Unit of Candle-power in White Light. 



to- 



direct step would have been so small, had the measurements 

 in this case also extended over a longer period. 



It will be seen that the largest difference o£ any observer 

 from the mean in the first result by the cascade method was 

 about 1*5 per cent, in the case of observer E. His mean 

 difference per step was, therefore, about 0*3 per cent. By 

 the direct method in one step the greatest difference was 

 again by observer E, viz., 1*2 4 per cent. 



An examination of the differences from the mean result 

 revealed by each observer for each successive step in the 

 cascade is facilitated by Table IV. 



Table IV. — Showing how much for each Step in the 

 Cascade each Observer differed from the Mean of all 

 for that Step. 



Observer 



Got 



mpansons. 



i pa 



Set 2 against Set 1 



,. 3 „ „ 2 



,. 4 .. ., 3 



„ 5 „ „ 4 



,, G ,, „ 5 



Mean 



Total ( cascade) 



Set G against Set 1 (direct; 



A. 



13. 



0. D. 



E. 



* 



P. cent. 



P.cent. 



I'. cent. P.cent. 



P.cent. 



P.cent 



-04. 



-0-5, 



-nl, -01 3 



+0-8 4 



4-0-3., 



-01.. 



-o-of 



-00 -fO'Og 



4-0-0, 



+ (>•<» 



-0% 



-(>•:;, 



-00 e +0-3, 



+ 0-4, 



-fO'Oj 



-0-2., 



-03 4 



-Ui». 4-OT, 



4-0'] 



+03- 



-o-i. 



-o-o, 



-04 „ 4-0-2, 



4-0 0, 



4-01 



-0-2, 



-0-2 



-0 0, +0-1, 



4-0-3,, 



+o-] : , 



-]'4 e 



-1'3 8 



-0-3, +0-7, 



+l'4 a 



4-09, 



-00 



-0-1- 



-07- -0-3, 



+1'2 4 



4-0-0, 



It will be seen from this table that there is, as would be 

 expected, a general tendency for each person's observations 

 to be either higher or lower than the average of all observers 

 throughout the whole series. The amounts by which the 

 determinations of any one observer are high or low for 

 the different steps varied slightly, but no more than would 

 be expected from Table V., showing the probable errors of 

 the comparisons. 



Observer E obtained rather a large difference from the 

 mean in the comparison of set 2 against set 1, but this 

 appears to be the only slightly anomalous result in all 

 the 30 cascade determination-. 



The table shows clearly that even the small colour differ- 

 ences represented by a change of temperature of the filament 

 of about 45° C. give rise to differences, though small, in the 

 personal judgment of the various observers. Furthermore,. 



