IT may be remembered that in 1890, the German Marine Observatory tested some three thousand running lights in use on board ships, and found two-thirds of them defective. Further tests of the visibility of lights of known candle-power were made by the German Committee last year, and some of the results obtained are noted in a leaflet just distributed to seamen by the U.S. Weather Bureau. The law of emission for a white light is that its visibility is proportional to the square root of its. candle-power, and the results of the experiments by the Com mittee closely follow the law, the departures being no greater than the estimated errors of position of the vessel. The mean, of a large number of observations gave as the distance at which a white light of one candle-power became visible 1˙40 miles for a dark clear night, and 1˙00 mile for a rainy one. Experiments undertaken in America, after the International Maritime Congress in 1889, gave the following results in very clear weather: A light of I candle-power was plainly visible at 1 nautical mile, and one of 3 candle-power at 2 miles. A 10 candle-power light was visible with an ordinary binocular at 4 miles; one of 29 candles faintly at 5, and one of 33 candles visible without difficulty at the same distance. On a second evening, exceptionally clear, a white light of 3˙2 candle-power could readily be distinguished at 3, one of 5˙6 at 4, and one of 17˙2 at 5 miles. The Dutch governmental experiments, conducted at Amsterdam, gave the following results: A light of 1 candle-power was visible at 1 nautical mile; 3˙5 at 2, and 16 at 5 miles. Experiments with green lights gave 0.80 as the distance in miles at which a green light of a single candle-power is just visible. The candle-power required for a green light to be visible at 1, 2, 3, and 4. nautical miles was 2, 15, 51, and 106, respectively. The American experiments before referred to give for green light: 3˙2 candle-power fairly visible at 1 mile, and 28˙5 clearly at 2 miles, these results being, however, from a limited number of experiments. The German trials were much more numerous. The extraordinarily rapid diminution of the visibility of the green light with the distance, even in good observing weather, and the still more rapid decrease in rainy weather of a character which will but slightly diminish the intensity of a white light, show that it is of the utmost importance to select for the glass a shade of colour which will interfere with the intensity of the light as little as possible. The shade recommended is a clear blue-green. Yellow-green and grass-green should not be employed, as they become indistinguishable from white at a very short distance. For the red, a considerably wider range is allowable, but a coppery-red is said to be the best.