208 



PROCEEDINGS OF THE AMERICAN ACADEMY. 



account, reckoning therefore the distance in any units of length, say 

 millimeters, from the normal curve to the one for m = 300 as our un- 

 known X, we shall find the whole set of equations giving a good average 

 value for x, and thus we may construct what might be called " the 

 normal curve based on in = 300." Now if we use only the curves for 

 250 to 150, so that our next x is the unknown distance from normal 



6 10 15 80 



Figure 6. [Table II.] 



Reversal magnetization curves for a Bessemer soft steel rod of diameter 

 0.6350 cm. 



curve to the curve of 250, we shall again find values for x which satisfy 

 all the equations moderately well. But the normal curve thus deter- 

 mined, which is the normal curve based on ui = 250, will lie slightly 

 to the right of the first one constructed, — at least every case tried gave 

 this result. Similarly, the normal curve based on ill = 200 will lie to 

 the right of the one based on in = 250, and so for the one based on 

 175. For higher values of B than 8000 the formula fails to hold at 



