34 ART. 4. — K. HOXDA AND T. TERAüA. 



The mean curves corresponding to these numbers are given 

 in Figs. 19, 20, 21, 22, 23, 24 and 25. They are drawn in a 

 very exaggerated scale. They show distinct hysteresis ; in three 

 of them, the ascending branch lies below the descending one, 

 while in the others, the case is reversed. In commercial nickel, 

 the increase of elasticity is the greatest ; in tungsten steel, it is 

 the least. In 50.72 % Ni. and 70.82 "/o Ni., the elasticity seems 

 to attain a vague maximum at a high tension. 



In the calculation of — yr- for different tensions, the values 

 of E for corresponding tensions were employed. This was abso- 

 lutely necessary, since for different tensions, E changed more 

 than 10 o/o in many of the sjiecimens. 



(t) Comparison of the results with those obtained 

 by the flexure method. 



In a previous paper by jMes.-iis. S. Shimizu, S. Kusakabe and 

 one of us, the change of elasticity of magnetic bars, as determined 

 by the flexure method is given. It is highly interesting to com- 

 pare the results with those obtained l)y our elongation method. 



In the flexure method, we observed a slight change of flexure 

 by magnetization when there was no suspended weight ; this per- 

 haps arises from the initial bending of the specimen due to its 

 own weight. In subsequent experiments of the kind, it was 

 thought preferable to subtract this initial change of flexure from 

 that of flexure due to ,a suspended load. We have therefore 

 recalculated tiie former results, as shown in Table XXIV and in 

 Figs. 2B, 27 and 28. 



