We ask ourselves to what it is due that this rolling lasts so long. There are several causes. The first is due to the length of lightning and the difference in speed between sound and light. Let us suppose, for example, a flash of lightning, AE, 11,000 metres long. The observer stationed at O, underneath extremity E of the lightning (which is one kilometre high), will see the lightning in its full length in one indivisible instant. The sound will form itself also at the same instant all along the line of lightning, but the sound-waves will only reach the ear of the observer at different times. That which starts at point E, the nearest, will arrive in 3 seconds, sound travelling about 337 metres a second. That which is formed at the same moment at point D, 2000 yards from point O, will take double the time to arrive. That which comes from point C will not arrive for 12 seconds. The sound formed at B will not arrive until the time necessary to cover 8 kilometres--that is to say, not for 23 seconds--and the sound formed at A will only reach after 32 seconds. Thus the rolling will have lasted more than half a minute from start to finish. [Illustration: DIAGRAM EXPLAINING THE DURATION OF THE SOUND OF THUNDER.] And if, which is very often the case, the astronomer is not exactly under one of the extremities of the lightning, but at some other point in its course, he first hears a clap, then an increased noise, then a diminution. In fact, in this case, the sound which leaves point D just overhead, which is 1000 metres off, arrives alone in 3 seconds, but the sounds formed from D to E on one side, and from C to D on the other, arrive at the same time, having joined each other, taking 9 seconds, which is the necessary time to come from 1000 to 3000 metres. The sounds beyond C arrive and depart according to distance, as in the preceding example, and the thunder has lasted 23 seconds instead of 32 seconds. [Illustration: COMMENCEMENT, AUGMENTATION, AND DIMINUTION OF THE INTENSITY OF THUNDER.] I must add that lightning is never straight, but always crooked. The length of time the thunder rolls has nothing to do with the distance of the cloud where the phenomenon begins. It is proportionate to the length of the lightning with which it is associated. The rolling is often still more prolonged by a succession of small discharges, which follow each other very rapidly between the stormy clouds; by the zigzags and ramifications of the lightning caused by the hygrometrical diversity of the different beds of air; by the echoes repeated by the mountains, the earth, the water, and the clouds themselves--to all which must be added also the interferences caused by the encounter of the different systems of sound-waves.