MAR. 4, 1923 CRITTENDEN: MEASUREMENT OF LIGHT 87 
to tungsten lamps, the advantage of the flicker lies not so much in 
greater precision as in the certainty that the results obtained in a 
particular case are consistent with those obtained at other times and 
that they will agree with results made by similar methods by a differ- 
ent group or a different laboratory. In other words, the variations 
shown by the flicker method are scattered about a definite point in a 
purely accidental fashion, whereas the measurements by equality-of- 
brightness by an experienced observer may be just as consistent among 
themselves and yet be greatly influenced by a permanent prejudice. 
The test-ratio scheme here illustrated is of course of limited use. 
Nevertheless it makes possible a decided improvement in the cer- 
tainty of the measurements which are now of most practical impor- 
tance, and the general principle of test measurements and of 
establishing normal values which will be accurately reproducible is 
TABLE II.—Correction or INDIVIDUAL OBSERVERS’ RESULTS TO NORMAL 
Color difference equivalent to step from 9.3 to 18 lumens per watt 
: PHOTOMETER RATIO 
DEVIATION 
OBSERVER NUMBER TEST RATIO pes ea eS 
Observed Corrected 
1 0.900 —0.087 0.2625 0.2648 
2 0.909 —0.079 25 46 
3 0.981 —0.006 47 49 
4 1.024 +0 .037 59 49 
5 1.110 +0.1238 86 54 
Mean 452 5520,-! 0.985 0.2648 0.2649 
applicable to all sorts of measurements. It must be admitted, however, 
that the principle of the flicker photometer appears to be much 
farther removed from the actual conditions under which light is used 
than the simpler photometers are. The general adoption of this 
method of comparison is therefore doubtful. 
Visibility Curves.—In discussing the definition of light it was sug- 
gested that we might consider each element of radiant energy to be 
tagged with a coefficient representing its relative visual effect. Fur- 
ther consideration has shown that the value of this coefficient is 
somewhat indefinite. Nevertheless if we agree on definite conditions 
and methods of observation, an average value can be given for this 
coefficient for each wave-length. A complete solution of the problem 
of photometry must include the determination of these relative visual 
coefficients, and when they are established all other comparisons 
might be based on them. It is, however, no simple matter to establish 
