Now, the single low value of which these authors find in the oil-drop work is obtained by computing from some twenty-five observations on the times of fall, and an equal number on the times of rise, of a particle which, before we had made any computations at all, we reported upon[117] for the sake of showing that the Brownian movements would produce just such fluctuations as Ehrenhaft had observed when the conditions were those under which he worked. When I compute by equation (29), using merely the twenty-five times of fall, I find the value of comes out 26 per cent low, just as Zerner finds it to do. If, however, I omit the first reading it comes out but 11 per cent low. In other words, the omission of one single reading changes the result by 15 per cent. Furthermore, Fletcher[118] has shown that these same data, though treated entirely legitimately, but with a slightly different grouping than that used by Zerner, can be made to yield exactly the right value of . This brings out clearly the futility of attempting to test a statistical theorem by so few observations as twenty-five, which is nevertheless more than Ehrenhaft usually uses on his drops. Furthermore, I shall presently show that unless one observes under carefully chosen conditions, his own errors of observation and the slow evaporation of the drop tend [Pg 169] to make obtained from equation (29) come out too low, and these errors may easily be enough to vitiate the result entirely. There is, then, not the slightest indication in any work which we have thus jar done on oil drops that comes out too small.