Thus, concluded Einstein, if the principle of equivalence holds in all of physics, light rays from distant stars that pass close to the sun on their way to the earth should bend toward the sun. This prediction was brilliantly confirmed in 1919 by a party of British astronomers observing a total solar eclipse in Africa. With the obscuring sunlight extinguished by the moon, stars near the edge of the solar disk were seen to be displaced about 1.75 seconds of arc away from the sun.
Let us next consider another type of accelerated motion-uniform rotation. (A body moving at constant speed on a circular path is accelerated because of its continuous change of direction.) Imagine a merry-go-round with a curtain around it so that people inside cannot tell by looking at the surroundings that it is rotating. If the merry-go-round is turning, the observers will be aware of centrifugal force, which pushes them out toward the rim. A ball placed on the platform will roll away from the center. The centrifugal force acting on any object on the platform will be proportional to the inertial mass of the object, so that here again the effect of accelerated motion can be considered as equivalent to that of a gravitational field. It is a peculiar field, to be sure; it is quite different from the field on the surface of the earth or of any other spherical body. The force is directed away from the center of the system, not toward it; and instead of decreasing as the square of the distance from the center, it increases proportionately to that distance. Moreover, the field has cylindrical symmetry around a central axis rather than spherical symmetry around a central point. Nevertheless, the equivalence principle holds, and the field can be interpreted as being caused by gravitating mass distributed at large distances all around the symmetry axis.
How will light propagate through this field? Suppose a light source that sends out rays in all directions is located at a point, A, on the periphery of the rotating disk, and is observed at a second point, B, also on the periphery. According to the basic law of optics, light always propagates along the shortest path. But what is the shortest path between A and B? To measure the length of various lines connecting the points A and B the observer uses the old-fashioned but always safe method of counting the number of yardsticks that can be placed end to end along the line.
As we watch the experiment from outside, we recall the special theory of relativity, which tells us that moving yardsticks shrink in the direction of their motion. Therefore we see that if the observer measures along the "true" straight line from A to B, his sticks will contract and he will need more of them to measure that line than if the platform were not moving. Now an interesting point arises. The closer a yardstick is to the center of the merry-go-round, the less its linear velocity and therefore the smaller its contraction. By bending the line of yardsticks toward the center the observer decreases the number he needs to go from A to B. Although the "actual" distance is somewhat longer, the increase is more than compensated for by the smaller shrinkage of each yardstick. A light ray following this shortest path, heading inward at the start of its journey and then bending outward, can be considered to be deflected by the apparent gravitational field, which is directed radially outward.
Before leaving the merry-go-round let us consider one more experiment. A pair of identical clocks are placed on the platform, one near the center and the other at the edge. As in the case of the yardsticks, the outer clock is moving faster than the inner one, and again special relativity predicts a difference in their behavior. In addition to causing yardsticks to contract, motion makes clocks run slow. Therefore the outer clock will lose time with respect to the inner one. Now the observer who interprets the acceleration effects in terms of a gravitational field will say that the clock placed in the higher gravitational potential (that is, in the direction in which gravitational force acts) runs slower.