If we consider only the earthquakes here described, we see at once how great is the diversity in the estimated velocity of the earth-waves. On the one hand, we have a value as high as 5.2 kms. per sec. for the Charleston earthquake, and, at the other end of the scale, a value of 0.9 km. per sec. for the Hereford earthquake. Between them, and equally trustworthy, lie the estimates of 3.0 km. per sec. for the Indian earthquake, and 2.1 kms. per sec. for the Japanese earthquake and its immediate successors. It is difficult to account entirely for such discordance. Errors of observation may be responsible for a small part of the differences. The initial strength of the disturbance appears to have some effect, and the nature of the rocks traversed must be a factor of consequence when the distances in question are not very great. In the Japanese and Hereford earthquakes, all three may have combined to produce the divergent results, the distance in these cases being only 275 and 142 kms. respectively. In the Indian and Charleston earthquakes, the distances are much greater (1944 and 1487 kms.), and the variety of rocks traversed must tend to give a truer average. In the former, the result obtained (3.0 kms. per sec.) agrees so closely with the velocity of the long-period undulations of distant earthquakes as to suggest that it was these waves that were timed at the stations west of Calcutta and disturbed the magnetographs at Bombay.[84] Omitting, then, the Indian estimate, we find that, for the Japanese and Charleston earthquakes, the velocity increases with the distance as measured along the surface. To a certain extent, such a result might have been expected, had we assumed the earthquake-waves to travel along the chords joining the focus to very distant places of observation.