(2) It is evident that since summer wood shrinks more than spring wood, this fact will interfere with the even shrinkage of the log. Consider first the tangential shrinkage. If a section of a single annual ring of green wood of the shape A B C D, in Fig. 38, is dried and the mass shrinks according to the thickness of the cell walls, it will assume the shape A' B' C' D'. When a number of rings together shrink, the tangential shrinkage of the summer wood tends to contract the adjoining rings of spring wood more than they would naturally shrink of themselves. Since there is more of the summer-wood substance, the spring-wood must yield, and the log shrinks circumferentially. The radial shrinkage of the summer-wood, however, is constantly interrupted by the alternate rows of spring-wood, so that there would not be so much radial as circumferential shrinkage. As a matter of fact, the tangential or circumferential shrinkage is twice as great as the radial shrinkage. [Illustration: Fig. 38. Diagram to Show the Greater Shrinkage of Summer Cells, A, B, than of Spring Cells, C, D.] Putting these two factors together, namely, the lengthwise resistance of the pith rays to the radial shrinkage of the mass of other fibers, and second, the continuous bands of summer wood, comparatively free to shrink circumferentially, and the inevitable happens; the log splits. If the bark is left on and evaporation hindered, the splits will not open so wide. There is still another effect of shrinkage. If, immediately after felling, a log is sawn in two lengthwise, the radial splitting may be largely avoided, but the flat sides will tend to become convex, as in Fig. 39. This is explained by the fact that circumferential shrinkage is greater than radial shrinkage. [Illustration: Fig. 39. Shrinkage of a Halved Log.] If a log is "quartered,"[A] the quarters split still less, as the inevitable shrinkage takes place more easily. The quarters then tend to assume the shape shown in Fig. 40, C. If a log is sawed into timber, it checks from the center of the faces toward the pith, Fig. 40, D. Sometimes the whole amount of shrinkage may be collected in one large split. When a log is slash-sawed, Fig. 40, I, each board tends to warp so that the concave side is away from the center of the tree. If one plank includes the pith, Fig. 40, E and H, that board will become thinner at its edges than at its center, _i.e._, convex on both faces. Other forms assumed by wood in shrinking are shown in Fig. 40. In the cases A-F the explanation is the same; the circumferential shrinkage is more than the radial. In J and K the shapes are accounted for by the fact that wood shrinks very little longitudinally. [Footnote A: See _Handwork in Wood_, p. 42.] [Illustration: Fig. 40. Shapes Assumed by Wood in Shrinking.]