the Pendulum vibrating Seconds in the Latitude of London, 423 
to conjecture why this is preferred to the sidereal day, a measure 
of time which marks a complete revolution of the earth, and is 
- readily obtained, being the interval between the returns of any 
fixed star to the meridian. . 
I shall now proceed to notice the sources of error which may 
be supposed to have affected the results of the preceding experi- 
ments. : 
These may be classed under the following heads : 
1. The measurement of the distance of the knife edges. 
2. The number of vibrations in 24 hours. 
3. The temperature, and 
4. The form of the knife edges. 
On the first, it is searcely necessary to offer any remark. Since 
the mean results of three several sets of measurements are within 
one ten-thousandth of an inch of each other, and the different 
methods employed, preclude, it may be presumed, any accidental 
coincidence, we may with confidence infer that the error in the 
distance of the knife edges cannot amount to one ten-thousandth 
of an inch. 
Among the number of vibrations in 24 hours given in the va- 
rious sets of experiments, there appear to be differences which 
amount in some instances to 16. These differences hewever do 
not influence the truth of the result, beyond a certain minute 
quantity, the extent and origin of which I shall proceed to ex- 
plain. 
In order to determine the vibrations in 24 hours, it is necessary 
to ascertain the number of vibrations and parts of a vibration 
made by the brass pendulum during a certain number of com- 
plete seconds; but the moment of observation being limited to 
that when the brass pendulum is at the lowest part of the are, 
the process is of necessity reversed, and the brass pendulum is 
observed to make a certain number of complete vibrations, during 
a certain number of seconds and parts of a second which consti- 
tute the interval. The disappearance of the disk can however 
be noted only to a single second, and the brass pendulum may 
arrive at the lowest part of the arc either precisely at this se- 
cond, or at any portion of a second preceding it. An error might 
possibly arise from this circumstance amounting to nine-tenths 
of a second, by which the interval deduced from observation would 
be less than the truth; and as an error of one second in the in- 
terval, occasions a difference of 0:63 in the number of vibrations 
in 24 hours, if 0°55 (the proportional part of 0°63) be divided 
by 4 (the number of intervals forming each set of experiments), 
we have 0-14 for the greatest error in defect in the number of 
vibrations in 24 hours which can arise from this cause, 
Dd4 On 
