242 MESSRS. R. THRELFALL AND J. A. POLLOCK 



of the temperature error, we have taken as possible errors quantities which could only 

 be realised by bad observing and by omitting to apply a level correction, which, 

 hitherto, we have not found requisite. 



Reduction of Observations. 



The readings of the instrument at any one place differ from day to day from two 

 causes from change of temperature and from the slow elastic after-working of the 

 thread and its supports. When a new thread is set up, or any alteration made in 

 the lever or supports, the first thing to be determined is the effect of temperature on 

 the readings. When a considerable number of daily observations have been collected, 

 in the first place, the maximum temperatures of each day's observation, as entered, 

 are corrected according to a scheme drawn up from the comparison of the coils of the 

 resistance box. The maximum readings of the position of the vernier arm are then 

 all reduced to one temperature by an assumed temperature coefficient, on the assump- 

 tion that the relation between the temperature and readings is a linear one. A 

 plot is then made with time as abscissa and readings as ordinates. If there appear 

 to be systematic errors connected with the temperature, a new coefficient is taken, 

 and the observations again reduced. This procedure is repeated until the systematic 

 errors cease to be apparent sometimes a lengthy process. If there was no elastic 

 after-working, the plot of the observations from day to day at any one place should 

 lie on a line parallel to the axis of time. If elastic after-working exists the line 

 joining the plots of the observations will be more or less sloped to the same axis. 



DISCUSSION OF RESULTS. 



There are two conditions which a balance of this kind must fulfil for it to be a 

 working instrument firstly, it must give accordant readings at any one place from 

 day to day ; and secondly, the readings must not be affected by the vibration 

 inseparable from transport. We shall adopt the historical method of treatment, 

 discussing the deviations from the rigorous fulfilment of the two conditions in the 

 order in which they were observed. 



The first journey made with the instrument (other than preliminary ones) was 

 commenced in June, 1897. The instrument was taken from Sydney to Melbourne by 

 train, and set up in a cellar of the Physical Laboratory of the University. It was 

 then taken to Hobart by steamer, observed in a cellar of the Museum and in the 

 University Physical Laboratory, then to Launceston by train and observed in the 

 strong room of the Custom House, then to Melbourne by steamer, and to Sydney 

 again by train. On this trip simultaneous observations of temperature and twist 

 were taken over long periods with the temperature rising or falling more or less 

 rapidly, as we had not at this time discovered that it was essential only to observe 



