It seems to have been a common requirement throughout the pueblo country that the kiva should be wholly or partly underground. So strong was this requirement in Tusayan that the occurrence of natural clefts and fissures in the rock of the mesa top has dictated the location of the kivas often at some distance from the houses. But in De Chelly there were some sites where the requirement could not be filled without extensive rock excavation wholly beyond the power of the builders. Here then it seems that other requirements were strong enough to overcome the ceremonial necessity for partly subterranean structures, for examples of that kind are comparatively rare. In all of the ruins on the canyon bottom the requirement could be filled, and as many of the villages on defensive sites were constructed after the site itself had been partly filled up with loose debris, it could also be filled in those cases. There are also instances where the bottom of the kiva rests directly on the rock, while outside the walls the site was covered deep with artificial debris. But it would be difficult to determine what was the surface of the ground when the kiva was in use. The size and character of the kivas in De Chelly, and their relations to the other rooms about them, are shown in the ground plans preceding. Some have walls still standing to a height of 6 feet above the ground, but this could not have been the total height. Dr H. C. Yarrow, U.S.A., in 1874 examined one of the five large circular kivas in Taos. He states[18] that it was 25 or 30 feet in diameter, arched above, and 20 feet high. Around the wall, 2 feet from the ground, there was a hard earthen bench, and in the center a fireplace about 2 by 3 feet. [Illustration: Fig. 70--Part of a kiva in ruin No. 31.] Entrance to the kivas is invariably from the roof by a ladder. This appears to be a ceremonial requirement. Doorways at the ground level are not only unknown, but also impracticable; but in De Chelly there are some puzzling features which might easily be mistaken for such doorways. The principal kiva in the ruin, which occurs at the point marked 10 on the map, and described above (page 123, figure 24), is on the edge of the ledge, and its outer wall is so close as to make a passage difficult, although not impossible. At the point where the curved wall comes nearest the cliff there is a narrow gap or opening, not more than 15 inches wide. In front of this there appears to be a little platform on the sloping rock, 2 feet long, 10 inches wide, and now about a foot high. At first sight this would be taken for a doorway so arranged that access to the kiva could be obtained only from below; but a closer examination shows that this was probably only what remains of a chimney-like structure, such as those described later. [Footnote 18: Wheeler Survey Reports, vol. VII, Archaeology, p. 327.]