Some Experiments on the Latent Heat of Steam. 39 



of Regnault's work. I therefore resolved to try and ascer- 

 tain the cause of the discrepancy we had observed, and 

 with this intention resumed the work last year. 



The results given by the five preliminary experiments 

 quoted in the paper (loc. cit.)* agreed well among them- 

 selves, though the duration of the experiment, the total 

 rise, &c, varied considerably. I thought it desirable,' 

 however, to see if a longer series of such experiments 

 agreed equally well among themselves, and made altogether 

 fifteen similar ones. Though most of these gave almost 

 exactly the same value of L as the ones quoted, I was 

 surprised to find one or two came out considerably higher, 

 giving a value approximating to that of Regnault. 



Considering the various sources of error in an experi- 

 ment of this kind, one of the first things one might expect 

 to be accountable for such differences is the thermometer 

 used. All the experiments, including those to be described, 

 were made with the same thermometer, one of French 

 hard glass, 'by Baudin of Paris (No. 12,771), similar to, 

 and made in the same batch as the one used in the 

 calorimeter by Schuster and Gannon in their determina- 

 tion of the Mechanical Equivalent of Heat (Roy. Soc. 

 Trans., 1895, Vol. 186). This thermometer had been most 

 accurately calibrated and compared by them with a Tonne- 

 lot Standard, whose constants had been determined at the 

 Bureau Internationale des Poids et Mesures at Sevres. 

 The scale of the thermometer I used extended from 

 o° — 13 , Dr. Schuster's from 12 — 25°, and by his kindness 

 I was allowed to obtain by direct comparison the correc- 

 tion of my thermometer at 12 . Thus from determinations 

 of the zero corresponding to this temperature, and a 

 calibration of the stem from o — 12, made with various 



* The highest value was 5259, the lowest 5236, or, excluding one 

 experiment done under abnormal conditions, 524- 1, the mean in that case 

 being 524- 8. 



