These consisted, first, in finding very exactly what is the coefficient of viscosity of air under conditions in which it may be treated as a homogeneous medium, and second, in finding the limits within which Stokes’s Law may be considered valid. II. THE COEFFICIENT OF VISCOSITY OF AIR The experiments on the coefficient of viscosity of air were carried out in the Ryerson Laboratory by Dr. Lachen Gilchrist,[46] and Dr. I. M. Rapp.[47] Dr. Gilchrist used a method which was in many respects new and which may fairly be said to be freer from theoretical uncertainties than any method which has ever been used. He estimated that his results should not be in error by more than .1 or .2 of 1 per cent. Dr. Rapp used a form of the familiar capillary-tube method, but under conditions which seemed to adapt it better to an absolute evaluation of for air than capillary-tube arrangements have ordinarily been. [Pg 94] These two men, as the result of measurements which were in progress for more than two years, obtained final means which were in very close agreement with one another as well as with the most careful of preceding determinations. TABLE IX for Air .00018227 Rapp, Capillary-tube method, 1913 (Phys. Rev., II, 363). .00018257 Gilchrist, Constant deflection method, 1913 (Phys. Rev., I, 124). .00018229 Hogg, Damping of oscillating cylinders, 1905 (Proc. Am. Acad., XL, 611) .00018258 Tomlinson, Damping of Swinging Pendulum, 1886 (Phil. Trans., CLXXVII, 767). .00018232 Grindley and Gibson, Flow through pipe, 1908 (Proc. Roy. Soc., LXXX, 114). Mean .00018240 It will be seen from Table IX that every one of the five different methods which have been used for the absolute determination of for air leads to a value that differs by less than one part in one thousand from the following mean value, . It was concluded, therefore, that we could depend upon the value of for the viscosity of air under the conditions of our experiment to at least one part in one thousand. Very recently Dr. E. Harrington[48] has improved still further the apparatus designed by Dr. Gilchrist and the author and has made with it in the Ryerson Laboratory a determination of which is, I think, altogether unique in its reliability and precision. I give to it alone greater weight than [Pg 95] to all the other work of the past fifty years in this field taken together. The final value is and the error can scarcely be more than one part in two thousand. III. LIMITS OF VALIDITY OF STOKES’S LAW