The fact seems to be that, as already stated,[260] the Arabs did not immediately adopt the Hindu zero, because it resembled their 5; they used the superscript dot as serving their purposes fairly well; they may, indeed, have carried this to the west and have added it to the [.g]ob[=a]r forms already there, just as they transmitted it to the Persians. Furthermore, the Arab and Hebrew scholars of Northern Africa in the tenth century knew these numerals as Indian forms, for a commentary on the _S[=e]fer Ye[s.][=i]r[=a]h_ by Ab[=u] Sahl ibn Tamim (probably composed at Kairw[=a]n, c. 950) speaks of "the Indian arithmetic known under the name of _[.g]ob[=a]r_ or dust calculation."[261] All this suggests that the Arabs may very {68} likely have known the [.g]ob[=a]r forms before the numerals reached them again in 773.[262] The term "[.g]ob[=a]r numerals" was also used without any reference to the peculiar use of dots.[263] In this connection it is worthy of mention that the Algerians employed two different forms of numerals in manuscripts even of the fourteenth century,[264] and that the Moroccans of to-day employ the European forms instead of the present Arabic. The Indian use of subscript dots to indicate the tens, hundreds, thousands, etc., is established by a passage in the _Kit[=a]b al-Fihrist_[265] (987 A.D.) in which the writer discusses the written language of the people of India. Notwithstanding the importance of this reference for the early history of the numerals, it has not been mentioned by previous writers on this subject. The numeral forms given are those which have usually been called Indian,[266] in opposition to [.g]ob[=a]r. In this document the dots are placed below the characters, instead of being superposed as described above. The significance was the same. In form these [.g]ob[=a]r numerals resemble our own much more closely than the Arab numerals do. They varied more or less, but were substantially as follows: {69} 1[267][Illustration] 2[268][Illustration] 3[269][Illustration] 4[270][Illustration] 5[271][Illustration] 6[271][Illustration] The question of the possible influence of the Egyptian demotic and hieratic ordinal forms has been so often suggested that it seems well to introduce them at this point, for comparison with the [.g]ob[=a]r forms. They would as appropriately be used in connection with the Hindu forms, and the evidence of a relation of the first three with all these systems is apparent. The only further resemblance is in the Demotic 4 and in the 9, so that the statement that the Hindu forms in general came from {70} this source has no foundation. The first four Egyptian cardinal numerals[272] resemble more the modern Arabic. [Illustration: DEMOTIC AND HIERATIC ORDINALS]