o'JO Professor J. A. Ewing [May 22, 



in tbe third stage (c), we sball find only a little more residual mag- 

 netism than we should have found by making the exjieriment at the 

 end of stage h. That jiart of the turning of the molecules which 

 goes on in stage a contributes nothing to the residual magnetism. 

 That })art which goes on in stage c contributes little. But that part 

 of the turning which goes on in stage h contributes very njuch. 



In some sjiecimcns of magnetic metal we find a much sharper 

 separation of the three stages than in others. By applying strain 

 in certain ways it is possible to get the stages very clearly separated. 

 Fig. 2, a beautiful instance of that, is taken from a paper by Mr. 



Fig. 2. 



Nagaoka — one of an able band of Jai)anese workers who are bidding 

 fair to repay the debt that Japan owes for its learning to the West. 

 It shows how a piece of nickel which is under the joint action of puU 

 and twist becomes magnetised in a growing magnetic field. There 

 the first stage is exceptionally j^rolouged, and the second stage is 

 extraordinarily abruj^t. 



The bearing of all this on the molecular theory will bo evident 

 when we turn to these models, consisting of an assemblage of little 

 pivoted magnets, which may be taken to represent, no doubt in a very 

 crude way, the molecular structure of a magnetisable metal. I have 

 here some large models, where the pivoted magnets are pieces of sheet 

 steel, some cut into short flat bars, others into diamond shapes with 

 ])ointed ends, others into shapes resembling mushrooms or nmbrellas, 

 and in these the magnetic field is produced by means of a coil of in- 

 sulated wire wound on a large wooden frame below the magnets. 

 Some of these are arranged with the pivots on a gridiron or lazy-tongs 

 of jointed wooden bars, so that we may readily distort them, and vary 

 the distances of the pivots from one another, to imitate some of the 

 ctfects of strain in the actual solid. But to display the experiments 

 U) a hii'gc audience a lantern model will fcrvc best. In this one the 



