442 THE SCIENTIFIC PAPERS OF 



eighth, and the steel about one hundred and twentieth of what 

 they respectively lost in the feed-water heater, the loss of ron 

 in this case being about two and half times that of the steel. 

 In Table IX., set 98, the Y steel produced by the Siemens 

 process gave a corrosion of 1,220 grains, and the DD. Yorkshire 

 iron 1,221 grains, in each case per square foot of surface ; the 

 corrosion in those instances being practically the same. In the 

 next set, 99, the Y steel gave a corrosion of 259 grains ; Stafford- 

 shire iron 269 grains, and DD. Yorkshire iron 260 grains ; showing 

 that the steel came out best in that series. In set 102, the Y 

 steel gave 109'5 grains (in rain water), and the DD. Yorkshire 

 iron 144 grains. And yet the author followed up these facts with 

 the conclusion that steel corroded on an average 64'8 per cent, 

 more rapidly than iron. He entirely objected to the mode of 

 reasoning adopted ; he contended that averages were only applic- 

 able to errors of observation. If an observer was not certain of 

 his weighings, and he made a hundred weighings of the same 

 piece of iron, he would be perfectly justified in taking the average. 

 But nothing could be more unscientific or erroneous than 

 averaging several materials, one group of which he chose to call 

 iron, and another which he called steel. It was as if a moral 

 philosopher wanted to find out whether fair complexioned people 

 were more virtuous than dark complexioned people, and were to 

 take six fair people and six dark people promiscuously ; then 

 finding that they were all very well behaved, except one of the fair 

 people, who happened to have just escaped from gaol, and had 

 committed six murders, he were to draw his average and say, 

 " I find that fair people have committed, on an average, one 

 murder each, and should therefore not be trusted." That was the 

 kind of argument which the author appeared to have adopted. 

 There were substances, compounds of iron, manganese, and silicon, 

 sold for steel, which no doubt corroded very rapidly ; but it was 

 fur the consumer not only to select the proper material, but also 

 to see that it was properly used. He believed it was in regard to 

 the proper selection and use of the materials that the enormous 

 discrepancies with which they had to deal would be found. The 

 author stated that there was more cinder in iron than in steel, and 

 therefore that there was prima facie ground for supposing that 

 iron would corrode more than steel. There was, however, an 



