An Experimental Study of Field Methods. 
549 
In Figs. 18 and 19 the x abscissas are the lengths of test 
sights in feet, while the y ordinates are the average or proba¬ 
ble error of such length sights, the heavy curve being simply 
an average curve. 
Fig. 18 records the work of midday. The decreased accuracy 
with longer sights, as shown by the rapid descent of the error 
curve to the x axis after 500 feet sights, is largely due to the 
excessive refraction at such times, increasing as the lower line 
was forced nearer the ground into the stratum of greatest re¬ 
fraction. 
The descent was arrested at the 1100 ft. sight, because this, 
and the remainder of sights, were read by half intervals on the 
upper and clearer portions of the rod, causing both lines of 
sight to traverse strata of nearly equal refracting power. 
Iggggg 
gggg 
—r:~K~ 
~S ~-\—— - 
B 
1 
1 
i 
-1 HHtrt'hi tiTHtTff II Iff +HH0 
iSSi KB 
apifftffiiijffffftffffif miiii 
iSS 
Fig. 19.—Error Curve for Midday Work (9 a. m.— 2:30 p. m.). 
204 test sights aggregating 41.3 miles , on a base line , using sights of 200-2000 ft. in length. 
East Troy , Wis., Aug., 1894. X and Y are the same as in fig. 18. 
Fig. 19, representing work of morning and evening hours, 
presents a striking contrast, for, during its whole length, the 
error curve varies but little from 1 in 1000. 
Not less interesting is the manner in which the error of these 
85 miles of stadia measurements accumulated. The 12 miles 
made in morning and evening hours, using an average length 
sight of 600 ft., show an error of 4- 1 in 2,685, while same class 
of work done in midday shows an error of — 1 in 655. Again, 
the 30 miles measured in morning and evening hours, with 
sights from 1,100 to 2,000 ft. long, show an accuracy of + 1 
in 1,741, while same class of work executed in midday shows an 
error of — 1 in 209. 
37 
