length of the pendulum vibrating seconds. 9 1 
of vibration is much reduced. It is therefore not impossible 
that an error of one second may sometimes, though rarely, 
occur in determining the time of this coincidence. This would 
occasion an error of about o ,63 in the number of vibrations 
in 24 hours, which divided by 4 as before, would influence 
the mean result 0,15 of a vibration. 
In estimating these errors, I have taken an extreme case, 
as it is probable they would in most instances be compensated 
by the succeeding intervals. Supposing them however to be 
combined, the greatest effect on the mean result of any one 
set of experiments might amount to about 0,3 of a vibration 
in 24 hours, and the difference between the number of vibra- 
tions in either position of the pendulum, might have been 
double this quantity, and yet when the great weight was 
below, not have differed from the truth more than 0,3 of a 
vibration. 
It appears then, that if the experiments have been con- 
ducted with sufficient care, no greater difference should be 
found between the mean, and any one of the resulting lengths 
of the pendulum contained in the preceding table, than might 
have been occasioned by a difference of 0,30 f a vibration in 
24 hours, and this is found to be about 0,0003 of an inch. 
In fact, on referring to the table we perceive that the expe- 
riments A and D, which differ most from the mean, give, the 
one, 00029 of an inch in excess, and the other ,00026 in defect. 
In considering the sources of error, it may not be unneces- 
sary to remark that had the bar of the pendulum been made 
too thick, and the knife edges not been placed accurately at 
right angles to it, an error, though very minute, might have 
arisen from the effect of the obliquity ip diminishing the dis- 
