Suddenly I wondered what those steam vents were doing. At first there had been one or two along the sea front; but now there were eight, ten, twenty, spurting high and scattered all over the volcano. A physician, Dr. Church, was standing near me, and we agreed that we disliked the outlook. Now there were forty jets, like so many ghostly locomotives run out from the Pelée roundhouse. Meanwhile, white-coated officers and scientists were scattered about in groups under the cliffs, some out of sight of Mount Pelée. We looked toward the USS _Potomac_; she had seen the steam, and her own white steam presaged quick, repeated toots of her fog horn. Pellmell the passengers came tumbling to the landing. The sailors had no sooner started the boats than two more white-coated figures appeared, and we had to put back for them. The mountain looked as though it were rifting in a hundred places preparatory to an outburst, and there were many stories of new craters forming. What we saw was actually the product of a smart rain shower, falling on red hot dry gravel; but we were to learn later about rain rill explosion. Wherever a stream rill runs down to such contact, a jet of steam forms at once. The main water gorge of the Pelée crater was blown clear of clouds as we steamed past, and we saw a cup under the summit amphitheater where a lake had been, with a pile of scaly looking hot boulders in its midst steaming violently. This crater extended into a deep gulch to the ocean, whence had come a disastrous mud flood on May 5 which buried a sugar mill. This had happened three days before the destruction of St. Pierre. Water preceded steam. The cracks under the gulch undoubtedly dipped away from the city, and from an unknown chasm athwart the gulch line ejected water and superheated steam toward the city, like a jet from a hose. This happened on May 8. The ejected material had been in dry steam, and red hot, accounting for early reports of lava at night. I saw molten rock five weeks after the _Potomac_ trip, when the crater cone was above the rim of the gorge, apparently large fragments of brown angular material resting on finer gravel. Cauliflower clouds of reddish dust spurted up the bed of the gulch below every half hour, and migrated down the gulch. This was followed by a low growl, perhaps from avalanches. The basin widened during the month, and the dome gained in height and breadth. A bright incandescent crack at night was seen to cross the heap obliquely. A sudden increase of glow was followed by a rumbling, as though the dome were heaving. Breadcrust bombs of andesite, cracked on their surface in deep gashes, and picked up on the mountain at both Pelée and Soufrière were pieces of the internal lava.