Aristoxenus begins his Harmonics by observing that there are two movements of the voice, not properly discriminated by any previous writer; namely, the continuous, which is the movement characteristic of speaking, and the discrete or that which proceeds by intervals, the movement of singing. In the latter the voice remains for a certain time on one note, and then passes by a definite interval to another. In the former it is continually gliding by imperceptible degrees from higher to lower or the reverse [53]. In this kind of movement the rise and fall of the voice is marked by the accents (prosôdiai), which accordingly form the melody, as it may be called, of spoken utterance [54]. Later writers state the distinction in much the same language. Nicomachus tells us that the two movements were first discriminated by the Pythagoreans. He dwells especially on the ease with which we pass from one to the other. If the notes and intervals of the speaking voice are allowed to be separate and distinct, the form of utterance becomes singing [55]. Similarly Aristoxenus says that we do not rest upon a note, unless we are led to do so by the influence of feeling (an mê dia pathos pote eis toiautên kinêsin anankasthômen elthein).
According to the rhetorician Dionysius of Halicarnassus the interval used in the melody of spoken utterance is approximately a Fifth, or three tones and a half (dialektou men oun melos heni metreitai diastêmati tô legomenô dia pente, hôs engista; kai oute epiteinetai pera tôn triôn tonôn kai hêmitoniou epi to oxy oute anietai tou chôriou toutou pleion epi to bary [56]). He gives an interesting example (quoted above on [p. 91]) from the Orestes of Euripides, to show that when words are set to music no account is taken of the accents, or spoken melody. Not merely are the intervals varied (instead of being nearly uniform), but the rise and fall of the notes does not answer to the rise and fall of the syllables in ordinary speech. This statement is rendered the more interesting from the circumstance that the inscription discovered by Mr. Ramsay (supra, [p. 89]), which is about a century later, does exhibit precisely this correspondence. Apparently, then, the melody of the inscription represents a new idea in music,—an attempt to bring it into a more direct connexion with the tones of the speaking voice. The fact of such an attempt being made seems to indicate that the divergence between the two kinds of utterance was becoming more marked than had formerly been the case. It may be compared with the invention of recitative in the beginning of the seventeenth century.
Aristides Quintilianus (p. 7 Meib.) recognises a third or intermediate movement of the voice, viz. that which is employed in the recitation of poetry. It is probable that Aristides is one of the latest writers on the subject, and we may conjecture that in his time the Greek language had in great measure lost the original tonic accents, and with them the quasi-melodious character which they gave to prose utterance.
In the view which these notices suggest the difference between speaking and singing is reduced to one of degree. It is analysed in language such as we might use to express the difference between a monotonous and a varied manner of speaking, or between the sounds of an Aeolian harp and those of a musical instrument.
(3) What has been said of melody in the two spheres of speech and song applies also mutatis mutandis to rhythm. In English the time or quantity of syllables is as little attended to as the pitch. But in Greek the distinction of long and short furnished a prose rhythm which was a serious element in their rhetoric. In the rhythm of music, according to Dionysius, the quantity of syllables could be neglected, just as the accent was neglected in the melody [57]. This, however, does not mean that the natural time of the syllables could be treated with the freedom which we see in a modern composition. The regularity of lyric metres is sufficient to prove that the increase or diminution of natural quantity referred to by Dionysius was kept within narrow limits, the nature of which is to be gathered from the remains of the ancient system of Rhythmic. From these sources we learn with something like certainty that the rhythm of ordinary speech, as determined by the succession of long or short syllables, was the basis not only of metres intended for recitation, such as the hexameter and the iambic trimeter, but also of lyrical rhythm of every kind.
(4) As to the use of the stress accent in Greek prose we are without direct information. In verse it appears as the metrical ictus or arsis of each foot, which answers to what English musicians call the 'strong beat' or accented part of the bar [58]. In the Homeric hexameter the ictus is confined to long syllables, and appears to have some power of lengthening a short or doubtful syllable. In the Attic poetry which was written in direct imitation of colloquial speech, viz. the tragic and comic trimeter, there is no necessary connexion between the ictus and syllabic length: but on the other hand a naturally long syllable which is without the ictus may be rhythmically short. In lyrical versification the ictus does not seem to have any connexion with quantity: and on the whole we may gather that it was not until the Byzantine period of Greek that it came to be recognised as a distinct factor in pronunciation. The chief elements of utterance—pitch, time and stress—were independent in ancient Greek speech, just as they are in music. And the fact that they were independent goes a long way to prove our main contention, viz. that ancient Greek speech had a peculiar quasi-musical character, consequently that the difficulty which modern scholars feel in understanding the ancient statements on such matters as accent and quantity is simply the difficulty of conceiving a form of utterance of which no examples can now be observed.