182 Trans. Acad. Sci. of St Louis. 



material in the case in which it was enclosed. On other 

 days the value of G was decreased by 16, 23, 27, 69, 77, 134 

 and 184 per cent of its initial value. It is not claimed 

 that these are results of any high degree of precision. 

 During the first week while the bifilar method above de- 

 scribed was being tested, the time of vibration during a 

 period of two hours was found to be very constant. For 

 example, in the first determination made on October 23, 

 the mean position of the suspended masses was deter- 

 mined from readings of the extremes for several vibra- 

 tions. The key was then snapped for six readings when 

 the mean reading passed the cross-hair of the telescope. 

 The mean reading was determined in each case by a com- 

 putation from previous readings. It was found to be very 

 constant. After lunch the same operation was repeated. 

 This gave six readings of the time interval of 7 complete 

 to and fro vibrations, with data for determining the 

 probable error of the mean. The result was 653.45± 

 0.062 seconds. The probable error was sometimes three 

 times as great as the above value and the mean value 

 varied from the above value on other days by between 

 one and two seconds. An error in the value of n might 

 amount to three or four per cent. The cause for these 

 variations proved to be a difficult problem. The silk 

 threads were very loosely wound, and before they were 

 put in place a weight equal to half the weight of the en- 

 tire suspended mass was hung upon each of them for sev- 

 eral days. They then seemed to be in permanent condi- 

 tion. The breaking stress was about five times that which 

 was thus applied to them. 



The causes for the variation in the time of vibration 

 were thought to be possibly due to a breaking of some of 

 the fibers in the threads which might result in an en- 

 tanglement between the two threads, or which might re- 

 sult in an unequal division of the load between them. It 

 was sought to eliminate the former source of error by ap- 

 plying vaseline to the threads. This did not appre- 



