the long notes of the ascending phrase, emphasized, moreover, by the accents of the xylophone, compel us to conceive the supposed up-beat as accented. This, no doubt, places us in the right attitude; but I have not been able to keep up this attitude beyond the choral passage; as soon as the principal theme (A) returns, the former conception re-established itself. Still, a comparison between the two passages shows that A is meant as an analogy to B (bars 3 and 4) and that the lower part is syncopated past our comprehension.

This example is instructive in more than one respect. The combination of binary and ternary time is characteristic of African metre in general. The principal division is twofold: a period breaks up into question and answer, tension and relaxation, arsis and thesis, which in antiphony are mostly distributed between solo and chorus, and, as regards tonality, represented, as a rule, by the contrast between dominant and tonic. We commonly express this by saying that the themes are of two bars. (This is true of all the musical examples quoted except the parlando rhythm of No. 7.) The next subdivision too is mostly twofold. To use the common term, the bar has two beats. (Nos. 1, 3, 5.) Only in further subdivisions groups of two are frequently replaced by groups of three triplets. These two kinds of groups are perfectly equivalent to each other, and evidently alternate in accordance with the text. (At a greater speed groups like ♪ see No. 1-orsee No. 5-are indistinguishable from

2

6 12

3

Whether in such cases, we write♫♫org (see No. 1) is quite arbitrary. In the same way 4, 3, 3, &c. measures, in the very frequent case of time-units being subdivided throughout into threes, are merely convenient modes of stating the fact that a period consists of (3+3)+(3+3) time units (Nos. 4, 8, 9).

African rhythm is ultimately founded on drumming. Drumming can be replaced by hand-clapping or by the xylophone; what really

។

matters is the act of beating; and only from this point can African rhythms be understood. Each single beating movement is again twofold: the muscles are strained and released, the hand is lifted and dropped. Only the second phase is stressed acoustically; but the first inaudible one has the motor accent, as it were, which consists in the straining of the muscles. This implies an essential contrast between our rhythmic conception and the Africans'; we proceed from hearing, they from motion; we separate the two phases by a bar-line, and commence the metrical unity, the bar, with the acoustically stressed time-unit; to them, the beginning of the movement, the arsis, is at the same time the beginning of the rhythmical figure; up-beats are unknown to them. To us, the simple succession of beats ♪♪ appears as syncopated, because we only attend to its acoustic aspect. In order to understand African rhythms as they really are, therefore, we must thoroughly change our attitude; and in order to write them down adequately we must place the bar-line before the rest or the up-beat (Examples 6, 9). (For the convenience of European readers, No. 7 is written in the non-African manner.) The same relation of stressed-unstressed (heavy-light) which exists in binary figures evidently dominates ternary figures as well: the elementary form of African rhythm is not but (see e.g. the triplets in 3 Id Ï No 3, the figure in No. 1 and others.) If the minim is subdivided we get, according to the same principle↓↓↓; and on sub

(P)

dividing it further: that is to say, exactly the figure of which the rhythmical phrase of the lower xylophone part in No. 9 is made up. The stroke on the third quaver() belongs to the 3 rhythm; and we can lay down as a general rule that strokes which are not syncopated belong to a superior division of time. Accordingly, the figure which takes up the second half of the bar (or the third and fourth crotchets of the 12 phrase) becomes

We may now proceed to test the proposition by a second and uncommonly intricate example, the Bahutu song (No. 6). The hand

clapping rhythm, which does not join in until the middle of the phonogram, is apt to surprise and disconcert European hearers. We understand the melody in the way indicated by the small bar-lines: the solo phrase as 4 and the choral phrase as JJJN. As the choral phrase immediately follows on the solo without a pause, its rhythm, in comparison to the solo rhythm, seems to be shifted by a quaver; the drum strokes seem to go on without taking any notice of the shift, and to be syncopated during the choral part, whereas they synchronize with the solo accents. But on analysing the drum rhythm in itself we find, according to the principle laid down before, that it consists of a period of two rhythmical phrases:

The bars have originated, as they frequently do in Ruanda songs, from bars, the two parts overlapping in antiphony by a crotchet. In consequence, this crotchet has a double function also as regards rhythm. It is the final unit of one and, at the same time, the first unit of the next phrase; as a final unit it has a drum stroke (syncopated by a quaver), as a first unit, it has none. (In notation the bars are indicated by [] and .) It can hardly be supposed that this double function is actually realized; if the parts actually overlap, the attention presumably concentrates on the new one, and the rhythmical concep tion is immediately diverted; the prematurely relinquished phrase remains unfinished, and in its state of truncation becomes a new and firm shape.

There are many instances even outside Africa of musical phrases having elongations (pauses) or truncations at the end. But the vital thing is the shape of the melody taken as a whole: and this is not impaired by liberties of the kind. Rhythmical shapes, too, present themselves to the mind as unities; they are performed, and are received by the ear, without the smallest time-units being counted out. In fact they need not fit into any rhythmic pattern at all.

Rhythmic patterns do not arise until people join to sing in chorus and use drum rhythm along with it. But even then, melody and drum rhythm often go their own separate ways; and again, each of the drums may have a rhythmical ostinato of its own to balance the chanted part. This brings about a rhythmical heterophony1 which Europeans, besides being altogether unprepared for it, find all the more difficult to disentangle as it lacks tone-quality. On the other hand, the variety of colours (timbre) brought about by beating the drum-head at different places, e.g. in the middle or on the rim, may be of some help in distinguishing the individual 'voices' of a polyrhythmic structure, at least for native hearers.

Melody and text, like music and bodily motion, originally form an inseparable unity. Except in late and advanced stages of civilization there is no such thing as spoken verse; there is only everyday language and song, i.e. poetry shaped tonally. Words and music come into the singer's mind as one; or rather, the tune is formed along with the first stanza, and remains the same for all the succeeding ones, varying only in so far as to suit changing text. We get the same kind of variation when new words are sung to an old familiar tune. (In primitive music the style is very narrow in range; all the songs seem to have the same melody and to differ only in words.)

Language and melody have a reciprocal influence on one another. But melody preponderates; language is confined by it, and has to yield. This is an experience which we are taught by our folk-songs: short syllables may be lengthened; unaccented ones stressed; slight ones suppressed; expletive syllables inserted; words and even fragments of words repeated. But even the intonation of language changes in singing. In fact, even apart from singing, speech is melodic in so far as it rises and falls in pitch. The pitch, and consequently the melody of the speech, rest on the vowels and sonants. But the individual sounds of speech are not stable in pitch; they glide, and

1 Examples from Togo in J. Schönhärl, Volkskundliches aus Togo, Dresden und Leipzig (Koch), 1909.

this is what makes speech different from singing. In singing every sound is fixed at a certain level, and so the size of the intervals becomes less undecided. Still the outline of the melody of speech, its rise and fall, might be retained (e.g. the rising tone at the end of an interrogative sentence in German). It is, in fact, frequently retained, but not always; chanted melody has to be more regular than speech for the sake of form. So a second phrase must, in its melody, adapt itself, as we have pointed out, to the first one which has fixed the flow of the chanted tune; or musical form may demand a descent to the tonic at the end of a section and thereby distort the natural melody of the spoken phrase (e.g. of a question).

It is particularly interesting to study the mutual influence between speech and singing in connexion with those languages in which changes of pitch and their direction determine the meaning of the words; of this the Sudan languages are an outstanding example. In this case it would primarily be expected that the tune was bound much closer to the words, considering that any change in the inflection of the voice alters or altogether destroys their meaning. In pursuit of this question, a number of Ewe songs were (in 1902-3) studied linguistically by Professor C. Meinhof and Professor D. Westermann, and musically by the late Dr. O. Abraham and the present writer. The following example is taken from this (unpublished) joint investigation. It is a woman's song deriding the town of Lome: 'Shall I marry a Lome man? [It is said that] they feed their wives largely on cassava flour [of inferior quality] and on [unripe] coco-nuts.' The songs were sung and spoken into the phonograph by an intelligent Ewe man, and both versions were turned into notation from the phonograms. The notes on the second staff do not, indeed, represent musical sounds like those on the first staff; they only indicate the approximate melodic course of the speaking voice. Even a spoken syllable, however, appears to have a distinct tone-quality when it is singled out and repeated on the phonograph; it is then possible to determine its pitch like that of a chanted note. In the transcription of the text the pitches of speech have been rendered by accents: 'for high-pitched tone,' for 1 C. Stumpf, Zeitschr. f. Psychol., 94, 1, 1923.

2 Translation according to Meinhof.