Efforts to test hypotheses about small-scale shifts in development (tinkering) that can only be observed in the fossil record pose many challenges. Here we use the origin of modern human craniofacial form to explore a series of analytical steps with which to propose and test evolutionary developmental hypotheses about the basic modules of evolutionary change. Using factor and geometric morphometric analyses of craniofacial variation in modern humans, fossil hominids, and chimpanzee crania, we identify several key shifts in integration (defined as patterns of covariation that result from interactions between components of a system) among units of the cranium that underlie the unique shape of the modern human cranium. The results indicate that facial retraction in modern humans is largely a product of three derived changes: a relatively longer anterior cranial base, a more flexed cranial base angle, and a relatively shorter upper face. By applying the Atchley-Hall model of morphogenesis, we show that these shifts are most likely the result of changes in epigenetic interactions between the cranial base and both the brain and the face. Changes in the size of the skeletal precursors to these regions may also have played some role. This kind of phenotype-to-genotype approach is a useful and important complement to more standard genotype-to-phenotype approaches, and may help to identify candidate genes involved in the origin of modern human craniofacial form. J. Exp. Zool. (Mol. Dev. Evol.) 302B:284–301, 2004. © 2004 Wiley-Liss, Inc.