408 Proceedings of the Royal Society of Edinburgh. [Sess. 
by black velvet, the diffuse reflection from which is considerably less. The 
following table gives some results then obtained : — 
T 
F 
C 
T + F-C 
Probable 
Error. 
752 
3906 
4714 
-56 
20 
825 
843 
1654 
+ 14 
12 
910 
23 
952 
-19 
9 
768 
823 
1552 
+ 39 
12 
832 
623 
1428 
+ 27 
12 
590 
779 
1376 
— 7 
3 
857 
260 
1175 
-58 
8 
Each of the above results is the mean of six determinations made at 
one sitting, which took about two hours. The error is in every case 
greater than the calculated probable error. F, though represented by 
different numbers, was approximately the same strength throughout. The 
S 
Exciting 
Rays 
Diagram 4. 
numbers are proportional to the squares of the tangents ; it was not 
thought necessary here to reduce them so as to represent F by the same 
number in each row. In the observations recorded in the sixth row, the 
cardboard square was used as a comparison source. 
A cadmium spark was tried as exciting source, but was not steady 
enough. 
The above determinations were made on the wave-length A = 554 fi /u. 
A strip of spectrum 38 jul/a wide was examined with an equally wide slit. 
For a given brightness of field, the light is purer when the slit and the 
strip of the spectrum examined are equally wide. Owing to using this 
wide slit, a slight difference of tint between the two halves of the field was 
at first visible in the F measurements. This difference of tint is referred to 
by Miss Wick and by Burke, who did not resolve his light spectrally at all. 
In order to test whether the difference of tint made an appreciable error, a 
piece of green glass was placed before the comparison source. This made 
the upper half of the field distinctly yellower. It had a distinct effect, 
giving a value of (T + F — C)/T = Tl, and on making the difference of tint 
more pronounced (T + F — C)/T was obtained = *22. 
The very slight difference of tint present originally was removed by 
