of every thousand rays emanating from the source, each screen transmits or stops the following quantities:
| Order. | Rays transmitted. | Rays stopped. |
| 1. | 484 | 516 |
| 2. | 380 | 620 |
| 3. | 303 | 697 |
By means of these data we obtain as the values of the calorific losses, considered with reference to the quantities of rays which present themselves successively to pass through the three equal layers into which we may suppose the last screen divided,
0·516 0·215 0·203.
These losses are still greater than those preceding, because of the badness of the material and the greater thickness of the layers, but they are still in a decreasing progression. Thus the diminution continues beyond 54 millimetres.
To compare this diminution with that which took place in the last screen in the preceding experiments we must multiply 0·012 (the difference between 0·215 and 0·203) by 2·068, and divide the product by 27. In this way we obtain the mean diminution for a thickness of 2mm·068 in passing from 54 to 81 millimetres, which is nearly 0·001; in the preceding experiment it was fifteen times as much while the rays passed through the same layer of 2mm·068 placed at a distance of 6 millimetres. The difference would be still greater if we had used very transparent layers of glass, such as flakes of the glass of a mirror attenuated.
Nevertheless I had some doubts as to the homogeneity of the glass: I was afraid that the striæ might not be equally distributed over all the points of the mass. But not being able to procure large pieces of this material entirely free from defects, I thought that analogous experiments performed with liquids might answer quite as well. In employing these instead of glass there was, in case of success, the additional advantage of extending the law of calorific transmission by making it independent of the physical constitution of the medium.
I procured therefore several copper troughs, of the same breadth but of different lengths, bounded at each end by a glass plate. These I placed successively between the perforated screen and the pile in such a manner that the anterior glass plate was quite near the screen, the distance of which remained constantly the same. The common section of the troughs was much larger than the central aperture of the screen; the reflexions on the lateral faces could not take place, and the only rays that entered a little out of the perpendicular direction reached the anterior surface of the pile. The lamp was moved up so near that the needle of the galvanometer exhibited a deviation of 30° through the two glass plates of each trough. The radiation was then intercepted, the trough filled with
