system is very high, of the order of +0.014 percent, owing to the 
great accuracy of the digital voltmeter and to the linearity of the 
amplifier. A correction curve could be obtained by using the 
whole series of calibrate resistors. 
The precision* with which the calibrating tank can be main- 
tained at an indicated temperature is +0.001°C. The precision of 
the measurement of the resistance of a thermistor is +0.09 ohm 
(equivalent to 0.0007°C); lack of precision is due in part to failure 
to maintain temperature constant. 
In addition, the value of the resistance of the equivalent 
normalized resistor at each of the calibration temperatures is 
taken as the mean of the resistance of the individual thermistors in 
the set. The probable error of these readings is taken as a quasi- 
precision of this resistance, and was found to be equivalent to 
+0.002°C. Note that this includes the two preceding precisions. 
The precision with which a given resistance can be deter- 
mined (this includes fluctuations in the bridge and in the digital 
data system) is obtained by repeated readings of a given calibrate 
resistor; this gives a probable error for the measurement of the 
resistance that corresponds to +0. 0025°C. 
The over-all precision of the temperature measurements is 
the square root of the sum of the squares of the above two individ- 
ual sources of probable error. The over-all precision is +0.0032°C. 
The errors in the temperature sensors introduced during the 
field trip are difficult to assess. Calibrations are made before and 
after the trip; the differences in calibration are usually within the 
probable error of the measurements. The effect of depth (i.e., 
pressure) on the calibration is difficult to determine, but tests 
elsewhere indicate that the effect is negligible at depths up to 
500 ft. ** 
*Precision here will be taken to mean the probable error in the 
value of a reading repeated a number of times. 
**Reference 4, page 55 
21 
