78 DONALD C. BARTON 
If the pendulum stations are placed closer together than a certain 
minimum distance, the “probable error” or the pendulum determina- 
tion of Ag between adjacent stations will be greater than the “‘prob- 
able error’ of the torsion balance determination of that Ag; and ad- 
justment of the torsion balance survey to the pendulum values of Ag 
will decrease the accuracy of the torsion balance survey. If the dis- 
tance between the adjacent pendulum stations is greater than that 
minimum distance, the ‘‘probable error” of the pendulum determina- 
tion of Ag between adjacent stations will be less than that of the 
torsion balance determination; and the accuracy of the torsion balance 
survey will be improved by adjusting its results to the pendulum 
results. 
The accuracy of the determination of relative gravity by the rela- 
tive pendulum varies. The “probable error’’ of the pendulum observa- 
tions of the latter part of the past century presumably was, in general, 
many milledynes. The probable error of the observations of the re- 
cent past in general has been +1 milledyne, although a few poorer 
but acceptable stations have had a “probable error” of +2 mille- 
dynes. The better modern observations are reported to have a “‘prob- 
able error” of +0.3 to +0.5 milledyne. The geophysical department 
of one oil company asserts that its pendulums are reoccupying sta- 
tions with a “probable error’’ of observation less than +o0.1 mille- 
dyne. The “‘probable error” of the determination of the elevation or 
relative elevation of the pendulum station is relatively unimportant 
in connection with the pendulum observations whose ‘‘probable error” 
is +1 milledyne or more, but is, of course, of very considerable im- 
portance in connection with observations whose “probable error” 
is less than +0.3 milledyne, for the variation of (g) with elevation 
is +0.308 milledyne per meter. The ‘probable error” in feet of first 
class (but not precise) levelling is approximately --o.o5 (number of 
miles), but only 1.5 miles of closed traverse, or 3 miles of unrepeated 
traverse can be run per day in levelling of that accuracy and, there- 
fore, most commercial levelling in connection with commercial geo- 
physical work probably will be speedier and less accurate and its 
“probable error” may have to be taken seriously into account. The 
“probable error” of the station determination of gravity will be the 
square root of the sum of the square of the “probable error” of the 
pendulum determination of gravity, plus the square of the “probable 
error” of the determination of the elevation. The ‘‘probable error”’ of 
the determination of Ag between two places will be the sum of the 
408 
