Go to National Library of New Zealand Te Puna Mātauranga o Aotearoa
Volume 56, 1926
This text is also available in PDF
(608 KB) Opens in new window
– 715 –

Determination of Word-rhythm.

[Read before the Canterbury Philosophical Institute, 5th November, 1924; received by Editor, 31st December, 1924; issued separately, 31st May, 1926.]

In my work on Latin Prose Rhythm I have endeavoured to establish (1) that words have an inherent rhythm, and (2) that in Latin this inherent rhythm is enforced by a stress accent. What I wish to do in this paper is to explain briefly the psychological foundations of my theory, and to support my general contention by particular illustrations taken from the classical languages.

It is necessary first to point out the hazy notion that generally exists as to what rhythm in language really is. Many suppose that rhythm is to be found chiefly, if not exclusively, in poetry; to speak of rhythm in prose seems to them absurd. Others admit the existence of prose-rhythm, but believe that when it occurs it approximates very closely to the regularly-recurrent rhythms of verse. Such misconceptions are widely prevalent, and show the need of a treatment that will get to the very roots of the matter. Far from believing that prose-rhythm can be understood in the light of verse technique, I am persuaded that the converse is true—the rhythms of verse can be properly understood only if we first grasp the rhythms of prose.

It will be convenient here to answer a question that the reader is doubtless asking: What exactly is the difference between rhythm and metre? Rhythm is an ordered succession of syllables grouped into units; but it is important to note that the grouping may be (a) natural, or (b) artificial. The natural grouping is found in normal prose utterance, and to a certain extent in verse; but frequently, especially in melic verse, the syllables are grouped in a way that would be impossible in natural speech. When this happens there is no distinction between the rhythm and the metre; for to read such passages as prose would be to destroy the rhythm intended by the poet. They must either be sung to music or be recited with “plasma.” But there are cases in which the metrical unit is not always the same as the rhythmical unit—viz., when a certain sequence of long and short syllables is preserved along with the natural pronunciation of the words. This happens in the iambic trimeter. Two lines quantitatively identical may contain quite different rhythms. Rhythm is produced by the subjective process of grouping syllables; metre is the term for a purely objective description of syllabic groups that may or may not correspond to a live reality. If they do not correspond, the real units are those of natural speech, and the metrical feet are convenient simply to describe the minimum that lines, rhythmically different, have in common. It is the same with English. Take a “regular” line from Milton's or Shakespeare's blank verse—i.e., one containing ten syllables. These the metrician divides into five “feet,” which, provided the accent falls on the second syllable, are strangely termed “accentual iambi.” But in very many

– 716 –

lines we find more syllables than ten and fewer accents than five, and, since rhythm in English verse depends largely on accent, this “metrical” description is often utterly inadequate for conveying any idea of the varied rhythms of different lines. Thus, in the line “Antonio, gratify this gentleman” (“Merchant of Venice,” iv, 1, 400), there are only three accents, and its rhythm is quite different from that found in “we all expect a gentle answer, Jew” (ibid., 34).

Again, if the accent falls on the first syllable of the so-called “foot” —e.g., “Bring me the fairest creature northward born,” the first (metrical) foot is not an “accentual iambus,” but an “accentual trochee.” In reality, the first rhythmical group is “Bring me the,” and the whole line is in descending rhythm. It is time we gave up the futile system that has so long been in vogue, and recognized that rhythm cannot be conveyed by the artificial dissection of a line into so many “feet,” that very often exist only for the eye. Such “paper rhythm” is divorced from reality. Rhythm, in prose and non-melic verse, is produced by the natural grouping of syllables as we hear them formed in ordinary speech.

Now, to confine our attention to the classical languages, most writers on the rhythms of classical verse and prose agree in believing that rhythm is simply a matter of long and short syllables. If we take a phrase the words of which are united into one rhythmical series, the only way (on this assumption) to describe the rhythm would be to write down the long and short syllables in order. To attempt to find “feet” on the basis of mere quantity is, as Laurand shows (Etudes sur le Style de Cicéron, p. 138), a hopeless task—you can find whatever “feet” you please. In verse the division by feet is easy, because the sequence of longs and shorts is pretty regular, or at least subject to well-defined variations. In prose, on the other hand, the feet are, in Cicero's words, “quasi mixti et confusi.” Laurand feels that the mixture and confusion is such that there emerges only the “règle négative déviter les vers et les parties de vers.” It is true that at the end of a sentence or phrase certain sequences are particularly common; but, even so, alternative scansions are possible.

This extraordinary view implies that words are really amorphous agglomerations of syllables, which make us sensible of a pervading rhythm, but refuse to be dissected or split up into groups. Can anything show more clearly how little the real nature of rhythm in language is generally understood? How can a mere succession of syllables of itself constitute rhythm? Professor Sirén says (Essentials in Art, p. 11): “Rhythm is essentially distinct from mere mechanical repetition. It indicates a rising or falling of certain units, and reveals thereby direction or intention. Rhythm may be even or uneven. In the former case the units follow one another at similar intervals; in the latter case the intervals are unequal.”

An important point to recognize is that without units rhythm cannot exist; it is “artistically controlled movement,” the “guiding principle in creation” (op. cit., pp. 11, 14). “It is a universal law,” writes Goodell (Chapters on Greek Metric, p. 65), “that man is a creature who rhythmizes … every kind of action that admits of it.” So in speech ‘there must be a τάξις, an arrangement of times inter se’ (op. cit., p. 61). This arrangement or grouping is necessitated by the irresistible rhythmizing impulse in man. In the case of English and German it is clear that the rhythmical units are determined mainly by accent, though the syllables composing a unit may exhibit infinite variations of quantity. In the case of the classical languages, investigators have been mostly content with

– 717 –

counting up syllables, and have regarded the rhythm as objectively existent on the printed page, oblivious to the fact that it is only with the subjective process of grouping the syllables that rhythm can be said to emerge.

In any polysyllabic word there must be a grouping of the syllables; the word has an “inherent rhythm,” for we surely cannot suppose that the same word is rhythmized now in one way, now in another.

Let me now summarize the psychological principles on which my theory is based :—


The rhythmizing impulse in man affects speech just as much as any physical movement; syllables cannot, even in the most artless speech, follow one another without being grouped in some way by the speaker.


This rhythmizing impulse causes us to group even mechanically produced sounds (e.g., the ticking of a clock, the stroke of an engine, &c.), though the sounds are qualitatively identical and occupy precisely the same time and occur at equal intervals.


Where some difference in time or quality is introduced, the grouping is forced upon the percipient. The natural tendency to rhythmize is now assisted and enforced by objectively existent distinctions of time or quality : e.g., if alternate taps are louder than the others, there will be a grouping by twos (or by fours, eights, &c., according to the rapidity of succession : in the case of such a larger group we are still conscious of the subordinate groups that compose it). Or, if the alternate taps are slightly longer in time, the same groupings will still be produced. (It is worth noting that psychological experiment has shown that a quantitative difference of a few thousandths of a second is sufficient to cause a grouping).

These are facts beyond dispute, and there is no reason to suppose that the tendency to rhythmize was different in the men of 2,400 years ago from what it is in the men of to-day. Two thousand years is but a fraction of the time covered by the evolution of mankind, and such rhythmical tendencies, though they now seem to be autonomous, are rooted in agelong physiological processes. It will follow that if we pronounce Latin and Greek words with due observance of quantity we should be able to rhythmize them in essentially the same way as did the ancients. This I believe to be the case, but I shall try to establish my thesis independently of this (no doubt debatable) conclusion.


Take a Latin word like attribuitis: the accent falls on the antepenult. Professor Sonnenschein believes that accent in Latin is a constitutive element in rhythm (Year's Work in Classical Studies, 1914, p. 63), and would agree that this word is to be rhythmized as trochee + tribrach (with a secondary accent on the first syllable). This means (1) that the speaker mentally groups the word into two units; (2) that this grouping is also made perceptible to the auditor, since a stress (according to 3 above) is an objective means of marking off one group from another. It will then be impossible to rhythmize attribuitis as dactyl + pyrrhic or as monosyllable +‿ ‿ ‿ ‿, just as in the English word “ambiguity” the rhythmical units are fixed by the accents (of course, the stress in English words is much heavier than in Latin). But the correct rhythmization is objectively indicated not only by the accent, but also by the slight increase in the time given to the second syllable of attribuitis. Let any one read attribu- as a dactyl and he will soon be conscious of the difference in the time of the second syllable.


In the Greek word ἐσκορόδισας the rhythmical units are the same as in the Latin word attribuitis, only we do not know of any stress being placed

– 718 –

on the third syllable. In Greek, then, the variation of time alone makes perceptible to the auditor the correct grouping.


But, while it may be agreed that attribuitis is - ‿ + ‿ ‿ ‿, it may not be conceded that ‿σκορόδισας is to be treated in the same way. There are two possible lines of argument—(1) the psychological; (2) the evidential, which adduces certain facts to be obtained from a study of the literature or the criticism of the ancients.


A full treatment of the psychological considerations would demand more space than can here be devoted to them, but the case may be briefly outlined as follows :—


The groupings referred to in (2) above are generally by twos, less commonly by threes, or exact multiples of these.


The rate of utterance in ordinary speech is favourable to the groupings by twos or threes.


A word, therefore, like συλλήψεσθαι would naturally be rhythmized as two spondees. Even if we regard it as one rhythmical whole, we are still conscious of the two groups composing it (cf. the dipodies of the metrists). In fact, we may say that the word has an inherent rhythm (for which see Latin Prose Rhythm, chapter iii)—“inherent” because the rhythmizing impulse could deal with it satisfactorily in only one way. A word composed of six long syllables (e.g., ἐσαλλαχθήσεσθαι) would be rhythmized as three spondees, because the rate of utterance is not great enough to allow of their being grouped in threes. On the other hand, a word like καταλαβόμενος would be rhythmized as two tribrachs, since the rate of utterance for each syllable has been practically doubled, thus enabling larger syllabic groups to be formed.


In the example just given the syllables are either all long or all short; when long and short syllables are intermingled, the correct grouping is still easier, if possible, to determine (see 3 above). Surely no one will argue that συλλαβέσται can be anything else than trochee + spondee, though the mere metrist would dissect it (in iambic verse) as long syllable + iambus + long syllable. So, ἡγεμονεύειν is dactyl + spondee; καταλαβέσθαι tribrach + spondee; Ἀθηναîοι iambus + spondee; ἀναβήσεσθαι anapaest + spondee.


If a word, as a word, possesses an inherent rhythm, it must end with a rhythmical unit. It is inconceivable, for example, that συλλαβέσθι, was rhythmized in ordinary speech as cretic + long syllable. To rhythmize a word correctly we must first know how the syllables at the end are grouped—“Respice finem.” Thus we shall see that in many words the rhythmical unit (or units) is preceded by an anacrusis : e.g., ἀπηλλάγησαν is anacrusis + two trochees, not, as the metrist would tell us, two iambi + short syllable. Tοξεύη is anacrusis + spondee. (See further below, 2 (a).)


So far we have been dealing with units containing three or four “morae”; there remains one other unit, the cretic, which contains five “morae” and no larger number can be rhythmized as a unit in one word. λαμβάνειν is a rhythmical unit, but συλλαμβάνειν is anacrusis + cretic. Similarly, ἡμετέρων is anacrusis + anapaest (the metrist calls it “choriambus,” and regards it as one unit; this is possible only through πλάσμα—e.g., when the word is sung to music). To rhythmize παρέρχεται as two iambi would be possible only on the assumption that the second syllable is slightly prolonged in time so as to make us conscious of such a division. In reality the antepenult is as closely attached to the penult as is the final syllable : for confirmation of this see below, 2 (c).

– 719 –

After what has been said I hope it will be evident that ἐσκορόδισας is to be rhythmized as trochee + tribrach. To pronounce it as though it were dactyl + pyrrhich would be possible only for βαρβαρόφωνοι, and would be as βάρρβαρον to the Greek ear as attribuitis would sound to the Roman ear if the penult instead of the antepenult were accented. Two shorts cannot be rhythmized as a unit any more than a long syllable, to which they are equivalent in time, unless the two shorts compose one word, as πάτερ, where the two shorts are not equivalent in time to the short syllables of, say, an anapaest, because the final syllable of a word always has its time prolonged (cf. the inane of Quintilian). ἐσκορόδισας exhibits precisely the same rhythm as ἠλλάγησαν, the second long syllable being resolved. That the word was not anacrusis + ‿ ‿ ‿ ‿ will be proved by what follows below, 2 (b).


These psychological considerations have in themselves a high degree of antecedent probability, which develops into certainty when taken in connection with the following facts obtained from a study of the Greek iambic trimeter of tragedy, and of an instructive passage of Quintilian, with which it is convenient to begin.


Quintilian's attempts at prose-scansion show that he had not got past the metrical point of view, though his common-sense and sensitive ear sometimes assert themselves against the imperfectly conceived metrical principles that were current in his day. Thus he was not consulting his ears when he thought that Brute dubitavi might be dactyl + bacchius. But on one occasion at least, his evidence is of capital value. He tells us (ix, 4, 64) that archipiratae, πολλὰ τοξεύη, and criminis causa all exhibit the same rhythm, the chief difference being that the first was molle, the last forte. There can be no doubt, then, that πολλὰ τοξεύη was rhythmized as cretic + spondee, though, if the words occurred in trochaic metre, the metrist would find − ‿ + − − + −. So we are bound to conclude that τοξεύη in prose utterance was rhythmized as anacrusis + spondee. (For further discussion see op. cit., p. 44 ff.)


In the iambic trimeter of Greek tragedy a word like ἐσκορόδισας * is never found in such a position in the line that the second and third syllables together are the equivalent of the usual long syllable, while such a word is found at another part of the line where the two shorts regarded as replacing the long syllable are the third and fourth syllables. Thus in the line

σύ μ᾽ ἐσκορόδισας, οὐκ ἐγώ σ᾽ ἐσκορόδισα

σύ μ᾽ ἐσκορόδισας is correct, but ἐσκορόδισα is not allowed. To my mind this remarkable limitation proves conclusively that ἐσκορόδισα could not be rhythmized as anacrusis + ‿ ‿ ‿ ‿, and that the third and fourth syllables did cohere, so that the word ends with a tribrach. Now, as the tribrach is really a resolved trochee, the line above given ends with the equivalent of a double trochee. Quid quaeris?


The observance of Porson's law bears out my general contention that words have an inherent rhythm. Why should ὦναξ Ἡσακλεις be avoided in tragedy? Surely because the last word was felt to be a unit standing by itself (“law of the final cretic”), and the separation of spondee and cretic has a harsh effect, especially with a trochee preceding.

[Footnote] * I do not take account of the first “foot,” in which all sorts of licences are allowed. It is, as Christ says, “ein ausnahmsfass,” on a totally different footing from the others.

– 720 –
  • ἐξορμώμενων and ὦναξ Ἡσακλεîς exhibit the same rhythm (spondee + cretic), but the former is molle owing to the feet being comprised in one word, while the latter is durum (cf. archipiratae and criminis causa).

The whole trouble has arisen because students of rhythm have often failed to distinguish clearly between metre and rhythm. They have been accustomed for centuries to dissect a line according to its metrical feet, and have regarded these feet as representing the ultimate rhythmical fact. The metrical foot (I am speaking of non-melic verse) is an inert and dead entity that owes its existence solely to the excogitatior of the metrician. It corresponds to no live reality. It may, properly understood, serve as a convenient means of describing what is common to a large number of lines, just as we may say of many of Shakespeare's lines that they contain ten syllables, and yet exhibit a great variety or rhythms owing to the fact that the accent does not always fall at the same place in the line. Metre simply tells us that a line is composed of so-many long and short syllables arranged in a more or less definite order, and conveys nothing of the wonderful variety of rhythms found in lines that have the same quantitative sequence. This variety is produced by the words, in which lie embedded the manifold rhythms that are woven into the metrical pattern. Take a line like

Καὶ πολλαὰ μισηθεîσα τᾖδε γᾖ πάρει.

Here the metrician sees − − | ‿ − | − − | ‿ − | ‿ − | ‿ −, and he sees correctly, as far as his sight goes; the long and short syllables are there and are heard, but we are not conscious of the grouping he has made. The rhythm of the line—i.e., the various units into which the syllables were grouped when uttered—is − | − ‿ | − − | − ‿ | − ‿ − | ‿ −, where the first syllable is an anacrusis. Of course, by no means every line begins with an anacrusis : e.g.,—

πάντων ὅσ᾽ ἔστι κτημάτων དπέρτατον

shows the same sequence of long and short syllables, but there is no anacrusis, while the rhythmical units, with one exception, are quite different.

The pure iambic line consists of twelve syllables which are alternately short and long; we could be conscious of six “iambic” groups only if the syllables were uttered with a mechanically precise observance of the time, as if they were so many taps. But words are entities as well as syllables, and the individuality of a word is shown by a slight prolongation of the time of the final syllable; otherwise words in combination would be a confused medley of syllables. This principle it is that prevents the line being a mere singsong—


The quantities are not grouped in any such lifeless and mechanical manner; they are the constants, but the groupings are the variables. ἠλάγησαν strikes the ear as two trochees;—αγη—is only a grouping formed on paper—it did not strike the ear as an iambus; a true iambus is seen (i) in iambic words—e.g., πόλει; (ii) in words compounded of two or more units—e.g, Ἀτηναîοι, διαφθείρεται (cf. oportebat, &c., in Latin). In Ἀθηναîοι the time of—η—is slightly prolonged, so that it is not joined rhythmically to—αî—to form one unit, as it is in Ἀθηναίοισιν. The ear cannot fail to perceive the different grouping.

In Latin hexameter verse the quantitative and accentual principles are blended, because the accent was not sufficiently strong to prevent the

– 721 –

[The section below cannot be correctly rendered as it contains complex formatting. See the image of the page for a more accurate rendering.]

speaker giving the long and short syllables their full time (in English the strong accent obscures the quantities to a large extent—i.e., it prevents our being able to divide syllables into two simple classes, in which the time-ratios are as near as possible 2/1; hence the futility of trying to write “quantitative” hexameters in English verse). But in the line

Italiam fato profugus Lavinaque venit

the accents are all heard, and the grouping of syllables is not

− ‿ ‿ | − − | − ‿ ‿ | − − | − ‿ ‿ | − ‿

but − | ‿ ‿ − | − − | ‿ ‿ − − | − ‿ ‿ | − ‿

We call it dactylic verse; but this description simply applies to the succession of long and short syllables; it does not give any indication of the rhythms—i.e., the groupings of syllables as they struck the ear. Itali-is not a true dactyl as it was pronounced, because the stress accent on the second syllable absolutely forbids such a grouping; the word is really − + anapaest. The first syllable is not joined to the two succeeding syllables to form a unit recognized by the ear, as is the case with the first syllable of the dactyl in the fifth foot; in this line -vinaque is the only true dactyl. To pronounce Itali-as a true dactyl is to employ πλάσμα and to omit the accent, both quite unwarrantable proceedings.

Note on Accent and Iiaa∑ma.

If the Latin was purely a pitch accent the grouping of syllables described above would be the same, but the accent would have nothing to do with making the correct groupings perceptible to the ear (cf. Greek); slight variations in time alone would suffice, as has been explained above for the rhythmization of Greek words.


IIλάσμα must be employed if we wish to convey, by reading a line, the rhythms that would be made clear if the words were sung to music; otherwise we should altogether miss the rhythms of the Greek choruses. The music does not care how the syllables are ordinarily grouped in prose utterance; it makes its own groupings, to which the singer must conform, though the time of the music is to a large extent determined by the quantities of the syllables.