§ IX. We have thus arrived at the idea of a vertical frame work of subordinated bars, supported by cross bars at the necessary intervals, and the only remaining question is the method of insertion into the aperture. Whatever its form, if we merely let the ends of the bars into the voussoirs of its heading, the least settlement of the masonry would distort the arch, or push up some of its voussoirs, or break the window bars, or push them aside. Evidently our object should be to connect the window bars among themselves, so framing them together that they may give the utmost possible degree of support to the whole window head in case of any settlement. But we know how to do this already: our window bars are nothing but small shafts. Capital them; throw small arches across between the smaller bars, large arches over them between the larger bars, one comprehensive arch over the whole, or else a horizontal lintel, if the window have a flat head; and we have a complete system of mutual support, independent of the aperture head, and yet assisting to sustain it, if need be. But we want the spandrils of this arch system to be themselves as light, and to let as much light through them, as possible: and we know already how to pierce them (Chap. XII. § VII.). We pierce them with circles; and we have, if the circles are small and the stonework strong, the traceries of Giotto and the Pisan school; if the circles are as large as possible and the bars slender, those which I have already figured and described as the only perfect traceries of the Northern Gothic.[58] The varieties of their design arise partly from the different size of window and consequent number of bars; partly from the different heights of their pointed arches, as well as the various positions of the window head in relation to the roof, rendering one or another arrangement better for dividing the light, and partly from æsthetic and expressional requirements, which, within certain limits, may be allowed a very important influence: for the strength of the bars is ordinarily so much greater than is absolutely necessary, that some portion of it may be gracefully sacrificed to the attainment of variety in the plans of tracery--a variety which, even within its severest limits, is perfectly endless; more especially in the pointed arch, the proportion of the tracery being in the round arch necessarily more fixed.