Hay 28 ,1917 Transpiration Rate of Medicago sativa 285 
of the four instruments was then taken. Since each of the instruments 
had been calibrated, the departure of the mean of this series of ratios 
from unity represents the error in the calibration coefficient. These 
departures, which are given in the third column of Table III, show 
errors ranging from +4 to —2 per cent for the white spheres, from +3 
to — 4 per cent for white cylinders, and from + 2 to — 2 per cent for the 
brown cylinders. 
Table III .—Calibration error and probable error of single observation for individual 
atmometers 
White cylinder: 
5 - 36 .... 
5-37 . 
5-57 . 
5-142. 
White sphere: 
16-16. 
i6-77. 
16-105.-- 
16-146. 
Brown cylinder: 
4 -i 7 . 
4-25-. 
4 - 4 i.•• 
Bell am plates: 
D-137. 
D-175. 
D-205.... 
D-206. 
Ratio to 
Observed 
error in 
coefficient. 
Probable 
error of 
mean. 
single ob¬ 
servation. 
Per cent. 
Per cent. 
I. 02 
+2 
± 3*7 
.96 
-4 
±1. 7 
l. 03 
+3 
±1.9 
.98 
—2 
±2. 7 
.98 
—2 
±2.3 
.98 
— 2 
±1.7 
1. 00 
0 
±1. 6 
l. 04 
+4 
± 3-2 
1. 01 
+ 1 
±2. 6 
.98 
—2 
±1.5 
1. 02 
+2 
±2. 5 
.98 
. OA 
—2 
±2. 6 
± 3 -o 
.96 
1-03 
1. 07 
±1.8 
±2. 6 
±3- 2 
The probable error of a single observation of each of the instruments 
has also been computed for each instrument. The results are given in 
the last column of the table. These errors, which do not include the 
errors in the coefficient, amount to from 2 to 4 per cent. It is possible 
from these data to obtain an expression of the degree of reliance to be 
placed upon individual observations. Determinations with a single 
atmometer at 2-hour intervals during the day thus appear to be subject 
to a probable error of about 5 per cent, 1 if the weighings are accurate 
to 0.1 gm. This probable error refers only to calibration uncertainties 
and to so-called “accidental” errors, as exemplified by the variation 
of the individuals in a group of atmometers of the same type. It does 
not include, for example, errors arising from the possible failure of the 
whole group of atmometers to respond freely to changes in their environ¬ 
ment. 
1 Based upon the square root of the sum of the squares of the mean values of the calibration error and 
of the observation errors, taken without regard to sign. 
