484 
PROF. J. W. NICHOLSON AND DR. T. R. MERTON ON THE 
and h = 198/33 = 6 mm. on the original photograph, whence I c is known on the 
same scale. The separations o-j, <r 2 are those of Stark reduced to millimetres on 
o 
the basis 10 A.U. = 1 mm. valid in these experiments, and become cr x = 0'26 mm., 
<r 2 = 0'62 mm. 
The heights of the two kinks are and h 2 , where 
I c e Al = 42’9e“ 16 ’ 5<ri + 2 {5 cosh 5o-i. e“ 5o ' l + 2 cosh 2<r\. 
l c e h ~ = 42‘9e _16 ’ 5<r2 +2 {5 cosh 5cr l ! e~ 5<r2 + 2 cosh 2<j 2 . e~ 2<r -}. 
Multiplication by 33 gives the heights on the enlarged photograph. The values of 
dy/dx are calculated in a similar way. 
In the resulting diagram, the kink at B is still a peak, but its height is very small. 
Down to the immediate vicinity of B, the curve is hardly distinguishable from a 
straight line, and its change to B is very abrupt. BC is again practically straight 
until the vicinity of C is reached, and C is not a peak, but a protuberance. The 
curve becomes practically straight again after C. The point B ceases to be a peak 
if qi is rather less than 5, and begins to resemble C. 
Perhaps the main points of interest about the curve are, next to its approximation 
to that for H a , the straightness of its branches and the smallness of the protuberances. 
But, in spite of their smallness, the method already described of using paper printed 
in a pattern of fine dots enables them to be detected readily by pricking out the 
final dot which is visible on the enlarged photograph. The distance between the 
protuberances on either side of the axis, divided by the magnification, gives at once 
twice the separation of the components from the central line on the original plate, 
and allows at the same time for the fact that the prismatic spectrum is not normal. 
It is to be noted that the ratios of the slopes of the nearly straight branches of the 
curve differ little from those of the values of q belonging to the components, when 
these values decrease so rapidly that the peaks become mere protuberances. 
Since the head of a protuberance on the curve lies necessarily on the axis of the 
component giving rise to it, the particular constant intensity which defines the 
contour is immaterial. If, therefore, a set of contours of various constant intensities 
are chosen on the plate, the heads of the various protuberances, one on each contour, 
due to any component, lie on a straight line parallel to the axis of the contours, or 
perpendicular to the base of the photograph. Protuberances of very small size can 
in this way be detected as such, without the risk of including slight irregularities 
in the contour which might be due to defects in the grain of the plate, and at 
the same time, the axes of the components can be defined with some degree of 
precision. 
We now come to the precise values of the separations determined from the 
photographs of H a . The contour exhibits three definite protuberances on each side 
at the same heights, and satisfying the conditions just specified. Oil one of the 
