OF ARTS AND SCIENCES. 97 



The calibration of the first four thermometers has been described. 

 The calibration of the Kew standard was almost perfect, and no cor- 

 rection was thought necessary. The scale divided on the tube was to 

 half-degrees Fahrenheit; but as the 32° and 212° points were not 

 correct, it was in practice used as a thermometer with arbitrary 

 divisions. The interval between the 0° and 100° points, as Welsh 

 found it, was 180°. 12, using barometer at 30 inches, or 180°.0o as cor- 

 rected to 7G0 mm - of mercury.* At the present time it is 179°.68,f 

 showing a change of 1 part in 486 in twenty-five years. This fact 

 shows that the ordinary method of correcting for change of zero is not 

 correct, and that the coefficient of expansion of glass changes with 

 time.t 



I have not been able to find any reference to the kind of glass used 

 in this thermometer. But in a report by Mr. Welsh we find a com- 

 parison, made on March 19, 1852, of some of his thermometers with 

 two other thermometers, — one by Fastre, examined and approved by 

 Regnault, and the other by Troughton and Simms. The thermometer 

 which I used was made a little more than a year after this ; and it is 

 reasonable to suppose that the glass was from the same source as the 

 standards Nos. 4 and 14 there used. We also know that Regnault 

 was consulted as to the methods, and that the apparatus for calibration 

 was obtained under his direction. 



I reproduce the table here with some alterations, the principal one 

 of which is the correction of the Troughton and Simms thermometers, 

 so as to read correctly at 32° and 212°, the calibration being assumed 

 correct, but the divisions arbitrary. 



* Boiling point, Welsh, Aug. 17, 1853, 212°.17; barometer 30 !n . 

 Freezing point, " " '* 32°05. 



Boiling point, Rowland, June 22, 1878, 212°.4(3; barometer 760 mm .. 

 Freezing point, " " " 32°.78. 



The freezing point was taken before the boiling point in either case. 

 t 179°.70, as determined again in January, 1879. 



t The increase shown here is 1 in 80 nearly ! It is evidently connected with 

 the change of zero; for when glass has been heated to 100°, the mean coefficient 

 of expansion between and 100° often changes as much as 1 in 50. Hence it 

 is not strange that it should change 1 in 80 in twenty-five years. I believe tliia 

 fact has been noticed in the case of standards of length. 



vol. xv. (n. 8. VII.) 



