( 554 ) 
obtained disagreed for the greater part of the time with the idea 
of a regular variation of the azimuths of the marks. 
9. That a small correction of the assumed constant of collimation 
for Polaris, which might give a regular variation of the azimuth 
during the whole period, is not impossible may be seen from 
what follows. This constant consists of two parts: the constant 
of collimation determined in the nadir by the reflection of the 
vertical thread on the horizontal mercury-surface, and the small 
correction for the flexure of the rotation axis expressed by the for- 
mula b(1-+ cos). The first part, the constant of collimation in the 
nadir, is probably very exact (only in the period 1884—1885 the 
degree of precision may be a little less, the level being less reliable), 
but the value of the flexure found by determining from time to time 
the constant of collimation in the horizontal position of the telescope 
by pointing at the marks before and after a reversal is less trust- 
worthy, especially as it appeared that it varies distinctly, when the 
position of the cell of the object glass was altered. 
Therefore I thought myself justified in deducing from the Polaris- 
observations and the readings of the meridian marks small corrections 
to the constant of collimation, which rendered the variation of the 
azimuth-correction more regular, and which, with a few exceptions 
were constant in the periods during which the cell of the object 
glass was in the same position. 
In order to be able to judge in how far the motion of the pole 
in the 14-monthly period deduced from the observations is dependent 
on this correction, I have computed this motion supposing: 1st that 
the constant of collimation is left unchanged, 2™¢ that the small 
correction mentioned has been applied to it. 
10. The values of o—c, during 14 years (1882 July to 1896 
July) corrected according to 7, for each observer separately, are 
divided according to the time of the observations into 12 groups, 
each enclosing 430 days (the assumed value of the period of the 
pole-motion) and each of these periods is subdivided into 43 periods 
of 10 days !). The mean values of o—c, during these 10 days for each 
culmination separately, were now formed and to each mean was 
given the weight 1, independent of the number of values. So we 
obtained a series of numbers (at a maximum of 24), representing 
the values of o—c in both culminations belonging to one and the same 
') In this computation the period is actually reduced to 439*/;, days, by causing 
a period of 430 days to succeed two periods of 431 days. In the periods of 431 
days one of the subdivisions consists of 11 days instead of 10 days. 
