as a Source of Small Constant Currents. 367 



the Rayleigh volt and the Congress volt, as we may very 

 well do. 



It is now left for us to determine what error a change of 

 temperature of 8° on either side of 15° C. is likely to cause. 

 In this country, New South Wales, we have to work pretty 

 often at temperatures as high as 23° C, and we imagine 

 that no one would work below 7° C. if it could be avoided ; 

 consequently we take these as our limits. 



We have already found that perturbations of the P.D. of 

 the cell itself may introduce an error in that quantity to the 

 extent say of 02 per cent, (the greatest we got by combining 

 the most aberrant observations of different days was 0*14 

 per cent.). 



On the strength of the experiment above 30° C. and the 

 experience with the small cells from 14° to 23° C, we assume 

 that no change in A will occur otherwise than that dependent 

 on the ordinary coefficient. This is not strictly legitimate 

 for lower temperatures, but it is difficult to obtain them much 

 lower here. 



We know that we shall have two causes at work tending to 

 change our current value ; these will be in the same direction. 



The E.M.F. will tend to diminish with increased tempera- 

 ture, and the resistance of the wire will increase. 



The cell-resistance change is of course included in varia- 

 tions of A, and these we have shown to be negligible for 

 practical purposes when the external resistance is large. 



By combining the effects of change of resistance and of 

 E.M.F. (using Lord Rayleigh's coefficient) we get a tempera- 

 ture-coefficient for the current yielded. 



The temperature change of resistance of platinoid wire is 

 about O022 per cent, per degree, and the change of E.M.F. 

 when the cell rises or falls 8 degrees from 15° C. is about 

 O0088 volt. Consequently the error arising from this 

 cause is about 06 per cent. Adding all the possible causes of 

 error together we find that the uncertainty may arise to 096 

 per cent., say 1 per cent., in the value of the current. It need 

 hardly be added that of these causes of error the two latter are 

 easily allowed for where accuracy is required, and the former 

 is very small. We do not think that in such a case an error 

 of more than 0*1 per cent, ought to be expected. 



For ordinary purposes for which the current may be used 

 an accuracy to within 1 per cent, is probably all that will be 

 required. 



We have to thank Lord Rayleigh and Prof. Ayrton for 

 some valuable suggestions as to the mode of statement of the 

 above results. 



