ending: thus, "life" may be either napsatu or napistu, “fear” may be pulkhatu or pulukhtu. Surd roots do not allow this omission of the -ă, as the final radical must be doubled: thus from sar "king," we may only have śarrătu "queen." A third mode of forming the feminine singular is by -itu, weakened from -ătu; e.g. elinitu "high." According to Dr. Hincks, this form is never used in the case of nomina agentis or with surd roots. The same rules that apply to the omission of the vowel of -ǎtu apply also here, except that surds always have -ătu. Thus we have binitu and bintu " daughter," saplitu and sapiltu "low," makhritu and makhirtu "former," tsikhritu and tsikhir'u "small." Words admit only this form, as elitu "high"; just as from dannu we can only have dannatu. Otherwise both forms are indiscriminately used, e.g. ilitu and ilătu "goddess," belitu and belătu “lady." 1

The addition of the feminine-terminations often causes a change in the last radical. N, d, dh, are regularly assimilated, as in limuttu "injuring" for limuntu, libittu "brickwork" for libintu, cabittu "heavy" for cabidtu. So 8, %, έ, and ts were generally changed to l. Thus we have mikhiltu "fortified," besides mikhitstu and mikhtsatu, marustu and marultu "difficult" (where ts has become s, as in risti for ritsti), lubustu and lubultu "clothing."

In one or two instances the feminine termination seems to have been contracted to a', as in Hebrew, Aramaic, and Arabic. Thus Dr. Hincks quotes the variant sukalula for sukalulat from Assur-nazir-pal.

1 This indiscriminate use of a and i in the feminine noun is analogous to the indifferent employment of sa and si for the feminine relative pronoun.

The origin of the feminine termination would take us back to the personal pronoun. The Assyrian, like Ethiopic, classical Arabic, Phoenician, and Sinaitic, preserves the archaic ǎt (u), which also appears in the Hebrew and the construct state, and in the Aramaic construct and emphatic states. In Berber the third personal pronoun is netta "he," netteth “she,” plural nuthni (masculine), nuthnet (feminine), and the accusative verbal suffix of the third person is -ith, -it, plural -ithen. So the demonstratives are wayyi "this" (masculine), theyyi (feminine), winna "that" (masculine), and thinna, thidhek or idhek (feminine). In Coptic nethof= "he," nethos="she," nethôû="they." The Assyrian enclitic -tu, -ti, which belongs to the pronouns (sunutu, yati, etc.), and is met with again in the Ethiopic wětu, yěti, ěmuntu, and, with the plural-ending affixed, wětomu, wětón, cannot be separated from the feminine abstract suffix -utu, or the ordinary feminine termination -ătu, -itu. These forms, accordingly, will be like iste, an emphatic reduplication of the demonstrative. We have already seen that the primitive Semitic recognized but one root for all the three persons (see p. 41).

The original plural-ending seems to have been -āmū, as found in old Arabic humu, antumū, kataltumū; Æthiopic hōmū, wětōmū, antěmū, nagarcymmū ; Aramaic himmo, himmón; Hebrew, -, etc. Arabic has shortened the final vowel, according to its general rule (e.g. ană "I,” hunnă, kataltů, kataltă by the side of Æthiopic gabarcu, etc.). So has Assyrian, as in sunů by the side of sunutu, khaltsănů by the side of khaltsānūm. Am has been changed to an in Assyrian, Ethiopic, Himyaritic, and Berber (just as the

mimmation becomes nunnation). So, too, in the Syriac anakhnan, hynan, "we." This change takes place in Assyrian even between two vowels, as in khaltsānu, sunu. Am, an, are weakened to im, in, in Hebrew and Aramaic; though the original form seems to be preserved in Hebrew D

'gnats." The Arabic -înă would display the same weakening; unǎ appears to be the result of a false parallelism with the singular case-endings, as though the nunnation were the same as the plural sign, and cannot be compared with the verbal -ūnā (with which compare Syriac nekdh‘lūnā-chon, nekdh'lūnai(hi), etc.). The dropping of the consonant in the Assyrian plurals succi, etc., or in the Hebrew construct, is parallel with the loss of the mimmation, or with the Assyrian verb-forms sacnu, sacna, iscunu, iscuna, for sacnunu, sacnanu, iscununu, iscunanu. The Assyrian dual in a, compared with the plural in -i, seems to have lost a final m,1 which is retained in Hebrew -áim, Aramaic -áin, Arabic -āni and -aini, Syriac en. The original dual was probably -a'amu, expressing by its long-continued reduplication of the pure primary vowel the reduplication of the object. So the Botocudos of Brazil extend ouatou "stream," into ouatou-ou-ou-ou" ocean," with the Chavantes rom-o-wodi "I go a long way," but rom-0-0o-o-wodi "I go an exceedingly long way," in Madagascar ratchi="bad," ra-a-atchi "very bad," and still more analogously among the Aponegicrans 6=itawuna, 7=itawu-u-una (Tylor, "Primitive Culture," vol. i. pp. 196, 197). Similarly, according to Schott, "six" in the Ural-Altaic languages is expressed by a modification of "three." Now a+a=either a or the gunated ai (p. 35). In Hebrew we 1 In Arabic n falls away in the dual before the pronoun-suffixes.

find Dothain becoming Dothan. The plural would have been formed upon the dual, with a contraction of the vowel-sound, as the idea to be expressed by the plural was less definite than that expressed by the dual. The m final, inclosing and strengthening the vowel, is to be compared with the mimmation, or with the accusative and neuter in Aryan nouns. We cannot follow the analogy of these, however, in holding that the plural -m was attached to the case-endings of the singular, or ever had a separate existence pronominal or otherwise. Here, as elsewhere, Semitic and Aryan procedure was contradictory. A double set of case-endings would have been unmeaning. The form in -un must be explained differently, as above. The plural imperfect follows in its vowelendings, not the cases, but the contrasted pronouns sunu and sina (sana). The feminine plural -ātu or -a'atu, Hebrew -6th (for -āwath=-āmath), is formed from the plural -ām, which indifferently denoted both genders, by the addition of the feminine termination, exactly as in the singular. Āt stands for -āmat or -āwat, m and v being interchangeable in Assyrian. (So amaru=, ma=1, etc.)

The forms ebirtān, etc., are of later growth, in which the plural termination has been attached to the feminine, instead of the converse. The same irregular formation appears in the Ethiopic wětōmū, wětón. This is another point in which Assyrian and Æthiopic grammar curiously The Ethiopic forms are even more exactly paralleled by the Assyrian demonstrative plural satunu, satina. For a Samaritan comparison see p. 116.

agree.

The Cases.-These are like the Arabic - nominative, - genitive, -ǎ accusative. Very frequently a final m is

added, lengthening the preceding vowel, similar to the nunnation in Arabic. The mimmation, as Dr. Oppert has happily termed it, becomes rarer in the later Assyrian inscriptions. The case-terminations are attached both to the singular and to the plural, to the masculine and to the feminine. They cause certain alterations in the vowels of many forms; and these are as follows. Whenever a long vowel precedes the last letter, or when the word is a monosyllable (provided it be not derived from a surd root), or when the last vowel, though short, is preceded by more than one consonant (as in sitcun, musascin, niscin), no change takes place. Thus we have 'ummanātu (construct 'ummanāt), mutu “man (construct mut), kitrubu "midst" (construct kitrub). When, however, a root ends in a weak letter, the latter is assimilated to the case-vowel. Thus, from atsi "going-out" (feminine atsitu), we have atsu'u. From agu "crown," Accadian ega, we get agu'u, agi'i or age'e, aga'a. So, again, we find pu'u, pi'i, pa'a.

In surd roots the construct form is a monosyllable. The case-ending, however, doubles the last consonant; e.g. śar, sarru; lib, libbu; 'um, 'ummu. This is really a Palel form of a biliteral; like the Palel triliterals agammu "lake" (agam), cidinnu "ordinance" (cidin), etc.

The vowel of the second radical is always omitted before the case-ending in sacan (but not in sacan), sicin, sucun,1 sicun, and in augmented forms like mustacin, where the second radical stands between two vowels, the latter of which

1 In Babylonian, however, instead of "uzna, the usual dual form, we have 'uzuna-su (W.A.S., I. 51, 1, 1, 4). As it occurs at the end of the line, the retention of й seems due to the pause and the naturally long syllable à.