456 PROFESSOR CHRYSTAL 



In regard to these the following precautions were used : — 



1 and 2. Marks were made across the continuous limnograms every twelve hours or 

 oftener, and the corresponding time by the observer's watch, with the date, was written 

 opposite. The rate of the driving clock was relied upon merely for the interpolation 

 between any required turning-point and the nearest time mark. This, of course, in- 

 volved a great deal of tedious calculation and measurement of the limnograms. To 

 give an idea of the irregularities possible from this source, it may be mentioned that the 

 extreme variation in the number of minutes per millimetre run of the recording paper 

 in the series of observations given in Table II. below was from "892 on 13th August to 

 •903 on the following day. For the most part, however, the variation of this number 

 in two or three days did not exceed one or two units in the third decimal place. The 

 effect of the errors arising from the present sources is, of course, much reduced by the 

 method of interpolation followed ; and it is still further minimised by taking the 

 weighted mean of a large number of observations. 



In order to test for any possible slipping of the paper between the draw-rollers, the 

 distance corresponding to ten revolutions of the clock arbor was always measured in the 

 neighbourhood of the part of the limnogram under consideration. In the series of 

 observations in Table II. the extreme variation of this number was from 651 "2 to 653*6. 

 As the variation in a period of several days was often less than a millimetre, and the 

 variation is affected by the hygroscopic as well as the slip error, probably the latter was 

 non-existent. 



3. All stretches of the limnogram which contained any sudden change of configura- 

 tion or other suspicious irregularity were rejected in the determinations of period. In 

 using long and apparently regular series of oscillations, the series was divided into two 

 or more parts, and the periods determined from these parts compared with each other 

 and with the period determined from the whole. As an example, attention may be 

 called to the fine series of 375 uninodal oscillations from 23rd to 27th September (Tables 

 II. and III. below), which was observed both at Picnic Point and near the E. Binode. 

 Alongside of this we may place the set of 95 uninodal oscillations observed on 

 4th September at the Binode (Table III.). In this last case the times of the initial and 

 final maxima were determined independently by means of an index limnograph, the 

 Sarasin being used merely as a counter. 



4 and 5. The errors from these sources were avoided by abstaining from the use of 

 turning-points showing zigzags or unusual flatness, and employing index limnograms 

 taken with a narrow access tube. 



6. As already explained in a former communication, errors of this kind were avoided 

 by attending to the configuration period, using turning-points where there was 

 symmetry in the limnogram, and, in the case of index limnograms, by using the process 

 of residuation.* 



7. The watches of the various observers were compared at short intervals with my 



* I.S.E., V . 382, 



