156 M. F. Reich's New Experiments on 



The orighial numbers themselves suggest the idea that the 

 two results marked thus * must be erroneous ; omitting these 

 fpom the calculation, we obtain as 



Mean 5*5712 with a 

 Probable error of 0-0113. 



The second series was executed with a copper wire 0*4 of a 

 millimetre thick and 620 millimetres long. The first wire was 

 too stiff to be stretched quite straight by the weight of the tor- 

 sion-balance attached to it, and it was a natural conjecture, that 

 the considerable alterations in the force of torsion, which pro- 

 nounced themselves both by differences in the magnitude of the 

 deflections and by differences in the times of vibration, might be 

 due to the imperfect tension of the wire ; this source of error 

 being excluded by making use of a thinner wire, a better coinci- 

 dence among the single results might be expected. The thinner 

 wire must, however, be taken shorter than the former one, in order 

 not to make the duration of a vibration inconveniently great. 

 This hope was not fulfilled ; for in the first series the time of a 

 double vibration varied from 688*6 to 722*0 seconds, and in the 

 second series from 497*0 to 512*6 seconds; the former of which 

 corresponds to a change of the force of torsion in the ratio of 

 1 : 1*099, and the latter to a change of the same force in the 

 ratio of 1 : 1*064; the results obtained in the second series are 

 also somewhat more discrepant than those of the first series, for 

 we found — 



TJie third series was executed with a bifilar iron wire ; the 

 wires being 4*2 milHmetres apart below, and 5 millimetres apart 

 above, and still 2270 millimetres in length, and I hoped thereby 

 to obtain more coincident results. This expectation seemed to 

 be justified by the fact, that after one or two days the wire 

 assumed an almost constant position, whereas a unifilar wire 

 requires several months before the tendency to turn to the one 

 side or the other is annulled. In this respect, therefore, the 

 bifilar wire is decidedly to be preferred. But unfortunately the 

 differences in the force of torsion were even more considerable 



