390 



Dr. Norman Campbell : Experiments on 



the oscillation method, and so far the theory does not seem 

 to work well ; but since the probable error of the deter- 

 mination is large it is worth while to proceed. We will use 

 the result only to give new " smoothed " values for u/Ci and 

 to extrapolate a value for L/ = '00025 ; these values are given 

 under u/Gi (calc). 



The next step is to use the minima and the other maxima 

 of the curves. The following table gives for each value 

 of L/, first Cx corresponding to the various maxima and 

 minima, second the corresponding value of m (easily deter- 

 mined by inspection or corrected later if an initial mistake 

 is made), third the corresponding value of u taken from 

 Table III., fourth the corresponding value of c taken 

 from figs. 14 or 15. 



Table IV. 



L/. 



o r 



m. 



u. 



c (obs.). 



c (calc.) 



00025 



•060 (max.) 



2 



•185 



•135 



119 





•017 (min.) 



2 

 3? 



•052 



•29* 

 •13* 





00050 



•045 (max.) 



2 



•155 



•175 



•185 





•017 (min.) 



2 



•058 



•24* 







•119 (min.) 



1 



•410 



•180 





•00075 



•035 (max.) 



2 



•134 



•210 



•240 





•071 (min.) 



1 



•273 



•240 





00100 



•024 (max.) 



2 



•100 



•240* 



•288 





•053 (min.) 



1 



•221 



•297 





•00200 



•017 (max.) 



2 



•096 



•245* 



•429 





•025 (min.) 



1 



•141 



•337 





There are considerable discrepancies between the values 

 of c determined for the same L/ ; but all those marked by 

 asterisks are very uncertain owing to the shape of the curves 

 from which they are determined; if these are neglected 

 some of the greatest differences disappear. Further, for all 

 the curves except IV = '00025 the value for the minimum is 

 much more certain than that for the maximum and will 

 alone be used in what follows. The case of L/ = '00025 is 

 peculiar ; c determined from the minimum agrees well with 

 that determined from the maximum if ra = 3, but the form 

 of the curve shows that m should =2 ; but it must be 

 remembered that the theory is already known to fail if L/ is 

 very small. 



We have now obtained c as a function of L/. According 

 to the theory 



L 12 . L 2 i 



1—cs- 



(L 1 + L 1 ')L 2 " 



so that if we plot 1/(1 — cs) against L/, we should get a 

 straight line of which the intercept on the vertical axis 

 is — L l5 the slope IV/L12 • L 21 . The points are denoted 

 by in fig. 16, s being taken as 1*025. The point for 



