OF ARTS AND SCIENCES. 1G9 



temperature as the water jacket, but the wheel was usually slightly 

 above it. At iir.st its temperature was noted by a thermometer, and 

 the loss to it computed separately; but it was found to be unnecessary, 

 and finally the whole was assumed to be a function of the tempera- 

 ture of tht; calorimeter and of the jacket only. 



At first sight it might seem that there was a source of error in 

 having a journal so near the bottom of the calorimeter, and joined to 

 it by a shaft. But if we consider it a moment, we shall see that the 

 error is inappreciable; for even if there was friction enough in the 

 journal to heat it as fast as the calorimeter, it would decrease the 

 radiation only seven per cent, or make an average error in the experi- 

 ment of only 1 in 700. But, in fact, the journal was very perfectly 

 made, and there was no strain on it to produce friction ; besides which, 

 it was connected to a large mass of cast-iron which was attached to 

 the base. Heuce, as a matter of fact, the journal was not appreciably 

 warmer after running than before, although tested by a thermometer. 

 The difference could not have been more than a degree or so at most. 



The warming of the wheel by conduction and of the journal by 

 friction would tend to neutralize each other, as the wheel would be 

 warmer and the journal cooler during the radiation experiment than 

 the friction experiment. 



The usual method of obtaining the coefficient of radiation would be 

 to stop the engine while the calorimeter was hot, and observe the 

 cooling, stirring the water occasionally when the temperature was 

 read. This method I used at first, reading the temperature at inter- 

 vals of about a half to a whole hour. But on thinking the matter 

 over, it became apparent that the coefficient found in this way would 

 be too small, especially at small differences of temperature ; for the 

 layer next to the outside would be cooled lower than the mean tem- 

 perature, and the heat could only get to the outside by conduction 

 through the water or by convection currents. 



Hence I arranged the engine so as to run the paddles very slowly, 

 so as to stir the water constantly, taking account of the number of 

 the revolutions and the torsion, so as to compute the work. As I 

 had foreseen, the results in this case were higher than by the other 

 method. At low temperatures the error of the first method was 

 fifteen per cent ; but at high, it did not amount to more than about 

 three to five per cent, and probably at very high temperatures it would 

 almost vanish. 



I do not consider it necessary to give all the details of the radiation 

 experiments, but will merely remark that, as the calorimeter was 



