Colonel Collinson, however, informs me that they bracket outwards, at the rate of 1 foot in 10, and he calculates that they would meet at a height of 30 feet so nearly that they could be closed by a single stone. He, however, overlooks the fact that all these horizontally-constructed domes, whether in Greece, or Italy, or Sardinia, are curvilinear, their section being that of a Gothic pointed arch, and consequently, if corbelling forward at the rate of 1 in 10 near the springing, they would certainly meet in this chamber at 15 or 20 feet from their base. When we recollect that before the Trojan war the Pelasgic architects of Greece roofed chambers 50 and 60 feet in diameter (vide ante, page 33), we should not be surprised at the Maltese architects grappling with apses of 20 feet span. This has generally been admitted as easy, but several authors have been puzzled to think how the flat spaces joining the two apses could have been so roofed. A careful examination of the plans of the Maltese building seems to make this easy. Looking, for instance, at the plan of Mnaidra, a retaining wall will be observed on the extreme right, which is a segment of a circle 75 feet in diameter, and continuing it all round, it encloses both chambers. If a similar circle is drawn round the left-hand chambers, it equally encloses them, and the circles osculate, or have one party wall at a point where there is the group of cells. This granted, it is easy to see that the external form of the roof was a stepped cone, covering the inner roofs, and so avoiding the ridges and hollows which would have rendered independent roofing impracticable. The external appearance of the building would thus have been that of two equal cones joined together, and rising probably to a height of 50 feet above their springing. To erect such a cone on an enclosing wall only 8 or 10 feet thick may appear at first sight a little difficult for such rude builders as the Maltese were when they erected these domes, but when we recollect that the cone was divided into two by a cross party wall, which may have been carried the whole height, all difficulty vanishes.
When we apply these principles to the ruins at Hagiar Khem, their history becomes plain at once. Originally the monument seems to have consisted of a single pair of chambers of the usual form, A and B of the accompanying plan; but extension becoming
