The realistic motion data such as the motion capture data and physically based motion are not suitable for cartoon animations, in which the narrative is emphasized by exaggerated physical motion. Recently, several methods have been developed for converting an input motion into a cartoon-style motion. Wang et al.  proposed an innovative method that generates anticipation and follow-through effects through the convolution of a Laplacian of Gaussians (LoG) kernel. Inspired by their work, we propose a new way of superimposing non-rigid squash-and-stretch effects [Lasseter 1987] on the motion of a rigid linkage. Our key idea is to apply a time-shift filter to position data of each joint of a skeletal character individually in order to allow its cartoon character to undergo squash-and-stretch.