DETERMINATION BY TORSION BALANCE 67 
were the small type of the Siiss ‘Original Edtvés” torsion balances 
and, on a few traverses, large Askania (automatic) instruments. The 
calculation of relative gravity around each traverse was made by a 
method in which superposition of a transparent graph over a gradient 
arrow gives the spacing of the isogams at that station; the isogams 
are sketched forward toward the next station and are bent as the 
experience of the geophysicist and the exigencies of the adjacent 
gradient arrows seem to demand. Single abnormal gradient arrows 
were neglected or given reduced weight according to subjective esti- 
mation of their value. The location of most of the surveys was in the 
Texas-Louisiana Gulf Coast within 150 miles of the coast, but a few 
of the surveys were farther inland and some of the surveys were 
located outside of Texas and Louisiana. Most of the terrane was good, 
but a few of the surveys were made in areas in which caliche was 
present in the subsoil and other surveys were made in river bottoms. 
The magnitude of the gradient most commonly was less than 20 &. 
The station sites on a few of the surveys were prepared with the 
meticulous care of the European custom, but on most of the surveys, 
slightly less care was used. Probably two-thirds of the observations 
were made during the day-time and one-third at night. The general 
rate of occupation of stations was two stations per day per instrument 
and two stations per night per instrument if observations were taken 
at night, although in about a quarter of the surveys, the rate of ob- 
servation was one station during the day and two at night per in- 
strument. These surveys were made in commercial oil prospecting in 
which, generally, speed and moderate accuracy are regarded as more 
important than high scientific accuracy, and the purpose of the sur- 
veys was qualitative or only roughly quantitative rather than ac- 
curately quantitative. The surveys were made by many different 
torsion balance parties. Most of the observers were not highly trained 
scientists. 
The precision of scientific measurements of such quantities as 
gravity in America is stated in terms of what is called the ‘‘probable 
error”; in Continental Europe, the ‘‘mean square error” is used; the 
“probable error” is 0.6745 times the-“mean square error.’ The term 
“probable error’’ is a technical term and is defined as the value such 
that half of the errors which are likely to occur in a large number of 
measurements of the quantity will be larger and half smaller; that is, 
if the determination of gravity at a station is stated in an American 
publication to be 978.569 +0.001 cm. sec.~’, it is understood that there 
397 
