Ellis' Accoiinl of Kilaitra. 



41 



the side walls of the dome yielding before the pressure of the enclosed lava, and quietly 

 allowing the passage of the molten current to the sea. It was, I confess, a little 

 puzzling to me, when I gave this opinion forty years ago, how the crater could be 

 leaking out to any extent and yet be so active on the surface as Ellis describes it, but 

 I have since seen more of the working of this wonderful place and am no longer 

 puzzled. I have seen the bottom drop out and leave the Halemaumau quite empty, 

 and the process was a silent one with no activity on the surface. I have again seen 

 the surface in the pit fall so rapidly that there could be no doubt that the supply was 



Fig. 36. VIKW OF KII.AUEA. ELI.IS, IN POLYNESIAN RESEARCHES. 



tapped, and some interruption intervening; the subsidence not only stopped but the 

 action on the surface became more violent than for a long time. This last seems to 

 have been the case in August, 1S23, when he saw what he describes as follows: 



Immediately before us yawned au immense gulf, in the form of a crescent, upwards of two 

 miles in length, about a mile across, and apparently eight hundred feet deep. The bottom was filled 

 with lava, and the .southwest and northern parts of it were one vast flood of liquid fire in a state of 

 terrific ebullition, rolling to and fro its fiery surge and flaming billows. Fifty -one craters of varied 

 form and size rose like so many conical islands from the surface of the burning lake. Twenty-two 

 constantly emitted columns of gray smoke, or pyramids of brilliant flame, and many of them at the 

 same time vomited from their ignited mouths streams of florid lava which rolled in blazing torrents 

 down their black indented sides into the boiling mass below. 



The sides of the gulf before us were perpendicular for about four hundred feet, when there 

 was a wide horizontal ledge of solid black lava of irregular breadth, but extending completely round. 

 Beneath this black ledge the sides sloped to the centre, which was, as nearly as we could judge, 



[419I 



