4 PROCEEDINGS OF THE AMERICAN ACADEMY. 



Richards and Lamb,* and the present paper is written to elaborate the 

 argument there presented, as well as to suggest new methods of over- 

 coming the difficulty. 



An analysis of the sources of uncertainty due to this cooling effect in a 

 reaction evolving heat easily reveals two chief factors : first, the deficiency 

 of the actual maximum temperature in the calorimeter due to loss of heat, 

 and, secondly, the inability of the thermometer to follow with adequate 

 speed the actual temperature. The first of these causes of error receives 

 a more or less satisfactory elimination by the methods of Regnault and 

 Rumford ; the second seems never to have been adequately treated. 



First, the former question may be briefly discussed. The method of 

 Regnault as formulated by Pfaundler depends upon Newton's law of 

 cooling, which is known to hold with reasonable accuracy for small tem- 

 perature intervals. | It must be noted, however, that this application 

 of Newton's law is not wholly free from assumption ; for the law has been 

 tested only during long-continued cooling, not when the temperature is 

 subject to rapid change. In this latter case it seemed quite possible that 

 the convection currents, to which the cooling is partly due, are not at 

 once started in full activity ; and hence that the rate of cooling after a 

 sudden change of temperature is not identical with that during a slow 

 change. 



Rumford's method of correction is subject to a somewhat similar pos- 

 sible uncertainty of unknown magnitude, and this method (of starting a 

 reaction as much below the temperature of the room as the reaction 

 finishes above the temperature of the room) is moreover only applicable 

 to reactions which progress evenly throughout their course. The consid- 

 eration of these circumstances led to a wish to devise a method free from 

 the objections to which these corrections for cooling are open ; and this 

 wish was intensified by the consideration of the second of the causes of 

 error already alluded to, — namely, the lag of the thermometer. 



Although this latter cause of error is not usually considered in the 

 treatment of calorimetric work, it necessarily enters to a greater or less 

 extent into all results in which the temperature is read on a moving 

 thread. 



The rate of cooling of the mercury in the thermometer, and therefore 

 the lag of its reading, varies very greatly with the construction, form, 

 and size of the thermometer, and also with the specific heat, conductivity, 

 and agitation of the surrounding liquid. Therefore it must be determined 

 for each thermometer under the exact conditions employed in each case. 



* These Proceedings, 40, 057 (1905). t Pogg. Ann., 129, 113 (1866). 



