accurately to a sharp and definite point. Perhaps the law
of refraction was not quite accurate, but only an approximation. So he
bought a prism to try the law. He let in sunlight through a small round
hole in a window shutter, inserted the prism in the light, and received
the deflected beam on a white screen; turning the prism about till it
was deviated as little as possible. The patch on the screen was not a
round disk, as it would have been without the prism, but was an
elongated oval and was coloured at its extremities. Evidently
refraction was not a simple geometrical deflection of a ray, there was a
spreading out as well.

[Illustration: FIG. 63.--A prism not only _deviates_ a beam of sunlight,
but also spreads it out or _disperses_ it.]

Why did the image thus spread out? If it were due to irregularities in
the glass a second prism should rather increase them, but a second prism
when held in appropriate position was able to neutralise the dispersion
and to reproduce the simple round white spot without deviation.
Evidently the spreading out of the beam was connected in some definite
way with its refraction. Could it be that the light particles after
passing through the prism travelled in variously curved lines, as
spinning racquet balls do? To examine this he measured the length of the
oval patch when the screen was at different distances from the prism,
and found that the two things were directly proportional to each other.
Doubling the distance of the screen doubled the length of the patch.
Hence the rays travelled in straight lines from the prism, and the
spreading out was due to something that occurred within its substance.
Could it be that white light was compound, was a mixture of several
constituents, and that its different constituents were differently bent?
No sooner thought than tried. Pierce the screen to let one of the
constituents through and interpose a second prism in its path. If the
spreading out depended on the prism only it should spread out just as
much as before, but if it depended on the complex character of white
light, this isolated simple constituent should be able to spread out no
more. It did not spread out any more: a prism had no more dispersive
power over it; it was deflected by the appropriate amount, but it was
not analysed into constituents. It differed from sunlight in being
simple. With many ingenious and beautifully simple experiments, which
are quoted in full in several books