The statement is that the part of the temple of Amen-Rā, the building of which blocked B, was commenced by Thothmes III., whose date, according to Brugsch, is 1600 B.C., and continued by Amen-hetep III. (1500 B.C.). Unless, then, some other provision was made, the observations of Canopus were not continued until another shrine was built. We know that another shrine was built, that of Seti II., and that its orientation gives a date of 1350 B.C. It might have been commenced by Seti I. after the Khu-en-Aten troubles, and finished by Seti II. One is therefore tempted to ask whether we have not here one of those crucial cases which Mariette himself contemplated, in which the true foundation is so far anterior to the last restoration or the last decoration, from which, for the most part, the archæologist gets his information, that one is absolutely misled by the restorations or decorations as to the true date of the original foundation of the shrine.[49] If the archæologists are right in attributing the granite temple of Osiris (?), near the sphinx, to a date anterior to, or even contemporaneous with, the second pyramid, we have evidence that in the early dynasties the temple building in stone, and even in granite brought from Aswân, was as perfect in the matter of workmanship as in the eighteenth dynasty; and that it was not then the fashion to inscribe walls, but only statues and stelas. May it possibly be that the fashion in question came in, or reached its greatest development, during the eighteenth dynasty, and that on this account so many temples are ascribed to that period, whereas they were actually in existence before? If the prior dynasties built no temples, why did they not do so? and if they did, where are they, if some of those _inscribed_ by the eighteenth dynasty be not they? In the absence of final archæological evidence--that is, admitting Mariette's own doubt as to the mere existence of inscriptions--are there any astronomical considerations which may possibly help us? Assuming that the temples were astronomically oriented, we have _one_ registering for us the time elapsed since the original direction of the axis was laid down, in terms of the change in the obliquity of the ecliptic. We have others registering time in like manner in terms of the change due to precession, if we can get any light as to the stars towards which the temples were oriented. I have already dealt with the temple of Amen-Rā in Chapter XI., and we found a foundation date of 3700 B.C. for the original shrine, so far as the rough observations already available can be trusted. Assuming the accuracy of this determination, it is clear that we must look for stars with appropriate amplitudes between that date and say 2500 B.C.