2. Words, Accents and White Spaces1
In adding performance indicators to the text it turns out I follow a most ancient practice in the textual history of Homer, to judge by the fragmentary remnant by which we reconstruct, and project, that history. Gregory Nagy and his Centre for Hellenic Studies has published, or is in the process of publishing online, the Venetus A manuscript of Homer hailing from the so-called ‘Byzantine’ (East Roman) era, the artefact of a manuscript tradition which in my view represents a significant culmination in the art of representing Homeric Greek to non-native Greek speakers. Homer would not exist for modernity, or for me, without this Roman-era advancement. But there is reason to view the achievement of these surviving manuscripts as a maturation of the grammatical art that also serves native speakers of Greek, and—above all—the original composers of the Homeric music. I shall make this case in the following sections.
Nagy points to a scholium on Dionysius of Thrace,2 which describes the office of a ‘corrector’ (διορθωτής). His skill was to put accent marks in the text where needed, so that a student could then take the prepared text to a different teacher, who would herself follow the corrected text in teaching the student how to recite the poetry. This correction was necessary or the student might fall into a ‘bad habit’ (κακὴν ἕξιν). The description seems to be about native speakers of Greek who are studying how to read Greek poetic writing aloud; it is worth noting that there is mooted a bad habit into which such a speaker might fall—though he was a native speaker, like a modern English speaker having a go at Chaucer or Shakespeare or even Keats—but this trap might be prevented by the use of signs, whose import was readily apparent to a different class of teacher, and perhaps the student as well. A modern Shakespearian must learn revénue to earn his revenue.
The first articulation of the parts of speech that inspired Greek grammar, and all sciences based on elements, is the alphabet itself: here was a quasi-phonemic representation of the sound of Greek, using Semitic signs, which itself exposed to consciousness the mostly meaningless atomic segments in speech that combined to form the larger molecules able to carry meaning. Earlier users of Greek writing, and most other users of writing, including Semitic writing, had managed with a syllabary. Some of the Linear B symbols were in fact ideograms, direct referents to things. After the introduced alphabet Greek speech (and poetry) could be preserved as consonants and vowels. But still this was Greek writing for Greek speakers—perhaps something of a shorthand.
The phonetic stream of speech came to be represented by a continuous stream of letters, in a style called scriptio continua: there was no division into words or phrases, and writing flowed in a continuous stream of letters, sometimes reversing directions when necessary, like a snake, or an ox turning the plough (boustrophedon). It seems it was at this stage of the graphic representation of Greek, the scriptio continua—before what we register visually as the separation of words—that the first deployment of the helpful accent marks occurred. In other words, ‘word-level’ accents were sometimes indicated in a style of writing where words themselves—as we (pretend) to understand them—were not visually demarcated.
The ancient Greek accent, however, was a word-level accent. It is essential that this descriptive fact be held in mind during what follows. It is supported by direct as well as comparative evidence. Such a description is, after all, the only rationally imaginable way to make sense of a recessive rule, defined in terms of the last syllable (ultima) of a word.
But what is a word? (Try giving a definition without using words—that’s not just a joke.) Dionysius of Halicarnassus passes on an apparent technical definition of what we call ‘word’, as part of his description of the ancient Greek accent: λέξις ἡ καθ’ ἓν μόριον λόγον ταττομένη. λέξις and λόγος, two words for ‘word’, both appear in the definition. Dionysius’ construction shows the influence of the earlier Dionysius of Thrace, and so there is good reason to think that the formulation arose for the grammarian in the context of scriptio continua, where the representation via Greek letters (grammata), which gave a name to his discipline, did not yet provide any obvious visual aid to describing the phenomenon in question. There is something of untranslatable jargon here; one modern translator (throwing his hands up) simply uses ‘word’ to render the whole Greek phrase; but what the expression tries to encompass is the ‘organisation of the continuous utterance of speech by distinct, rational, unitary parts.’
The generality of such a description, however, can allow it to apply equally, in context, to a phoneme or a phrase as well as to what we call a word. The complex locution produced by the acuity of ancient grammarians in addressing what appears to be the same general concept as ours, ought to give us pause. To what extent, for example, does the practice of leaving spaces between words in writing tend to reify the things, without our having adequately grasped their nature and substance—either what they are or that they are?
For Homer the unitary idea applies best to ἔπος in the sense ‘line’; what we call word division only has to occur for Homer between hexameter lines. Viewed as scriptio continua, the metrical Homeric text observes diaeresis every six feet, without any exception whatsoever—there are no hyphens across lines. The practice of printing Homer by lines perhaps obscures from attention this most fundamental metrical fact in the continuous utterance of his poetry: the musical unity of the Homeric line is absolute and more fundamental than the unity of the word, for all that it presupposes the latter and is itself marked by word division. Aurally, diaeresis demarcates the Homeric unit as surely as does printing stichic verses visually.
In modern grammar we speak also of ‘parts of speech,’ but not in reference to material, phonetic elements of speech. This rather highlights a problem with the attempt at a definition of ‘word’: from the perspective of syntax and morphology, words have wildly different capacities. Some Greek words, for example, are complete sentences in themselves. Others (like articles and particles) can seem as arbitrary as phonemes. In some respects it is only from the late practices in Greek writing that we might see a Greek verb and a Greek particle on some kind of equal footing, as demarcated, separate graphic objects. But even when written in a continuous stream, one imagines a native speaker or his listener would readily separate parts out as he read aloud, audibly or in his mind, if he intended to be understood or to parse statements as sentences.
It turns out a key marker that individuates words, without regard to morphology or meaning, is in fact their prosodic accentuation. (It is perhaps not an accident, therefore, that Dionysius of Halicarnassus invokes the definition of ‘word’ in the context of describing Greek accent.) At some point in their history—prior to Homer in the Greek case—the prosodic pattern in Greek, Latin and Sanskrit became recessive with respect to the ultima, rather than free as in Indo-European; but with different rules in each case. Whenever the Indo-European accents happened to have been located within the permitted final syllables under the recessive regime, they appear to have persisted in those locations without further receding. It could be claimed as a matter of description that prosody mediated between the stream of sound as such (represented by scriptio continua) and its morphological groupings (roots, prefixes and suffixes, thematic vowels etc.), so as to demarcate ‘words’ of widely varying semantic valences; from particles to whole sentences, each and all were, prosodically, ‘words’.
The notion ‘word’ is more of an ideogram in this sense than other entries in the English dictionary, pointing to the jumble-load of things that, since the era of the East Roman texts, we have printed as separate entities in Greek. But the denominator in common is the fact that Greek speech naturally divided into culminative prosodic units, characterised by recessive limitations on the onset of accentuation defined by the nature of the final syllable. These prosodic aural units correspond to what we register visually as ‘words’, with the proviso that polysyllabic prefixes and enclitics may sometimes encompass more than one accentual event; and in certain environments not prior to a pause, none at all—not in the sense that the word’s prosody does not there exist, but that it is suppressed. Suppression can also be seen as a speech act that marks a boundary.
The Greek word-level accent could not cross word boundaries: this demonstrates its culminative function in demarcating words aurally. As I have argued, following the description of W. Sidney Allen, the Greek accent was a ‘contonation’ which rose then fell in pitch.
[T]he Greek accent may be considered essentially as a ‘contonation’, comprising the high pitch and the falling pitch which immediately follows it; this contonation may be either monosyllabic (in the case of the compound accent) or disyllabic; but in either case not more than one vowel-mora (= short vowel) may follow the contonation.3
Where the received accent rose on the ultima of a word, it had nowhere to fall so as to complete the contonation. The deployment of the grave accent mark in writing appears to indicate that such oxytone words had their prosody suppressed: what cannot come down within the word, must not be allowed to rise. But the written practice suggests that pauses following such oxytones, usually at line end in verse, or at mid-line when pauses are indicated by punctuation, freed the pitch to rise; this functionality of pauses in releasing a sort of prosodic momentum in syllables and words, is observed in empirical phonetic analysis. Following enclitics also allowed the release of the ultima rise in pitch on otherwise suppressed oxytones, by allowing the down-glide to occur on these enclitics.
As a corollary, whenever one sees the acute accent mark deployed rather than the grave in ancient Greek—on any syllable—this implies that the contonation is completed, whether on the subsequent syllable in the word, or within the syllable itself if it is closed, or in the first syllable of a subsequent enclitic, or via the physical mechanism of a pause. In the first three cases, this completion is audible, as it is in the case of syllables bearing the circumflex, as a down-glide in pitch.
The key to my new account of the Greek accent is that when this down glide falls on a heavy syllable—within a closed syllable or long vowel, or on the immediately subsequent long vowel or closed syllable—this down-glide in pitch, over two moras, registers as the most dynamically prominent event in the word: its syllable bears the ‘heavy’ (βαρύς) accent. The rise in pitch in the contonation, by contrast, only ever occurs over one mora. In such words, except in the case of all the words bearing a circumflex, this most prominent syllable is not marked; it is, counterintuitively, not the one marked by the acute accent which is most prominent, but the following one. Such long syllables were said to be pronounced ‘with the heavy accent,’ and the words bearing them are (I claim) the original referent for the term ‘barytones’ (βαρύτονοι). μοῦσα, πολλῶν, ἔπερσεν, ἑταίρων, ἱέμενός περ.
On the other hand, it is the syllable bearing the rise in pitch that registers as most prominent when it is not followed by a heavy syllable: that rise is called ‘sharp’ or ‘acute’ (ὀξύς). Words where the acute rise is prominent in relation to the following down-glide—on whichever of the final three syllables this prominence occurs—were originally called ‘oxytones’ (ὀξύτονοι) before the classical tradition garbled the distinction. ἔννεπε, ἀρνύμενος, ἴδεν, νόστον, θυμόν, βουλή, ἱέμενός περ. (Currently we are taught that any word with an acute on the final syllable is an ‘oxytone’, while any word with no accent mark on its final syllable is a ‘barytone’—a distinction without a descriptive purpose, ideal for schoolboys and their idle masters.)
It turns out that these sharp and heavy prominences, so defined, combining the pitch changes of the two segments of the contonation with the relative quantities of the associated syllables, serve to reinforce the Greek poetic metres in such a way as to reveal, for the first time, the dynamic musical pattern that could not otherwise be found for Greek in the relationship between the accentual prosody of her words and the ictus of her metrical feet. There was no such problem for the received Latin stress rules and Latin metres—and most of the latter are either borrowed or cognate with Greek ones. But my work reveals that underlying this perceived Latin stress is also Allen’s contonation: the rule is simply that the voice must rise, where possible, two moras before the ultimate Latin syllable. It is also a recessive rule that concomitantly serves to demarcate the boundaries of Latin words. The key is that unlike in ancient Greek, the rule is indifferent in Latin as to the quantity of that ultima. Whether the ultima is short or long, Latin words are ‘sharp’ on the antepenult when the penult is short, but ‘heavy’ (circumflex) on the penult when the penult is long.
Note that it is a circumstance of the Latin rule that this dynamic tonal prominence, which came historically to register as stress, allows it to be perceived as a monosyllabic accent. One is taught which syllable in a Latin word ‘takes the stress.’ There is evidence that Greek as well, at the time the East Roman manuscripts were produced, had already transitioned to a monosyllabic stress accent, located on the syllable bearing the part of the contonation where the voice rose in pitch. It will prove a key to assessing the achievement of manuscripts like the Venetus A, that it still reflects the classical accentuation which was very often disyllabic.4 There were not only oxtyones and circumflexes, but barytones as well. Ubiquitous in verse and prose cadences are unmarked down-glides on long syllables following the marked acute accent: the voice rose in pitch on the marked syllable, but fell emphatically on the following long, the ‘post-acute barytone’.
But as we’ve noted, native users of Greek once saw no need to demarcate individual words at all. Accent marks only appeared in the Alexandrian period, an innovation ascribed to Aristophanes of Byzantium. About the definition of words and the formulation of rules for word-level prosodic units, the following observation may be instructive:
In their field work among the guslari, the traditional oral singers of Yugoslavia, for example, Milman Parry and Albert Lord discovered that these singers thought in terms of sound groups, not individual words; the two do not necessarily coincide. Only after a Serbian or Croatian guslar had seen a printed copy of his dictated song on a sheet of paper, with the individual words set apart by the white space between them, could he understand what Parry and Lord meant by the term “word.”5
In light of this, perhaps we should take note of the historical development in the case of Greek manuscripts: the development of a notation to indicate prosodic contours preceded what we nowadays interpret as the division of words in texts. Attention to prosody would inevitably have led to identifying the natural ‘sound groups,’ defined in a way that demarcates specific syllables as ultimas—as final syllables of such groups—in relation to which the beginning of the prosodic contonation was recessive according to rule. It is the separating of these ultimate syllables from what follows, and recognising that consequent initial syllables were also thus defined—into which the contonation was not allowed to cross—that led to separating these ‘sound groups’ in their graphic representation. It was the articulated nature of the things themselves that forced the ‘word division’: the prosodic habits and traditions of classical speakers, as they came to be better understood, self-organised the scriptio continua into sound groups with final and initial syllables. From a literate point of view, it was therefore likely that the study of recessive prosodic rules in ancient languages led to the habit of writing that underlies our conception, now second nature, of words as distinct beings.
Professional scribes and native readers evidently did not originally need to record word divisions, or the prosodic features that helped define these divisions. In general modern native speakers do not need accent marks in their writing systems to instruct them which syllables to stress or how to intone their words, phrases or sentences. Such instructions are absorbed osmotically, or if you prefer, ‘transmitted orally.’ Context is adequate to suggest when to read rébel as a noun and rebél as a verb. For the sake of composed music, or language set to music, a separate notational system has been developed for this purpose.
So why is it that a text of Homer might need ‘correction’, in the form of accent marks, for a Greek speaker to learn how to perform the verse? If he already knows the words in their prosody, what would be the point? What are the sources of possible danger whereby, left to themselves, students who knew how to read Greek might all the same fall into a ‘bad habit’ when reciting Greek poetry from a written text? And on what grounds do I myself presume to highlight a text—to ‘correct’ it prosodically, in the ancient sense—that itself exhibits the separation of words, their primary accents, and a computer-generated typescript, which each reflect quite late developments in the history of Homeric publication?
In beginning to tackle these questions, it may be helpful first of all to recognise the beachhead established by the manuscripts like the Venetus A. It is possible to see as an entirely natural consequence of the maturing study of Greek prosody at Alexandria, that ‘sound groups’ would start to define themselves in terms of their final syllables; and that an inevitable consequence of their intuition, in those scholars whose purpose was to both preserve and explicate an ancient prosody, would be to separate these accented sound groups graphically. Hence there were developed in the manuscripts conventions of word division, prosodically significant punctuation, and so on.
This finding must be contrasted as starkly as may be with the view that the distinguishing features of East Roman manuscripts, in relation to Alexandrian and earlier texts—the division of the scriptio continua into separate words, and the systematic deployment of word-level accent marks—were late developments that distort their sources or reflect a late prosodic reality in relation to the original compositions. Not a bit of it. I have demonstrated that it is the disyllabic contonation described by Allen, with its rising and falling segments, that reinforces the metres not only of Greek verse, but Latin poetry as well.6 The achievement of these manuscripts is to preserve or even restore the historical reality of classical Greek word-music, in the face of the historical reality of language change.
If we remember the befuddlement of the guslari at the concept until they saw the ‘white spaces’ in the transcriptions, it is impossible to underestimate the influence of this visual division of sound groups in the development of the modern literate and linguistic conception of the ‘word’. In this sense, the notion of a ‘word-level accent’ remains a true description; but one which implies an historical/genetic development that is opposite to the truth. It is likely their prosody that determined the individuality of words, not the other way around.
Some questions have been left hanging like unripe fruit. To be continued …
Originally from A. P. David, ‘Singing Homer’s Spell: The Disyllabic Contonation and the Proposition Made by East Roman Manuscripts’, Dramaturgias: Revista do Laboratório de Dramaturgia da Universidade de Brasília, 19:7 (2022), 805-66; since revised.
Gregory Nagy, ‘Traces of an ancient system of reading Homeric verse in the Venetus A’, in Casey Dué (ed.), Recapturing a Homeric Legacy, Cambridge MA: Center for Hellenic Studies, 2009, 134-5.
W. Sidney Allen, Vox Graeca, 3rd edn., Cambridge: Cambridge University Press, 1987, 124.
Unlike Greek, Latin can only stress a final syllable if it is a monosyllable! But there is one very important case of a possible disyllabic contonation in Latin: these involve disyllabic iambs. A famous example is cánō in the first line of Virgil’s Aeneid. The instruction to stress the penult here is dead wrong: Virgil is reinforcing, not violating, the opening cadence at the opening caesura of his poem. The voice rises in pitch on the penult, but falls with emphasis on the long ultima, on the ictus of the third foot. This deployment of a disyllabic iamb by Virgil reveals that the Latin accent was just as much a contonation as the Greek.
A. B. Lord, The Singer of Tales, Cambridge, Mass.: Harvard University Press, 1960, 25, 99-123; in Steve Reece, ‘Some Homeric Etymologies in the Light of Oral-Formulaic Theory’, The Classical World 93:2, 1999, 189-90.
see, e.g. David, ‘The Classical Interlude: A Law of Recessive Tonal Accent for “Classical”-Era Indo-European’, Dramaturgias: Revista do Laboratório de Dramaturgia, 4:2 (2017), 229-37.