5. Aristarchean Questions: Emerging Certainties about the Finality of Homer’s Text*

5§1 The New Companion to Homer, edited by Ian Morris and Barry Powell (1997), is bad company, for the most part. [1] Such is the opinion of Richard Janko in his review of this book, which had been meant to replace A Companion to Homer, edited by Alan J. B. Wace and Frank H. Stubbings (1962). [2] Preferring in many ways the “old” Companion, I too have my own negative opinions about various aspects of the New Companion, as is evident from my remarks in my earlier work. [3] But these aspects are not the same as those with which Janko is occupied. Aside from these differences, I disagree with much of what Janko has to say about my chapter in New Companion, “Homeric Scholia.” [4] My concern here, however, is not to air all my disagreements with Janko’s views. Rather, I propose simply to challenge certain assertions that he makes about Aristarchus as an editor of Homer, specifically with reference to my chapter “Homeric Scholia” (hereafter = HS), {110|111} which is linked to my book Poetry as Performance (1996; = PP). Essentially, Janko expresses doubts about the usefulness of Aristarchus as a source for genuine Homeric variants. To counter Janko’s doubts, I return here to my ongoing project of defending the reliability of Aristarchus as a definitive source for establishing the history and prehistory of the Homeric textual tradition.
5§2 Janko criticizes me for claiming something that I know I did not claim in HS (or in PP): that Aristarchus and the two other major Alexandrian editors of Homer, Zenodotus and Aristophanes, never made emendations of the received text. What I do argue is that Aristarchus did not put his own emendations into the received text of Homer, confining them instead to his commentaries or hupomnēmata. Further, I have argued that the emendations of Aristarchus were generally based on variant manuscript traditions, not on his own conjectures.
5§3 In this context, it is important to distinguish between emendation and conjecture: you can emend a text (for example, by choosing one attested variant over another) without necessarily making conjectures (that is, proposing a form that is unattested in the manuscript evidence). Janko is saying “emendation” where I mean “conjecture.” He writes: “To sustain the authenticity of all Alexandrian readings, … N. has to claim that they never emended the text (p. 114).” In my original version of HS in the New Companion, I never use the word “emend,” nor do I make the claim attributed to me. [5] Instead, I criticize the stance of Homeric scholars who consistently dismiss variant readings attributed to Aristarchus on the grounds that they are “conjectures.” [6]
5§4 Earlier in his review, Janko captures the essence of my argument when he says: “N. argues that the extent of rhapsodic variation in the text of Homer is so great that we cannot accept either an Aristarchean quest for the original reading, or Wolf’s distrust that Aristarchus could recover authentic readings in general.” But then he goes on to overstate my case, in order to create a foil for his own arguments, by paraphrasing me with these words: “all his [Aristarchus’] variants are authentic.” To back up this overstatement, Janko quotes me as saying: “there is no reason to doubt that any Homeric variant attributed to Aristarchus can be considered an authentic reading.” [7] The quotation is taken out of context. I repeat here my actual formulation: {111|112}

To be sure, we may disagree fundamentally with the premise of Aristarchus, who searched for variants in Homeric textual transmission in order to find in each case the authentic variant. Instead, we may wish to argue for an evolutionary model, accounting for a plethora of different authentic variants at different stages (or even at any one stage) in the evolution of Homeric poetry as an oral tradition; variations in the textual tradition would reflect different stages in the transcribing of this oral tradition. Such a model is fundamentally at odds with the theories of Villoison, who puts his trust in Aristarchus, validating that Alexandrian scholar’s case-by-case search for the authentic reading in the text of Homer.
Such a model is also at odds with the theories of Wolf, who distrusts Aristarchus’ ability to recover authentic readings in general. Whereas Aristarchus—and Villoison—may have gone too far in positing the authentic reading in any given case throughout the Homeric text, there is no reason to doubt that any Homeric variant attributed to Aristarchus can be considered an authentic reading. For Wolf to cast general doubt on variant readings attributed to Aristarchus may well be going too far in the opposite direction. [8]

The emphatic an in the last part of my formulation was meant to contrast with the emphatic the in “the authentic reading.”

5§5 My position on Aristarchus’ editorial methods converges for the most part with that of Michael Haslam, whose chapter “Homeric Papyri and the Transmission of the Text,” precedes my “Homeric Scholia” in New Companion. [9] The one major point of divergence between our positions is that Haslam subscribes to Janko’s theory “that the Homeric texts were indeed written down from dictation during the eighth century.” [10] By contrast, I argue for an evolutionary model. [11] To sum up my position, I can simply highlight from what I have said above: this evolutionary model accounts for a plethora of different authentic variants at different stages (or even at any one stage) in the evolution of Homeric poetry as an oral tradition; variations in the textual tradition would reflect different stages in the transcribing of this oral tradition.
5§6 Here is the crux of my disagreement with Janko’s dictation model. In terms of my evolutionary model, the variant readings adduced for the Homeric text by the Alexandrians, most prominently by Aristarchus, can be viewed as reflexes of formulaic variation in an oral tradition. The {112|113} variants themselves may in theory date from a wide range of chronological points extending from the second millennium BCE all the way to the era of Aristarchus. In reality, though, we may expect most of the surviving variants to stem from the latest recoverable phases of the oral transmission, especially from the timeframe spanning the last 500 or 600 years before Aristarchus. Even within this timeframe of half a millennium or so, we may expect the degrees of variation to drop off drastically during the last few centuries, as the Homeric oral tradition becomes ever less flexible while various political controls over performance conventions become ever more rigid.
5§7 In terms of Janko’s dictation model, by contrast, the variant readings adduced by the likes of Aristarchus cannot really be products of an oral tradition, on the grounds that Homeric poetry had already become, by the eighth century BCE, a fixed text. At that point in time, we would still be well over 500 years away from Aristarchus. For over half a millennium before Aristarchus, according to Janko’s explanation, Homeric poetry would have survived primarily as a text, not as a performance tradition. This is why Janko is forced to dismiss as “conjectures” some of the most significant surviving variants adduced by Aristarchus. This is why Janko must disagree with Haslam’s analysis of such variants.
5§8 Janko highlights what Haslam says about the variant readings adduced by Aristarchus: “the prevailing opinion is that Aristarchus invented them, that is, conjectured them.” [12] Calling Haslam’s statement an “overstatement,” Janko reacts by saying: “It is certainly not my opinion, only that he [Aristarchus] did venture some conjectures in his commentary (about thirty-three in the c. 3,000 lines of Iliad XIII–XVI).” Janko’s reference makes it clear that he is relying here on his own 1992 commentary on Iliad XIII–XVI (hereafter = IC). Later on, when he is reacting to my position, Janko refers me as well to the same commentary, specifically to a specific page where he cites thirty-three “conjectures” supposedly made by Aristarchus. [13] In disputing me, however, he refers to the “thirty-three cases” as “emendations,” whereas he had called these same cases “conjectures” in his earlier dispute with Haslam.
5§9 The basis for Janko’s using this word “conjecture” is made clear when he says: “I agree with van der Valk and Kirk [vol. I p. 43] that most readings where the Alexandrians lack support in the papyri and early codices are conjectures.” [14]
5§10 Prominent in Janko’s sample list of Aristarchean “conjectures” is Iliad XIII {113|114} 423. [15] In this Homeric line, as Janko argues, “Aristarchus altered στενάχοντα (read by Zenodotus and the vulgate) to -οντε, so that the bearers groan instead ([Aristonicus, Didymus]/A); a papyrus and some codices adopt this solution.” [16] Hypsenor cannot be ‘groaning’, στενάχοντα, if he has already been killed at XIII 402–423. If we were to believe that the Iliadic text as we have it, with στενάχοντα at XIII 423, is the result of a performative blunder, then the short distance between XIII 423, where Hypsenor is badly wounded, and XIII 402–423, where he is already killed, makes the hypothetical blunder of the performer all the more pronounced.
5§11 But why reject the possibility that Aristarchus found textual evidence, inaccessible to us, that supported the reading στενάχοντε at Iliad XIII 423? That reading is in line with the context of XIII 402–423. Moreover, the form στενάχοντες in the hexameter at Odyssey xiv 354 is in fact attested in precisely the same metrical slot that στενάχοντε occupies in the hexameter at Iliad XIV 423. In other words, an argument can be made that the formulaic system of Homeric poetry, as attested in the surviving Homeric text, can generate στενάχοντε at precisely the point where Aristarchus adduces such a variant. [17]
5§12 To be sure, Aristarchus is not thinking in terms of an oral formulaic system. Instead, his editorial agenda are based on presuppositions of an original text that was supposedly written by Homer. [18] And yet, he adduces a form that could indeed have been generated by such an oral formulaic system. Here as elsewhere, the value of Aristarchus’ testimony is not in his theories per se but in his editorial methods, which could occasionally yield information that would otherwise be lost. As I have argued in earlier work, Aristarchus “respected the reality of textual variants” because, from the standpoint of his own working theory, any one of these variants “could have been the very one that Homer wrote.” [19]
5§13 From an evolutionary point of view, in terms of the model that I have already outlined, it would suffice to say that Iliad XIII 402–423 is incompatible with any version of Iliad XIII 423 that features στενάχοντα instead of στενάχοντες or στενάχοντε. Instead, Janko invokes his theory that the Homeric Iliad was dictated in the eighth century BCE, which would be the point of origin for the “blunder” of στενάχοντα. [20] For nearly half a {114|115} millennium, according to Janko’s explanation, XIII 402–423 happily coexisted with the version of XIII 423 that featured στενάχοντα—until Aristarchus in the second century BCE finally offered his “solution.” According to Janko, “such blunders decisively support Lord’s view that the Iliad is an oral dictated text.” [21] I venture to ask, however, whether Lord’s view is being used here and elsewhere to support Janko’s view—rather than the other way around. [22]
5§14 I see a general problem with Janko’s references to Albert Lord in this regard. Janko’s own view of an oral dictated text of Homer is much narrower than Lord’s. So too are various other views that I have criticized elsewhere. [23] Here I confine myself to pointing out some of the differences between Lord’s and Janko’s views. Lord’s dictation theory stems from his article “Homeric Originality: Oral Dictated Texts,” [24] republished (along with a 1990 addendum) in his book Epic Singers and Oral Tradition. [25] In this article, Lord does not speculate on a specific time (or place) for his heuristic model, which is essentially comparative in nature. Further, Lord’s model is not based on the theory of an eighth-century Homer. Nor is it tied to theories of a textual transmission that somehow persists for several centuries without the possibility of any further significant contact with oral transmission. Nor does it depend on theories about Aristarchean “conjectures.” For all these reasons and more, I resist Janko’s identification of his model with that of Lord.
5§15 In his review, Janko extends the theory of an orally dictated text of Homer, “which was absolutely fundamental to Albert Lord,” to Milman Parry: “Indeed, Milman Parry never even considered any other explanation for the origin of the Homeric texts.” In support of this assertion, Janko cites a single remark of Parry, recorded in fieldnotes published by his son Adam Parry in The Making of Homeric Verse: The Collected Papers of Milman Parry (1971). [26] We find Parry writing these informal notes to himself:

I even figure to myself, just now, the moment when the author of the Odyssey sat and dictated his song, while another, with writing materials, wrote it down verse by verse, even in the way that our singers sit in the immobility of their thought, watching the motion of Nikola’s hand across the empty page, when it will tell them it is the instant for them to speak the next verse. [27] {115|116}

If Janko supposes that this thought of Parry amounts to the formulation of a theory—let alone a unique explanation for “the origin of the Homeric texts”—he should read on: Parry’s linked thoughts, extending through the rest of the paragraph, need to be considered in their entirety. Parry’s next sentence, for example, reads: “The reasons I have for such an opinion are many, some of them still very vague, some very exact.” [28] I stand by what I said in Bryn Mawr Classical Review 1997: Milman Parry never formulated a “dictation theory.” [29]

5§16 On the next page of this printed version of Parry’s informal notes to himself, a different but related dimension of his thinking emerges: “The whole problem of the transmission of the poems once composed is also one which must be considered in detail.” [30] Parry goes on to ponder “the alterations made to the Southslavic texts” by “unscholarly collectors and editors.” At this point in his thinking, Parry is undecided about the relevance of these Southslavic typologies to the history of the Homeric textual tradition, though he seems to be leaning in the direction of discounting the variants: “A methodological study along such lines will probably show us much about the sources of the variants of the texts such as Ludwig [sic] and Allen give them in their editions, about the longer and shorter papyrus texts, and the action of the early editors.” [31]
5§17 I note en passant that Adam Parry’s edition of his father’s work allows the name of Arthur Ludwich to be misspelled here. Moreover, Ludwich is missing altogether from the index of Parry’s edition. Nor does this index record the reference by Milman Parry to T. W. Allen’s work on the text of Homer. [32] It is not that I blame Adam Parry. My point is far more simple. I see here a minor symptom of a major trend: Homeric studies in the era after Milman Parry—an era that extends into the present—have tended to neglect the work of Ludwich, despite the signals that Parry had left behind. In this and in many other respects, I believe, things would have turned out differently if Albert Lord had been invited to edit Milman Parry’s collected papers. My sense is that Lord’s editing in such contexts would have been different: he would have followed up on all of Parry’s inchoate signals, which lead in a variety of different directions—including the lines of thought represented by the likes of Ludwich.
5§18 The neglect of Ludwich’s work is in turn symptomatic of another {116|117} major trend: recent Homeric studies have tended to slight Aristarchus’ contributions to the textual history of Homer, treating the variants that he adduces as mere conjectures. My book Poetry as Performance and Haslam’s chapter for the New Companion seek to reverse this trend. [33]
5§19 Here I return to a point I made earlier about Ludwich: he, along with Karl Lehrs, was a premier defender of Aristarchus’ editorial methodology against the attacks of Friedrich August Wolf, as Rudolf Pfeiffer observes in his History of Classical Scholarship. [34] With reference to Wolf’s position, I applied the term “Wolfian vulgate,” as used by Michael Apthorp, [35] to what I describe as “post-Wolf Homer editions that tend to discount the judgments of Alexandrian critics.” [36] Janko objects to my use of the term, particularly with reference to the recent editions of the Homeric Iliad and Odyssey by Helmut van Thiel, [37] claiming that Wolf did not produce his own editions of Homer. [38] My point, in any case, is that the concept of a “Wolfian vulgate” is antithetical to the concept of a unified edition. Moreover, it is antithetical to the concept of an Aristarchean edition, as envisioned by Ludwich: for Ludwich, the Alexandrian “edition” of Aristarchus represents a quantum leap beyond the pre-Alexandrian “vulgate”; for van der Valk, by contrast, the pre- and post-Alexandrian “vulgate” text is relatively superior to the Alexandrian “edition” of Aristarchus, which may not even be deserving of the term “edition.” [39]
5§20 Milman Parry writes elsewhere about Ludwich’s editions of Homer: “for my purposes the traditional text is that of Ludwich.” [40] He is careful to add: “‘traditional text’ is of course a relative term.” [41] Parry’s understanding of this “traditional text” of Homer is germane to the methodology of Aristarchus:

We must go back to the principle of Aristarchus of getting “the solution from the text,” but we must enlarge it until it covers not only the meaning {117|118} of a verse or passage but the poems entire, and lets us know why the poet, or poets, of the Iliad and Odyssey made them as they are, or as they were at first. Whatever feature of poetic art we may study, we must follow it throughout the traditional text, and try to see it clearly and fully; but our hope will not be to find places out of harmony with one another, but instead, after finding all the elements of the poems which bear upon that feature, to draw from them when we can, but from them only, a new idea of poetic artistry [emphasis mine]. [42]

This methodology is Parry’s incipient answer to the Homeric questions that he says the “scholars of our time” are unable to answer. He describes these scholars as neo-unitarians who have succeeded in refuting the analysts—but who fail to give satisfactory answers to his questions, which he formulates as follows:

… what reasons have they had for passing over the fact pointed out by Wolf that a limited use of writing for literary purposes, which is the most one can suppose for Homer’s age, must have made for a poetry very unlike ours? What source have they given for the tradition [Pausanias 7.26.13; Josephus, Against Apion 1.2.6] that Homer was recorded only at a later time? How have they explained the unique number of good variant readings in our text of Homer, and the need for the laborious editions of Aristarchus and of the other grammarians, and the extra lines, which grow in number as new papyri are found? Finally, have they shown why the poems should be of such a sort as to lend themselves to the many attempts to show the parts of which they were made, and have they told why these attempts were often made by men of the best taste and judgment? [43]
5§21 Parry’s own answer, to repeat his wording, is that “we must look back to the principle of Aristarchus of getting ‘the solution from the text’.” My book, Poetry as Performance, follows up on this principle by considering the “traditional text” in the relative sense of the term as Parry applies it, not in the absolute sense of Aristarchus. [44] Here I return to my earlier formulation, but now with three points of emphasis instead of two: “Whereas Aristarchus … may have gone too far in positing the authentic reading in any given case throughout the Homeric text, there is no reason to doubt that any Homeric variant attributed to Aristarchus can be considered an authentic reading.” [45]
5§22 To test this formulation, I return to Janko’s sample list of thirty-three {118|119} Aristarchean “conjectures,” minus the lead-off example that I have already questioned, Iliad XIII 423. [46] In each of the thirty-two cases I am about to study, I will argue either (1) that the given form adduced by Aristarchus can be considered an authentic reading or (2) that the given form was only mentioned but not actually proposed by Aristarchus:

1. Iliad I 423-424 (IC p. 26n30): ἐπί vs. μετά at 423 and ἐπί vs. κατά at 424. Janko’s observations here are most useful for arguing against the assumption of Aristarchean conjecture. It is clear from the verbatim quotations of Aristarchus at Didymus/A that he simply glossed μετά and κατά by way of ἐπί, while the paraphrase of Didymus/T misrepresents Aristarchus as if he had argued for ἐπί as a genuine variant. I agree with Janko’s inference that the scholia often give “a false impression of his work” and, a fortiori, of the work of his predecessors. The same inference could have been invoked, I contend, in several other cases in Janko’s present list. I invoke it also in general against some other claims that Janko adduces. [47] Aristarchus and his predecessors deserve the benefit of the doubt whenever they are paraphrased by indirect sources on the subject of variant readings.
2. Iliad IX 222 (IC p. 26): ἂψ ἐπάσαντο or αἶψ’ ἐπάσαντο vs. ἐξ ἔρον ἕντο. Here we see a case for arguing the second of the two alternatives, namely, that the given form was only mentioned but not actually proposed by Aristarchus. The scholia make it explicit that Aristarchus chose not to adopt in his base text such a different reading, ἂψ ἐπάσαντο (Didymus/A): ἄμεινον εἶχεν ἄν, φησιν ὁ Ἀρίσταρχος, [εἰ] ἐγέγραπτο “ἂψ ἐπάσαντο”· ἀλλ’ ὅμως ὑπὸ περιττῆς εὐλαβείας οὐδὲν μετέθηκεν, ἐν πολλαῖς οὕτως εὑρὼν φερομένην τὴν γραφήν ‘It would have been better, says Aristarchus, if it had been written “ἂψ ἐπάσαντο” or “αἶψ’ ἐπάσαντο”; nevertheless, because of his extreme caution, he changed nothing, having found in many of the texts this attested way of writing it [= “ἐξ ἔρον ἕντο” instead of “ἂψ ἐπάσαντο”]’. The wording assumes that some of the texts did indeed feature ἂψ ἐπάσαντο instead of ἐξ ἔρον ἕντο. [48] (We may compare the parallel morphology of ἂψ ἐγένοντο at Odyssey x 395; also the ἂψ vs. αἶψ’ variation at Odyssey x 405 and viii 92.) I infer that Aristarchus ‘changed nothing’ (οὐδὲν μετέθηκεν) even though he could have made a change on the basis of manuscript attestations of a variant reading. Moreover, he is quoted as considering the variant reading as a contrary-to-fact proposition. Accordingly, it seems unjustified to describe such readings as his own editorial conjectures. {119|120}
3. Iliad IX 394 (IC pp. 26–27): γε μάσσεται vs. γαμέσσεται. I am not sure whether Janko was counting this case as one of his thirty-three examples, or whether he counted as two examples the case of I 423–424. In any event, it is not justifiable to assume that Aristarchus adduced μάσσεται mainly to avoid a violation of “Hermann’s Bridge,” as Janko claims. (Hermann’s formulation dates back to 1805.) Also, the morphology of μάσσεται could have been generated by the formulaic system. Compare for example ἐπιμάσσεται at IV 190.
4. Iliad XIV 125 (IC p. 165): εἰ vs. ὡς, where ὡς is read “rightly” by the dēmōdeis (δημώδεις) ‘popular’ (Didymus/A), the papyri, and the codices. But how can Janko be sure that the reading ὡς is “right”? Granted, “ὡς is common after ἀκούω,” but ἐτεόν, as here, is attested elsewhere with εἰ twenty-one times, by Janko’s own count. Thus there are valid reasons to justify either reading, εἰ or ὡς, in terms of the formulaic system that generates Homeric diction. Janko’s position is that he has to choose one or the other variant, and he has to discredit the variant(s) that he does not choose, because his dictation theory depends on one single text stemming from the eighth century. In Poetry as Performance, I questioned this position of Janko and summarized my counter-position as follows: “the empirical methods of comparative philology and the study of oral tradition can be used only to defend a variant reading as traditional, not to establish it as the superior reading—let alone the correct reading.” [49]
5. Iliad XV 197 (IC p. 248): βέλτερον vs. κέρδιον. Both are attested in manuscripts, but Janko claims that κέρδιον is “confirmed” on the grounds that he sees an “echo” at XV 226 (another κέρδιον, but in a different verse-position) and that this word is “formular in this phrase, unlike the scholars’ emendations.” Although I find no instances of βέλτερον εἴη elsewhere in the Homeric corpus, there are attestations of βέλτερόν ἐστι at Iliad XVIII 302, XXI 485, and Odyssey xvii 18 (all before the trochaic caesura, in a position that accommodates cognate phrases in verse-final position; all with infinitive constructions similar to the one at XV 197).
6. Iliad XVI 5 (IC p. 315): θάμβησε vs. ᾤκτιρε. Although θάμβησε in this verse is not attested in the manuscripts known to us, it is attested elsewhere, as at XXIV 483, where Achilles is looking with wonder at Priam just after the old man has kissed his hands at 478—and just before Priam begins to speak at verse 486, appealing to Achilles to pity him. The exegetical reasoning of Aristarchus, as reported in Didymus/T, is called by Janko a “misjudgment,” which “removes the central theme of pity from this central scene of the Iliad.” I see no such removal and no such misjudgment as I read the contexts {120|121} evoked by θάμβησε. The purported “misjudgment,” according to Janko, “confirms that Aristarchus could emend on improper grounds.” But we do not know his grounds, since we do not know of his manuscript evidence. Here and elsewhere, Janko has not been able to prove that Aristarchus would ever consider an emendation on the basis of content alone.
7. Iliad XVI 50 (IC p. 322): εἴ vs. ἥν. “Aristarchus altered ἥν to εἴ (in no ms).” I agree with Janko that ἥν as in ἥν τινα οἶδα can be justified on the basis of parallels as at Odyssey i 415 and ii 201; still, εἴ as in εἴ τινα οἶδα can also be justified on the basis of parallels as at Odyssey viii 145–146: πείρησαι ἀέθλων | εἴ τινά που δεδάηκας· ἔοικε δέ σ’ ἴδμεν ἀέθλους. Again, there are valid reasons to justify either reading at XVI 50 in terms of the formulaic system that generates Homeric diction.
8. Iliad XVI 638 (IC p. 392): Σαρπηδόνι δίῳ vs. Σαρπηδόνα δῖον. Janko assumes that Aristarchus emended from accusative to dative, without manuscript evidence. Again, an argument from silence. Also, compare συμφράδμονες plus dative at Iliad II 372; also συμφράζομαι plus dative at I 537, 540; IX 374; Odyssey iv 462.
5§23 Janko’s secondary list begins here:

9. Iliad XIII 28 (IC pp. 46, 122): ἠγνοίησαν vs. ἠγνοίησεν. There are other instances where a neuter plural subject takes a plural rather than singular verb, as at Iliad XII 159, and I see no reason to assume that Aristarchus “normalized” from singular to plural on the basis of such examples. So I disagree with Janko’s premise when he says: “it is wrong to normalize an oral dictated text.” [50] I have already indicated my opposition to Janko’s invoking “an oral dictated text” as a premise for his rejecting one given variant in favor of another. [51]
10. Iliad XIII 191 (IC p. 71): χρόος vs. χροός. I agree with Janko that Aristarchus interpreted χρόος as a diectasis of χρώς. It does not follow, however, that the form χρόος is an editorial conjecture in the sense of an alternative reading proposed by the editor to replace a supposedly false form as transmitted in the received text. For Aristarchus, accents were not part of the textual tradition of Homer. [52] At best, we may consider χρόος an exegetical reconstruction, however flawed, that Aristarchus must have considered in his commentary.
11. Iliad XIII 384 (IC p. 96): ἦλθ’ ἐπαμύντωρ (also XV 540) vs. ἦλθεν {121|122} ἀμύντωρ. Despite the parallel attestation in Odyssey xvi 263, Janko thinks that Aristarchus’ reading “should be rejected,” in view of ἦεν ἀμύντωρ at XV 610. But compare the verse-final placement of the verb ἐπαμύνω as at Iliad VI 361 and XII 369. I suggest that neither reading should be “rejected.” Both forms could have been generated from the formulaic system of Homeric diction.
12. Iliad XIII 449 (IC p. 104): ἴδῃ vs. ἴδῃς. Aristarchus’ adducing of the variant ἴδῃ alongside ἴδῃς here and elsewhere does not necessarily mean that he is “standardizing.” Janko claims: “The Alexandrians wrongly standardize one way or the other.” Rather, it may simply be a matter of consistently reporting such variants. Compare my remarks on “normalization” at Iliad XIII 28. Again, both forms could have been generated from the formulaic system of Homeric diction.
13. Iliad XIII 584 (IC p. 118): ὁμαρτήδην vs. ὁμαρτήτην. Janko claims that “this is a conjecture to avoid having two main verbs.” But compare the interaction of adverbial -δην with other verbs, as in the case of such forms as κλήδην, ἐξονομακλήδην, ἐκ δ’ ὀνομακλήδην at Iliad IX 11, XII 415 and Odyssey iv 278, xii 250.
14. Iliad XIII 599 (IC p. 120): ἐυστρεφεῖ vs. ἐυστρόφῳ. Given that the two forms are both morphologically predictable in Homeric diction, as Janko points out, I prefer to treat them as two interchangeable variants in the formulaic system. I disagree with Janko’s description of the variant adduced by Aristarchus: “He is conjecturing to impose homogeneity.” See also my remarks on XIII 28 and XIII 449.
15. Iliad XIII 810 (IC p. 145): αὔτως vs. οὕτως, attested in a papyrus and in “the good codices.” I question such criteria involving “better” vs. “worse” manuscript readings. [53]
16. Iliad XIV 72 (IC p. 158): ὅτε vs. ὅτι: “this needless conjecture has weak ms support.” Janko argues that ὅτε “tidies up the syntax without altering the sense.” Why assume, though, that it is the editor who tidies up? The formulaic system can generate either ὅτε or ὅτι in this context, and one of these alternatives happens to be more tidy than the other from Aristarchus’ point of view. The adducing of the form by Aristarchus could simply be added to the manuscript evidence, however weak in this case, which points toward the existence of two variants in this context. Janko cannot prove that ὅτε is not an authentic variant. Giving Aristarchus the benefit of the doubt, I prefer to argue that he had access to two variants ὅτε vs. ὅτι, not that he conjectured ὅτε in order to oust a supposedly exclusive ὅτι that he found in the manuscripts. Aristarchus would then proceed to choose one variant {122|123} over the other, on the basis of the internal evidence. Janko’s preference for the other variant, by contrast, is based on external considerations prompted by his theory of an eighth-century archetype that was dictated by Homer. In terms of such a posited archetype, ὅτι seems the plausible choice for Janko, since it seems to him the lectio difficilior; but you need to make a choice between the variants precisely because you are positing such an archetype. From an evolutionary point of view, by contrast, you do not need to choose one or the other variant as the true form. Rather, the choice is relative—depending on the given time and place in the history of the paradosis. In the case of Aristarchus, to repeat, his own need to make a choice in such cases is based on his theory of an archetype written by Homer.
17. Iliad XIV 173 (IC p. 176): κατά vs. ποτί: “but Aristarchus’ alteration is unjustified, since we are dealing with a misused formula.” Rather, I argue that it is unjustified to claim an “alteration.” Further, it is unjustified to claim that the expression ποτὶ χαλκοβατὲς δῶ is “misused” in this context, vs. the other contexts at Iliad I 426 and 438, XXII 505, Odyssey viii 321 and xiii 4 (in the last case, the δῶ is that of Alkinoos, not of Zeus). The reading ποτὶ χαλκοβατὲς δῶ may be less “tidy” (to invoke Janko’s criteria as applied to the previous case) than the reading adduced by Aristarchus, κατὰ χαλκοβατὲς δῶ, but it is still justifiable in terms of the formulaic system that generates Homeric diction. More important, expressions involving ποτὶ δῶμα (verse-final at Odyssey iii 488 and xv 186) and ποτὶ δώματ’ (Odyssey vi 297) are parallel to those involving κατὰ δῶμα (verse-final at Iliad XXII 442 and 478) and κατὰ δώματ’ (Iliad XXIV 512, Odyssey xxi 372) in the formulaic system—both in terms of positioning within the hexameter and in terms of traditional themes at work in the given contexts. Finally, the thematic contexts of κατὰ δῶμα / δώματ(α) in verses like Iliad XIV 257 and Odyssey iv 44 (cf. 72) are evidently cognate with the thematic context of Iliad XIV 173 (about the wonders of the palace of Zeus).
18. Iliad XIV 235 (IC p. 188): τοι χάριν εἰδέω vs. τοι ἰδέω χάριν (scanned – ) in papyri and some codices; also vs. τοι εἰδέω χάριν in the dēmōdeis (δημώδεις) ‘popular’ texts (Didymus/A) and in our “vulgate.” Janko says that Aristarchus’ reading “removes the hiatus [between τοι and ἰδέω] and synizesis [the εω in ἰδέω],” citing Odyssey xvi 236. Actually, the prevalent manuscript reading there is ὄφρ’ εἰδέω ὅσσοι τε …, which scans as – – – – – . The synizesis there [the εω in εἰδέω] suggests to me that Aristarchus’ reading could also feature synizesis: that is, τοι χάριν εἰδέω could scan as – – – as well as – (with non-synizesis of εω but with correption of ω before the following vowel). The placement of χάριν εἰδέω before the bucolic diaeresis may be compared with the analogous placement {123|124} of χάριν ἴδε (scanned as ) at Iliad XI 243. Note too the placement of εἰδώς before the bucolic diaeresis at Odyssey iv 818 and v 250.
19. Iliad XIV 485 (IC p. 220): Ἄρεω ἀλκτῆρα vs. ἄρεως ἀλκτῆρα. The latter “vulgate” reading, with synizesis of εω, is parallel to ἄρεως ἀλκτῆρες at Iliad XVIII 213, again with synizesis; but ἄρης ἀλκτῆρα at XVIII 100. Janko notes: “Aristarchus read Ἄρεω in all three places, but this too [like Zenodotus’ reading ἄρης] is conjectural.” How can Janko be sure? He explains thus about Ἄρεω: “this Ionic form, absent from the mss, first occurs in Archilochus [F 18].” But how can he be sure that such an Ionic form is excluded from Homeric diction? Janko continues: “The truth is surely as follows.” He proceeds to argue that ἄρη (short α, as distinct from the long α of ἀρή ‘curse’) became “confused” with Ἄρης. “The poet let the barely intelligible formula [ἄρης ἀλκτῆρα] stand at [XVIII] 100, but here [XIV 485] and at [XVIII] 213 he substituted Ἄρεος, a normal epic genitive of Ares, found in a few mss; because of the substitution, it has to be scanned (uniquely) with synizesis.” Finally, “Ἄρεως will then be a superficial Atticism, also found as a variant at [Iliad XIX] 47.” From an evolutionary point of view, by contrast, the formulation could be simplified: ἄρης ἀλκτῆρ- can coexist with an Ionicized variant Ἄρεω ἀλκτῆρ- as well as an Atticized variant Ἄρεως ἀλκτῆρ-.
Addendum: I agree with Janko (IC p. 37) that “the superficial Attic traits in the epic diction do prove that Athens played a major role in the transmission, and this must be related to the Pisistratids’ patronage of Homeric poetry.” But I disagree with Janko’s linked idea (IC p. 37) that the Peisistratids “probably procured the first complete set of rolls to cross the Aegean.” From an evolutionary point of view, the Attic phase of Homeric transmission was still a performative phase, not a textual phase (as required by Janko’s theory of an eighth-century dictation). In such an Attic phase, it is important to note, we may expect the evolution of hyperionisms in terms of the performative tradition. That is, hyperionisms could be generated by performances in an Attic-speaking context. From Janko’s point of view, by contrast, hyperionisms must be considered only in terms of the textual tradition. A case in point is the set of hyperionic forms adduced by Zenodotus for the Homeric text, as listed by Janko (IC p. 24). Janko dismisses all these forms as spurious editorial conjectures. In a forthcoming work, I will counterargue that such hyperionic forms are authentic performative variants stemming from an Attic phase of the performance tradition of Homer. [54] {124|125}
20. Iliad XV 82 (IC p. 237): μενοινήῃσι (in a few manuscripts) vs. μενοινήσειε (in most manuscripts, also in a papyrus). “An opt. is odd after an aor. subj., but so is a pres. subj., especially one formed like this …; we are surely dealing with a conjecture.” Janko cites Chantraine to justify his description of μενοινήῃσι as “odd.” [55] But Chantraine in fact defends the authenticity of the form, in terms of the productivity of the ending -ῃσι in the Dichtersprache. [56] For a case in point, consider δώῃσιν at Iliad I 324, XII 275. We may note such other “odd” forms as ἐπιπνείῃσιν at Odyssey iv 357, which scans like μενοινήῃσι, – – – ; both forms are located immediately after the trochaic caesura. Note too μενοινάᾳ at Iliad XIX 164: it scans – – – and it too is located immediately after the trochaic caesura.
21. Iliad XV 114 (IC p. 241): δ’ ἔπος ηὔδα vs. δὲ προσηύδα in most manuscripts. Similarly at Iliad XV 398 and Odyssey xiii 199. According to Janko, the Alexandrians “surely abandoned our vulgate δὲ προσηύδα (with its papyrus support) on the ground that it lacks an addressee in the acc., but this can be supplied from the context (cf. e.g. [V] 871).” I can understand how the Alexandrians could have used this kind of reasoning, but it does not follow that they should have conjectured δ’ ἔπος ηὔδα. I maintain that they would have “abandoned” δὲ προσηύδα only if they had δ’ ἔπος ηὔδα available as a textual variant. I am not persuaded by Janko’s argument that verse-final ἔπος ηὔδα, as attested at Iliad XII 163, could have been a source for conjecturing δ’ ἔπος ηὔδα as an alternative to δὲ προσηύδα. Rather, I view this attestation of ἔπος ηὔδα as a formulaic cognate of δ’ ἔπος ηὔδα. Janko adds that ἔπος ηὔδα “occurs 12x elsewhere, but its ϝ- is never ‘neglected’.” But the “neglect” of ϝ- in δ’ ἔπος ηὔδα does not make this sequence any less formulaic than ἔπος ηὔδα. We may compare the notorious “neglect” of ϝ- whenever a female speaker speaks ‘winged words’: feminine φωνήσασ’ ἔπεα πτερόεντα προσηύδα vs. masculine φωνήσας ἔπεα πτερόεντα προσηύδα. [57]
22. Iliad XV 252 (IC p. 253): ἵξεσθαι vs. ὄψεσθαι. Janko himself compares an interesting variation, attested in the manuscripts, between verse final ἵκηαι vs. ἴδηαι at Odyssey xvii 448. I contend that both pairs of variants, ἵξεσθαι vs. ὄψεσθαι and ἵκηαι vs. ἴδηαι, reflect a functional variation within the formulaic system of Homeric diction. Janko thinks that ὄψεσθαι “accords better with the stress on sight” in the present context. Well and good. But such an editorial preference for one variant over the other does not discredit {125|126} the other variant’s authenticity. From an evolutionary point of view, I contend that both variants are authentic. See also my comments at Iliad XIII 810 on “better” vs. “worse” manuscript readings.
23. Iliad XV 714 (IC p. 305): πέσον vs. πέσεν. Again, a case of neuter plural subject with a plural vs. singular verb. Janko refers back to his discussion of ἠγνοίησαν vs. ἠγνοίησεν at Iliad XV 28, and I in turn refer back to my comments on that case.
24. Iliad XVI 35 (IC p. 320): ὅτε vs. ὅτι: “a needless change lacking ms support.” See my comments on Iliad XIV 72.
25. Iliad XVI 53 (IC p. 322): ὁππότε τις vs. ὁππότε δή. In support of his claim that “Aristarchus altered δή to τις,” Janko says that there are twelve attestations of “ὅπποτέ (κεν) δή.” I find, however, only three other cases of plain ὁππότε δή: Odyssey xx 386, xxiii 345, xxiv 344. In each case, the verb is not in the subjunctive (two indicatives, one optative). In the present case, we see the subjunctive: ὁππότε τις [or δὴ] τὸν ὁμοῖον ἀνὴρ ἐθέλῃσιν ἀμέρσαι. I find two cases of verse-initial ὁππότε τις, and both feature the subjunctive: Iliad XIX 201 and XXI 112. The first of these two cases is strikingly parallel in syntax to the present case: ὁππότε τις μεταπαυσωλὴ πολέμοιο γένηται. The parallelism is in terms of “deep structure,” not “surface structure,” and it would be implausible, I think, to claim that Aristarchus was inspired by a verse like Iliad XIX 201 in preferring ὁππότε τις to ὁππότε δή at Iliad XVI 53. I infer instead that Aristarchus had manuscript evidence for the reading ὁππότε τις alongside the “vulgate” reading ὁππότε δή. From an evolutionary point of view, however, there is no need to justify Aristarchus’ preference, as opposed to Janko’s preference. There is only the need to justify the authenticity of Aristarchus’ reading, alongside the authenticity of the “vulgate” reading (as justified by Janko).
26. Iliad XVI 106 (IC p. 330): καὶ φάλαρ’ vs. καπ’ φάλαρ’ (all manuscripts and all papyri). Janko claims that the καί “is plainly a conjecture,” because it turns the phrase βάλλετο δ’ αἰεί at the end of the preceding line into a “parenthesis.” We may restate Janko’s claim this way: καί is an optional connector with the syntax of βάλλετο δ’ αἰεί, while καπ’ is an obligatory connector. But there are formulaic parallels to the “parenthetical” syntax of βάλλετο δ’ αἰεί (if followed by καί): within the same “Adonic clausula” of the hexameter, scanned – – –, I find such constructions as τείρετο δ’ αἰνῶς at Iliad V 352. Compare also ἵετο δ’ αἰεί at Iliad XIII 424, which is not followed by “necessary enjambment” in this context, as opposed to the context of Iliad V 434, where we do find “necessary enjambment.” Similarly with βάλλετο δ’ αἰεί, we find absence vs. presence of “necessary enjambment” when followed by καί vs. καπ’. As for καί vs. καπ’, compare the reverse situation in Odyssey {126|127} iv 72, where the manuscripts have καὶ δώματα ἠχηέντα vs. κατὰ δώματα ἠχηέντα in the scholia T at Iliad XXIV 323. Compare verse-initial καὶ κεφαλῆς at Odyssey xviii 355, where one of the manuscripts (Allen’s “R12”) reads κἀκε…, leading to the modern emendation κακ’ κεφαλῆς.
27. Iliad XVI 227 (IC p. 347): ὅτι μή vs. ὅτε μή. Janko says of ὅτι μή: “a common idiom in Herodotus and later, has no Homeric parallel.” But ὅτι μή at this verse is attested in some manuscripts, so that it cannot simply be assumed to be non-Homeric. The four cases of ὅτε μή at Iliad XIII 319, XIV 248, Odyssey xvi 197, xxi 185 do not disprove the potential presence of ὅτι μή in XVI 227. Those four cases of ὅτε μή (aside from XVI 227) introduce a verb in the optative, whereas we find no verb introduced by ὅτι μή / ὅτε μή at XVI 227. For Janko to say that the expected verb “is easily supplied” does not explain why the verb is missing only at XVI 227 but not elsewhere. The attested Ionic constructions introduced by ὅτι μή, which are regularly without a verb (compare Herodotus 1.18.3, 1.143.2, etc.), could supply an answer.
28. Iliad XVI 252 (IC p. 351): σάον vs. σόον. “Aristarchus [Didymus/A] wavered [διχῶς] between σάον and σόον,” while “the mss rightly read σόον.” But compare σαόφρονα at Iliad XXI 462 and σαόφρων / σαοφροσύνης / σαοφροσύνῃσι at Odyssey iv 158 / xxiii 13 / xxiii 30. Janko explains that epic forms in σο- “arose by diectasis when the vernacular had contracted *σάϝος to σῶς, just as φόως replaced *φάϝος after it became φῶς.” Still, as we see from Chantraine’s Homeric grammar, φόως occurs when the second syllable is long by position (Iliad VIII 282, XI 2, etc.) but φάος is the regular form when the next word begins with a vowel (unless a caesura intervenes, as at Iliad XV 741). [58] Moreover, there are residual manuscript attestations of φάος before consonant, at Odyssey xviii 317 and xix 34.
29. Iliad XVI 504 (IC p. 381): ἔχοντο vs. ἕποντο. Janko rejects the form adduced by Aristarchus, saying: “it lacks ms support.” But there are other kinds of support: for example, compare the syntax of the verse-final expression ποτὶ δὲ φρένες αὐτῷ ἔχοντο—if we admit ἔχοντο here as an authentic variant—with the syntax of the verse-final πρὸ δὲ δούρατ’ ἔχοντο at Iliad XVII.355. I submit that the two constructions are cognate. Janko adds that “Aristarchus [Didymus/T] and nearly all mss read unmetrical ποτί in [XVI] 504; προτί is a facile normalization of the rough-hewn text.” But why should brevis in longo need to be “normalized” at a penthemimeral caesura? On this point, Parry’s observations are telling. [59]
30. Iliad XVI 522 (IC p. 383): οὗ παιδὸς ἀμύνει vs. ᾧ παιδὶ ἀμύνει. Janko contends that “this effort to emend away a hiatus is in no good ms.” But why {127|128} assume that hiatus was Aristarchus’ main concern? I suggest that he was interested in the lectio difficilior of the genitive vs. the dative with ἀμύνω. Janko himself cites Iliad XIII 402–403, Ζεὺς κῆρας ἄμυνε | παιδὸς ἑοῦ (vs. dative constructions at Iliad XVI 265 and 512).
31. Iliad XVI 668 (IC p. 396): Σαρπηδόνι vs. Σαρπηδόνα. “Aristarchus [Didymus/A] read Σαρπηδόνι, but verbs of cleansing can take a double acc. ([Iliad XI] 572, [XVIII] 345).” In this case, Aristarchus may be opting for the lectio facilior. Also, Janko compares Aristarchus’ “change” at XVI 638. But see my comments on that verse. [60]
32. Iliad XVI 775 (IC p. 408): ὁ δ’ ἐν στροφάλιγγι κονίης vs. ὁ δὲ στροφάλιγγι κονίης. Janko describes the form adduced by Aristarchus as “a facile emendation.” But note the prepositional construction at Iliad XXI 503: μετὰ στροφάλιγγι κονίης. The “deep structure” of the syntax in this case helps explain the ἐν in the other case. Further, the “surface structure” of μετὰ στροφάλιγγι κονίης seems to me too opaque to motivate ἐν στροφάλιγγι κονίης by some sort of analogy.
5§24 We have reached the end of Janko’s list of his best-case arguments for doubting the testimony of Aristarchus. Having offered a counter-argument in each case, I conclude that Aristarchus deserves the benefit of the doubt.
5§25 There remains the theoretical possibility that a variant adduced by the Alexandrian scholars cannot be validated on the basis of what we know about the formulaic system inherited by Homeric poetry. [61] Even if we found such a case, it would not prove that the given variant had to be a conjecture. It could mean simply that we do not have enough data about the formulaic system. We know the workings of that system only to the extent of the surviving texts generated by the system. Even in the Homeric text as we have it we can find numerous unique occurrences, hapax legomena. A “hapax” occurrence in the Homeric text can be just as formulaic as multiple occurrences. The same principle holds for a “hapax” occurrence that exists not in the received text but as a variant adduced by an Alexandrian editor.
5§26 I close by signaling my intention to pursue further the rehabilitation of (1) the concept of an Aristarchean edition of Homer and (2) the importance of variant readings adduced by Aristarchus and other Alexandrian editors of the Homeric textual transmission. {128|131}
[[Left blank are pp. 129–130, marking the transition from Part I to Part II = from Text to Language, so that Ch.6 starts at p. 131.]]

Footnotes

[ back ] * The original version of this essay is N 1998b.
[ back ] 1. Morris and Powell 1997.
[ back ] 2. Janko 1998a; Wace and Stubbing 1962. Since this review by Janko (1998a) was published in an electronic journal, I cannot assign page numbers wherever I quote him.
[ back ] 3. N 1997b = HR 4–7.
[ back ] 4. N 1997d = pp. 101–122 in Morris and Powell 1997 = Ch.1 in this volume.
[ back ] 5. HS 114.
[ back ] 6. Just to be double-sure, I ran a word-check on “emend” and “conjecture” in PP, and I found that there too as in HS I am consistently careful in maintaining a distinction between these two concepts.
[ back ] 7. HS 111.
[ back ] 8. HS 110–111; see p. 13 in the present book.
[ back ] 9. Haslam 1997.
[ back ] 10. To quote from Janko 1998a; see Haslam 1997:80–81.
[ back ] 11. HS 111 = p. 13 in the present book. See also Ch.2 above.
[ back ] 12. Haslam 1997:72.
[ back ] 13. IC 26.
[ back ] 14. IC 24–25. See van der Valk 1963/1964; Kirk 1985 I 43.
[ back ] 15. Janko lists this Iliadic passage first at IC 26n30 (XIII 28, XIII 191, and XIII 384 go into a secondary list), perhaps intending it as a premier case in point (cf. IC 37).
[ back ] 16. IC 99.
[ back ] 17. See also HR 67–68.
[ back ] 18. See Ch.1, p. 11.
[ back ] 19. PP 151.
[ back ] 20. IC 99–100 (also pp. 37–38).
[ back ] 21. IC 99–100.
[ back ] 22. Again, see also HR 67–68.
[ back ] 23. N 1997b.
[ back ] 24. Lord 1953.
[ back ] 25. Lord 1991:38–48.
[ back ] 26. Hereafter cited as MHV.
[ back ] 27. MHV 451.
[ back ] 28. MHV 451.
[ back ] 29. N 1997b.
[ back ] 30. MHV 452.
[ back ] 31. MHV 452.
[ back ] 32. MHV 452.
[ back ] 33. N 1996a = PP; Haslam 1997.
[ back ] 34. Pfeiffer 1968:215–218, with bibliography.
[ back ] 35. Apthorp 1980:xviii.
[ back ] 36. HS 115; see also above at p. 17; cf. PP 135n121.
[ back ] 37. See van Thiel 1996 and 1991 respectively.
[ back ] 38. But see now his retraction of that statement, published in Bryn Mawr Classical Review 98.6.17, = Janko 1998d. In fact, Wolf did indeed publish his own editions of the Iliad and Odyssey, in 1804 and 1807 respectively.
[ back ] 39. HS 114–115. See again Ch.1.
[ back ] 40. MHV 269n5. The editions of Ludwich are Homeri Odyssea I/II (Leipzig: Teubner 1889/1891) and Homeri Ilias I/II (Leipzig: Teubner 1902/1907). (As in the other instance that I have already noted, this reference to Ludwich is likewise omitted in Adam Parry’s index of MHV.)
[ back ] 41. MHV 268n5.
[ back ] 42. MHV 268.
[ back ] 43. MHV 268.
[ back ] 44. PP Ch.5.
[ back ] 45. HS 111. See again p. 13 above.
[ back ] 46. See again Janko’s list in IC 26.
[ back ] 47. I have in mind Janko 1998a.
[ back ] 48. See again my discussion in Ch.3 and Ch.4 of the scholia at Iliad Ι 222.
[ back ] 49. PP 117–118, 132–152; the summary is at p. 133.
[ back ] 50. Janko IC 122.
[ back ] 51. On the principle of lectio difficilior in the analysis of variants stemming from an oral tradition, see PP 129n99. In the same note, I also adduce data collected by Ludwich to argue that Aristarchus’ editorial priorities did not rank internal logic ahead of manuscript evidence.
[ back ] 52. PP 128–132.
[ back ] 53. See above p. 61; also PP 148–149.
[ back ] 54. With reference to an Attic phase, see also PP 134–136 on Zenodotean vs. Aristarchean editorial preferences concerning Iliad I.5.
[ back ] 55. Chantraine GH I 77.
[ back ] 56. Here and elsewhere, I use “Dichtersprache” as a shorthand way of referring to poetic language as a meta-language that evolves from the grammar of everyday language but develops a distinct grammar of its own.
[ back ] 57. Parry MHV 397.
[ back ] 58. Chantraine GH I 81.
[ back ] 59. Parry MHV 213–216.
[ back ] 60. At p. 121 above.
[ back ] 61. I am grateful to one of the two anonymous referees who alerted me to the necessity of addressing this theoretical possibility.