Temperature on the Viscosity of Gases. 205 



Since the variation to be studied in the present investigation 

 is one of quite small amount, and is determined as the result 

 of a somewhat complex measurement, any systematic error of 

 even quite small magnitude might mask the quantity sought. 

 It is essential therefore to show how far such errors of sensible 

 magnitude have been avoided or eliminated. A discussion * 

 of the various possible sources of error, of their influence on 

 the final result, and of the probable magnitude of the error in 

 the measurements as actually performed, indicates that it is 

 quite unlikely that known sources could have introduced 

 accumulated errors into that result exceeding 0*5 per cent. 

 even at the highest temperatures, and the means from which 

 the equations are deduced are probably less than that amount 

 in error. 



The cathetometer and barometer used were carefully cor- 

 rected for instrumental errors, and the temperatures are 

 properly corrected and reduced to the air-thermometer by 

 tables of errorsf determined by a comparison with the air- 

 thermometer. For the latter purpose the comparator and 

 air-thermometer employed by Prof. H. A. Rowland for the 

 same purpose in his investigation of the Mechanical Equiva- 

 lent of Heat was used between the limits of 0° and 100° C. 

 The results of the comparison are given in full in his memoir 

 on that subject %. The comparisons above 100° were made by 

 me in his laboratory in another comparator designed by Prof. 

 Rowland, and described in full in my paper just referred to. 



The advantages of the method arise from its simplicity and 

 directness. The only measurements of precision required are 

 the temperature of the second capillary, and of the lengths 

 of several mercury columns. No measurements of volume 

 are involved. The two capillaries through which identically 

 the same mass of gas is successively transpired, are as nearly 

 as possible alike, and under the same conditions except as to 

 temperature. The gas is transpired through each under 

 nearly the same pressure-difference, and at a pressure not far 

 from one atmosphere. The flow of the gas is perfectly uni- 

 form, except for slight accidental fluctuations, throughout the 

 whole of a measurement or set of measurements, tending thus 

 to the elimination of a class of errors possible in such methods 

 as that of oscillating plates, or of transpiration when the 

 driving-pressure falls from a given initial to a less final 

 amount. The disturbing effect of eddies, or other special 



* Holrnan, Proc. Amer. Acad. Arts and Sci. xxi. pp. 25-44 (1885). 

 t Holrnan, Proc. Amer. Acad. Arts and Sci. xxi. pp. 40, 41 (1885). 

 \ Rowland, ibid. xv. 117 (June 1879). 



