Abstract: As an attractive scenic spot, the Jiudongtian cave hall in Dafang County, Guizhou province should be worth world praise and notice. It is thought of great academic significance by foreign and domestic scientists. The Jiudongtian (nine sky holes) sinking stream is located along a nick point in the middle reaches of the northern branch of Wujiang river, Yangtze Basin. It is named “Jiudongtian” by local people because there are nine karst windows along its 2.8 km stream course. In 1987 the sinking stream was first explored by a China-New Zealand speleological expedition led by P.W. Williams and the professors from the Guizhou Normal University. P.W. Williams believed that: “Jiudongtian not only belongs to Guizhou or China but also to the world; It is not inferior to any well-known karst spot in the world……; specifically, it represents a nice natural model for combined formation of surface and subsurface karst landform.” Later on, Prof. Yang Mingde from the Guizhou Normal University presented a classic study on the Jiudongtian karst. But both of them did not address much to origin of those cave halls. This paper will add to their ideas with a demonstration showing the role of the Coriolis force and dominant erosion to the wall on the right of the cave hall.（1）This paper give some new ideas about cave hall formation. First, cave hall must originate from a prior phreatic cave,as believed W.M.Davis. Second, owing to that the moving energy and dissolution capacity of flood water are of severel orders of magnitude larger, the zone of seasonal water level fluctuation becomes a place on the walls where the most intensive denudation occurs. Third, as a result notches are forming deeply cutting into both the cave walls. Fourth, at least 3 sub-processes and 5 sub-processes of lower order are subsequintly involved in formation of cave hall. Fifth, coupling of sub-prosesses,3 types of molti-interrelation have already been recognized.（2）Reloading fissures alter the opening state of fissure system, making its permeability more even. This in turn leads to even distribution of speleothems along top and walls of the cave. There are some obvious evidences in morphology of the cave halls for dominant rightward erosion (in Northern hemisphere) by the Tiudongtian subterranean stream. Firstly, there is a unique pattern of bedrock outcrop and speleothem distribution， 90% area on the left wall is covered with speleothems, more than 90% on the right wall with bedrock; while in the middle of the right wall there is a branch cave passage full of speleothems. Secondly, rock collapse mainly appears on the right wall. Thirdly, a vertical fissure-shaped cave can be found on the bottom or at the top of some halls along the wall on the left hand. All of these strongly infer that a vertical phreatic fissure-shaped cave occurred there originally near the left wall. When evolving and entering into the zone of seasonal fluctuation in groundwater table the fissure-shaped cave must has been excavated intensively by flood water, which is estimated as large as 400 m3/s in discharge. The Coriolis force generated by such a flood flow would be 10～100 times larger than that of an underground river of 40 m3/s or 4 m3/s in discharge respectively. It must force the right wall moves backward to form the huge space of the hall. The rightward erosion for a gently dip bedding plane cave is slightly different to the vertical fissure-shaped one. Direction for cave floor extension is controlled by two factors, namely the rate of rightward erosion and the rate of descending of the local denudation base. When denudation base is descending rapidly, both sorts of caves might form and layered cave system may takes place. While in the case of slow and intermittent tectonic descending, a cave with slight-dipping bedding will form a gently inclined floor along the bedding, as is seen in the Fengyu cave, Guilin. (3) Presented above is a preliminary result and needs to be verified by detailed study in future.