242 MESSRS. W. R BOUSFIELD AND W. ERIC BOUSFIELI) 



as to be easily reproducible in any well-equipped physical laboratory. The present 

 mean calorie from C. to 100 C., which is sometimes used, cannot be accurately 

 determined without extraordinary precautions. If the mean calorie from 15 C. to 

 55 C. were taken as the standard practical unit, it would have the following practical 

 advantages : 



1. The Imvrr limit of 15" C. would enable the unit to be .readily realised with 

 convt-iii'Micf at any time of the year. 



2. The upper limit of 55 C. is so low that no trouble would arise from dissolved 

 air, and the vapour-pressure of the water would be so small that effective precautions 

 against error on this score could be easily taken. 



3. The range of 40 C. would embrace the flat portion of the curve, and allow the 

 temperature interval to be easily ascertained with mercury thermometers with an 

 error not greater than 0'01 C., or 1 part in 4000. 



4. The value of the mean calorie from 13 C. to 55 C. is nearly the same as the 

 value of the 1 5 calorie. When the data are ascertained with sufficient accuracy, it 

 would be easy to adjust the interval so as to make the mean calorie over the interval 

 exactly equal to the 15 calorie. 



Received January 30, 1911. Additional Note on the Thermoid Effect. 



Further investigation appears to indicate that the thermoid effect is not dependent 

 upon the conductor being made of an alloy, but that it occurs with pure metals. 

 Where the temperature coefficient of a resistance is positive, so is the coefficient of 

 the thermoid effect ; and where negative, negative. 



Sir J. LARMOK suggested to us that the effect might be due to strains in the wire 

 caused by the steep temperature gradient from the centre line of the wire to its outer 

 surface. To test this we have taken three wires of pure platinum of different 

 diameters and measured the coefficients. Platinum was chosen because the hysteresis 

 effect with platinum is very small, and one has therefore to deal with a simpler case 

 than manganin. For currents not so heavy as to strain the wire unduly, we have 

 found that, if R is the resistance in ohms of the wire at a given temperature, and E 

 the resistance of the wire at the same temperature whilst a current of C amperes is 

 passing through the wire, we have 



H = Ra(l+aC 2 ). 



Further, if we take as the unit of current density that which exists when a current 

 e ampere is passing in a wire of one square millimetre section, then, if r = radius 

 of wire in centimetres and D = current density, 



D = C/lOO^r 8 , 

 and we have 



R = Ro(l+/3D 2 ). 



