The truth of the matter seems to be that, after all their experience, the architects had got back to precisely the point from which they started, namely, the necessity of a square space for the erection of a satisfactory intersecting vault. The Romans saw this, and never swerved from it. The side-aisles of all cathedrals and all cloisters adhered to it throughout; and, when it was departed from in the wider central aisles, it always led to an awkwardness that was hardly ever successfully conquered. In some instances, as in the retro-choir at Peterborough (1438-1528), two windows are boldly but awkwardly included in one bay (Woodcut No. 818), and the compartments are so nearly square that the difficulty is not very apparent, but it is sufficient to injure considerably the effect of what would otherwise be a very beautiful roof. In Henry VII.’s Chapel the difficulty was palliated, not conquered, by thrusting forward the great pendants of the roof and treating them as essential parts of the construction, and as if they were supported by pillars from the floor instead of by brackets from the wall. By this means the roof was divided into rectangles more nearly approaching squares than was otherwise attainable; but it is most false in principle, and, in spite of all its beauty of detail, cannot be considered successful. [Illustration: 819. Choir Arches of Oxford Cathedral. (Cath. Hb.)] Strange as it may appear from its date, the most satisfactory roof of this class is that erected by Cardinal Wolsey in the beginning of the 16th century over the choir of Oxford Cathedral. In this instance the pendants are thrust so far forward and made so important that the central part of the roof is practically quadripartite. The remaining difficulty was obviated by abandoning the circular horizontal outline of true fan-tracery, and adopting a polygonal form instead. As the whole is done in a constructive manner and with appropriate detail, this roof— except in size—is one of the best and most remarkable ever executed. The true solution of the difficulty, in so far as the vault was concerned, would have been to include two bays of the side-aisles in one of the centre; but this would have necessitated a rearrangement of both plan and exterior to an extent the architects were not then prepared to tolerate, and it never was attempted, except perhaps in the instance of the retro-choir at Peterborough (Woodcut No. 818). Had it been done in King’s College Chapel at Cambridge (Woodcut No. 846), it would have been in every respect an immense improvement. At present the length of King’s College Chapel is too great for its other dimensions. Had there been six bays instead of twelve, its apparent length would have been considerably diminished, and the variety introduced by this change would have relieved its monotony without detracting from any of the excellent points of design it now possesses.