132 Joty—On a Method of Photography in Natural Colours. 
comparing the height DG with the height of the green sensation curve at HK. At 
D we are, in fact, not only projecting the amount of red sensation DR, but 
adding the large quantity of red sensation involved in the quantity DG of E 
green, as well as adding some violet. The image of the spectrum will, accordingly, 
be much too orange at this point. It is evident that the error arises from 
the compound nature of the sensations attending the excitation of the green 
sensation, over which we have no control. Hence it is that deductions from the red 
and violet curves are necessitated whenever green light is required in the synthesis ; 
that is, wherever the green-sensation curve overlies the red or the violet sensation 
curves. Maxwell constructed curves to which the red and violet constituents are 
diminished by the amount of red and violet sensations conveyed in the green 
chosen by him as the fundamental green, and by him assumed to excite only the 
green sensation. 
As the results of trials on Koenig’s curves made with the object of ascertaining 
the green wave-length involving least negative red and violet, the accompanying 
figure 2 shows in the full lines derived curves which, in their general location, 
resemble Maxwell’s, but are based on a more correct choice of the fundamental 
colours, and connect our photographic process directly with the colour sensa- 
tion curves. It is seen that the violet curve is but little altered, and the 
red considerably altered. The amount of negative colour (which cannot be 
realised) is inconsiderable. The outcrop of the red-sensitive curve in the 
violet is not difficult to realise in the use of orange dyes for the red-taking 
light filter, which in most cases show some transmission of the violet wave- 
lengths. 
Although it is possible that the compound nature of our green sensations 
may deny absolute accuracy to this method of colour photography, still my own 
results on the curves just described, and the results of Ives and others on modified 
Maxwell curves, appear to show that a degree of accuracy baffling the criticism 
of the ordinary untrained eye may be attained, and that in the reproduction of 
the most complex tints. 
The symmetry of the derived curves renders their application easy. The 
transmission of light through a pigment is not limited generally to a small group 
of predominant wave-lengths, but falls off uniformly at either side in the direc- 
tions of longer and shorter waves. If we choose the pigments used on the 
analysing screens so that their predominant transmissions are at three points in 
the spectrum indicated by the axis of symmetry in the three curves, these being 
nearly symmetrical, very accurate results are obtained. The positions of these 
axes of symmetry are shown by the vertical dotted lines. Accordingly, I make 
the colour of the red-taking screen that of the spectrum at a point displaced to 
