We find also attū-ni, attū-cunu; and the other persons might be restored, attū-ca, atṭū-ci, attu-su, etc. See p. 15.

THE NUMERALS.

These have two forms, masculine and feminine, as in the other Semitic tongues, and show the same peculiarity of using the masculine of the numerals from 3 to 10 with feminine nouns and the feminine of the numerals with masculine nouns. Originally the numerals seem to have been abstract substantives, like Tpiás, and could take either a masculine or a feminine form. The feminine was most commonly employed, and so became associated with nouns of the predominant masculine gender. In Ethiopic (and vulgar Arabic) the feminine is almost exclusively used.

The forms of the Semitic numerals early became fossilised, and hence are almost identical in the various dialects. Notwithstanding this, the Assyrian cardinal-numbers are more closely connected with the Hebrew than with those of the cognate languages. Estin "one" is found in the Hebrew 'ny; there are no traces of the Æthiopic cal'a “two”; and the numeral for "six," like Hebrew, omits the dental, which appears in Arabic and Æthiopic, while the Aramaic consonantal changes in, etc., find no place in Assyrian.

20. esra'a (Hebrew □); 30. selasa'a; 40. irbahā; 50. khansaʻa ; 60. sisa'a, sussu; 70. sibba'a; 80. [samna'a]; 90. [tissa'a]; 100. mih (Hebrew ); 1000. alapu ().

The words in brackets have not yet been found in the inscriptions. Generally the cardinals are denoted by symbols; "one" is an upright wedge, "two" two wedges, and so on. Ten" is expressed by <; 11 by <1; 20 by <<, and so on.

The masculine numeral estin is important as throwing light upon the Hebrew in 11, which does not appear in the cognate languages. Besides akhadu, a theme khad seems to exist, which shows itself in the adverb edis "only," edis-su "by himself." We also have instances in which the Accadian id "one" is used, apparently with the value of khad or ed, as both masculine and feminine, singular and plural. Now kh and e are interchangeable (see pp. 28, 29) in Assyrian, especially in the case of foreign words, and the Semite often tried to represent the rough Turanian vocalisation at the beginning of an Accadian vocable by the guttural

1 Sh in Ethiopic.

kh (as in Idiklat, p). Id or kat in Accadian-meant "hand" primarily, so that we are taken back to the time when the savage signified

"" one by holding up his hand. As in Hebrew and Arabic, irba' interchanges with reba'. The form khamisserit shows that the Assyrian could contract its numerals like vulgar Arabic, or New-Syriac.2

The origin of the Semitic cardinal numbers is a matter of some difficulty, Ewald and others, struck by the superficial resemblance of one or two, shesh, sheba', etc., to the corresponding Aryan numerals, have imagined common roots. But this proceeds upon the assumption of the common parentage of the two families of speech; and even were this granted, we should have no Grimm's Law upon which to base our comparisons. Moreover, there are several numerals which are confessedly unlike in the two classes of languages; and the resemblances in the case of those which are most like are not greater than between shesh and the Basque sei, or irba and the Mongol durban. Nothing, again, is more usual among savage tribes than to adopt different roots at different times to express the same numerals. Thus in English we have "first," second,” “ace,” 66 'tray"; and among the Semitic languages themselves, the only trace which Æthiopic presents of the ordinary numeral for "two" is in the words sanuy and sānet, while it has taken another root, cal'a "to divide," to express the idea of duality. The same holds good of estin and 'ashte. The whole theory, however, has been disposed of by an analysis of the Aryan numerals, which

3

1 See my paper on Accadian in the Journal of Philology, vol. iii., No. 5 (1870), p. 39.

2 Nöldeke, Neusyr. Gramm., p. 152.

3 Ct. Tylor, Primitive Culture, vol. i., pp. 231, 233.

demonstrates that the original forms of the numbers were widely different from those required to bring them into relationship with the Semitic. Thus "six" (which a similar analysis applied to the Semitic languages shows was primarily shadash in them) had originally a guttural at the beginning of the word, now preserved only in the Zend kshwas. Professor Goldstücker, taking this word as the starting-point of his investigations, has obtained the following results from an analysis of the numerals. "One" is the demonstrative pronoun "he"; 2 is " diversity" (Stá, dis-, zer-); 3="that which goes beyond" (root tar, whence trans, through, etc.); 4" and three,” i.e. “1+3” (cha-tur); 5="coming after" (pan-chan, quinque); 6="four," i.e. "(2) and 4" (kshwas for ktwar); 7="following" (saptan, eπw, etc.); 8 "two fours "(dual ashṭau, ỏÊтw, with prosthetic ă, 0); 9="that which comes after" (same root as navas, novus); 10="28" (da-san, de-cem).

These results are in full accordance with the facts presented by the Turanian and Allophylian languages generally, and, in short, by all those modern savage dialects which still bear on their surface, unobscured by decay, the primitive machinery of language and calculation. Analogy would lead us to infer that the Semitic tongues formed no exception to this mode of forming numbers, which, so far as it can be analysed, is found to be universal. Calculation is an art slowly acquired; many modern savages cannot count beyond "two" or "three," and we find that this was the case with the ancestors of the highly-gifted Aryan race itself. Once acquired, however, calculation is continually needed: no words are more used than those which denote the numerals ;

and consequently no words are more liable to be contracted, changed, and, in short, to undergo all the phenomena of phonetic decay. If we apply this test to the Semitic tongues, we shall find that they fully submit to it. Not to speak of instances like khamisserit, or vulgar Arabic sette "six," a more pertinent example would be shesh for shadash. The Aramaic tĕrên shows how an often-repeated word could change its primitive form, and the Ethiopic cal'a and Assyrian estin remind us of the possibility of co-existing roots. Then another element has to be taken into consideration. We have seen how many words, not to speak of an alphabet, the Semites could borrow from their Turanian neighbours, more especially words like sabar "copper" which signified objects communicated by the civilized Accadian to the rude Bedouin tribes. Now the Accadians had attained a high degree of knowledge of arithmetic and astrology; the great libraries of Huru and Senkereh, formed in the sixteenth century BC., contained tablets of square and cube roots, a developed sexagesimal system, observations of eclipses, and a symbolic numeration. We may therefore expect to find among Semitic loan-words Turanian numerals. Comparative instances among other nations warrant, I think, the following analysis of the Semitic numerals.

Akhadu, found in Assyrian in akhadi—akhadi “ the one— the other," has already in historic times undergone contraction in the feminine ikhitu, akhat for ikhidtu. The stronger masculine a has been weakened into the feminine -i, and this has affected both vowels, according to the vowelharmony of all savage people. Now by the side of akhad we have Aramaic and Targumic khad, and Assyrian ed(u) and