The evidence thus obtained was rendered still more conclusive by the discovery on the under side of several fragments of paintings which had evidently been intended for the decoration of a ceiling."[200] It is clear that these curvilinear and frescoed blocks were fragments of a tunnel vault that had fallen in; and their existence explains the great thickness given by the Assyrian constructor not only to his outer walls, but to those that divided room from room. The thinnest of the latter are hardly less than ten feet, while here and there they are as much as fifteen or sixteen. As for the outer walls they sometimes reach a thickness of from five and twenty to thirty feet.[201] The climate is insufficient to account for the existence of such walls as these. In the case of the outer walls such a reason might be thought, by stretching a point, to justify their extravagant measurements, but with the simple partitions of the interior, it is quite another thing. This apparent anomaly disappears, however, if we admit the existence of vaults and the necessity for meeting the enormous thrust they set up. With such a material as clay, the requisite solidity, could only be given by increasing the mass until its thickness was sometimes greater than the diameter of the chambers it inclosed. M. Place lays great stress upon the disproportion between the length and width of many of the apartments. There are few of which the greater diameter is not at least double the lesser, and in many cases it is four, five, and even seven times as great. He comes to the conclusion that these curious proportions were forced on the Assyrians by the nature of the materials at their disposal. Such an arrangement must have been destructive to architectural effect as well as inconvenient, but a clay vault could not have any great span, and its abutments must perforce have been kept within a reasonable distance of each other. Taken by itself, this argument has, perhaps, hardly as much force as M. Place is inclined to give it. Doubtless the predilection for an exaggerated parallelogram agrees very well with the theory that the vault was in constant use by Mesopotamian architects, but it might be quoted with equal reason by the supporters of the opposite hypothesis, that of the timber roof. Our best reason for accepting all these pieces of evidence as corroborative of the view taken by MM. Flandin, Loftus, Place, and Thomas is, in the first place, the incontestable fact that the entrances to the town of Khorsabad were passages roofed with barrel vaults; secondly, the presence amid the debris of the fragmentary arches above described; thirdly, the depth of the mass of broken earth within the walls of each chamber; finally, the singular thickness of the walls, which is only to be satisfactorily explained by the supposition that the architect had to provide solid abutments for arches that had no little weight to carry.