s, taking the Riffel as ten. It will be seen
that on December 23 we have really very little violet and less than half
the green, although we have four fifths of the red.

The next diagram before you shows the minimum loss of light which I have
observed for different air thicknesses. On the top we have the calculated
intensities of the different rays outside our atmosphere. Thus we have
that through one atmosphere, and two, three, and four. And you will see
what enormous absorption there is in the blue end at four atmospheres.
The areas of these curves, which give the total luminosity of the light,
are 761, 662, 577, 503, and 439; and if observed as astronomers observe
the absorption of light, by means of stellar observations, they would
have had the values, 761, 664, 578, 504, and 439--a very close
approximation one to the other.

Next notice in the diagram that the top of the curve gradually inclines
to go to the red end of the spectrum as you get the light transmitted
through more and more air, and I should like to show you that this is the
case in a laboratory experiment. Taking a slide with a wide and long slot
in it, a portion is occupied by a right angled prism, one of the angles
of 45 deg. being toward the center of the slot. By sliding this prism in
front of the spectrum I can deflect outward any portion of the spectrum I
like, and by a mirror can reflect it through a second lens, forming a
patch of light on the screen overlapping the patch of light formed by the
undeflected rays. If the two patches be exactly equal, white light is
formed. Now, by placing a rod as before in front of the patch, I have two
colored stripes in a white field, and though the background remains of
the same intensity of white, the intensities of the two stripes can be
altered by moving the right angled prism through the spectrum. The two
stripes are now apparently equally luminous, and I see the point of
equality is where the edge of the right angled prism is in the green.
Placing a narrow cell filled with our turbid medium in front of the slit,
I find that the equality is disturbed, and I have to allow more of the
yellow to come into the patch formed by the blue end of the spectrum, and
consequently less of it in the red end. I again establish equality.
Placing a thicker cell in front, equality is again disturbed, and I have
to have less yellow still in the red half, and more in the blue half. I
now remove the cell, and the inequality