It is impossible to determine exactly at what distance from the earth's surface the air ceases altogether, for it grows continually more and more rarefied. There are, however, two distinct methods of ascertaining the distance beyond which it can be said practically not to exist. One of these methods we get from twilight. Twilight is, in fact, merely light reflected to us from those upper regions of the air, which still continue to be illuminated by the sun after it has disappeared from our view below the horizon. The time during which twilight lasts, shows us that the atmosphere must be at least fifty miles high. But the most satisfactory method of ascertaining the height to which the atmosphere extends is from the observation of meteors. It is found that these bodies become ignited, by the friction of passing into the atmosphere, at a height of about 100 miles above the surface of the earth. We thus gather that the atmosphere has a certain degree of density even at this height. It may, indeed, extend as far as about 150 miles. The layer of atmosphere surrounding our earth acts somewhat in the manner of the glass covering of a greenhouse, bottling in the sun's rays, and thus storing up their warmth for our benefit. Were this not so, the heat which we get from the sun would, after falling upon the earth, be quickly radiated again into space. It is owing to the unsteadiness of the air that stars are seen to twinkle. A night when this takes place, though it may please the average person, is worse than useless to the astronomer, for the unsteadiness is greatly magnified in the telescope. This twinkling is, no doubt, in a great measure responsible for the conventional "points" with which Art has elected to embellish stars, and which, of course, have no existence in fact. The phenomena of _Refraction_,[13] namely, that bending which rays of light undergo, when passing _slant-wise_ from a rare into a dense transparent medium, are very marked with regard to the atmosphere. The denser the medium into which such rays pass, the greater is this bending found to be. Since the layer of air around us becomes denser and denser towards the surface of the earth, it will readily be granted that the rays of light reaching our eyes from a celestial object, will suffer the greater bending the lower the object happens to be in the sky. Celestial objects, unless situated directly overhead, are thus not seen in their true places, and when nearest to the horizon are most out of place. The bending alluded to is upwards. Thus the sun and the moon, for instance, when we see them resting upon the horizon, are actually _entirely_ beneath it. When the sun, too, is sinking towards the horizon, the lower edge of its disc will, for the above reason, look somewhat more raised than the upper. The result is a certain appearance of flattening; which may plainly be seen by any one who watches the orb at setting.