426 ^ Transactions of tJie Boyal Society of South Africa. 
became apparent that the observations, now taken by the Eepsold method 
with hand guiding, could not be reconciled within reasonable limits with 
earlier observations taken by the older methods. Though instrumental 
defects were suspected, which, however, could not be located, it was fully 
apparent that the discordances were in a large measure of a personal 
character. 
The attempt to follow up these discordances was abandoned at the 
time, it being thought that the discrepancies could be more effectively 
dealt with when a suitable mechanical movement, which was at the time 
in course of construction, was available. The experience, however, led 
me to doubt whether the Eepsold method, either with or without clock- 
work movement, could be effectively used for accurate determinations of 
the Right Ascension of slow-moving stars. In the case of equatorial stars, 
the observer at the telescope need only concentrate his attention for a 
short interval, say some eight or ten seconds at a time, whereas the 
observation by strictly similar means of the closest circumpolar on our 
working list would need highly concentrated attention over some ten 
minutes. 
This, as may be readily judged, either put an inordinate strain on the 
observer, or resulted in very unreliable observations. 
The plan adopted for the removal of this difficulty was to revert in the 
case of circumpolar stars to the older methods of chronograph and key 
adopted for clock stars. The chief difficulty in relation to the circumpolars 
is that, though they are relatively free from apparent motion, they are not 
absolutely so, and the personality seemed to be mainly due to this small 
residual motion which they possess. Of course they could not be either 
hurried or reduced to complete rest, but an alternative plan was available 
of utilising the mechanical movement provided for observing clock stars 
to give a definite and comparatively rapid motion to the travelling wire. 
It thus became possible to accurately observe the time of transit of the 
moving wire over the relative fixed star instead of the moving star over 
the jfixed wire. Of course the personal errors with which I have been 
dealing to-night would affect such an observatioa, but there is an impor- 
tant differentiating feature. 
A wholesome rule which has been laid down for all measurements of 
precision for the elimination of systematic errors is ''reverse wherever 
possible." Even unrecognised sources of error may often be eliminated 
by the adoption of this expedient, and practically all the sources of error 
that I have been discussing to-night could be thus eliminated if we could 
include under the heading of possibility some means of reversing the 
direction of rotation of the earth itself. 
When, however, the significant part of the apparent motion is not that 
due to the earth, but to that mechanically imposed on the moving web in 
