426 
CAPTAIN ABNEY AND MAJOR-GENERAL FESTING 
The most difficult part of the spectrum to measure is close to the maximum, as in 
this case at eacli end of the oscillations the shadow illumined by the candle is too 
light. It is well at this part to oscillate the comparison-light gently backwards and 
forwards on the scale (the slit slide being stationary) until a balance is obtained, and 
to compare the result thus obtained with that by oscillating the slit slide. 
By commencing with a narrow slit in the collimator the above method answers 
admirably as far as FfG in the spectrum of a brilliant source of light, and by opening 
the slit to a moderate degree the intensity of the spectrum as far as G in the violet 
can be measured. Beyond G the method fails to give very reliable results unless the 
patch of light be diminished very considerably by substituting a lens of short focus 
for the re-combining lens, and using a much thinner rod to cast the shadows. It will 
be seen further on that the slope of the curve in the violet is very gentle. It is, 
therefore, practicable to use for this part of the spectrum a method which is not well 
adapted for the part near the limit of the red for reasons which will presently appear. 
The spectrum itself is thrown on a white screen by a lens of long focus—that which 
we used gave a visible spectrum 8 inches long. The violet portion is sufficiently 
brilliant to be measured with accuracy. By oscillating a thin rod in front of the 
screen at different parts of the spectrum the same alternations of “too light” and 
“too dark” of the candle-illumined shadow are observed, and the part of the spectrum 
where there is a balance for each position of the candle is determined with very close 
approximation down to the limit of visibility of the spectrum. 
It might be feared that the white light which illuminates the prism might vitiate 
the results; but this is not so to any measurable extent, as was proved by using a slit 
and a second prism with which to disperse the white light. This last plan answers, 
therefore, perfectly for the whole of the spectrum on the green side of the maximum 
On the red side the case is different. As a reference to our figures will show, the curve 
here is very steep, consequently there is a perceptible gradation of luminosity in the 
length of the spectrum covered by the shadow thrown by the candle, and the com¬ 
parison is therefore liable to error. By changing the relative positions of the candle, 
rod, and spectrum, so as to reverse the relative positions of the shadows, we have found 
that a slightly different value is obtained, but the mean of the two agrees very closely 
with the value obtained by the other method of comparison. 
§ III. Finding the Position of the Spectrum. 
We fixed our position in the spectrum by means of the bright lines produced by 
burning different salts in the arc. Those of magnesium and lithium we found most 
convenient; both of these which we used give copious indications of sodium, and we 
were thus able to fix the positions on the scale of the red and blue lithium lines, the 
green magnesium line, and the orange and yellow sodium lines. When necessary, 
calcium chloride gave us a means of fixing H. A piece of ground glass being placed 
