There is another point, however, of which much is often made as a plea for the substantial accuracy of the scientific laws and generalisations, namely that they enable us to _predict_ events. But this need not detain us long. J. S. Mill in his "Logic" has pointed out--and a little thought makes it obvious--that the success of a prediction does not prove the truth of the theory on which it is founded. It only proves the theory was good enough for that prediction. There was a time when the sun was a god going forth in his chariot every morning, and there was a time when the earth was the centre of the universe, and the sun a ball of fire revolving round it. In those times men could predict with certainty that the sun would rise next morning, and could even name the hour of its appearance; but we do not therefore think that their theories were true. When Adams and Leverrier foretold the appearance of Neptune in a certain part of the sky, they made a brief prediction to an unknown planet from the observed relations of the movements of the known planets; that does not show, however, that the grand generalisation of these movements, called the "law of gravitation," is correct. It merely shows that it did well enough for this very brief step--brief indeed compared with the real problems of Astronomy, for which latter it is probably quite inadequate. Tycho Brahé, excellent astronomer as he was, kept as we saw to the epicycle theory. He imagined that the moon's path round the earth was a fixed combination of cycle and epicycle. Kepler introduced the conception of the ellipse. Later on the motion of the perigee and other deviations compelled the abandonment of the ellipse and the supposition of an endless curve, similar to an ellipse at any one point, and maintaining a fixed mean distance from the earth, but never returning on itself or making a definite closed figure of any kind. Finally the researches of Mr. George Darwin have destroyed the conception of the fixed mean distance, and introduced that of a continually enlarging spiral. Certainly no four theories could well be more distinct from each other than these; yet if an eclipse had to be calculated for next year it would scarcely matter which theory was used. The truth is that the actual problem is so vast that a prediction of a few years in advance only touches the fringe of it so to speak; yet if the fulfilment of the prediction were taken as a proof of the theory in each of these different cases, it would lead in the end to the most hopelessly contradictory results.