86 JOURNAL OF THE WASHINGTON ACADEMY OF SCIENCES VOL. 13, NO. 5 
The densities of the two solutions are so chosen that for the average 
observer they transmit equal percentages of light from the standard 
carbon lamp. In general, different observers find different values for 
the transmission of each of these solutions under standard conditions, 
and it is convenient to use the ratio of the two transmissions as a 
rough indication of the characteristics of the observer. The use of 
such tests is not merely a matter of finding “normal” and ‘‘abnormal”’ 
observers, for we find that the characteristics of observers extend 
over a considerable range with a distribution roughly approximating 
that represented by the ordinary “curve of errors.” It should be 
noted that this distribution does not represent error of measurement 
since each observer will repeat his values with a certainty represented 
by a very small part of the range of distribution. In other words, 
the curve represents the real distribution of individual characteristics, 
not the errors in determining those characteristics. 
This characteristic ratio for a given observer is closely correlated 
with the results which the observer will obtain in comparing the 
light emitted by radiators at different temperatures, such as lamps 
of different efficiency. It is therefore possible by the use of these 
test measurements to reduce the results obtained by any small group 
of observers to a normal value with a considerable degree of certainty. 
An illustration of such correction is shown in Table Il which repre- 
sents an actual case where five observers determined a ratio equivalent 
in color to the step from vacuum tungsten to gas-filled standards. 
It happens that at the time this work was undertaken the group 
of observers available showed a wide variation in test ratios as given 
in column 2. Consequently their deviations from the average were 
very large, as shown in column 4. Nevertheless when corrected 
they gave an agreement which is practically perfect for such work, 
the average deviation of the individual observers being less than one- 
tenth of one per cent from the mean. In this particular case it 
happens that the observers are so distributed that their uncorrected 
mean is equally good, but this condition can not be depended upon. 
Ordinarily a group of observers will show much less scattering in 
their characteristics so that the individual corrections will be less 
important than they were here, but one can not depend upon a small 
group being symmetrically placed with reference to the mean as 
this one was. 
It should perhaps be noted that for relatively small color differences 
the flicker photometer is not as precise as the ordinary form. For 
the moderate color differences represented by the step from carbon 
