86 Transactio7is of the Boyal Society of South Africa. 
with a view to estimating the correction for variations of altitude of land 
and depth of sea ; it was concluded that the attraction could safely be 
taken as that of an indefinite plateau of height 55 m. and density 
about 2. 
Criticism of Method. 
It is unnecessary to describe the method of using the Sterneck appa- 
ratus, which has already been described more than once ; but some 
remarks on it have occurred to the writer during the course of the work. 
Three pendulums — numbered 23, 44, and 62 — were provided, and two 
stands, one for use on a pillar, the other against a wall ; of the two stands 
the former has been used, for the most part. The corrections to be applied 
are for — 
(a) Eate of clock. 
{b) Amplitude. 
(c) Temperature. 
{d) Density of air. 
(e) Oscillation of support. 
(a) In order to render this as certain as possible, it is desirable to 
extend the observations as nearly as possible over twenty-four hours, so 
as to make use of the average rate of the clock determined by observations 
of the stars. In the case of a good observatory clock it is probable that 
no sensible diurnal variation of the rate occurs ; though this is a point 
not quite satisfactorily settled ; whilst the variation in rate of a good 
clock from one day to another is hardly sufficient to affect the value of 
by more than one in a million. Chronometers, however, usually 
show a diurnal variation — an effect that has been but little studied ; and 
when moved about from place to place undoubtedly suffer irregular 
changes of rate; hence it is all the more important in this case to 
distribute the observations evenly between the star-transits. 
(b) Presents no difficulty. 
(c) The pendulums are old ones, made of brass. An error of degree 
corresponds to one-millionth in the value of " g." The correction, as 
given in the tables, appears very large, sometimes exceeding 1,000 units, 
but this is because the reduction is made to 0° C. The limits of 
temperature in the South African observations are about 12° to 24°, 
and it would be preferable to make the reduction to a standard tem- 
perature such as 20°, when its true value would be apparent. When 
working in a cellar there is no difficulty in finding the temperature 
correctly to degree or less. At the same time it would be much 
better to use pendulums of invar. 
* The unit is the seventh decimal of the (single) period in seconds. It is about 
1/5,080,000 part in the period or 1/2,540,000 in " 
