118 
PRESTON. 
lum swings but little over one-third as long. Now it is well 
known that the best made clocks give variations from the 
mean daily rate of one or two tenth of a second for differ¬ 
ent hours when the temperature changes more than a few 
degrees. Admit one and one half tenths for 15,000 oscilla¬ 
tions and we have a discrepancy of 20 between different 
swings. This is slightly greater than those usually found, 
but we have supposed the variations on each side of the 
15,000 oscillation period to be equal and in opposite direc¬ 
tions, which is an exceptional case. At the greatest ele¬ 
vation in India, the English had variations of temperature 
of 26° daily, and a two-hourly rate variation of the clock 
from the mean that would amount to 24 seconds per day, 
which would give a range in the discrepancies for a 15,000 
oscillation period of about 70. 
The extreme range of temperature at a Coast Survey sta¬ 
tion during a period of eleven days, 10,000 feet above sea 
level, was 16° centigrade, and we have a range of about 40 
in the observed discrepancies. This range of temperature 
of 16° was not that to which the pendulums were subjected 
during the experiments; it is that of the air surrounding the 
chronometers. The variation in the pendulum house was 
about 11° centigrade. On the other hand, experiments made 
by the same observer and with the same pendulums only show 
a range of 4 in the discrepancies at the Lick observatory. 
Here however every condition with regard to the time deter¬ 
mination was of the most favorable. The Dent clock, which 
was on the chronograph most of the time for the pendulum 
work, in a seven-day period only varied from its mean rate 
two-tenths of a second. The variation of one of the Dutch 
clocks, which was afterwards substituted for the Dent, was 
even less. Mr. Keeler gives as the probable error of a clock 
correction 0 S .02. At the mountain station the chronometer 
varied a second from a uniform rate in a seven-day period, 
and as the star observations show that a change of nearly 
two seconds. has occurred between the rates of consecutive 
days we may expect discrepancies of 20 in the separate 
