256 .MHssljs. 14. THRELFALL AND J. A. POLLOCK 



00154 inch. Also the twist was far less than our previous experiments had 

 indicated as allowable. Thus, we thought that we ought to have been allowed to 

 put in a twist of one turn per centimetre, and we only put in a twist of "204 turn 

 per centimetre. This showed that either we had got hold of a bad sample of quart x. 

 in spite of all efforts to the contrary, or that the conditions of continuous stress 

 obtaining in a balance affect the elastic properties in some hitherto unrecognised way. 



Now the thin thread which did so badly from June till August, 1894, had a 

 diameter of "0004 inch or '00102 centim. The length of all the threads was 12 

 inches. We concluded that the diameter of the thread should lie between '0035 and 

 "001 centim., using levers of the mass hitherto employed, so as to avoid breakage^on 

 the one hand or irregularity on the other. The cause of breakage was also seen to 

 be related to the sag produced by the use of too heavy a lever. It was at this time 

 that we made the comparative examination of different samples of quartz referred to 

 in the text. We did not terminate these investigations till the end of 1894, and it 

 was not till March, 1895, that we succeeded in obtaining a thread to satisfy us. The 

 diameter of this thread lay between '0014 and "0015 inch, and was made of our 

 most infusible quartz and mounted with a very light straight wire lever. We 

 observed this thread till September, 1895, finding, amongst other things, that the 

 flotation correction was very small, as had been anticipated. The viscous yielding 

 was large, and we suspected that the silvering had not been removed sufficiently close 

 to the coppering (from fear of the nitric acid getting in and gradually loosening the 

 thread) ; this was tested by a reapplication of nitric acid, and the reading promptly 

 changed by 40 of twist. 



Finding the great weight of the instrument a drawback, we re-made the base and 

 replaced the theodolite circle at the twisting end by a sextant arc. By November we 

 had traced such irregularity as still persisted, to the sticking of a barrel spring, 

 which caused the tension of the thread to vary irregularly whenever the temperature 

 changed. We therefore designed and made the " rosette " spring referred to in the 

 text, but were so unfortunate as to break the thread in dismounting it ; we spent no 

 less than two months in getting a fresh thread to satisfy the conditions. On 

 January 22nd, 1896, this thread was mounted, with about three whole turns at each 

 end. It was observed continually until July 9th, when it was pleased to break. The 

 daily rate of subsidence of this thread was 6 sextant minutes in March, and fell to 

 2 '5 in the latter part of June. There appeared to be some effect depending on the 

 time which elapsed between the releasing of the lever and the time at which the 

 reading was taken. We did not find the reason of this at the time, but we now know 

 that it was due to the heating up of the balance by the observer. However, we 

 wasted some time over it. 



A new thread was got in September, 1896, after some weeks' shooting. This 

 thread is still in use and is the best we ever got. We had a great deal of trouble in 

 stopping leaks in the apparatus, which was beginning to show signs of wear in 



