and in compounds as aπúyovos, (compare the Arcadian κατύ) &c. The adverbs άμυδις and ἄλλυδις seem to be Ionic as well as Lesbian. övvua is vouched for by grammarians, but does not occur. But the compounds ενών μos, vávvμos, &c. are general. This change is the furthest limit of the tendency which we have already seen in the Lesbian to change a to o. It was however probably not so extensive as the change from a to in Boeotian.

(iii) Weakening of v to ..

CH. VII.

This, though not at all an unnatural change, is not (iii) U=. common in Greek. From φυ we have ὑπερφίαλος : compare iπep-pu-ns and the Latin super-bus, where the b corresponds regularly to as the representative of original bh: σίαλος seems to come from συ-ς. φίτυς and φιτύω must certainly be referred to pv. In iupós from ulos we have apparently dissimilation, caused by the v of the suffix: μîoos and μúoos are probably not akin. The Grammarians give ἱπέρ, ἴπαρ, ἴψος, ἔψηλος as Aeolic: the last does occur in Sappho: but there is no more evidence for any of them: and they do not occur even in Boeotian, which had much more tendency to the i-sound. The change can be regarded as only a distant indication of the passage into which became general at a much later period, according to Curtius not before the eleventh century. The rarity of it deserves notice when we estimate the value of modern Greek pronunciation as a guide to the sounds of the older language.

1 Ahrens, 1. 81, &c.

2 Frag. 119.

3 Erlaüt. p. 22.

CH. VII.

Peculiar weaknessof the Latin

vowelsystem.

6. Further Vowel Substitution (Latin).

This has found place in Latin to a much greater extent than in Greek. We have seen above how fully, even down to their most flourishing period of literature, the Greeks had preserved their original store of diphthongs, whilst the Romans at the age of Plautus had retained only one. We have also seen how vividly the distinction remained in the Greek mind of the three different vowel-scales, by the insignificant list of transitions from a to i or u, of which the more important have been given in the preceding section. The same precision must not be looked for in Latin. It has been seen indeed already that the distinction of scales was certainly received by the Italians from their forefathers of the Graeco-Italian age: for some of the traces of modification of the vowels, each in its own scale, have been given above. But that elaborate method could not be maintained in a language which suffered nearly all its diphthongs to degenerate into single sounds. Indeed the most striking characteristic of the Latin language is the exceeding weakness of its vowel system. The vowels have no longer any life in them. They are often the mere servants of the consonants to which they cling, and from which they take their tone: never (as in the Greek) do they expel the consonants by their own fuller life and energy1.

1 In my first edition I quoted here Corssen's connection of the vowel-degradation with the gradual weakening of the Roman character. This connection was objected to by Prof. Mayor, I think with justice; and I have cancelled it. It is doubtless wrong to argue from a portion only of the phonetic system, whatever light the whole may throw upon the character of a nation. I am not quite sure that Prof. Mayor has escaped the same error; for he goes on to give his own interpretation of the fact. He considers the contempt of vowel sound to be a mark of the "strength and energy" of the Romans; I really do not see why; and of their "inartistic nature." Did they then restore their lost vowels as they grew more artistic? they went on corrupting them more

The steps of this vowel-degradation have been arranged by Corssen' in the following table, which is also given by Dr Donaldson":

CH. VII.

From which table we see that while a retains its position
as the primary vowel, never derived from any other, and
while o is only derived from a, the other vowels u, e, and i
are constantly substitutes of stronger sounds, not indeed
indifferently, but in accordance with no law of vowel-
scales. We shall see hereafter that the decision, which
vowel shall be taken, rests generally with the following
consonant. The vowel-change does not originate with
the consonants: it is caused, as has been already insisted
on, by weakness of articulation. But the direction which
that change follows does generally rest with the conso-
nants. This will be clearly seen in the section on Assi-
milation, where I shall describe that vowel-change which
arises from weak articulation, but is modified by the
affinities between particular vowels and consonants. At
present I shall describe such change as is due to simple
weakening, where the effect of neighbouring sounds is
than ever.
Prof. Mayor also cites "the elaborate vowel-system of the
Indians as a mark of the indolent and unpractical life of the dreamy
Oriental." But surely the Indian vowel-system is much less elaborate
than the Greek, and even than the old Latin: they had no ĕ or Ŏ, and by
consequence form fewer diphthongs; their only gains were distinct
symbols for the long original vowels, and also for the so-called vowel-
sound of r (i. e. the glide), all of which sounds the Graeco-Italians doubt-
less had, though not the symbols. The Sanskrit lri existed in the
brains of grammarians and perhaps in one verb.

1 Ausspr. 1. 299, ed. 1. A fuller table exhibiting the results of assimilation as well as substitution is now given by him in his second edition (II. 334) as follows:

CH. VII.

at least not distinctly traceable. As however I have contrasted the vowel-system of the Greek and Latin, between the I may in passing give a few instances where assimilation

Contrast

Greek and the Latin.

is u.

has been at work, in order more fully to shew the dif fering genius of the two languages. They are cases where the Latin has borrowed from the Greek, and has changed the word, after it had become naturalised, to suit its own phonetic laws. They are taken from different parts of Corssen's chapters on "Umlaut1." He has treated the subject so fully as to leave little else to be done but to select examples from his stores. From them will be seen how rigid and lifeless, how dependent on neighbouring sounds, is the vocalism of the Latin, as compared with the Greek. Take the five words, Hecuba2, crapula, catapulta, triobulus, epistula. The penultimate vowel in each But when written in the original language—'Exάßŋ and κραιπάλη, καταπέλτης, and ὄβελος, ἐπιστολή—we see three vowels, a, e, o. The reason is, as will be hereafter shewn in the chapter on Assimilation, that a labial and (more especially) I have a prevailing tendency in Latin to fix the preceding vowel (when weakened by some other cause) at u. Similarly i has an affinity to n, and e tor: μηχανή and βαλανείον become machina and balineum; páλapa and Téσoapa become phalerae and tessera. Indeed, before the suffix -ro- hardly any other vowel but e is found, as in libero, aspero, &c. Contrast with this the varied abundance of the Greek καθαρο-, φοβερο-, άλμυρο, πονηρο-, &c.3 When, for ease of utterance, a vowel is inserted between two following consonants of a borrowed Greek word, the vowel is determined by the following consonant. Thus Ennius wrote Пaтρокλns as Patri-c-oles, because in the older language o shewed the same affinity to las u does afterwards; compare the real Italian Hercoles or Hercolus: but 'Аσλýπios becomes

1 II. 60-333.

2 There was an older mid-form Hecoba: Quint. 1. 4. 16.
3 Corssen, II.' 199.

Aesculapius: δραχμή and 'Αλκμήνη become drachuma and Alcumena in Plautus, because of the labial nasal m. But uva and réxvn become mina and techina', because of the following n.

I proceed to give examples of vowel-weakening, independent in the main of assimilating tendencies, under three principal heads-in formative elements (both formative and inflectional suffixes), in composition, and in reduplication.

CH. VII.

(i) Weakening of

First then, in formative elements, we may see in the nom. sing. o the Graeco-Italian termination of the base formative (itself weakened from Indo-Eur. a) in classical Latin syllables. weakened to u. We have filios Barbati―with the o—on the epitaph of Scipio. This change was complete about the end of the Second Punic war: in the Edict of L. Aimilius (189, B.C.) the u instead of o is regular2. The o was retained always in -ōs (equivalent, as Corssen thinks, to as with vowel intensified), e.g. clamos, arbos, honos: and the vowel sank no further even when the s became r. The neuter nom. ended in os, like the Greek genos, then genus': opos is found on a statue3, the o is still seen in the gen. of many nouns, as corporis (for corpos-is): though others have weakened it to e, as generis for genos-is. This os can also be traced in the case-endings. Thus in the Ep. de Bacch. we find senatu-os (weakened through senatuis to classical senatus), and corpor-us, which speaks of the older corpos-os; compare yéves-os afterwards yévous. Both the Ep. de Bacch. and the Edict of Aimilius shew also u before m in the accusatives. That the u of the genitive plural is weakened from o is shewn by the form duonoro (m) = bonorum on Scipio's tomb: and the tenacity with which the Italian provincials still clung to this, as to other old vowel-sounds, is shewn, e.g. by the "loro" illorum of modern Italian. As Corssen well says*: “the

1 Plaut. Capt. 641.

2 Corssen II. 90. The edict, discovered in 1867, is given in Roby's Grammar, p. 419.

Corssen, II. 87.

4 1. 246 (ed. 1).