Determining the connection between ontogeny and phylogeny continues to be a major theme in biology. However, few studies have combined dissection of pattern and process that lead to transformation of complex morphological structures. Here we examine the patterns and processes of shape change in a model system-the gastropod radula. This system is a simple one having only two processes: initial secretion and postsecretional movement of teeth. However, it produces a tremendous amount of shape variability and fusion patterns. To determine both pattern and mechanism of shape change in an evolutionary context, we use three complementary approaches and datasets. First, we use a phylogenetic hypothesis to determine the polarity of developmental events. Second, we perform a morphometric analysis of shape change using relative warp analysis that allows us to locate and compare the direction and magnitude of ontogenetic and phylogenetic shape divergence. These comparisons are the basis for testing hyptheses of heterochrony and heterotopy, and we show how our results do not conform to expectations of pure heterochrony. The rejection of heterochrony as a hypothesis is based on empirically demonstrating (1) initial shape differs in each taxon; (2) a single dimension of shape variability does not simultaneously describe ontogenetic and evolutionary shape changes; and (3) a significantly different shape and size covariance between taxa. This rejection is probably based on spatial changes in initial conditions and not spatial changes caused by the process itself. Finally, we construct a mechanistic model that explains how shape change happens based on the sequence of events during ontogeny. By using the parameters in the model as characters in the phylogenetic dataset, we show that different parts of the system have arisen at different times and become co-opted into the process. By integrating our analyses together we show that spatial process parameters can be responsible for our nonspatial patterns and that different ontogenetic processes can create similar end morphologies.