128 PROCEEDINGS OF THE AMERICAN ACADEMY. 



at 32.38°, the preparation of the salt in a perfectly pure state was a 

 serious task. Therefore, instead of always using sodic sulphate of the 

 greatest possible purity, only moderate precautions were used, and the 

 presence of slight impurities was corrected for by observing the tempera- 

 ture of the melting mass on the standardized Baudin thermometer num- 

 ber 9390. A very small amount of impurity did not affect the constancy 

 of the point, but only its absolute position. The zero point of this 

 thermometer was determined immediately afterwards in each case by 

 plunging it into ice, which had been prepared similarly to that for the 

 bath itself ; thus was the actual interval over which the gas had been 

 heated recorded on the thermometer. 



In order to determine the exact value of the temperature interval 

 thus found on the international scale, the thermometer was afterwards 

 standardized by reading it in the purest sodic sulphate many times 

 recrystallized and then in purest ice. The following changes in tempera- 

 ture were observed : 32.580 — 0.103 = 32.477 ; 32.583 — 0. 107 = 32.476 ; 

 32.579 — 0.101 =32.478; average, 32.477. 



Because in these tests exactly as much of the thermometer was ex- 

 posed to air at 20° as in the actual measurement, the reading 32.477 

 signified the transition temperature of pure sodic sulphate, a temperature 

 of 32.383°. In order to obtain the true temperature, 0.094° must be 

 subtracted from the reading of the thermometer. This was done in 

 each case. 



In a few cases the temperature is recorded as 32.384°, or 0.001°, 

 above the maximum temperature given by the purest sodic sulphate. 

 This difference was probably due to a slight error in thermometry ; 

 but since it may have represented the real interval, — the excess being 

 due to a trace of impurity in the ice, — the reading is recorded as found. 



Measurements of Pressure. 



The standard of length, in terms of which pressure was measured, was 

 kindly furnished us in terms of the international meter by the National 

 Bureau of Standards in Washington, by their calibration of our Geneva 

 brass meter-bar with silvered scale, at 20.5° C. The scale was found to 

 be amply accurate enough for our purpose, no single centimeter varying 

 more than 3 microns from the mean, and very few varying more than 

 1 micron. No length of 20 centimeters on the scale varied from the 

 average value by more than 0.01 millimeter. The actual readings, 

 however, were recorded in terms of the dial divisions of the micrometer 

 screw which carried the cross-hair in the eyepiece of the telescope. 



