KV1I.S OF i.'ASTlNU TKST UAKS K1.AT. 489 streng'tll could be obtained than if the bar was turned the reverse side up. I have found in experimenting" with a larg-e number of bars one-half inch square, one inch square, and one and one-eighth inches diameter, with supports twelve inches apart, that I obtained on an average 30 pounds more strength for a one-half inch square bar, 100 pounds in the one-inch bar, and 150 pounds in the one and one-eighth inch round bar. I wish these figures to be accepted only as an average of many tests of bars of the respective si/cs g'iven, and with which, as a rule, the results have been very erratic. I have found in a one-half-inch square bar as much as 50 pounds difference in testing" the two sides and in the one-inch square and one and one-eighth inch round 1 have found a few bars which showed from ^oo to 400 pounds difference, thereby presenting' proof that casting" flat any form of size of bar admits of errors and jug'g'lery and is wholly wrong'. I would state that in experiment ing* with testing" on the lower and upper sides of test bars, the}-should always be moulded in the same flask, poured from the same ladle and from the same gate. To prove my position on this question, I would lirst rail attention to conditions which can be found by any who arc sufficiently interested to experiment in this line. In Pig;. 109, next pag'e, is shown a side elevation of a bar resting* on pointed supports A B, 12 indies apart, the distance which the author used in his experiments. The point of load is shown at, I >, The position of the bar is the same as when east or lying" in its mould. In exam inę ing such a bar it will be found that the metal at the lower side or shell 1C E is generally denser, or of a closer gram, than that composing' the upper half of the Chapter I,XX,, par,*-*; :;y.| in r,.s.j. and attacked the usual formulae for loaded beams asound burs..........................*fi$7 "