Latin grammarians borrowed terms from Greek ones to describe their native prosody; so did Latin poets use Greek metres. This is not because they admired and fetishised the ancient Greeks. No doubt they did indulge both those impulses, to admire and to emulate. But the main reason they borrowed the Greek accentual descriptors and metres is because they worked for Latin. The nature of the prosody of Latin and Greek was almost identical: a contonation of changing pitch. It is false to claim that classical Latin had a stress accent, except as a byproduct of pitch contours married to quantities. This is, at last, to be true to the direct testimony of Quintilian. And this is why Latin prosody could reinforce the same quantitative foot-based metres that were, to be sure, developed natively for Greek.
1. The theory
A point of departure for the new theory of the Greek accent was W. S. Allen’s comparison of the Vedic udatta-svarita system with the classical Greek descriptions and prosodic notation. The heuristic basis of such comparison is already the comparative-reconstructive framework of Indo-European linguistics. In light of such a context it is not strictly necessary, but all the same desirable, that comparisons between particular cognates yield fruit in the descriptive analysis of other cognates further afield. The procedure is not deductive but inductive, and not subject to prediction, so the interconnectedness and interoperability of models proposed in a particular comparison, applied successfully also to other cognates, is a welcome self-buttressing to historical theory. Sometimes the whole process is prone to disparagement as ‘circular reasoning’; it is difficult not to be biased in the search for corroboration of a proposal.
Induction of descriptive formulae, as an intellectual process, is, however, circular, and must not apologise for participating in the divinity of that figure. The only sort of ‘proof’ that a descriptive principle is in fact a principle, lies in its having distinct empirical consequences—that is, in this case, manifestations that solve or otherwise illuminate the description of known cognates. This is what we shall find, resolution and illumination, when we apply the model of the Vedic contonation to the quantities of Latin, and re-examine the question of Latin stress.
A multiplicity of distinctions and terms, produced separately by philologists, linguists and metricians, can be confusing. Allen’s term ‘contonation’ refers to the combination of a rising tone and a falling tone, in that order, over one syllable or two successive ones. The Vedic rising tone, udatta, occupies only one vowel mora, but the svarita or down-glide could occupy both moras of a following syllable. A ‘mora’ indicates an element of vowel quantity; its existential status depends largely on its utility in elegant descriptions. A short vowel has one mora (níhil), a long one has two (râri).
Metricians distinguish between ‘heavy’ and ‘light’ syllables. Syllables containing long vowels are heavy, but closed syllables with short vowels also contain the equivalent of two moras, and are considered heavy from the perspective of metre. Philologists somewhat confusingly refer to these vowels as ‘long by position’. Short vowels followed by a mute and a liquid, however, are considered ‘doubtful’ or ‘common’; this is because such syllables can be either heavy or light, depending on their placement in the thesis (another ambiguous philological term, by which I refer to the ictus-bearing downbeat of a foot) or the arsis (the shorter or weak part of the foot, or upbeat). This ability stems from the fact that a mute + liquid can be seen either as divisible, closing one syllable and opening the next, or together as the initial plosion of the ensuing syllable. One should note that on my account, the ‘doubtfulness’ of such syllables closed by a mute and liquid is not a reflection of anything necessarily intrinsic, but only of the convention in Greek and Latin of using such syllables in either the arsis or the thesis of a foot, and so to be performed either as short or as long by a poet’s arrangement.
I have argued that udatta corresponds to Greek ὀξύς and svarita to Greek βαρύς.1 The circumflex denotes a situation where the pitch rises on the first mora of a long vowel and drops on the second, udatta + svarita on one vowel. Thus the Greek prosodic notation is consistent in marking the mora where the voice rises, but only in this one instance (the circumflected vowel) does it indicate the following drop in pitch. Allen suggests that the two elements fused in these situations, with the down-glide predominating; he cites Sanskrit grammarians who describe the cases corresponding in Vedic to the Greek circumflex simply as svarita.2 Meanwhile, it is important to note that despite its name, the ‘grave’ sign in Greek written texts serves not to mark the ‘heavy’ svarita, which is an automatic down-glide following the ὀξύς-udatta; rather, it marks the suppression of the pitch-rise on the ultima of a non-prepausal word. In Greek sandhi, if the voice does not have ‘room’ to descend in pitch within the word, it is not permitted to rise: what cannot come down, must not go up. This ‘border rule’ makes the terminal contonation definitive of a Greek prosodic unit, which for lack of a better term we may call a ‘word’.
Hence the Greek version of the svarita is only indicated in the circumflex, where the pitch-rise occurs on the first mora of a long vowel. An acute sign on a long vowel simply indicates a pitch-rise on the second mora; any subsequent down-glide on the following syllable, whether the vowel is long or short, in this situation or any other, is left unmarked in writing. It should be noted that trochaic shapes with an initial closed syllable, containing a short vowel, are marked with an acute sign, on the only vowel mora available to be marked (e.g., ἄνδρα). Textual and metrical evidence suggests that the contonation was completed within such closed penults in trochaic disyllables, in that a new contonation may begin on the immediately following syllable if an enclitic adjoins, e.g. ἄνδρά μοι. But this need not happen in polysyllables — in ἄνθρωπος, e.g., the down-glide has to occur on the penult, not within the antepenult, as only one vowel mora may follow the end of the contonation in Greek.3
The new theory of the classical accent depends on the idea that accentual prominence derives not just from rising pitch, but from the combination of pitch change with quantity. Hence a down-glide over a closed syllable or a vowel of two moras would be more prominent (‘barytone’) than the preceding rise on a single mora. If, however, the syllable following the rise was short, or there was a following pause, the syllable containing the rise would itself register as more dynamically prominent (‘oxytone’). Although the combination of pitch change and duration is a feature of stress, along with intensity, in these contexts the weakening of adjacent vowels that is also characteristic of stress does not occur. Hence such phrases as ‘dynamic tone’ or ‘tonal prominence’ are in order for this peculiar prosodic phenomenon.
It was found that positions of tonal prominence described by the new theory corresponded exactly with the positions predicted for an hypothesised stress in ancient Greek by Allen’s study of the ends of lines of stichic verse. His rationale for the study follows:
Since Greek metrical patterns, unlike those of classical Latin, were, so far as we know, evolved specifically for Greek, it is likely that they represent, in Meillet’s terms, ‘a stylization of normalization of the natural rhythm of language’. So it is probable that any such patterns of metrical reinforcement would tend to agree rather than conflict with any similar patterns in speech. If this were so, then one might expect that particular syllabic word-patterns would tend to be placed in particular relationships to the strong/weak positions of the verse, even though their purely quantitative structure might qualify them for other placings. And conversely, if one were to discover a strong tendency of this type, it would suggest the presence, in both verse and speech, of some factor additional to quantity—whatever the nature of that factor might be.4
I describe his result in this way:
The study generates a remarkable formula that neatly reveals the ‘preponderant tendencies’ of correspondence between particular syllables and the strong positions of feet; [Allen] claims that these tendencies ‘approach complete regularity.’ … As to the nature of this prominence, Allen is obliged to rule out both high pitch and length; on his understanding, the former belongs to the accent, while the latter is an independent phonemic variable. He concludes: ‘of the three common prosodic parameters … this then leaves only the dynamic, i.e., stress.’5
A. M. Devine and L. D. Stephens described Allen’s study as ‘the first work in the field of Greek metre that can truly be said to understand the requirements of scientific method and theory construction.’6
My own theory for Greek accent happily follows, entirely from two strands of Allen’s work not connected by him:
It turns out, however, that the first four of Allen’s [five] prominence rules constitute the rules for locating either the ὀξύς or the βαρύς accent as I have explained them, in all classes of Greek word with the characteristic recessive pitch accent, as well as two other types (in the traditional nomenclature, long-final oxytones and perispomena). In all of these cases, the syllable primarily stressed according to Allen’s rules is also the primarily accented syllable according to my theory, whether ὀξύς or βαρύς. The only exceptions to this correlation—the only cases in which the stress rules do not predict the location of the ὀξύς accent in an ὀξύς [‘oxytone’] word, or the βαρύς accent in a βαρύς [‘barytone’] word—also involve [certain] exceptions to the recessive accent rule. The true nature of Allen’s prominence is thus revealed. There appears to be a direct connection between the hypothetical stress and the traditional pitch system.7
2. The Latin case
No argument is needed, however, for the idea that Latin had a stress accent. This is the received wisdom. Can any insight into this supposed Latin stress accent be gained from the comparison between Vedic and Greek, by applying the model of the contonation?
The Latin rules are straightforward: disyllables are stressed on the penult, polysyllables on the penult when the penult is long, but on the antepenult when the penult is short or ‘common’ (closed by a mute and a liquid). According to Gildersleeve and Lodge (sec. 15), enclitics ‘are said’ to shift the accent of words accented on the penult to the ultima (prior to the enclitic); but in their own estimation, it is ‘more likely’ that they maintained their ‘ordinary’ location prior to the enclitic.8 We should consider this enclitic environment in light of the contonation, in case the re-analysis helps us decide on the choice of accentuation.
Gildersleeve and Lodge give the following samples, with acutes here marking the stressed syllables: équus, mandáre (to commit), mándere (to chew), íntegrum (mute+liquid), circúmdare, supérstitēs. Note that on the traditional understanding, accent on the ultima never occurs in Latin as it does in Greek; hence the contonation is always completed within the word, and its onset is never suppressed, as is the case with ultimas marked grave in Greek writing.
It turns out that formulating a rule in terms of a Latin contonation is also straightforward: the contonation is universally recessive, and must begin (that is, the pitch-rise must occur), where possible, on the second mora before the ultima. Unlike in the Greek version—crucially—the rule is indifferent as to the quantity of the ultima. Tonal prominence in Latin then becomes an automatic consequence of the possible conjunctions of pitch change and quantity. Once again we should expect two kinds, corresponding to Greek oxytone and barytone, where ‘barytone’ also includes the special case of the circumflex. When the down-glide occurs over more than one mora, it is prominent; otherwise the pitch rise is prominent. This prominence is interpreted as the received Latin stress. Let us take up the examples.
Equus has a pitch pattern that is oxytone on the penult: équùs. The acute in my notation indicates rising pitch; the grave does not represent a suppressed acute, as in Greek, nor does it mean ‘unaccented’ as in a prevailing interpretation, but literally the ‘heavy’ down-glide in pitch that immediately follows the rise. In this case, it occurs over only one mora (-quùs). Bold print indicates prominence according to the proposed rule. Clearly, we have a match for the stress rule. Disyllables appear to present no issues, but the iambic disyllables, short-long, do present a conundrum. We shall reserve its discussion to the end.
Mandāre, with its long vowel in the penult, would require that the contonation begin at the beginning of that vowel. Hence we could represent its prosody with a circumflex, rise + fall: mandâre. Again, a match for the stress rule.
In mandere, on the other hand, two moras receding from the ultima land on the antepenult, with a pitch-rise followed by a light syllable for the down-glide: mándère. In the other examples, similarly we have pitch-rise, prominent because immediately followed by a light syllable, occurring on the same syllable as expected from the stress rule: íntègrum, circúmdàre, supérstìtēs. All these prominent syllables are the same as the ones supposed to be stressed. Hence it is reasonable to conclude that a pitch accent in the shape of a recessive contonation, when applied to the quantities of a Latin word, causes the attested stress pattern and received rule.
There exists, however, a different way to formulate the contonation rule which equally well replicates the received stress rules. It is as follows: the contonation is universally recessive, and must begin (pitch-rise must occur) on the antepenult wherever possible. When the penult is short or light, this rule predicts the same prosodic shapes and locations for prominence as the stress rule; the difference in the analysis comes when the penult contains a long vowel. Mandâre becomes mándàre, where the second version contains what I call in the Greek analysis a ‘post-acute barytone’, a svarita over two moras, rather than a circumflex. The circumflex would then only occur in Latin on disyllables like rârī. The first formulation produces many more instances of the circumflex in stressed penults; the second predicts post-acute barytones in the same locations. Both formulations result in tonal prominence on the same syllable that is supposed to be stressed. Is there any historical evidence that can help decide between the two, between circumflex and barytone, and the formulation of the recession in terms of moras rather than syllable position?
[Part 2 follows.]
A. P. David, The Dance of the Muses: Choral Theory and Ancient Greek Poetics, Oxford: Oxford University Press, 2006, 55 ff.
W. S. Allen, Vox Graeca, 3rd edn., Cambridge: Cambridge University Press, 1987, 122.
see David, 65.
Allen, 132.
David, 69.
A. M. Devine and L. D. Stephens, Language and Metre, Chico, Calif.: Scholars press, 1984, 26.
David, 72.
B. L. Gildersleeve and G. Lodge, Latin Grammar, London: St. Martin’s Press, 1895, 8.