ISOLATION OF AN ION MILLIKAN. 253 



ter of fact is not appreciably affected thereby, but solely because of 

 the experimental uncertainties involved in work upon either exceed- 

 ingly slow or exceedingly fast drops. When the velocities are very 

 small residual convection currents and Brownian movements intro- 

 duce errors, and when they are very large the time determination 

 becomes unreliable, so that it is scarcely legitimate to include such 

 observations in the final mean. However, for the sake of showing 

 how completely formula (9) fits our experimental results throughout 

 the whole range of the observations of Table XIII, figure 3 has been 

 introduced. The smooth curve in this figure is computed from (7) 

 under the assumption of e=4.89lXlO~ 10 and the experimentally de- 

 termined values of e x are plotted about this curve, every observation 

 contained in Table XIII being shown in the figure. 



The probable error in the final mean value 4.891 X 10 -10 , computed 

 by least squares from the numbers in the last column, is four 

 hundredths of 1 per cent. If there is an error of as much as 3 per 

 cent in the determination of A the final value of e would be affected 

 thereby by only about 0.2 per cent. Since, however, the coefficient of 

 viscosity of air is involved in the formula, the accuracy with which e 

 is known is limited by that which has been attained in the measure- 

 ment of this constant. There is no other factor involved in this 

 work which has not been measured with an accuracy at least as 

 great as 0.2 per cent. 



The value of /* 15 which has been used in the computation of all of 

 the preceding tables, viz, 0.00017856, is in my judgment the most 

 probable value which can be obtained from a study of all of the large 

 mass of data which has been accumulated within the past 40 years' 

 upon this constant. It represents not only the result of what seems 

 to me to be the most reliable single determination of )x which has 

 thus far been made, viz, that of Stokes and Tomlinson 1 who deduced 

 it from the damping of oscillating cylinders and spheres, but it is ex- 

 actly the mean of the three most recent and very concordant values 

 obtained by the outflow method (Table XV), and it is furthermore 

 the mean of all of the most reliable determinations which have ever 

 been made. These determinations are as follows: 



[The discussion of the determinations of the coefficient of viscosity 

 of air is here omitted.] 



We have devised two modifications of this method of determining 

 e which do not involve the value p.. It is scarcely likely, however, 

 that the necessary experimental error in these methods can be re- 

 duced below the error in/* It is probable, therefore, that any in- 

 creased accuracy in our knowledge of e is to be looked for in in- 

 creased accuracy in the determination of }jl. 



1 Stokes, Math, and Ph5's. Papers, v. 5, p. 181. 



