556 PROCEEDINGS OF THE AMERICAN ACADEMY. 



this for a massive steel magnet, though several observers have obtained 

 extremely low coefficients for very slender steel wires, and even negative 

 coefficients for comparatively weak magnets made of some alloys. 



Using such chilled magnets as I have described, and employing com- 

 posite galvanometer coils of manganine and copper, with permanent 

 manganine shunts, it is not difficult to make a cheap fixed amperemeter, 

 the indications of which shall be almost wholly independent of the room 

 temperature. In the case of a d'Arsonval galvanometer of the usual 

 form, slight temperature changes in the torsional rigidity of the suspen- 

 sion wire have to be taken into account. 



Jefferson Physical Laboratory, 

 December, 1902. 



