^{page 388 note 1} See note on next page.

For convenience of reference, I give here the table of the verbal voices of the principal Semitic dialects, following the classification of Renan, M. (Mémoires de la Société de Linguistique de Paris, vol. i. p. 97)Google Scholar and taking as type the stem qabal.

Remarks.—Assyrian.—As this dialect does not possess an Aorist-Past proper, I give here as model the infinitives. In the secondary and tertiary voices of the Niphal, the characteristic n disappears in the infinitive and is assimilated to the t in the other tenses. There are few examples of a niphal with the second radical doubled naqabbulu, which recalls to mind the rabbinical niphael. There are also traces of few other forms.

Hebrew.—In this dialect are found traces of a tiphel and shaphel, and also a miphel, this last being, however, very doubtful. The verbs having a vowel as middle radical form regularly a palel. A passive by vocalisation is found for few voices: Mqbil pass, hoqbal, qibbel pass, qubbal, hitqabbel pass, hotqabbal, qÔbel pass, qÔbal.

Aramaean.—Syriac and Chaldee have traces of many other voices formed by inserting u or w, y, m, and r, by prefixing m, ś, and t, and by doubling as a palel, palpel, this last voice is much more used than in Hebrew. In every case these voices have as passive a secondary or t-voice, though there are also few traces of a passive by vocalisation.

Arabic.—There are, besides the following voices, 'iqbanlay (15), 'Hqbanlala(14), 'iqbawbala(12), 'Hqbawala(13), 'Hqballa(9), 'Hqba'lla(11), which are rarely used. The numbers given between brackets are those of the order adopted by the Arab grammarians. All the Arabic voices form regularly a passive by change of vocalisation.

Ethiopian.—By analogy, the following voices have been developed:— 'aqabbala (aphel of pael), 'astaqabbula(pael of istaphel), 'aqa'bala (aphel of pdel), 'astaqa'bala (pôel of istaphel), which are little used; there are traces of few other voices, and of a passive formed by vocalisation.

^{page 389 note 1} Lee's Hebrew Grammar, § 157, et seq. (third edition, 1841).

^{page 389 note 2} Journal of the R.A.S. Vol. XIV. p. 105.

^{page 389 note 3} It may be said that all initial vowels are always accompanied by an aspirate; when two vowels meet, if they do not coalesce, the speaker must separate them with an aspirate. This seems evident in English, where the hiatus is admitted in poetry. The fact that in the Anglo-Saxon alliterated poetry words beginning by a vowel rhyme together, their vowels being generally not the same, shows that what rhymes is not the vowels, but the unwritten aspirate, so in this verse of the Beowulf:

unriht æfnde, óð ðæt ende becwóm

the alliterated words begin by an aspirate ‘unriht, 'æfnde and 'de. In the tongues where the aspirates are nearly altogether lost, all the words of a sentence are united together as forming one word as in French, or the vowels when meeting are fused together and form complicated diphthongs as in Italian.

^{page 390 note 1} In his grammar Prof. Lee, after connecting the two prosthetic letters, Aleph and He, seems to infer that all these verbs have the same origin. It is natural to suppose that, as in other cases, the different forms were distributed and localized to express various and different meanings.

^{page 390 note 2} The words formed by prefixing yod, as yzhar ‘oil’ from zahar ‘to shine,’ are very few, and might be even considered as third persons of verbs used substantively like the French vaurien, fainéant.

^{page 390 note 3} Similarly the French have derived from gauche a verb gauchir.

^{page 390 note 4} In Assyrian there are verbs formed by a prefix placed before adjectives, as a-tābu from tābu ‘good.’

^{page 391 note 1} The long ōis represented by æ when the Cholem is placed over a quiescent vowel.

^{page 391 note 2} It must be remembered that according to the theory of the Semitic grammarians all nouns are derived from verbs, and when verbs do not exist these are invented. Several modem scholars have protested against such an unscientific process. See Lee's Grammar, p. 83; Nominal Origin of the Semitic Verbs, by ProfToy, C. H., Transactions of the American Philological Association, 1877, p. 18Google Scholar.

^{page 391 note 3} With this root is connected nūh ‘to rest, to sit down.’ As we shall see further on, in many tongues the two ideas of ‘to be’ and ‘to stand,’ esse and stare, are confounded in some cases, and in others derived from the same root.

^{page 393 note 1} The doubling of a consonant properly indicates sometimes that the accent is on the preceding Towel (see Proceedings of the Society of Biblical Archeology, Nov. 1882); it is the case now in Italian cedralo and cedratto for eedráto, and also no doubt in Anglo-Saxon recceleas for recceleas; but in Assyrian, as in our modern tongues, there was a tendency to confound length with accent.

^{page 393 note 2} Lexicon, p. 523 (Tregelles' edition).

^{page 393 note 3} I may notice here that the Hebrew pronoun of the 1st p. plur. 'anōky is a weakened form of the singular 'anōky with the old plural suffix nū; the other form 'anū is the word ‘being’ with the possessive suffix of the first person plural, and must be for 'annu; the Assyrian anini seems to prove that there were primitively two ns. As for the second person anta for antak, see my previous paper.

^{page 393 note 4} So in Armenian the prefix z came to indicate the accusative.

^{page 394 note 1} Few examples are sufficient here; the reader may refer to the list given by grammarians, and will find that all the words formed by prefixing t can be explained in the same way.

^{page 395 note 1} Diet. p. 371.

^{page 395 note 2} The objective force always carried by the prefix s explains also why few or no nouns are formed with it. If Assyrian, however, has applied the pronoun formed by it to the subjective third person, it may be due to some foreign influence.

^{page 396 note 1} ummā ana sari belia ama ‘Be it so to the king, my lord.’—Pinches's Babylonian Texts, p. 7. It is possible that the adverb umma ‘thus, so‘ may be connected with this primitive verb.

^{page 397 note 1} It may be, however, that the two prefixes have really the same origin, the idea of locality is easily developed from that of being. We have ‘State,’ a constituted Society, used often as the land or country.

^{page 397 note 2} This root 'amah seems to have been confounded at an early time with em ‘mother,’ and the etymologists derive words sometimes from one, sometimes from the other. In Hebrew 'amah has been localized to mean ‘handmaid,’ and is even by some considered as a derivative of 'em; 'amahwould be ‘a small mother,‘ ‘a secondary wife.’

^{page 398 note 1} It is forms of this kind which made some suppose the existence of an aphal and ataphal in Assyrian, but, though traces are found in the nominal formation, such voices do not seem to have ever been in use among the Assyrians or Babylonians.

^{page 399 note 1} The hypothesis is also supported in regard to sakānu by the fact that this verb generally takes, as vocalisation of the aorist, the vowel u, which is the one especially affected to the secondary voices. We will see further on why these voices took by preference this vowel.

^{page 401 note 1} Actes de la Société Philologique, vol. iv. part 1Google Scholar, La hi fondamentale de la formation trilitaire, Paris, 1874Google Scholar.

^{page 401 note 2} The hard guttural h of the Hebrew is in many cases weakened into h in Assyrian, and sometimes disappears altogether.

^{page 402 note 1} It must be noticed here that M. Ancessi did not hold the same theory as I On the triliteral formation; these examples cannot consequently have been chosen to suit my views.

^{page 402 note 2} To dispel all misunderstanding it is necessary to remind the reader that the infixes are really prefixes placed before the second element of a compound stem, so that the secondary voices are parallel voices formed as to the kal, but with a modified stem; the secondary voice of the aphal (Syriac) 'ettaqbel is really the aphal of the modified stem taqbel, as the aphal 'aqbelis formed from the kal qěbal; the secondary voice of the shaphel (Assyrian) sutaqbulu, is really the shaphel of the modified stem ^*tuqbulu, as the shaphel suqbulu is formed from the kal. The arrangement and classification of the forms is the work of time, that is why so many are wanting.

^{page 403 note 1} I take this opportunity to declare clearly that I do not believe in the existence of a primitive Semitic settled tongue, as some are inclined to do, but that the various Semitic dialects developed slowly, and that the process of development which has formed the main points and given a special character to these languages was still applied in a certain measure at all times. The Semitic tongues have been subject to the same laws of development as the others; as for the Latin tongues grown out of Latin, they must have grown out of a primitive stock, with this difference that the primitive Semitic was not settled as the Latin, and, as in all unsettled states of language, it had no regular formation, all the processes which later on were used being employed no doubt indifferently.

^{page 403 note 2} Pinches's Babylonian Texts, p. 20, 1. 15. There are two or three other examples.

^{page 404 note 1} Egyptian gives us the same development: ūn ‘to be,’ an expletive, n preposition.

^{page 406 note 1} We have here an example of vocalic harmony, the infixes taking e with esiru.

^{page 406 note 2} The great numher of quadriliterals in Arabic beginning with kar might incline one to see in them compounds of a verb kar with nouns more or less abbreviated.

^{page 407 note 1} I say “the supposed verb” because I believe that the I and r are two forms of the originally same formative, these two letters interchange easily.

^{page 407 note 2} The supposed verb must have expressed an idea of motion, and would have formed such prepositions as the Chaldee 'arū, the Arabic and Hebrew I, the Assyrian lu, etc., and it might be detected in a few triliteral roots. It may be noticed here that Assyrian possesses about half a dozen quadriliteral verbs: naplusu ‘to perceive,’ napruku ‘to discontinue,’ which are niphal verbs, and sarbutu ‘to spread,’ a shaphel verb, the others are only used in the niphal, and are formed by the insertion of r or I after the first radical; natalkutu ‘to transgress,’ naparsudu ‘to flee,’ napahuhu ‘to weigh.’ There would be an interesting study to make on the origin of the prepositions, for we see that each, formative has one or more prepositions corresponding in meaning and form.

^{page 408 note 1} There are a few examples where we see the other infixes inserted (i.e. prefixed to the second element of the stem), as m in the Syriac hamśen and 'etsmhar; but as these examples may be contested, I purposely leave them out.

^{page 408 note 2} On the verb-reduplication as means of expressing the completed action, by Williams, Alenzo, in the Transactions of the American Philological Association, 1875, p. 54Google Scholar.

^{page 408 note 3} At the outset it is necessary to make a distinction between the two kinds of doubling, one which is, as will be shown, a short way to express the reduplication of tbe whole word, and the other which merely indicates the accent, or rather stress. This distinction is most essential.

^{page 409 note 1} It is to be noticed that the reduplicated forms of the Akkadian verb are generally translated in Assyrian by this pael.

^{page 409 note 2} In Syriao the palpel often answers to the Hebrew palel. In Hebrew there are a few examples of papel at the time of the prophets (especially in Isaiah).

^{page 409 note 3} The Xllth Arabic form 'iqbawbala gives us an analogous formation, but the second element only of the verbal stem is doubled balbala, no doubt for bal-bala.

^{page 409 note 4} Képulym.

^{page 409 note 5} alfihlu 'la'iammu.

^{page 410 note 1} In Syriac there is even the secondary voice 'ethramrem.

^{page 410 note 2} Often the š of the shaphel is changed into ś, it is the case in Arabic, for s and ś are confounded; in Syriac we find such forms as śarheb, and the secondary voice 'eśangad.

^{page 410 note 3} Prov. xxviii. 22.

^{page 410 note 4} It has been noticed that certain weak verbs form in Hebrew the pael (that is the reduplicative voice) by repeating the last letter, this form is called palel, and that these same verbs in general are fully reduplicated in Syriac, the pael thus becoming a palpel. This seems to come from the fact that Syriac is a more decayed dialect than Hebrew, and had entirely lost conscience of the compound origin of the words, so one of their elements could no more be repeated alone to give the strengthened idea.

^{page 411 note 1} This name, which is the first one adopted by our great Master, Hinoks, has the advantage of expressing by a new term a formation unknown in the other Semitic dialects, and is for this reason used here; the name of presentfuture has, on the contrary, the defect of being that of a different form well known in Arabic, Hebrew, etc.

^{page 411 note 2} It seems that only through foreign influence have the Semites ever been induced to precise the value of their tenses as regards time; that is why so many different methods have been used; in Assyrian reduplication, in Syriac use of auxiliaries, in modern Arabic addition of complementary words or of prefixes.

^{page 411 note 3} This is, I grant, shifting the difficulty, as it might be asked the reason by which the infinitive took a and the participle i.

^{page 411 note 4} The Mutative corresponds generally to a lengthened form of the verb in Akkadian; the use of the reduplicated Mutative is therefore due perhaps to foreign influence, but it cannot be said that its formation is.

^{page 411 note 5} See the paper by Alenzo Williams already quoted on verb-reduplication, where it is demonstrated that all the Aryan strong perfects are due to reduplication of the verbal stem, as lĕgo perfect lēgi for lĕlĕgi; such forms as cucurri, pepuli, etc., are uncontracted perfects. See also Some Irregular Verbs in Anglo-Saxon, by Francis, A.03, in the Transactions of the American Philological Association, 1872, p. 110Google Scholar, and two other papers in the same publication.

^{page 412 note 1} To be well treated this important question would require a whole paper. See Guillard, S., who studied the question as regards Arabic poetry, Journal de la Societe Asiatique, série 7, vol. vii. p. 214Google Scholar; see also Proceedings of the Society of Biblical Archeology, November, 1882.

^{page 412 note 2} This is illustrated in many modern tongues. In Italian, for instance, to pronounce a closed syllable, as in accento, the first sound is uttered, and the mouth is then placed in position for the next sound, so that we have ᾰcěnnto; the final consonant of the closed syllable, though the mouth is in position before the silence or rest, is pronounced only with the next syllable. The study of phonetics in the Cuneiform texts seems to prove that such must have been the Assyrian pronunciation. It may be added that in Italian the doubling of a letter is often, but not always used to express the accent on the preceding vowel. As a curiosity it is interesting to notice that through the character of the pronunciation of the double letters, they have been used in some tongues to express sometimes that the preceding vowel was long and sometimes that it was short (see Guest, E., History of English Rhythms, edited by Skeat, V. W., p. 104, et. seq.)Google Scholar; we notice in early English the same confusion, as in the Semitic tongues, of accent, length, stress, etc.

^{page 412 note 3} This is the term used by M. S. Guillard, though he does not give perhaps to it exactly the same interpretation, but his views seem to practically confirm mine.

^{page 413 note 1} The forms limim and lumnu are also found, but never lamin. This difference of vocalization as distinction of meaning has been noticed already by Dr. Oppert (Grammaire, p. 52); in the Aorist, active verbs taking a, and neuter u; the same verb is found with both according to the sense in which it is taken.

^{page 413 note 2} This is the opinion of most scholars of the old school, Gesenius, Ewald and others.

^{page 414 note 3} See de Rougés Grammar.

^{page 414 note 4} This vocalic harmony, contrary to the forward process of the Altaic tongues, proceeded backward from the final to the inner vowels. This seems to have been so clear to the Egyptian scribe's mind, that he took the habit of expressing the vowel ontside the words; we have a striking example in the Semitic word raiš ‘head,’ which the Egyptian scribe transcribed ršai, though this is no doubt a graphic phenomenon, and the vowels expressed outside must have been pronounced, as in Hebrew, between the two consonants.

^{page 414 note 5} There is something similar in the use of the reflective pronoun in French as It s'est coupé le doigtfor ‘he has cut his finger.’

^{page 414 note 6} The same has happened in English where words altered by vocalic harmony have lost the influencing syllable, as in ‘men,’ plural of ‘man.’

^{page 415 note 1} In Latin the flections were lost at an early date in the vernacular tongue, and already in the classical language the accusative in m had become a mere mimmation; in Italian it is entirely lost; it may be noticed that the article, which takes to a certain extent the place of the flection in the Latin tongues, has been derived from a demonstrative pronoun.

^{page 415 note 2} Oppert, Gramm. Assyrienne, 2e édition, p. 13, et seqGoogle Scholar.

^{page 415 note 3} In the same way in Arabic the primitive flection weakened into a nunnation has disappeared entirely in the modern dialects.

^{page 415 note 4} In the modern European tongues, as French, Italian, etc., the definite article is really a mere head-flection denoting only gender; in some languages the article, as in Danish, appears as end-flection. Though it is inferred by this paragraph that the Semitic flection has grown out of a demonstrative pronoun, as did the articles in the Latin.and Teutonic languages, the proof would require a development which would carry me beyond my present subject. I will oaly notice now that the full form umma, imma, amma, found in some old Assyrian texts, is also preserved in the paragogic or energetic future of the Arabic verbs under the form of anna and anni, the m being weakened into n, as the mimmation is into nunnation.

^{page 415 note 5} See Origin and Primitive Home of the Semites, Journal of the Anthropological Institute, vol. xi. p. 423. In many instances Ethiopian seems to have retained this vocalic harmony to a greater extent than Arabic.

^{page 416 note 1} Especially in Assyrian, the only dialect where the vowela are clearly expressed; in the imperative, which gives the stem without addition, the vocalization is uniform, so we have accordingly the words a, u or i: sabat, piqid, sukun. There are in many cases examples of false concord caused by vocalic attraction, often the vocalization of the penultimate syllable is influenced by the case-ending, as in salelti for salalti, etc., and also cases of forward vocalic attraction, as in eresu for erasu, etc. In Arabic vocalic attraction is evident, to quote only one example, with the pronoun-suffixes. The vocalization in Hebrew has been too much tampered with by grammarians, commentators and text correctors, and Syriac is too much decayed a dialect, and has too much suffered under Greek influence to enable us to detect easily the laws of vocalization, it would require a long and laborious study. I can here only hazard an hypothesis, which appears to me justified at first sight. It may be added that the vocalic harmonism is a process which has, in spite of the Ougro-Alta'ic scholars, been more or less employed in all families of tongues, Teutonic, Latin, and others, and seems inherent to mankind.