28 



PROCEEDINGS OF THE AMERICAN ACADEMY 



numerical values of AiT, A/), A t, &c., deduced from the experimental 

 data of the second series for carbonic acid. The first column gives 

 the tabular number of the measurements from which the data were 

 taken. 



TABLE VIII. A!/ = 0.001!/. 



Probable Effect on y of Instrumental Errors and Errors of Obser- 

 vation on the Values found and used for K, p^, p.^, etc. 



Since the accuracy in the determination of K depends upon that 

 of p^, ]}.„ and jOg, when both tubes are in ice, and thus under identical 

 conditions, the latter quantities will be first considered. 



I. Ap. 



Constant errors in p^, p^, p^, may be due to instrumental error of 

 the barometer, or of the cathetometer which is used in the measure- 

 ment of the mercury columns. Of course the fractional precision of 

 the pressure measurements would be increased by increasing the driving 

 pressure, but the possible deviation from the assumed laws of trans- 

 mission of the gas through the tubes and of the increased liability to 

 leakage with great rarefaction renders the use of a small pressure 

 desirable. The cathetometer error has been reduced by the applica- 

 tion of corrections determined for each decimeter of the scale by Prof. 

 Wm. A. Rogers of Cambridge, and it was found not difficult to read 

 the length of the mercury columns in the gauges with a probable error 

 of less than 0.02 mm. The barometer was No. 835, made by James 

 Green of New York, and had a scale of millimeters with vernier 

 reading to 0.1 mm., and read by estimation as nearly to 0.01 as possible. 

 Its instrumental error after careful refilling was determined by tests 

 with the cathetometer and by indirect comparison with instruments of 

 the U. S. Signal Service, and was probably under 0.3 mm., possibly 

 considerably under that amount. As will be seen flxrther on, a con- 

 stant error of this magnitude would have no sensible effect. 



