Calorimetric Thermometers. 339 



the thermometer employed, but the reading-error will be 

 smaller in direct proportion to the delicacy of the instrument. 

 As an estimation-figure with No. 62839 represents about ^ of 

 what it does in the case of the instruments to which these 

 numbers refer, the reading-error with it ought to be only ^, 

 or about *9 cal., giving a total error for each experiment of 

 + (-9 + 7-59 = )8-49. 



Now the average difference of two duplicate experiments 

 with this thermometer was found to be 14*84 cal. (deduced 

 from 38 experiments), giving an error of only +7*42 * cal. 

 for each experiment, a quantity even smaller than that which 

 would have been anticipated, proving conclusively that such 

 an instrument can be worked with just as much certainty as 

 the less delicate ones. The fact that the total error here is 

 even less than the manipulation-error alone in the other ex- 

 periments is due to greater care being taken in manipulation, 

 a larger number of intervals being allowed for the determi- 

 nation of the rate of cooling. It must be remembered, 

 however, that this diminution of the difference between two 

 duplicate experiments by no means represents the total advan- 

 tages accruing from the use of this thermometer, for, the 

 same portion of the instrument being used for experiments 

 with the same substance at all the different temperatures, these 

 will be on the same footing as duplicate experiments them- 

 selves. We shall, therefore, have to make no addition to this 

 error when comparing different experiments in a series, and 

 we thus get rid at once of the additional error of +5'56 cal., 

 which had to be admitted with the other instruments f . 



The extra time required for experiments with such an in- 

 strument is considerable ; the determination itself occupies 

 two or three times as long as with an ordinary thermometer, 

 and the preliminary adjustment of the mercury J and the 



* I am probably very much overstating the error here ; this number is 

 deduced from all the experiments in which this thermometer was used. 

 Several of these were performed before the peculiarities of the instrument 

 were thoroughly known, and contained known sources of error ; eliminating 

 these, the average error of the remaining experiments (24 in number) is 

 only +3-33 cal., a surprisingly small quantity. 



f From what has been mentioned in the Phil. Mag. March 1886 as to 

 the error of standardizing this thermometer, the divergence of any 

 measurement from the absolute value will probably not be more than 

 •00045° 0. for each degree C. measured, or +3-4 cal. 



X In one instrument which I possess a small particle of dust has become 

 fixed in the tube just below the upper chamber, the result of which is 

 that the column of mercury, whenever it has passed a few millimetres 

 beyond this point, breaks off there and shoots up into the upper chamber. 

 If it were possible to introduce such an obstacle at will, the time expended 

 in separating the mercury by heat, and the clanger thereby incurred, would 

 be obviated. I am not without hopes of succeeding in effecting this. 



2B2 



