162 BRIDGMAN. 



Consider the upper curve in Figure 8. A needle was stuck rhrougli 

 the paper at the point .1 and another at B. The spring was passed 

 under A and over B, and under a third needle C which was moved 

 up and down the axis of ordinates at the same time that slight trial 

 readjustments were made in the positions of A and B until that com- 

 bination was found at which the spring coincided most closely with 

 the observed points, particularly at the high frequencies. It will 

 be seen that under these conditions the spring crosses the axis in a 

 straight line, which is one of the demands. The spring also passes 

 through ^ as a straight line, which it theoretically should not do. 

 Any error due to this cannot have a perceptible effect on the extra- 

 polated value. A number of times I made independent extrapolations 

 by the free hand method and by the use of the spring, with essentially 

 the same results. 



Additional evidence of the correctness of the extrapolations is the 

 fact that these were all made independently. The extrapolations 

 were first made, and then afterwards the factors were computed by 

 which the extrapolated differences in cm. of bridge wire were reduced 

 to fractional parts of resistance, ^^hen the extrapolations were made 

 I had no knowledge of whether the results would conform to the 

 results with other specimens or not. 



Successful results were obtained with gold S X 10~* and l.()7 X 10"^ 

 cm. thick, and with silver 2 X 10"^ thick. 



The collected results for the eleven samples of 8 X 10"" gold (that 

 is, extrapolated difference between A.C. and D.C. settings as frac- 

 tional parts of the initial resistance plotted against current density) 

 are shown in Figure 10. The breadth of these samples varied from 

 0.066 to 0.226 mm., and both gold leaf and spattered gold are included 

 in the results. If the effect is genuine, there should of course be no 

 relation with the breadth. We have seen that there is a striking 

 dependence of heating effects on the breadth, but it is evident that 

 there is no such correlation here, and that within the limits of error 

 points for all samples lie on the same curve. This, I lielieve, consti- 

 tutes rather telling evidence for the genuineness of the effect. With 

 regard to the magnitude of the experimental error, it is true that the 

 points are perhaps more scattered than one could wish, but when we 

 consider that we are dealing with an effect whose exiscence even has 

 not hitherto been established, although much sought for, I do not 

 believe that the scattering is more than one might expect. 



The collected results for gold 1.67 X 10"^ thick are shown in Figure 

 11. This comprises results on four different samples. The results 



