i^nicicu^- tit »i.i euui.il cxiicmcK ui two iiasits, 72 ins. or 4.13 per cent., nut the greatest difference in one flask, tests N"os. 15 and 16, and which is the way Table 58 should be shown, is but 1.03 per cent. Tested by Thos. D. West, October 23, 1894, assisted by C. B. Kantner, at Sharpsville, Pa. TABLE 104. Chemical Silicon. 1.4S analysis < Sulphur. .019 >f ]>!£*• iron charged. Chemi Silicon. i-38 cal analys Sulphur. .038 Is of test bars. Mang. •35 Plios. ,097 Mang. •31 Phos. .099 A study of the tests on page 493 shows that the greatest difference in one flask of the strength extremes of the bars cast on end is but 1.03 per cent., compared with 11.59, 11.99 and 20.07 per cent, found in the bars cast flat. It maybe well to mention again the fact that all the bars were poured out of the same ladle and that the flat cast bars were all moulded and. cast together in one flask, giving them a much better chance to be uniform than the bars cast on end, as the latter were cast in separate flasks. When a system is obtained,where with two bars cast together, there will only be three pounds of difference in their breaking loads per square inch, as is found with tests Xos. 13 and 14, Table 103, the author has a suspicion that it is about time some were making a stuch' of the elements bringing about such close results.* The difference of 72 pounds between the two flasks poured on end, shown in Table 103, could be charged to the difference in the fluidity of the metal, which existed through lapse of time in pouring the two moulds, a quality affecting the strength, etc., of test bars more fully defined on pages 37.: and 5.»<>. Additional information on easting test bars on end will be found on pages 50,'*; anil 51.*. *ThrSft\v<» trslri a IV i;Kvil lllrivly to showthr rlfisr ivsults thai, ran be obtaim'd, in a v,rn<-ral practice, inudi better with round bars than with square »»nis.burs..........................*fi$7 "