When light is thus refracted, or has taken a new direction, it then proceeds invariably in a straight line till it meets with a different medium,[7] when it is again turned out of its course. It must be observed, however, that though we may by this means cause the rays of light to make any number of angles in their course, it is impossible for us to make them describe a curve, except in one single case, namely, where they pass through a medium, the density of which either uniformly increases or diminishes. This is the case with the light of the celestial bodies, which passes downwards through our atmosphere, and likewise with that which is reflected upwards through it by terrestrial objects. In both these cases it describes a curve of the hyperbolic kind; but at all other times, it proceeds in straight lines, or in what may be taken for straight lines without any sensible error. There are two circumstances essential to refraction. 1. That the rays of light shall pass out of one medium into another of a different density, or of a greater or less degree of resistance. 2. That they pass in an _oblique_ direction. The denser the refracting medium, or that into which the ray enters, the greater will be its refracting power; and of two refracting mediums of the same density, that which is of an oily or inflammable nature will have a greater refracting power than the other. The nature of refraction may be more particularly explained and illustrated by the following figure and description. Let ADHI fig. 2, be a body of water, AD its surface, C a point in which a ray of light BC enters from the air into the water. This ray, by the greater density of the water, instead of passing straight forward in its first direction to K, will be bent at the point C, and pass along in the direction CE, which is called the _refracted_ ray. Let the line FG be drawn perpendicular to the surface of the water in C, then it is evident that the ray BC, in passing out of air, a _rare_ medium, into a _dense_ medium, as water, is refracted into a ray CE which is _nearer_ to the perpendicular CG than the incident ray BC, and on the contrary, the ray EC passing out of a denser medium into a rarer will be refracted into CB, which is _farther_ from the perpendicular. [Illustration: _figure 2._]