36 EEPOUT— 1901, 



five or six results at points between 66° and 92°, -NV'liich ar6 represented 

 within one part in 10,000 by the linear formula 



s=l + -00014 (<- 60) . . . . (4) 



This formula gives a value nearly 1 in 1,000 lower than (3) at 90°, 

 but it cannot be reconciled with Regnault's observations between 110° 

 and 190° C, and it would therefore probably be better to retain (3), 

 since it is likely that the specific heat would increase more rapidly at 

 high temperatures. 



Although the actual observations at these higher jjoints agree with 

 formula (4) much more closely than 1 in 1,000 it is conceivable that they 

 might contain a constant error of this order at 90^. 



More complicated formulse are given by Dr. Barnes,^ but since the 

 whole variation of the specific heat is so small it does not seem worth 

 while to change the simpler formulje already published in the ' British 

 Association Beport,' 1899, which represent the observations equally well. 



Comparison tvith Liidin. 



» 5 » 



The results of the observations of Liidin by the method of mixtures 

 are given in Table II. for comparison. They agree very well below 20°, but 

 show a minimum at 25° C. Above this point they increase rapidly to a 

 maximum at 8.5° C, which is 1 per cent, greater than the value found by 

 Barnes when expressed in terms of the same unit. This rapid increase 

 may possibly be explained by radiation error from the hot-water supply. 

 The subsequent diminution between 85° and 100° may be due to 

 evaporation of the boiling water on its way to the calorimeter. These 

 errors ai-e peculiar to the method of mixtures, and are completely 

 eliminated in the electrical method. Moreover, the quantity measured 

 in the method of mixtures is not the actual specific heat at the higher 

 limit t, but the mean specific heat between t and the temperature of the 

 calorimeter. Tlie values of the actual specific heat at t, which depend on 

 difierentiating the curve of mean specific heat, are thus rendered 

 extremely uncertain near the extremities of the range. The electrical 

 method avoids this uncertainty, since it directly measures the rise of 

 temperature produced by the same quantity of energy at difierent points 

 of the scale. 



Correction for Variation of Temperature Gradient in the Flow-tube. 



If E is the difference of electric potential in volts between the ends 



of the conductor ; 

 C, the current in amperes through it ; 

 J, the number of joules required to raise 1 gramme of water 1° C. 



at the mean temperature of the experiment \ 

 Q, the water-flow in grammes per second ; 

 / 6, the rise of temperature ; 



/t9, the loss of heat by radiation, tfec, in joules per second, 

 we have the simple equation 



EC==JQ9-H/i^ .... (5) 



If we assume that the heat-loss hQ is the same for two difierent flows, 

 provided that the electrical current is regulated so as to secure the same 



' Froc. R S, 1900. 



