glasses in Newton's experiment was 1/250,000 of an inch, but Sir Wm. Thomson has shown that the separation is caused by shreds or particles of dust. However, if this separation is equal throughout, we have the phenomena as described; but if the dust particles are thicker under one side than the other, our phenomena will change to broad parallel bands as in Fig. 8, the broader the bands the nearer the absolute parallelism of the plates. In Fig. 7 let _a_ and _b_ represent the two plates we are testing. Rays of white light, _c_, falling upon the upper surface of plate _a_, are partially reflected off in the direction of rays _d_, but as these rays do not concern us now, I have not sketched them. Part of the light passes on through the upper plate, where it is bent out of its course somewhat, and, falling upon the _lower_ surface of the upper plate, some of this light is again reflected toward the eye at _d_. As some of the light passes through the upper plate, and, passing through the film of air between the plates, falling on the upper surface of the _lower_ one, this in turn is reflected; but as the light that falls on this surface has had to traverse the film of air _twice_, it is retarded by a certain number of half or whole wave-lengths, and the beautiful phenomena of interference take place, some of the colors of white light being obliterated, while others come to the eye. When the position of the eye changes, the color is seen to change. I have not time to dwell further on this part of my subject, which is discussed in most advanced works on physics, and especially well described in Dr. Eugene Lommel's work on "The Nature of Light." I remarked that if the two surfaces were perfectly _plane_, there would be one color seen, or else colors of the first or second order would arrange themselves in broad parallel bands, but this would also take place in plates of slight curvature, for the requirement is, as I said, a film of air of equal thickness throughout. You can see at once that this condition could be obtained in a perfect convex surface fitting a perfect concave of the same radius. Fortunately we have a check to guard against this error. To produce a perfect plane, _three surfaces must_ be worked together, unless we have a true plane to commence with; but to make this true plane by this method we _must_ work three together, and if each one comes up to the demands of this most rigorous test, we may rest assured that we have attained a degree of accuracy almost beyond human conception. Let me illustrate. Suppose we have plates 1, 2, and 3, Fig. 11. Suppose 1 and 2 to be accurately convex and 3 accurately concave, of the same radius. Now it is evident that 3 will exactly fit 1 and 2, and that 1 and 2 will separately fit No. 3, _but_ when 1 and 2 are placed together, they will only touch in the center, and there is no possible way to make three plates coincide when they are alternately tested upon