1016 
Fig. 4. 
and it is no. better when we choose another point instead of P. 
The curves 2 and 3 however are in good agreement with the 
situation of the. points, taking into consideration that the deviations 
are naturally considerable, quite, apart from errors. 
It cannot be decided with certainty whether 2 or 3 should be 
preferred. It seems to me that 2 is slightly more satisfactory. At 
ai 1 
any rate the supposition that A? should be proportional to ms is 
contradicted by the experiment, whereas HEINSTEINs formula, at 
iy 
least as far as the connection between A? and a is concerned, is 
confirmed. ’ 
