As to the numerals above three there have been very many conjectures. The figure one of the Demotic looks like the one of the Sanskrit, the two (reversed) like that of the Arabic, the four has some resemblance to that in the Nasik caves, the five (reversed) to that on the K[s.]atrapa coins, the nine to that of the Ku[s.]ana inscriptions, and other points of similarity have been imagined. Some have traced resemblance between the Hieratic five and seven and those of the Indian inscriptions. There have not, therefore, been wanting those who asserted an Egyptian origin for these numerals.[105] There has already been mentioned the fact that the Kharo[s.][t.]h[=i] numerals were formerly known as Bactrian, Indo-Bactrian, and Aryan. Cunningham[106] was the first to suggest that these numerals were derived from the alphabet of the Bactrian civilization of Eastern Persia, perhaps a thousand years before our era, and in this he was supported by the scholarly work of Sir E. Clive Bayley,[107] who in turn was followed by Canon Taylor.[108] The resemblance has not proved convincing, however, and Bayley's drawings {31} have been criticized as being affected by his theory. The following is part of the hypothesis:[109] _Numeral_ _Hindu_ _Bactrian_ _Sanskrit_ 4 [Symbol] [Symbol] = ch chatur, Lat. quattuor 5 [Symbol] [Symbol] = p pancha, Gk. [Greek:p/ente] 6 [Symbol] [Symbol] = s [s.]a[s.] 7 [Symbol] [Symbol] = [s.] sapta ( the s and [s.] are interchanged as occasionally in N. W. India) Buehler[110] rejects this hypothesis, stating that in four cases (four, six, seven, and ten) the facts are absolutely against it. While the relation to ancient Bactrian forms has been generally doubted, it is agreed that most of the numerals resemble Br[=a]hm[=i] letters, and we would naturally expect them to be initials.[111] But, knowing the ancient pronunciation of most of the number names,[112] we find this not to be the case. We next fall back upon the hypothesis {32} that they represent the order of letters[113] in the ancient alphabet. From what we know of this order, however, there seems also no basis for this assumption. We have, therefore, to confess that we are not certain that the numerals were alphabetic at all, and if they were alphabetic we have no evidence at present as to the basis of selection. The later forms may possibly have been alphabetical expressions of certain syllables called _ak[s.]aras_, which possessed in Sanskrit fixed numerical values,[114] but this is equally uncertain with the rest. Bayley also thought[115] that some of the forms were Phoenician, as notably the use of a circle for twenty, but the resemblance is in general too remote to be convincing. There is also some slight possibility that Chinese influence is to be seen in certain of the early forms of Hindu numerals.[116] {33}