518 Jlf?' Campbell, Discontinuities in Light Emission. 



to obtain series of 200 (13 minutes) without some sign of such 

 change. Since the calculations gave incidentally a computed 

 value for the drift, all series were rejected in which that value 

 showed any material change in consecutive series. 



Blank experiments were performed in which the fluctuations 

 were observed while no light fell on the cells. For such blank 

 experiments ^y^ had a remarkably constant value, 0'8 scale 

 divisions. The value seemed quite independent of the values of 

 a, h, p, G and R. The fluctuations in such experiments arise 

 probably from two sources. (1) From the fact that the scale 

 was read only to the nearest division, so that, even if the spot 

 was moving perfectly uniformly there would be apparent fluctua- 

 tions. It is easy to show that for fluctuations arising from this 

 cause dj!^ should have the value ^. (2) From chance mechanical 

 disturbances in a building which is by no means completely free 

 from vibration. The independence of the blank fluctuations of 

 the electrical constants of the system seems to show that electrical 

 disturbances were sufficiently excluded. A correction equal to the 

 value of the fluctuations found in the blank experiments was sub- 

 tracted from all other values of the fluctuations. 



It was remarkable that consistent results could be obtained 

 for the value of the mean fluctuation of less than a single scale 

 division : but it must be remembered that the means are never 

 taken for less than 300 observations — sometimes for more than 

 1000. 



§ 7. The first point to be investigated is the variation of the 

 fluctuations with the intensity of the light: this intensity is 

 measured by Nco. Table I gives some of the results. 



The variation of the fluctuations with the intensity of the 

 light is important because it might throw some light upon the 

 two theories, the consideration of which is the chief object of 

 the work. On the " spherical wave " theory, N, the number of 

 light impulses falling upon the cell in unit time, is independent 

 of the distance of the source from the cell : the decrease in 

 illumination with increase of distance is due to a decrease in «o, 

 the number of electrons liberated by each impulse. On the 

 " bundle of energy " theory, it is N which changes with the 

 distance, while co remains constant. Accordingly, if we fcake 

 only N and to into account, equation (9) of the previous paper 

 shows that on the first theory 6^^ should be proportional to co^, 

 i.e. proportional to the square of the intensity of the light, whereas 

 on the second theory ^y^ should be proportional to N, i.e. pro- 

 portional to the intensity simply*. 



* A similar argument for the case of y rays has been given by E. v. Schweidler 

 (Phys, Zeit. xi. 1910, p. 225). The considerations about to be given apply also to 

 that argument. 



