122 PROCEEDINGS OP THE AMERICAN ACADEMY 



water. The method was that of mixture in an open vessel, where 

 evaporation might interfere very much with the experiment. No 

 reference is made to the thermometer, but it seems not improbable 

 that it was one from Geissler ; in which case the error would be very- 

 great, as the range was large, and reached even up to 70° C. The 

 error of the Geissler would be in the direction of making the specific 

 heat increase more rapidly than it should. The formula he gives for 

 the specific heat of water at the temperature t is 



1 -f- .000302 t. 



Assuming that the thermometer was from Geissler, the formula, 

 reduced to the air thermometer, would become approximately 



1 — .00009 t + .0000015 f. 



Had the thermometer been similar to that of Recknagel, it would 

 have been 1 -f- .000045 t -f .000001 t\ 



It is to be noted that the first formula would actually give a decrease 

 of specific heat at first, and then an increase. 



As all these results vary so very much from each other, we can 

 hardly say that we know anything about the specific heat of water 

 between and 100°, though Regnault's results above that temperature 

 are probably very nearly correct. 



It seems to me probable that my results with the mechanical 

 equivalent apparatus give the variation of the specific heat of water 

 with considerable accuracy ; indeed, far surpassing any results which 

 we can obtain by the method of mixture. It is a curious result of 

 those experiments, that at low temperatures, or up to about 30° C., 

 the specific heat of water is about constant on the mercurial ther- 

 mometer made by Baudin, but decreases to a minimum at about 30° 

 when the reduction is nude to the air thermometer or the absolute scale, 

 or, indeed, the Kew standard. 



As this curious and interesting result depends upon the accurate 

 comparison of the mercurial with the air thermometer, I have spent 

 the greater part of a year in the study of the comparison, but have 

 not been able to find any error, and am now thoroughly convinced of 

 the truth of this decrease of the specific heat. But to make certain, 

 I have instituted the following independent series of investigations 

 on the specific heat of water, using, however, the same thermome- 

 ters. 



The apparatus is shown in Fig. 4. A copper vessel, A, about 20 cra * 



