necessarily must reject it out of hand. [47] However, to repeat, my point is that if we look closely at the internal and external evidence for each reading, we shall find that both of them have good evidence in their support. [48] There is no need to choose one reading and reject the other (apart from the editorial requirements of privileging one reading at the expense of all others [49] ): problems arise when we treat the Homeric material as if it were a fixed written text. [50] West characterizes Zenodotus’ δαῖτα as “apparently a variant familiar to Aeschylus ... but that does not make it the original reading; πᾶσι is good idiom” [my emphasis]. [51]

Still on the same passage, I note the comments of Walter Leaf in his commentary, which pre-dates Parry by several decades. Leaf describes πᾶσι as “a perfectly natural expression.” As for Zenodotus’ reading, it is only in Athenaeus, “on whom no reliance can be placed.” However the reading is in itself “vigorous and poetical.” Leaf thinks that “the metaphor is so natural that we cannot even argue with confidence that Aeschylus (Supp. 800) had δαῖτα before him when he wrote.” But after mentioning passages from Euripides (Hec. 1077, Ion 505), and Sophocles (Ajax 830), he concludes that “in all these cases there is an apparent echo of the present passage, and δαῖτα if a real variant is much older than Zenodotus.” “On the whole δαῖτα seems intrinsically a better reading, but we have no right to leave the uniform tradition of the mss.” [52] Perhaps this scholar, if he had lived to benefit from Parry’s researches, would have agreed that in a case like this, where two variants both have legitimate claims to authenticity, the best treatment is one which grants such authenticity to both readings. In the words of a post-Parryan scholar, “Both variants are traditional multiforms. In a multitext format of editing Homer, we would have to take both forms into account.” [53]

While we are in the proem of the Iliad, I note that there are other significant textual variations which deserve consideration; I mention two of them briefly here. Compare:

ii) In addition the manuscript tradition includes lines 4 and 5, whereas Zenodotus (according to schol. A—again Aristonicus) athetized these two lines. This is one of several cases where one of the Alexandrian scholars athetizes a line as well as offering a variant reading for that same line. This indicates clearly that athetesis is not nearly as drastic as outright deletion of a line.

Shipp [56] points out that there is no real parallel involving the opposition of ψυχή ‘soul’ and αὐτός ‘body’; also that ἑλώρια is a hapax legomenon in Homer. This would seem to be evidence against ψυχή in line 3, and against lines 4 and 5. However, Apollonius Rhodius (Argonautica 2.264) has the only other occurrence of ἑλώριον, as well as δαῖτα in the previous line, making an apparent allusion to the two lines in Homer. Shipp also notes that without lines 4 and 5 either of the two readings in line 3, ψυχάς or κεφαλάς, is satisfactory; with these lines only the former is possible, as one cannot contrast head and body. Thus one can explain a change from κεφαλάς to ψυχάς in order to accommodate lines 4–5, but not a change in the opposite direction. [57]

Bolling [58] is unsure as to whether Zenodotus read ψυχάς or κεφαλάς; but the fact that the scholia stress repeatedly the impossibility of κεφαλάς along with lines 4–5 suggests that it was Z.’s reading. Bolling feels that the shorter text is “superior,” giving a better sentence structure: ἐξ οὗ is now closer to μῆνιν ἄειδε, on which it depends (loosely; the former phrase is usually translated ‘from the time when’). I draw attention to this statement: “In line 3, when freed from interpolation [i.e. lines 4 and 5], either ψυχάς or κεφαλάς will conform to Homeric usage, so that it is impossible to determine which was the ‘original’ reading.” [59]

I agree that without lines 4–5, either of the two readings in line 3 is satisfactory; however I would rather say that both are acceptable; moreover, the external evidence points also to the authenticity of the version which includes lines 4–5 along with the reading ψυχάς in line 3. If we consider these variants from the point of view of oral composition and performance, it seems natural that each reading, given that each has evidence in its favor, can be considered “authentic,” and the efforts to determine “originality” become irrelevant.

I give below some of the various possibilities discussed above, with relevant differences highlighted:

By seeing each of these six “variants” simultaneously, the reader can make his or her own decision as to preference, [60] and also get a sense of some of the history of the transmission of the text over time. Such is rather more difficult to do with the traditional layout, with the “main” text at the top of the page, and the variants, usually in fragmented form, in the apparatus criticus at the bottom. [61]

I have cited this passage and these scholars’ opinions concerning the variants at some length, in order to show once again how prevalent are the assumptions behind judgments about the correctness or otherwise of variant readings in Homer. I next consider conventional views concerning the fixation and subsequent transmission of the Homeric text, and offer my own suggestions in the light of my understanding of oral composition, performance, and transmission.

Current views on the transmission of the Homeric text generally include the assumption that an archetype of some sort was written some time between the eighth and sixth centuries BCE, perhaps by means of dictation by an ἀοιδός, and that all our extant papyrus and manuscript copies derive from this archetype. [62] These views generally must assume that the practices of oral performance and “re-creation” either died out around the time this archetype was written down, [63] or that such practices continued but were irrelevant to the subsequent history of the written Homeric text, i.e. that they had no perceptible effect on it. Along these lines Janko has proposed 750–725 BCE and 743–713 BCE for the dates (of the final written form) of the Iliad and Odyssey respectively, based on his counts of certain linguistic criteria, as well as comparisons with Hesiod and the Homeric Hymns. [64]

Figure 2. The conventional view of Homeric transmission.

Figure 2 above illustrates this conventional view of Homeric textual transmission: the top row of vertical lines represents fluid oral transmission up until the eighth (Janko), seventh (M. West [65] ), or sixth (Jensen [66] ) century BCE. At this point oral transmission fades or ceases altogether, but first an archetype (Ω) is written down, by means of dictation or otherwise. [67] From here on the Homeric text behaves exactly like a written text, and we can imagine a stemma which theoretically explains all surviving manuscripts, inasmuch as they are descended from this archetype. This archetype is equivalent to Bolling’s “fountainhead”; [68] Greek letters stand for non-extant and Roman letters for extant manuscripts, as in “regular” stemmata. Even to think of Homeric transmission in these terms—as most scholars apparently do, at least subconsciously—necessarily leads one to want to find an archetype for the Homeric text, as well as to feel constrained to choose between “genuine” and “spurious” readings whenever the need arises.

Needless to say, I believe that the conventional model is illogical. There is evidence that Homeric performance traditions continued into Ptolemaic times, [69] and indeed did affect the written Homeric texts, some small portion of which we possess today. Nagy adduces iconographical evidence which supports his theory of an evolutionary process of textualization. In particular he discusses sixth-century vases which show versions of Iliadic themes that vary from our Iliad, [70] and also mentions non-Homeric verse inscriptions which act not as transcripts of a performance but as equivalent to the performance itself; [71] such evidence points to the extreme unlikelihood of there having been a fixed written text before this time. He suggests that such text-fixation cannot have happened until the time when such variations have faded: around 530 B.C. may be a possible terminus post quem for the textualization or quasi-textualization of the Iliad and Odyssey. [72] However this date does not indicate a sudden fixing of the text, but rather the transcribing or recording of performances; such transcriptions would in no way have replaced live performances. [73]

Figure 3. An alternative view of Homeric transmission

Figure 3 illustrates a scheme which takes into account the continuation of oral performance traditions (time moves from top to bottom): we have a period of fluid transmission lasting well into the Ptolemaic period and beyond. During this long period, Homer’s poetry is performed, and texts in the sense of “transcripts” are written down. [74] These non-extant (or possibly extant) transcripts are represented by Greek letters; from them our “eccentric” manuscripts, such as the Ptolemaic papyri (represented by letters A–G and Q–W), will have been copied. Those classical authors who quote lines of Homer may well not have depended on written texts at all; their versions of Homer will have derived from one or other of the various and varied performance traditions with which they will have been familiar. There may therefore be no obvious or necessary distinction between the “eccentric” text of a Ptolemaic papyrus and the divergent text of a Homeric quotation in Plato or Aeschines [75] : both will derive, at not too many steps removed, from a “current” performance.

Towards the end of this period of textual fluidity, at around 150 BCE, the standard text or “vulgate” comes into existence, and the majority of our medieval manuscripts (letters H–P) reflect this “textus receptus.” Now if we consider the dates of our “best” medieval manuscripts, such as Venetus A, we see that they were written centuries later than the date of the performances from which they derive. On the other hand, such texts as the Ptolemaic papyri, inasmuch as they were written while performance traditions were still “alive,” represent texts very close in time to their “originals.” And yet we shall see that even medieval manuscripts like Venetus A can preserve ancient readings which provide evidence of much earlier performance traditions.

Because we can never identify a single unique “original” version of Homer’s epic, and because new “authentic” versions were being created and recreated in performance throughout the period of oral transmission, it follows that any written records that survive from this period when oral transmission was alive and well are in all likelihood records (whether firsthand or a few times removed) of these authentic versions, and hence should be given the weight due to documents that are close in time to their “original source,” i.e. the Homeric performance from which they derive.

I compare this situation with that of modern (and a few ancient) authors who have made alterations to their own works, either before or after publication, with the result that more than one “authentic” version gets into circulation. Examples include Matteo Maria Boiardo’s Amorum libri [76] and William Wordsworth’s The Prelude. [77] This situation has been handled by printing the multiple versions in parallel, a procedure which allows the reader to compare versions and trace developments in the author’s thought and style over time. This is what the Homer Multitext project aims to do in the case of the Iliad and the Odyssey.

If indeed our documentary evidence is so close to its “original” performances, then it may be more appropriate to treat it in the way that more modern editors treat their texts. Tarrant discusses the differences in approach between classical and modern editors, describing how in the case of the former, the original text, inasmuch as it is felt to be distant from its editors, both in time and in generations of copies, tends to gain a sort of idealized perfection, and correspondingly the manuscripts themselves are liable to be depreciated. [78] For the modern editor, on the other hand, because the documents are felt to be so much closer to the original work, their importance is magnified: indeed different authorial versions as well as versions produced by copyists are given high regard, and we find statements such as: “a work encompasses all its authorial versions and . . . all of them should be read in order to experience a work fully.” [79] I suggest that at least some of our texts of Homer deserve to be treated in precisely this same way—that they are in effect “versions” of Homeric poetry, and all deserve to be read in order for the scholar to experience Homer fully. The difference, of course, is that instead of assuming the existence of a single author called “Homer,” [80] we are viewing each ἀοιδός as an authentic composer/performer of Homer’s poetry. [81]

While on the topic of “authorial versions,” I refer again to Martin West’s treatise on textual criticism. [82] In his discussion of the ways in which textual variation arises in a text, he states that the first such way “is that the author himself may change it, after copies have already gone into circulation.” [83] Examples given include Aristophanes, Apollonius Rhodius, and Ovid. West states that this type of variation can be evidenced by “major divergences between different branches of the tradition, if both versions convince the connoisseur of their authenticity.” [84] Further on, West gives guidelines for determining if there is unlikely to have been a single archetype: firstly, “from the presence in the medieval tradition of many pairs of variants known to be ancient,” and secondly, “from the presence of divergences so substantial ... or so early ... that one cannot believe them to have arisen in the short time available or under the conditions that prevailed after the end of antiquity.” [85] In other words, in the kinds of texts envisaged by West, it is extremely unlikely either that one variant arose from the other by some form of corruption, or that the two variants share a common origin, with both being derived from a third, original reading. West also states that the aim should be “to determine which of the manuscripts or manuscript families are most independent of each other, for these must go back most directly to the earliest phases of the tradition that we can reach, and they must be the most fruitful sources of ancient readings.” [86]

It seems to me that when West’s criteria are used, a large number of sets of variant readings preserved in the Homeric manuscripts and other sources do indeed point to there being no single archetype for the Homeric text. We have already seen at the beginning of the Iliad a pair of variants, both “known to be ancient,” and we will see many more examples of “major divergences” that are both substantial and ancient; in addition it is clear that many of the sources of these sets of variants are independent of each other. I cite T. W. Allen’s attempts to create families of Homeric manuscripts, as part of his effort to trace back the lines of textual transmission. In his endeavors to discover the history of the Homeric text, he is unable to construct a single stemma, and instead creates twenty-four textual families or “sub-stemmata,” by comparing variant readings shared by various uncials, minuscules, and papyri. [87] And even these attempts are deemed unsuccessful (except for one of Allen’s families—h) by later scholars. [88]

Further on in his book, West lists the requirements that the “correct” reading must fulfill (assuming there is one “correct” reading). I draw attention to his third requirement: On the next page he points out that plausibility is no guarantee of genuineness, since there is a multitude of places (in classical literature) where more than one of the surviving variants is plausible. [90] Indeed, apart from the minority of cases where a scribe appears to have copied something unintelligible (to him as well as to us), [91] I suspect most variants survive precisely because they are plausible. [92]

Of course the problem with so many Homeric variants is precisely that: it is impossible the see how “the presumed original reading could have been corrupted” into one or more of the other variants. To quote Stephanie West once again, variants in the Ptolemaic papyri “tend to differ markedly from the traditional text in a way which cannot be explained by the processes of merely mechanical corruption.” [93] And although a few pages later she concludes that there is no clear connection between the extent of interpolations in these papyri and the “long oral tradition of the poems,” [94] in a later work she says that even the papyri dated after 150 BCE “offer too wide a range of variants to allow the hypothesis that they might all be copies of a single edition.” [95] This is after claiming that a presumed sixth-century recension “must be regarded as the archetype of all our Homeric manuscripts and of the indirect tradition represented by ancient quotations and allusions.” [96]

For a similar example of seeming ambivalence between the single archetype hypothesis and a willingness to consider the effects of oral transmission, I cite Richard Janko’s statement: I contrast this statement with his earlier remarks about the extent of textual variation in the text of Homer and the Homeric Hymns, as evidenced by a quotation in Thucydides: Further on, But then he is constrained to make a choice between these variants, and we find, for example, where the two versions of the Hymn to Hermes (Hy 4 and Hy 18) present different readings, “thus Hy 4 is better here.” [100]

Mention of the two versions of the Homeric Hymn to Hermes brings me to another important concept—that of expansion and contraction (or compression). Hymn 4 with 580 lines stands in significant contrast to Hymn 18, which has only 12 lines. Nagy, in discussing the long and short versions of the Homeric Hymn to Hermes, notes that the expansion and compression involved in such differing treatments is a clear indication of oral poetics. [101] I refer also to the words of Parry quoted at the beginning of this chapter, where he describes how a singer sings: “he will leave out and add whole passages as the time and the mood of his hearers calls for a fuller or a briefer telling of a tale or of a given part of a tale.” [102] Lord too, throughout his work, stresses the ability of the singer to “lengthen or shorten a song according to his own desires and to create a new song if he sees fit.” [103] At the end of one chapter he lists six ways in which the singer can change his story: these include having fewer or more lines, “expansion of ornamentation,” and addition or omission of material. [104]

I see this principle of contraction and expansion applying on two related levels. First, we will see shortly examples of type-scenes within the Iliad and Odyssey that are widely used and that can be changed in length as the context demands. The poet can tell of a sacrifice or an arming in more than way and at greater or lesser length, depending on poetic context. Second, as the Homeric poems were performed and transmitted orally over time, shorter and longer versions must have been heard and eventually written down, leaving us with texts that appear to have “plus verses” and also those with “minus verses.” However, scholars such as Bolling and Apthorp, looking at shorter and longer versions of passages of Homer, choose to see only expansion, never contraction, leading them to generally label all lines that are in one text and not another as interpolations. [105] While this may be applicable for works that were never transmitted orally, it is an unwarranted assumption in the case of Homer. Even when Apthorp says that he is allowing for the operation of oral transmission, for him this seems to mean merely that a fixed number of lines, which everyone knew, was sung by singers and passed on to other singers, with the occasional lapse of memory causing a line or two to be left out from time to time. [106] This is hardly the way true oral transmission takes place.

It is true that over time the story as told by poets expands overall, with the Iliad and Odyssey being, as Nagy says, the “ultimate expansion.” Indeed this monumental expansion makes it difficult to appreciate the phenomenon of compression, but it is there nevertheless, and we should not constrain ourselves into seeing either the expanded or the compressed version as “basic.” [107]

At this point I return to the text of Homer and look at further examples of passages where, to use Martin West’s words, there are “major divergences between different branches of the tradition, [and] both versions convince the connoisseur of their authenticity.” [108] These examples will also serve as illustrations of compression and expansion. [109]

Since Zenodotus is the earliest of the three great Alexandrian scholars, I consider first another example involving his readings. [110] He has the reputation of being the most “conservative” of the three (in the sense of having the shortest text; those who consider him as a “hacker of the text” would see him as the least conservative). His “alterations” to the text can be the most minor details (e.g. νῶιν for νῶϊ at Iliad VIII 377); they include variations in one word (e.g. δαῖτα for πᾶσι at Iliad I 5 [111] ); and extend even to the rejection of whole groups of lines (e.g. Iliad I 396-406). One of the more significant ways in which Zenodotus’ readings differ from those of the “vulgate” is in his treatment of groups of lines. Sometimes he appears to contract two lines into one, e.g. Iliad I 219f., 446f., etc.; in Iliad II 55 he expands one line into two; in Iliad II 60–70 he contracts eleven lines into two; and so on. [112] In cases where there is no extant manuscript support for his readings, scholars have often assumed that he was indulging in pure conjecture. [113] In cases where there is some support, one might suppose that it was Zenodotus’ reading which has given rise to the (later) manuscript evidence. There are a few cases, however, where the only support for a Zenodotean reading is to be found in a relatively early papyrus, thus raising the possibility that Zenodotus had available to him a text with the given reading already in it, in other words, reversing the direction of the influence.

In Iliad IV 88f., where the “vulgate” has the two lines: Zenodotus [114] has the single line: This variant is found in no extant manuscript, except for the Ptolemaic papyrus P41. P41, which contains portions of books III, IV and V of the Iliad, is dated by Allen to the third century BCE, [115] and more specifically by S. West to 280–240 BCE. [116] West considers the possibility that the papyrus reading is due to the influence of Zenodotus’ text, but thinks it more likely that the debt was the other way around: “On the other hand, Zenodotus must have had MS. support for some, if not for all, of his readings.” [117]

Both readings are clearly “acceptable” in terms of grammar and flow of thought; the “vulgate” includes epithets complimentary to Pandarus, which Zenodotus (and thus presumably his source) omits. The variation in each version could easily be explained as a difference in emphasis, arising from different versions. As well as calling Pandarus αἰσχροκερδής (‘sordidly greedy of gain’, LSJ) and ἡ πάντων ἀρά (‘the curse of all’), the scholia discuss how the longer version appears to be overly “anthropomorphic”: why would the goddess Athena need to look around for Pandarus, as if she didn’t know where he was? In line with Zenodotus’ supposed concept of τὸ πρέπον (‘what is appropriate’), Leaf says that “Zenodotus was offended at the doubt which he thought was expressed as to the certainty of the goddess finding him [Pandarus].” [118] On the other hand, the other point of view is also mentioned in the scholia, namely that when a god or goddess takes human form, he or she has to resort to human activities. [119] Thus each reading can be supported by both internal and external evidence. To repeat, both the short and the long version have ancient support, and both can be justified on the grounds of language and narrative flow. Rengakos notes that Apollonius Rhodius in Argonautica 3.113f. conflates the “vulgate” and the Zenodotean text, proving that Zenodotus’ reading, far from being a conjecture, must have depended on documentary material. [120]

I mention for comparison a line from book xii of the Odyssey. To put the line into context: Odysseus and his men have anchored their ship on the island of the sun god; whilst Odysseus is elsewhere on the island, his men, desperate for food, have decided, against the orders of Odysseus, that it would be better to kill and eat the cattle of the sun god and risk sudden destruction, than to starve to death slowly. So, after rounding up the cattle, they prepare to sacrifice them. Usually such a sacrifice would include the sprinkling of white barley; but as none was available, oak leaves had to suffice. Thus we get the following line: The more “usual” sacrifice formulation has two lines, which include the sprinkling of the barley, as well as the drawing back of the victims’ heads:

I highlight the pieces of the two lines from the “expanded” version which have been used for the “shorter” version. In the Odyssey passage, the lack of barley grains leads to the omission of the second half of the first line, and thus also the pulling back of the victims’ heads in the first half of the second line. In other words, we have here an “abbreviated” version of part of the regular sacrificial form, based on differences in the narrative context. Thus we note the existence of a “long” and a “short” version of a particular element in a type-scene, each tailored to fit into its respective context. So when we find evidence for both a “long” and a “short” version belonging to the same passage, we must at least allow for the possibility that both are valid and “original.”

In looking at other examples of “typical scenes,” in particular of sacrifices, I notice that there is considerable flexibility in retaining or omitting “essential” elements. For instance, in the two sacrifice episodes of Iliad I and II (I 458–469 and II 421–432), ten of the twelve lines are identical, and in the same order. However when we move to Odyssey iii (447–473), we get only five of these same lines, and in a much longer passage overall. By way of contrast, in Odyssey xii (359–365) we find a considerably shorter version, but still with six of these lines. In each of the Odyssey passages a flexibility of composition is exhibited which should lead us to treat with a more open mind passages, like that discussed above, where more than one version of an episode is preserved. I note too that Edwards, in a useful article summarizing scholarship on the subject of “typical scenes,” points out that “use of type-scenes is probably a better test for orality, at least in Greek poetry, than use of formulae.” [121] He also indicates that more work needs to be done on how type-scene structure relates to oral versus written style. [122]

The passage from Iliad IV brings up the subject of interpolation: one of the characteristic features of the Ptolemaic papyri (to be discussed in the next chapter) is that they often possess “plus verses,” and many if not most scholars generally judge these to be both “spurious” and “inferior.” During the twentieth and early twenty-first centuries two scholars in particular, Bolling and Apthorp, have devoted themselves to the question of interpolation in Homer. In three books published in 1925, [123] 1944, [124] and 1950, G. M. Bolling sought to rid our text of Homer of all “spurious” lines, and his studies culminated with his edition of the “Athenian Homer”—the so-called Π-text, containing about 14,650 lines, reflecting the Iliad that he believed must have been current in sixth-century Athens. [125] Between the sixth and second centuries, approximately one thousand “extraneous” lines were added to this text, with the result that Aristarchus’ text—which Bolling calls the “Alpha text”—contains about 15,600 lines. [126] Each of these “interpolated” lines was in Aristarchus’ text, but many were marked with the obelus, indicating that Aristarchus considered the external evidence for them to be too weak to establish their authenticity. [127] In many cases, however (e.g. Iliad I 4–5, II 60–70, etc.), Aristarchus judged as genuine lines which Zenodotus had athetised. Bolling generally excludes these lines as well, looking for the “lowest common denominator” in order to establish his sixth-century “archetype”; thus if he finds two manuscripts covering the same passage, but which do not both include the same number of lines, i.e. one omits some lines while the other omits different lines, Bolling will only judge as belonging to his Π-text those lines which are in both manuscripts. [128] There are also a few cases where Bolling leaves in the text lines which Aristarchus had athetised (in book I lines 29–30 and line 96).

Bolling’s language throughout indicates that he is conceiving of a purely written transmission; in addition he frequently stresses his assumption that texts always expand and never contract. Also, And Bolling also approvingly gives this quotation from Leaf: Bolling rescues Zenodotus from the charge of “hacking at the text,” made by scholars who have assumed that the longer text is original, and states that Two further quotations illustrate clearly how completely bound Bolling is to the concept of a written archetype, with no allowance for the mechanics of oral performance and transmission: And In fairness one should note that these statements were published in 1925, before the appearance of Parry’s seminal work; however, Bolling in 1950 (well after Parry’s death) could still write: The second of the two scholars, M. Apthorp, has continued the work of Bolling, using recent papyrus discoveries to reinforce Bolling’s arguments, and sometimes re-stating points which he feels Bolling did not make forcefully enough. [134] Apthorp gives some credence to oral performance and transmission, and yet he appears to misunderstand how it would work: “even when due allowance has been made for the operation of oral transmission (and thus omission through a reciter’s forgetfulness), in conjunction with the written transmission, ... the lines omitted by Aristarchus were spurious.” [135] In other words oral transmission means little more than singers simply memorizing and “reciting” songs; in the course of reciting a song some lines may be left out due to forgetfulness and hence get left out of one or more manuscripts.

Apthorp does explicitly allow that Parry’s theory of oral transmission and the way it can explain divergences in the written transmission “remains plausible”; [136] and yet imagining that the chief way for a line to be omitted is through “a reciter’s forgetfulness” shows that he too is thinking in terms of an archetype, ultimately written. His use of terms such as “spurious” leads him, as it did Bolling, to reject all those “additional lines” which occur in the Ptolemaic papyri, in the scholia, and in the quotations, because of their weak attestation in surviving manuscripts, and because (in most cases) Aristarchus does not seem to know of them. [137] This use of Aristarchus as the sole “filter” through whom alone all “genuine” lines had to pass seems to be a problem with the approach of Bolling and Apthorp. [138] There seems to be no good reason why “genuine” lines could not have “got around” him and made it into the later manuscript tradition without ever coming to his attention. [139] And we should never rule out the possibility that Aristarchus “made mistakes”—a possibility both Bolling and Apthorp seem very reluctant to admit.

In seeking to show that various variant readings and “plus verses” are in fact authentic, one may use the tools of comparative philology and formulaic analysis. [140] Nagy uses the former to show that certain readings are ancient and thus authentic (but not “superior”). [141] In addition, such evidence can be used to show that oral performance traditions continued far later than is generally accepted. I refer to some cases where even seemingly “insignificant” variants can provide important evidence about the transmission of the Homeric text. Following a conjecture by Wackernagel, Nagy adduces the following sets of variants. Of each pair of variants, the first is transmitted in the majority of manuscripts, but its accent is “anomalous”; the second has the “regular” accent and is reported as a variant. In each case the word with the anomalous accent can be shown by comparative Indo-European linguistics to be archaic. [142] Nagy argues that this kind of accentual variation was “not derived by the Alexandrian exegetes from spoken dialectal pronunciation, but rather, from intense research in recited rhapsodic pronunciation.” [143] Since accents do not appear to have been considered to be an integral part of the written text by Aristarchus and other Alexandrians, the variants given above suggest that such information about them could not have depended solely upon written exemplars—rather there was another kind of evidence available, namely the evidence of performance traditions, which were thus still current at that time.

Another example of an ancient form being preserved by scholia and also in one or more medieval manuscripts involves the pair of variants τ᾽ ἄρ and ταρ. In most modern editions the beginnings of these three lines from the first book of the Iliad are printed as follows: The German linguist Wackernagel had proposed the readings: The first two of these readings occur in the manuscript Venetus A (in 93 Ven. A has the unusual οὔτὰρ), and are explicitly commented upon there by the scholiast—generally thought to derive from Herodian. Subsequently comparative evidence—from the Luvian language (Luvian kuiš=tar is equivalent to Homeric τίς ταρ), which was unknown to Wackernagel—has come to light which provides strong support for the antiquity of the reading ταρ. [144] Thus we have yet more evidence for the survival of ancient forms in later manuscripts, [145] and more evidence that oral performance traditions continued for a lot longer than is commonly believed.

As an illustration of how the technique of formulaic analysis can be used to demonstrate the authenticity of more than one variant in a given passage, I cite Iliad VIII 526, in which the following three variants are reported by various papyri: External evidenc e: most mss.; a scholion in Venetus A—Erbse’s A^im—the authority is stated to be “οὕτως ἡ γραγή”—which appears to mean Aristarchus; also P486a—a third-century CE papyrus. External evidence: some other mss.; another scholion in Venetus A; Plutarch Vita Homeri ii 118. External evidence: 7 mss., according to Allen’s editio maior.

In the course of a study of the uses of εὔχομαι in Homer, [148] Leonard Muellner deals with these variants, [149] and after detailed argument, concludes that εὔχομαι ἐλπόμενος can possibly be genuine epic diction, although (he claims) it is somewhat more awkward; Zenodotus’ reading ἔλπομαι εὐχόμενος is much more easily “defended” by use of parallels. Thus both variants have a right to be considered authentic, and it is not a case of “finding the right one” and discarding the “wrong one.” [150]

As an example of how one might deal with variants which are treated as being of equal authenticity, I quote A. di Luzio, who, after examining closely the variants of the Ptolemaic papyri, concludes that they reflect a text not yet normalized, and closer to the fluctuating state of the rhapsodic epic when transmission and “reproduction” of the text were still under the influence of “recitation.” This in contrast with the seemingly normalized “vulgate” text, coming from a later time when transmission was due more to the letter and less to the hearing. Di Luzio envisions a critical edition of Homer in which

To return for a moment to biblical textual criticism, I illustrate below (Figure 4) how a similar scheme has been proposed for handling Greek New Testament variants, with the goal of comparing variants with each other, rather than with any supposed “standard.” [152] This diagram deals with the variants for the Gospel of John 7:12. (I omit the mss. evidence, which is listed under each reading in the original scheme.)