During his tour in the epicentral area in the winter of 1897-98, Mr. Oldham had many opportunities for observing these earth-sounds. They were, he says, close to the lower limit of audibility, less a note than a rumble, and very like distant thunder, though sometimes they consisted of a rapid succession of short sounds, such as is caused by a cart when driven rapidly over a rough pavement. "As a rule, they began as a low, almost inaudible rumble, gradually increasing in loudness, though to a very varying degree, and then gradually dying out after having lasted anything from 5 to 50 seconds. It cannot be said that there was any connection between the duration and the loudness of the sounds, some of the most prolonged never becoming loud, and some of those which lasted a shorter period being as loud as ordinary thunder at a distance of two or three miles." Mr. Oldham records an interesting fact in connection with the distribution of the earth-sounds. At Naphak, in the Garo hills and about five miles south of Samin, 48 distinct rumbles were heard during 23 hours on January 21-23, 1898, only seven of them being accompanied by a perceptible shock. At Samin, which was visited next, they were much less frequent, not more than 8 or 10 a day, and most of them attended by tremors. At Damra, a few miles to the north-east, they again became frequent; while, in the Chedrang valley, very few were heard, and only a small proportion of them were unaccompanied by sensible shocks. In the next section, it will be seen that the most conspicuous fault-scarps known in the epicentral area pass close by Samin and along the Chedrang valley. Thus, though the statement perhaps requires further confirmation, it would appear that earth-sounds were more common where the surface of the ground had been merely bent than where fractures extended right up to the surface. STRUCTURAL CHANGES IN THE EPICENTRAL AREA. We come now to the important features which assign the Indian earthquake to a small class apart from nearly every other shock. Most earthquakes are due to movements that are entirely deep-seated. If strong enough, they may precipitate landslips or fissure the alluvial soil near river-channels. In the Neapolitan, Andalusian, and Charleston earthquakes, there were many such effects of the shock within the meizoseismal areas. In all three, however, the disturbances produced were superficial; no structural change, no fissuring that did not die out rapidly downwards, was in any place perceptible. In the Riviera earthquake, the seismic sea-waves point to a small displacement of the ocean-bed; but it is only in the long fault-scarp of the central Japanese plain that we find a rival of the mountain-making movements that gave rise to the Indian earthquake.