Reading Greek Poetry Aloud: Evidence from the Bacchylides Papyri

Gregory Nagy
[This essay was originally published in 2000 in Quaderni Urbinati di Cultura Classica 64:7–28. The page numbers of the original publication have been placed within braces '{ }', so that {7|8} indicates the page break between p. 7 and p. 8.]
Ancient scholarship on the songs of Bacchylides, as revealed by the visual formatting of these songs in papyri, reveals much that has been neglected by modern Classical scholars. Most of these revelations concern the mechanics and aesthetics of reading aloud the compositions of Bacchylides. The formatting of the Bacchylides papyri features a variety of visual aids for reading aloud: scriptio continua, colometry, and selective markings of accentuation and quantity. How the songs of Bacchylides were read out loud is an indirect indication of how these songs had once been performed in the earlier era of their composition. Further, what we can learn about performing the songs of Bacchylides is applicable to the performance traditions of ancient Greek song and poetry in general.
Papyri containing victory odes and dithyrambs of the poet Bacchylides, as published by F.G. Kenyon in 1897, have considerably supplemented what little has survived of the corpus of nine canonical lyric poets as edited by the scholars of Alexandria in the Hellenistic period. In form and in content, the songs of Bacchylides are closely related to those of Pindar, whose victory odes have for the most part survived by way of the medieval manuscript tradition, along with considerable fragments of his dithyrambs and other compositions. This close relationship between the media of Bacchylides and Pindar has led to a gradual “pindarization” of the text of Bacchylides in modern Classical scholarship over the last century, as is evident from a comparison of the original Teubner edition by Friedrich Blass in 1898, one hundred years ago, and today’s Teubner edition by Herwig Maehler, “post Brunonem Snell” (hereafter abbreviated as SM). My argument here is that the ongoing quest of modern Classical scholars to recover the “real” text of Bacchylides has led to the neglect of insights provided by ancient Classical scholarship—as represented by the Bacchylides papyri, stemming mostly from the 2nd century CE but reflecting a stretch of scholarship that goes back to the 2nd century BCE (and, most {7|8} likely, earlier). [1] This essay highlights some of those insights from the ancient world, which center on a basic question: how to read poetry aloud?
In the Appendix to my essay [see the pdf of pp. 26–27 of QUCC 64 (2000)], I juxtapose a column of lines chosen at random from the Bacchylides papyri, as reproduced in the first Teubner edition of Blass (pp. 46 and 48, lines 59–91; I add the carry-over from line 58) with the corresponding lines in the most recent edition of Snell/Maehler (pp. 17–18, again lines 59–91). The juxtaposition is meant to show how the information inherent in the text of the Bacchylides papyri has been successively eroded by the modern editions. This information, stemming from the scholarship of Hellenistic Alexandria and earlier, involves the following features in the visual formatting of the Bacchylides papyri: (1) scriptio continua, (2) colometry, (3) selective marking of accents, and (4) selective marking of brevis. My essay is divided into four parts, addressing each of these features in the order just given.

1. scriptio continua.

In the Bacchylides papyri, word-endings within the individual lines are not indicated by spaces. This visual format of avoiding word-divisions is known as scriptio continua, and it characterizes ancient Greek writing in general.
In the Bacchylides papyri, the presence of a word-ending within the line is occasionally indicated in indirect ways, by way of special signs. For example, the sign of the “apostrophe” (as at lines 60, 61, 62, 65, 71, 74, 76, 80, 84, 89 in the Appendix [see the pdf of pp. 26–27 of QUCC 64 (2000)]) directly indicates elision—and thereby indirectly indicates word-ending. Similarly, the sign of the “raised dot” (as at lines 62, 67, 70, 73, 76, 77, 78, 83, 84 [2x], 85, 86, 88, 89, 91 in the Appendix) directly indicates syntactical pause—and thereby indirectly indicates, again, word-ending. In this context, I use “pause” merely in a compositional rather than a performative sense. (On distinctions between performative and compositional perspectives in the application of the term “pause,” see Nagy 1998.)
{8|9} The occasions of these signs, as we will see, are significant in and of themselves, and this significance is stripped when modern editors discontinue the format of scriptio continua, printing the lines in a visual format that shows the words separated from each other by way of spaces that were not there in the original written text.
In elaborating this point, I need to consider within a broader scope the general phenomenon of scriptio continua in ancient Greek script traditions, which took shape in an era of expanding literacy, as early as the Classical period of the 5th century BCE, and which persisted all the way through the 9th and 10th centuries CE. [2] A major question is, why was scriptio continua a basic feature of ancient Greek literacy for a period that covers well over a thousand years?
Within this broader scope, I hope to show that the format of scriptio continua is not a disadvantage but an advantage for the mechanics of reading, especially reading aloud. As we will see from the Bacchylides papyri, this advantage of scriptio continua is further enhanced by the usage of such signs as I have just described informally as the “apostrophe” and the “raised dot.”

2. colometry.

In the Bacchylides papyri, line-divisions generally show cola, not periods. The line-divisions in the papyri and in the medieval texts of Pindar also generally show cola, not periods. The line-divisions in modern editions of Pindar, however, show periods, not cola.
This “modern” practice started with the Pindar edition of August Boeckh (Leipzig 1811/1821), which reshaped the line-numbering on the basis of a larger metrical unit that Boeckh defined as the “period” (or “verse,” as Laetitia Parker calls it in the Oxford Classical Dictionary, p. 975). [3] In the next paragraph, I offer a working definition of the period.
While the colon is a smaller metrical unit, the period is a larger metrical unit containing cola. A colon-ending may or may not coincide with a word-ending, but a period-ending must coincide with both a colon-ending and a word-ending. (Occasionally, a period may consist {9|10} of only one colon.) As Boeckh showed in his 1809 book on Pindaric meter, a word-ending that coincides with a period-ending has special features, namely, the potential for hiatus in the case of a word-final vowel and “brevis in longo” in the case of a word-final syllable. [4]
Compare the formulation of West 1982.5: “The period is the fundamental self-contained unit in metrical composition. It is analogous to the sentence in discourse: the sentence is a segment within which there is syntactic continuity and at the end of which syntactical connection is interrupted, the period in metre is a segment within which there is prosodic continuity and at the end of which prosodic connection is interrupted.” [5]
The lower numbers that we see in the line-numbering of the left-hand margins in modern editions of Pindar, like that of Snell/Maehler, go back to the edition of Boeckh, who used his then-revolutionary period-counting format. The higher numbers that we see in the right margins go back to C.G. Heyne’s edition (Göttingen 1798), who had in general used the earlier colon-counting format of the medieval manuscripts. (These higher numbers are useful for cross-references to the Pindaric scholia, since Drachmann’s edition of these scholia uses Heyne’s edition of Pindar as its point of reference.) The coexistence of these higher/lower numbering systems can easily be missed in the more recent editions of Pindar: Snell, for instance, spells it out only on p. 2 of the SM edition, at the beginning of the text of Olympian 1, but there is no indication thereafter that the numberings go back to Boeckh and Heyne respectively. The Loeb edition of William Race, as the editor himself informs the reader (p. vii), omits the right-hand numbering of Heyne altogether.
And yet, the older colometric system of Heyne is closer to what is represented in the Bacchylides papyri, as distinct from the newer colometric system of Boeckh. That is to say, the colometry of Pindar’s poetry as transmitted by the medieval manuscript tradition—and as approximated by earlier editors like Heyne—is cognate with the colometry of Bacchylides’ poetry as transmitted by the papyri. Moreover, the colometry of the medieval Pindar manuscripts matches closely the colometry of the occasional Pindar papyri that have come to light, and {10|11} it also matches the testimony of the ancient metrical scholia to Pindar. [6] Why is it, then, that the publication of the Bacchylides papyri in 1897 did not lead to a decisive reversion—from the period-format of Boeckh’s 1811/1821 edition back to the older colon-format of earlier editions like that of Heyne?
The answer is simple: the period, as formulated by Boeckh, was recognized as a working principle in metrics, not just a visual scheme, whereas the earlier colometry of the cola was generally thought to be an arbitrary—and useless—set of segmentations, even though this colometry dated back to the world of ancient scholarship, in particular to Aristophanes of Byzantium. [7]
The reaction of L.R. Farnell is typical: “Unfortunately we have no external evidence to guide us; for our existing mss. have no ultimate authority on this matter, nor does the discovery of the papyrus of [Bacchylides]..., showing the short line as the metric standard, supply us with evidence of any avail for Pindar, as Kenyon supposes it may [p. xvii].” [8]
Farnell was influenced by the discoveries of metricians like Paul Maas concerning the principle of responsion. This principle, where a given strophe rhythmically matches other strophes—as well as its antistrophe—and where a given epode matches other epodes, establishes the period as the viable unit of rhythmic replication: responsion involves not just the meters but also the “pauses” that delineate these meters, and this sort of delineation can only be understood in terms of the larger unit of the period, not in terms of the smaller unit of the colon. [9] Maas inferred that ancient scholars were unaware of the {11|12} period, on the grounds that they refer to responsion in terms of cola. [10] Here is the way Maas puts it (1962 par. 6): “The colometry of Pindar’s poems in ancient manuscripts, together with the scholia, shows that ancient scholars did not even try to find out where pauses [my emphasis] occurred.” In speaking of “pause” here, Maas refers to his par. 45, where it becomes clear that he means the end of a period. [11] In this context, Maas in using “pause” merely in compositional rather than performative terms.
From Farnell’s point of view, the more “severe” you are about responsion, the less confidence you have in the old colometry. Farnell p. xxiii says that Maas is most severe, while Ulrich von Wilamowitz-Moellendorff and Otto Schroeder are less severe in restoring responsion, that is, in proposing emendations that result in perfectly or near-perfectly matching responsions. [12] C.M. Bowra in the preface to his edition of Pindar (1935; rev. 1947) professes severity, in deference to Maas, though his stance is defensive: at p. viii, he declares that he follows the colometry of Boeckh (note especially the convention of making the period-beginnings flush with the left-hand margin); at p. ix, he says of the textually-transmitted colometry: “haec divisio, ut utilis est atque nonnihil fert oculis venustatis, ita nullam habet apud codices auctoritatem”; wherever possible, as Bowra says (ibid.), he tries to make the concentus to be absolutus—though he hedges about following Maas completely.
The less “severe” Schroeder is a particularly interesting case in point: though he adopted the period-counting format for his 1900 edition of Pindar, he switched to the older colon-counting format in a later edition (1913: see Farnell p. xxii). Another interesting case is Snell himself, whose Teubner edition of Pindar (1943/1953), as his 1964 “Praefatio” declares, is a continuation of the work of Schroeder. When it comes to “correcting” the Alexandrian tradition, Snell says (p. v) that Schroeder was more confident than he is. Still, Snell adopts the period-counting format for his edition of Pindar, though he retains the colon-counting format for his edition of Bacchylides. [13]
{12|13} In response to such controversies over the colometry transmitted by way of ancient scholars like Aristophanes of Byzantium, I propose that the colon, as it figures in this colometry, is in fact a working principle in metrics, just like the period. Even the shapes of the cola as demarcated by the ancient colometry correspond to what we can reconstruct, by way of comparative metrics, as the functional building-blocks of the period. [14] Second, I propose that even the formatting of the colon, as demarcated in the ancient colometry, constitutes a functional building block of the period, even if the periods themselves are not explicitly indicated by way of demarcating their endings. There is, after all, an explicit marking of the non-endings of periods, by way of a mechanism that I am about to describe as hyphenation.
Let me anticipate my conclusions: hyphenation assumes the existence of the period. Moreover, as I will now argue, hyphenation serves as a mechanical aid for fluency in reading aloud. More specifically, it serves as an aid for reading aloud the whole period.
The cola, as we see them in the visual format of the Bacchylides papyri, sometimes do and sometimes do not coincide with word-endings. This format’s non-observance of word-ending is what I am calling hyphenation (as at lines 58, 60, 68, 69, 74, 75 in the Appendix [see the pdf of pp. 26–27 of QUCC 64 (2000)]). Hyphenation is a mechanism that can serve as an aid to reading aloud. It is easier for the reader to develop a “feel” for a period-ending, that is, for a colon-ending that is also a period-ending, simply by developing a “feel” for all the places where hyphenation can happen, that is, where colon-endings need not coincide with word-endings. This mechanism of hyphenation is an aspect of the overall mentality of scriptio continua: it is easier to develop a “feel” for a period-ending, which is followed by a pause, simply by developing a “feel” for all the places where a word-ending must not be followed by a pause. In this context, I am using “pause” in performative rather than compositional terms.
As I tell my beginning students when I initiate them into the reading of dactylic hexameter: (1) try not to stop between words until you reach the end of the line; (2) if you have to stop in order to catch your breath before you reach the end of the line, allow yourself to do so only {13|14} at a caesura or at the diaeresis. The reason for not allowing any place to stop other than the caesuras or the diaeresis is simple: if you do stop anywhere other than those places, you risk breaking the rhythm. [15] Let me restate my formulation by translating it into the metrical terms of the medium now under consideration, the poetry of Bacchylides: (1) try not to stop between words until you reach the end of the period; (2) if you have to stop in order to catch your breath before you reach the end of the period, allow yourself to do so only where the colon-end happens to coincide with a word-end. [16]
With regard to the distinction between performative and compositional “stopping” or “pausing,” it is important to stress that such a distinction is generally not overt in Alexandrian scholarship on punctuation. In the Tekhnē Grammatikē of Dionysius Thrax, a contemporary and student of Aristarchus, there is a concept known as the diastolē, which indicates a separation between words in terms of content. The diastolē (as discussed in section 4 of the Tekhnē) has been interpreted by modern scholars as marking a pause that can be measured in relatively longer vs. shorter units of time. [17] It is clear that Nicanor, an Aristarchean scholar who flourished in the era of Hadrian and who is the author of a treatise on Homeric punctuation (the fragments have been edited by Friedländer 1850), does indeed think of diastolē in terms of relatively longer vs. shorter units of time in physically pausing between words. [18] There is room for doubting, however, whether Nicanor’s notion of diastolē is equivalent to the corresponding notions of his predecessors, especially in the case of Aristarchus and his contemporary, Dionysius Thrax.

3. selective marking of accents.

It is evident from the Bacchylides papyri that each colon, as marked by the colometric descriptions attributed to Aristophanes of Byzantium, has a melodic contour, which is generally marked by one or two accent-signs that indicate the peak or peaks of this melodic {14|15} contour (one accent only at lines 60, 61, 62, 63, 66, 69, 70, 73, 74, 76, 77, 81, 82, 83, 85, 87, 88, 89, 90 in the Appendix [see the pdf of pp. 26–27 of QUCC 64 (2000)]; two accents only at lines 64, 67, 80; no accent at lines 59, 65, 68, 71, 72, 75, 78, 79, 84, 86, 91; three or more accents at no line). [19]
This feature of selective accentuation in the visual formatting of the Bacchylides papyri is linked to the two other features that we have already considered, scriptio continua and the colometry itself. A pioneer in the analysis of selective accentuation is Bernhard Laum (1928).
Laum’s work has not received the attention it deserves. References by later scholars tend to focus on details that need to be corrected. [20] Not enough has been written, however, on the actual data and arguments that Laum contributed.
One of these contributions was Laum’s highlighting of a text that tells how the Alexandrian critic Aristophanes of Byzantium supposedly “invented” the traditional notation-system for ancient Greek accents (p. 62). In his book (pp. 100–102), Laum prints the relevant text of chapter 20 of Pseudo-Arcadius’ epitome of Herodian’s Katholikē prosōidia from the Paris manuscripts 2603 and 2102, and he traces the testimony of this text, however flawed, to Theodosius of Alexandria, whose floruit he dates at around 400 CE. [21] {15|16} Pfeiffer is justified in objecting to the notion that Aristophanes actually “invented” the marking of accents, since there are traces of accentual notation going back to the era of Aristotle. [22] Still, the testimony assembled by Laum demonstrates that Aristophanes had a role in the systematization of accentual notations in the era of Alexandrian scholarship, and that the patterns of selective accentuation that we find in the papyri are related to his system. [23]
Another contribution of Laum was his demonstration that Aristarchus, that pre-eminent Alexandrian editor of ancient texts who flourished in the middle of the 2nd century BCE, did not provide visual formatting for accents in his editions of central poetic texts such as the Homeric Iliad and Odyssey. A key to Laum’s demonstration was the comparative evidence of the Bacchylides papyri, which reflect a selective notation of accentual patterns. These accentuations, as Laum argued, are derived from a system of notations that is independent of the actual lettering of the text of Bacchylides. So also with the accentual traditions that survive in the Homer papyri: the marking of accents is patterned on principles that are analogous to the selective accentuation of the Bacchylides papyri.
For Aristarchus and his immediate predecessors at the Library of Alexandria, questions of Homeric accent were addressed not in the diorthōsis—in the sense of ‘edition’—but in the hupomnēmata or ‘commentaries’ that accompanied the edition as a separate text. [24] Laum successfully challenged the assumption that the Homer edition of Aristarchus entailed the placement of accent-signs over each word in any given Homeric verse. [25] Aristarchus’ information on Homeric accentuation did not get systematically transferred into the texts of the {16|17} Alexandrian Homer editions. [26] It can be said in general for Greek literature that only in the Byzantine editions of the ninth and tenth centuries did it become a regular practice to mark the accent on each word in a given text. Significantly, this newer practice coincides with the discontinuation of scriptio continua.
The older practice, as we see it clearly attested in the Homer and Bacchylides papyri, was the selective placement of accent-signs (and other diacritics, such as breathings). This practice served a practical purpose: the readers of these papyri were concerned not with the accents of individual words per se, which had been the primary research interest of Aristarchus and later Aristarcheans, especially Herodian, but with the correct pronunciation of the colon, which is the equivalent of the entire “verse” in the case of lyric poets like Bacchylides and of part of the verse in the case of Homer.
To cite an example: in the Homer papyri, there is a tendency to signal an acute accent belonging to only one word within a given string of words, instead of signaling all the acutes belonging to all the words. [27] Consider the phrase
τηιδ᾿ετερηισακοσευρύγερον in Pap.Oxy. III 448,
which corresponds to
τῆι δ᾿ ἑτέρηι σάκος εὐρὺ γέρον… in Odyssey 22.184, as we see it spelled in modern editions. To mark the one acute is to mark the highest point of the melodic contour. [28]
It may be possible to compare this kind of pattern with what we find in the Homeric scholia, which frequently refer not to individual words but to strings of words (e.g. Laum 1928.379), reflecting a practical mode of commenting on texts that had once been spelled without word-divisions. In the scholia, there is a tendency to comment on only one accent belonging to only one word within a given string of words instead of commenting on all the accents belonging to all the words (e.g. Laum p. 143). [29]
{17|18} Laum argues that the Byzantine conventions of marking accents go back to Theodosius, whose orthographic system reveals some surprising divergences from the earlier accentual patterns attested by Herodian and his Aristarchean predecessors. It is not the accents of individual words that turn out to be different in the earlier sources: rather, it is the accentuations of word-combinations. Modern editors of ancient Greek texts anachronistically obey the Byzantine accentual orthographic system, to be traced back only as far as Theodosius. Thus they bypass the tradition represented by the earlier Herodian, not to mention the earlier testimony of papyri featuring marked accents. To cite one example:
Though modern editors print a polysyllabic oxytone word consistently with a grave accent when that word is followed by another word without an obvious intervening syntactical break, the evidence of the papyri and of the Homeric scholia indicates that the accent in this context could in fact be acute, not grave: see Laum 1928.152, 159, 161. ... I say “could,” not “should,” because Moore-Blunt 1978 has found several instances of papyri dated earlier than 400 CE where we do see the spelling of grave as well as acute in this same context. Laum treats the earlier pattern of acute spellings as a constant, whereas in fact it is a gradually disappearing tendency. The point remains—and Laum says this just as effectively as Moore-Blunt—that earlier patterns of ancient Greek accentuation are conditioned by the melodic contour, as it were, of the overall syntax. [30]
We may compare the formulation of West 1992.199 concerning a general tendency in ancient Greek melodic traditions: “when the accent [is] on the final syllable of a word, and is not circumflex, and not succeeded by a grammatical pause, then the melody does not fall again until after the next accent.”
For readers of papyri, questions of “melodic contour” were simply a matter of getting the pronunciation right. [31] The scholia to the Tekhnē Grammatikē of Dionysius Thrax put it this way (12.3f, ed. A. Hilgard 1901):
πρὸ μὲν τοῦ ἄρξασθαι τὸν νέον ἀναγινώσκειν ὁ διορθωτὴς λαμβάνων τὸ {18|19} βιβλίον διωρθοῦτο αὐτὸ ἵνα μὴ ἐπταισμένον αὐτὸ ἀναγνοὺς ὁ νέος εἰς κακὴν ἕξιν ἐμπέσῃ· μετὰ δὲ ταῦτα λαβὼν ὁ νέος τὸ βιβλίον διωρθωθὲν ἀπῄει πρὸς τὸν ἀναγνωστικὸν τὸν ὀφείλοντα αὐτὸν διδάσκειν ἀναγινώσκειν κατὰ τὴν διόρθωσιν τοῦ διορθωτοῦ
Before the student would begin to read, the corrector [ diorthōtēs ] would take the book and correct [ diorthoûsthai ] it so that he [the student] would not read it wrong and thus fall into a bad habit. Afterward, the student would take the book, as corrected [ diorthoûsthai ], to a reading-teacher [anagnōstikos] who was supposed to teach him how to read according to the correction-work [ diorthōsis ] of the corrector [ diorthōtēs ]. [32]
In my previous work, I have focused on the inherited melodic contours of the Homeric hexameter, the “melody” of which was reduced in comparison to the lyric meters, with their overt melodies:
In Nagy 1990a.20–28, there is an extended discussion of the phenomenon that I call reduced melody or recitative in hexameter traditions as performed by rhapsodes. For more on the melodic contours of the hexameter, see West 1986.45, who argues that the epic singer of the eighth century “followed the contours given by the word accents”; also, that “this tradition was perpetuated by the rhapsodes, but in a gradually decaying form,” and that “the rhapsodes preserved many archaic accentual features of Homeric Greek into the Hellenistic age”; cf. also West 1981.114 and 1992.208–209. I agree with most of these formulations, though I resist the idea of a “decaying form.” On the concept of recitative, see van der Werf 1967. In the traditions of the Old French chansons, as he argues, there are cases of distinctly recitative melodies and distinctly arioso ones but there are other compositions where “we can no longer discern whether the original of a given line was a recitative on d or an arioso melody with d as a tonal center” (van der Werf 1967.234). In other words, there are instances where “we cannot conclude from the preserved music whether a manuscript gives us a simplified variant of an arioso original or an ornamented variant of a strict recitative” (ibid.). It is clear that “a trouvère recitative could easily be transformed into a trouvère arioso, or an arioso transformed into a recitative” (ibid.). Though it is impossible at times to determine in which direction the shift is headed, whether it is from arioso to recitative or vice versa, it is clear that these two styles were not “two rigorously separated styles for the jongleurs, notators, and scribes at the end of the thirteenth century” (ibid.). We may compare the ancient Greek traditions associated with {19|20} the “lyric” Stesichorus and the “epic” Homer, as discussed in Nagy 1990a.49–51. [33]
These reduced melodic contours, as I have argued extensively, aided in preserving archaisms in the pitch accentuation—archaisms that were otherwise leveled out in everyday Greek. [34]
The phenomenon of melodic contouring is relevant to a rule in traditional Greek music, to the effect that unaccented syllables should not have higher pitch than the acute-accented syllable. [35] Within the melodic framework of such a rule, embedded patterns of archaic accentuation could be preserved. [36] In the Delphic Hymns, for example, syllables having acute or even grave accent in any given word consistently avoid any pitch that is lower than the other pitches assigned to the other syllables in the same word. [37]
Here I disagree with the opinion of West 1992.198–199 concerning choral poetry (including that of Bacchylides): “But in strophic compositions, ... correspondence of accents and melody could only have been achieved if each strophe sung to a given melody had been so composed as to have the same pattern of word accents. So far as we can see, this was never attempted.” West’s concept of responsion is overrestrictive, as if “a given melody” had no flexibility. [38]
If indeed the selective accentuation of the colon, as we see it {20|21} marked in the Bacchylides papyri, reflects the traditional patterns of melodic contouring, then the modern editorial practice of assigning accents to every word in a string of poetry amounts to stripping the melodic contour. Accenting each word is like separating each word from the next one: in one case, you strip the melody, while in the other, you strip the rhythm. [39]
I close this part of the essay by offering a general observation from a diachronic point of view: cola were meant not only to be read but also to be performed, and the performance included the dimension of dance, not just song. [40]

4. selective marking of brevis.

In the Bacchylides papyri, the brevis mark is found over some short vowels that bear an acute accent (as at lines 61 and 70 in the Appendix [see the pdf of pp. 26–27 of QUCC 64 (2000)]). [41]
The significance of this pattern becomes evident when we consider a salient fact about the history of the ancient Greek language: already by the time of Aristarchus, whose floruit was the middle of the second century BCE, unaccented vowels were shortened while accented vowels were lengthened. That is, the accent-system of the Greek language had already shifted to the pattern that we find to this day in Modern Greek. Vital evidence is presented by Horrocks 1997.67, in analyzing a sample papyrus dated ca. 152 BCE (Pap.Par. 47/UPZ 70). In this papyrus, the patterns of confusion in spelling long vowels as short (notably, omicron instead of omega) show that “vowel-length oppositions had already disappeared, a change that is directly correlated with the shift from the classical pitch accent to an accent characterized primarily by greater loudness” (Horrocks ibid.).
{21|22} The implications are far-reaching: even for a knowledgeable scholar like Aristarchus, it seems that reading the meter no longer comes naturally. The Greek language no longer lends itself to quantitative meter. [42]
In this light, the brevis+acute markings of the Bacchylides papyri provide a remarkable confirmation of the accuracy of a notation system that describes a phase of the Greek language that is no longer current at the time that the actual notations are entered.
If indeed the Greek language was no longer compatible with quantitative meter already by the time of Aristarchus, it is easier to understand the limitations of the punctuation-system of the scholars who postdate Aristarchus, most notably Nicanor (floruit in the era of Hadrian), who freely prescribes pauses in performative terms that contradict the inner rules of quantitative meter. [43] It seems to me fair to say, then, that you cannot read Homer in terms of quantitative meter if you follow the punctuation system of Nicanor, which requires the rendering of diastolē in terms of performative pauses.
More important, it is also fair to say that you will find it far easier to read Homer—and Bacchylides—out loud if you follow the visual format inherited from ancient Classical scholarship.{26–27} Appendix [pdf of pp. 26–27 of QUCC 64 (2000)]{26} On this page I reprint a papyrus version of a part of Bacchylides Ode 5, lines 59–91 as printed in the first Teubner edition of Blass (1898 pp. 46 and 48).{27} On this page I reprint a modern reformatted version of the same lines 59–91 as printed in the most recent edition of Snell/Maehler (pp. 17–18). {27–28} What follows is a commentary on the papyrus version (via Blass 1898) of the same part of Bacchylides Ode 5, lines 59–91.
Line 60: The “apostrophe” establishes word-division
Line 61: On the brevis+acute over the i of aïda, see the last part of my essay.
Lines 60–63: For three lines in a row, we see just one acute in each line.
Line 64: Note the two acutes, typical of a “longer” line. Sometimes there are other reasons for two acutes, as at Ode 3 line 2.
{28} Line 65: note the breathing for hoía
Line 65: The “apostrophe” makes it clear that the a of phullanemos goes with anemos and not phulla ; cognitively, the apostrophe gives new information, after the brain has already processed phulla , reassigning the last vowel to the next word. So we see here a prospective mechanism (“this vowel belongs to the next word”), not a retrospective one (“the last vowel of the word that you have just read has been elided”). See also line 60, where the elision mark is even more vital, since the next word after the elision is enjambed into the next line.
Line 70: On the brevis+acute over the i of porthanida, see the last part of my essay.
Line 71: The diaeresis over the iota indicates a new word. See also line 75.
Line 74: There is a closed syllable –ok– in khalkeok-ranon, as indicated with a special mark under –ok–.
Lines 74–75: Note the tmesis of ex-eileto between the end of the antistrophe and the beginning of the epode.
Lines 75–76: The colometry is “off,” by way of the division of anaptu-xas instead of the “expected” ana-ptuxas, if we were to demand “absolute” responsion.
Line 76: Note the syntactical break, marked by the raised dot.
Lines 80, 88, 90, 91, 92: Note the macra over alpha: there is great concern, it seems, about keeping the length of unaccented vowels. Note the macron over unaccented omega at 5.52.


Allen, W. S. 1973. Accent and Rhythm. Prosodic Features of Latin and Greek: A Study in Theory and Reconstruction. Cambridge.
Bakker, E. J. 1997. Poetry in Speech: Orality and Homeric Discourse. Ithaca.
Blank, D. L. 1983. “Remarks on Nicanor, the Stoics, and the Ancient Theory of Punctuation,” Glotta 61:48–67.
Boeckh, A. 1809. Über die Versmasse des Pindaros. Berlin.
Cohn, L. 1888. “Konstantin Palaeokappa und Jacob Diassorinos.” Philologische Abhandlungen für Martin Hertz 123–143. Berlin.
Comotti, G. 1989. “Melodia e accento di parola nelle testimonianze degli antichi e nei testi con notazione musicale,” Quaderni Urbinati di Cultura Classica 32:91-108.
Devine, A. M., and Stephens, L. D. 1994. The Prosody of Greek Speech. Oxford.
Erbse, H. 1960. Beiträge zur Überlieferung der Iliasscholien. Zetemata 24.{22|23} München.
———, ed. 1969–1988. Scholia Graeca in Homeri Iliadem I-VII. Berlin.
Farnell, L. R. 1932. The Works of Pindar II: Critical Commentary. London.
Feaver, D. D. 1960. “The Musical Setting of Euripides’ Orestes.” American Journal of Philology 81.7-15.
Fleming, T., and Kopff, C. 1992. “Colometry of Greek Lyric Verses in Tragic Texts.” Studi Italiani di Filologia Classica 85:758-770.
Friedländer, L., ed. 1850. Nicanoris Περὶ Ἰλιακῆς στιγμῆς reliquiae emendatiores. Königsberg.
Gentili, B. 1988. “Metro e ritmo nella dottrina degli antichi e nella prassi della ‘performance’.” La musica in Grecia (edd. B. Gentili and R. Pretagostini) 5-16. Roma/Bari.
———. 1992. “Dibattito” (with reference to Fleming and Kopff 1992). Studi Italiani di Filologia Classica 85:771-773.
Giessler, J. 1923. Prosodische Zeichen in den antiken Handschriften griechischer Lyriker. Dissertation: Gießen.
Hilgard, A., ed. 1901. Scholia in Dionysii Thracis Artem Grammaticam. Grammatici Graeci Part I Vol. I. Leipzig: Teubner, 1901.
Horrocks, G. 1997. Greek: A History of the Language and its Speakers. London.
Householder, F. W., and Nagy, G. 1972. Greek: A Survey of Recent Work. The Hague.
Irigoin, J. 1952. Histoire du texte de Pindare. Paris. Kenyon, F. G., ed. 1887. The Poems of Bacchylides, from a Papyrus in the British Museum. London.
Laum, B. 1928. Das alexandrinische Akzentuationssystem unter Zugrundelegung der theoretischen Lehren der Grammatiker und mit Heranziehung der praktischen Verwendung in den Papyri. Paderborn.
Law, V., and Sluiter, I., eds. 1995. Dionysius Thrax and the Technē Grammatikē. Münster.
Mazzucchi, C. M. 1979. “Sul sistema di accentazione dei testi greci in età romana e bizantina,” Aegyptus 59:145-167.
Montanari, F. 1979. Studi di filologia omerica antica I. Pisa.
———. 1988. “Filologia omerica antica nei papiri.” Proceedings of the XVIII International Congress of Papyrology, Athens 25-31 May 1986 I 337-344.
Moore-Blunt, J. 1978. “Problems of Accentuation in Greek Papyri.” Quaderni Urbinati di Cultura Classica 29:137-163.
Nagy, G. 1990. Pindar’s Homer: The Lyric Possession of an Epic Past. Baltimore.
———. 1996a. Poetry as Performance: Homer and Beyond. Cambridge.
———. 1996b. Homeric Questions. Austin.
———. 1996c. “Metrical Convergences and Divergences in Early Greek Poetry and Song.” In Struttura e storia dell’esametro greco I/II, edited by M. Fantuzzi and R. Pretagostini, II 63-110. Rome.
———. 1998. “Is there an etymology for the dactylic hexameter?” In Mír Curad: Studies in Honor of Calvert Watkins (ed. J. Jasanoff, H. C. Melchert, L. Oliver) 495-508. Innsbruck. {23|24} Recast in Nagy 2004.144-156, “An Etymology for the Dactylic Hexameter” (Ch. 8).
———. 2004. Homer’s Text and Language. Urbana IL.
Nickau, K. 1977. Untersuchungen zur Textkritischen Methode des Zenodotos von Ephesos. Berlin.
Pfeiffer, R. 1968. History of Classical Scholarship: From the Beginnings to the End of the Hellenistic Age. Oxford.
Robins, R.H. 1993. The Byzantine Grammarians: Their Place in History. Trends in Linguistics: Studies and Monographs 70. Berlin.
Saenger, P. 1997. Space Between Words: The Origins of Silent Reading. Stanford.
Salomons, R. P. 1984. “Use and Meaning of the Middle Point in the Hawara Homer.” Atti de XVII Congresso Internazionale di Papirologia II 249-253. Naples.
Scheller, M. 1951. Die Oxytonierung der griechischen Substantiva auf -ια. Zürich.
Severyns, A. 1933. Bacchylide: essai biographique. Liége/Paris.
Tessier, A., ed. 1989. Scholia Metrica Vetera in Pindari Carmina. Leipzig.
Turner, E. G. 1987. Greek Manuscripts of the Ancient World. 2nd ed. by P. J. Parsons. Institute of Classical Studies, Bulletin Supplement 46. London.
Uhlig, G., ed. 1883. Dionysii Thracis Ars Grammatica. Grammatici Graeci Part I Vol. I. Leipzig.
Wahlström, E. 1970. Accentual Responsion in Greek Strophic Poetry. Commentationes Humanarum Litterarum 47:1-23. Helsinki.
West, M. L. 1981. “The Singing of Homer and the Modes of Early Greek Music.” Journal of Hellenic Studies 101:113-129.
West, M. L. 1982. Greek Metre. Oxford.
West, M. L. 1986. “The Singing of Hexameters: Evidence from Epidaurus.” Zeitschrift für Papyrologie und Epigraphik 63:39-46.
West, M. L. 1992. Ancient Greek Music. Oxford.
Werf, H. van der. 1967. “Recitative Melodies in Trouvère Chansons,” Festschrift Walter Wiora, 231-240. Kassel.
Wilamowitz-Moellendorff, U. von. 1878. Bakchylides. Berlin. {24|25}


[ back ] 1. Ptolemaeus, a contemporary or pupil of Aristarchus, is recorded as making commentary on Bacchylides (Pap.Oxy. XI 1361.5, 13; = scholia to fr. 20A line 19 at SM p. 94); also mentioned as commenting on Bacchylides: Aristarchus, Callimachus, Dionysius Phaselita (scholia to Ode 23 at SM p. 128). Didymus was author of a hupomnēma ‘commentary’ on the victory odes (see SM apparatus at p. 2 for Ode 1 line 8; cf. SM p. xv on Papyrus M). Cf. also SM p. 132 no. 10 and Severyns 1933.147.
[ back ] 2. Note the striking distinction between ancient Greek and Jewish script conventions: in the latter, word-endings are consistently indicated by spacing. Cf. Saenger 1997.11. Cf. also the forthcoming work of Alexander J. Beecroft, “Splitting the Difference: Word-Separation and the Reception of Hellenistic Homeric Scholarship in the Scholiasts to Dionysius Thrax.”
[ back ] 3. Cf. Fleming and Kopff 1992.759.
[ back ] 4. For important adjustments to the formulation of Boeckh, with reference to both hiatus and “brevis in longo,” see Gentili 1992.771–772.
[ back ] 5. I have problems with West’s concept of “interruption” in this context. Instead, I prefer an unmarked alternative, like “pause.” See Nagy 1998. There are also problems with West’s universalizing assumptions about the “sentence”: I prefer the approach of Bakker 1997, especially ch. 3.
[ back ] 6. Irigoin 1952.101: the colometry of Pindar’s victory odes, up to the time of Boeckh, goes back to the metrical scholia, which go back to Aristophanes of Byzantium. This old colometry is faithfully reproduced in e.g. the edition of Heyne. See also Irigoin p. 88, who points out that the colometry of the 2nd c. CE papyrus of Pindar Olympian 2 matches closely the analysis of the metrical scholia. For a new edition of the metrical scholia, see Tessier 1989. See also Irigoin p. 51 on the testimony of Hephaistion p. 74 ed. Consbruch, who seems to be making a distinction between a “current” (νῦν) Aristarchean edition of Bacchylides, as opposed to the Aristophanean.
[ back ] 7. See Irigoin 1952.44ff, who provides a revealing collection of testimonia on Aristophanes’ colometry, including an instance where the scholar obelizes a superfluous colon but scrupulously retains it in his text. See also Irigoin p. 46 on the opinion of Wilamowitz, who thought that Aristophanes’ colometry was worthless; Irigoin himself on p. 47 comes close to renouncing its value. I subscribe to the efforts of Gentili 1992.772–773 in developing a more positive appreciation of the work of Aristophanes.
[ back ] 8. Farnell 1932.xxii.
[ back ] 9. In this essay, I use the words “rhythm” and “meter” interchangeably with reference to archaic in classical songmaking and poetry. On the later distinctions between rhythm and meter, as developed by Aristoxenus and thereafter, see the clear summary of Gentili 1988.13–14.
[ back ] 10. For an example involving Aristophanes of Byzantium himself, see above.
[ back ] 11. For criticism of the formulation developed by Maas, see Gentili 1988.12.
[ back ] 12. For more on the position taken by Wilamowitz, see Gentili 1992.773.
[ back ] 13. Relevant to my argumentation are the comments of Snell in SM pp. xi–xii on Laum 1928. For more on Snell’s attitude toward Alexandrian scholarship on Bacchylides, see also SM p. xxxvi.
[ back ] 14. Extensive discussion in Nagy 1996c, including an inventory of major colon-shapes at pp. 77 (Aeolic meters) and 81 (“dactylo-epitrite” meters). On colometric principles in general, I am in substantial agreement with the views of Gentili 1988 and 1992. I especially agree with Gentili’s appreciation (1988.13) of the valuable work of Wahlström 1970 on traces of melodic/accentual responsion between strophe and antistrophe. See also Allen 1973.231–234, followed by Nagy 1990.39n113. Allen p. 232 acknowledges Wahlström’s research prominently. I cannot find any mention of Wahlström in Martin West’s books on Greek meter (1982) and Greek music (1992).
[ back ] 15. Cf. Nagy 1998, especially p. 499n10.
[ back ] 16. For a reassessment of the dactylic hexameter itself in terms of a period containing cola (from both synchronic and diachronic perspectives), see Nagy 1996c and 1998.
[ back ] 17. But see Saenger 1997.86. Cf. also Beecroft, as cited above.
[ back ] 18. On convergences between the diastolē-system of Nicanor and the punctuation of the Hawara Papyrus of Homer, see e.g. Salomons 1984. Cf. Nagy 1998.499n10.
[ back ] 19. This same pattern holds throughout the Bacchylides papyri: (1) frequently no accent or one accent for each colon, (2) less frequently two accents, and (3) almost never any case of more than two accents. The third of these three features is particularly remarkable. The only exceptions I can find are 5.15 (one circumflex and two acutes), 11.51 (three acutes), 15.48 (two acutes and one circumflex), 17.25 (three acutes), 17.89 (circumflex and two acutes), 18.24 (circumflex and two acutes), 19.11 (three acutes). (I am not counting instances of consecutive graves + further accentuation, as at 1.44, 9.15, 9.29, 10.19, 11.14, 11.44, 13.230, 16.20, 17.91, since the marking of syllables with the grave accent indicates simply the postponement of acute accent rather than any accent per se).
[ back ] 20. See the bibliography in Turner 1987.159. Cf. Erbse 1960.371–406.
[ back ] 21. The reliability of the text as printed by Laum is called into question by Pfeiffer 1968.179 on the grounds that the text of Parisinus 2102 stems from “a disreputable forger of the sixteenth century, Jacobus Diassorinus.” Pfeiffer at p. 179n1 cites as his authority Cohn 1888, who proved that 2102 was handwritten by Diassorinus. What Pfeiffer does not mention, however, is that Laum himself at p. 99 actually cites Cohn 1888 and that he acknowledges Cohn’s arguments concerning false interpolations by Diassorinus. But then Laum goes on to counterargue that the actual wording of the passage, even if it was falsely interpolated by Diassorinus, still goes back to the authorship of Theodosius; the same passage, as Laum emphasizes, is found in 2603 (minus the introductory formula concerning Aristophanes’ “invention” of the accentual system). Pfeiffer (p. 179n1) does not address Laum’s counterargument when he dismisses Laum’s printed text simply on the grounds that Laum “unfortunately mixed up the text of Par. gr. 2603 with the forgery of Diassorinus in 2102.” By “mixed up” he evidently means “merged.” Even Cohn (p. 142n1), I should stress, acknowledges that 2603 is independent of 2102. Also, it is essential to reassess Pfeiffer’s notion of “forgery.” As Cohn himself admits about such manuscripts as 2102 (p. 142), “Fälschungen waren sie nur insofern, als sie mit falschen antiken Autornamen ausgestattet wurden; im übrigen sind sie für jene Zeit anerkennenswerte gelehrte Leistungen, die kaum weit hinter den Arbeiten eines Moschopulos oder Thomas Magister zurückstehen.” The history of the life and times of Diassorinus is of considerable interest in and of itself (cf. Cohn, pp. 137–143; especially p. 139 on the execution of Diassorinus in 1563 on charges of plotting to oust the Venetians from the island of Cyprus, where he had founded a school).
[ back ] 22. Pfeiffer 1968.179–180; cf. Nagy 1996a.125–132.
[ back ] 23. Pfeiffer 1968 concedes this much, at p. 180.
[ back ] 24. Nagy 1996a.125.
[ back ] 25. Nagy 1996a.125, with reference to Laum 1928.60.
[ back ] 26. Nagy 1996a.126, with reference to Laum 1928.327.
[ back ] 27. Nagy 1996a.132. Cf. Laum 1928.164.
[ back ] 28. For more on the ancient practice of marking polysyllabic oxytones with acute rather than grave (note the modern spelling εὐρὺ in the example above) in some clause-medial situations, as here in the case of εὐρύ, see further below.
[ back ] 29. Nagy 1996a.133n113.
[ back ] 30. Nagy 1996a.126–127n87. The formulation of Moore-Blunt 1978.147 provides a most revealing indirect confirmation: “the accentuation may be governed by the phonetics of the sentence rather than by position in the line.” For a negative assessment of Laum’s argumentation, see Erbse 1960.371–406, esp. p. 371n2 (where he follows Giessler 1923.34).
[ back ] 31. Laum 1928.63. On “melodic contour,” see also Nagy 1996a.132n113.
[ back ] 32. Cf. Laum 1928.53.
[ back ] 33. Nagy 1996a.131n108.
[ back ] 34. Nagy 1996a.125–132. Besides accent, there are also variations in breathings that must have survived by way of performance traditions: see Householder and Nagy 1972.66 on such Homeric contrasts as ἁμός vs. ἄμμι, ὑμός vs. ὔμμι. Cf. Laum 1928.365 on the spelling άμμι in Papyrus A for Bacchylides 17.25.
[ back ] 35. Scheller 1951.9n3. See also Gentili 1988.13, with reference to Wahlström 1970.
[ back ] 36. Scheller, ibid. For more on the relationship of pitch accent and melody in ancient Greece, especially on the more archaic style where the melodic patterns are conditioned by the accentual patterns, see Nagy 1990a.39 and n113, following Allen 1973.231–234. Cf. Devine and Stephens 1994.160, 162, 167–168.
[ back ] 37. See Comotti 1989.91–92, who argues that the Delphic Hymns are musically far more conservative than the lyric compositions of, say, Euripides. Cf. Erbse 1960.376–377. See also the helpful account of Gentili 1988.12 on “la spiccata tendenza di Euripide a subordinare la parola al mélos,” as evidenced by the melodic notations in papyrus fragments of Euripides Orestes 140–142 / 153–155 (strophe/antistrophe; = no. 2 in the inventory of West 1992.277ff; cf. Dionysius of Halicarnassus De compositione verborum 11.63) and of Iphigeneia in Aulis 784–792 (epode; = no. 4 in West).
[ back ] 38. I note with interest, however, the concession by West 1992.199n17, citing Feaver 1960. West could have also cited Comotti 1989.97 and Allen 1973.231–234, who both rely on the important observations of Wahlström 1970—again, not cited by West (see above, also with reference to West 1982). Cf. also Devine and Stephens 1994.169.
[ back ] 39. We may compare the concept of the lēmma, derived from lambanein ‘take’: when you cite a lemma, you are “taking,” literally, the given word out of its metrical and melodic context.
[ back ] 40. See also Irigoin 1952.50 on Apollonius Eidographos, who was the successor of Aristophanes of Byzantium. Given that Apollonius classified the odes of Pindar according to modes, e.g. Dorian, Phrygian, Lydian, etc., it is likely that Aristophanes, as his predecessor, had access to a melodic tradition when he was working on the accents of Bacchylides. In the case of the melodies of classical Attic tragedy, Fleming and Kopff 1992.762–763 argue for traditions of melodic transcription that go back to the fifth century.
[ back ] 41. The examples in the Appendix involve brevis+acute over iota, but there are examples of other vowels in the Bacchylides papyri: e.g. brevis+acute over alpha (3.1, 4.16, 5.36) and brevis+acute over upsilon (5.22, 5.145, 11.12).
[ back ] 42. In this connection, we must keep in mind that Aristarchus was not interested in questions of performance in the first place: see Nagy 1996a.130, 150.
[ back ] 43. Cf. Nagy 1998.499n10.