For at least a century biologists have been talking, mostly in a black-box sense, about developmental mechanisms. Only recently have biologists succeeded broadly in fishing out the contents of these black boxes. Unfortunately the view from inside the black box is almost as obscure as that from without, and developmental biologists increasingly confront the need to synthesize known facts about developmental phenomena into mechanistic descriptions of complex systems. To evolutionary biologists, the emerging understanding of developmental mechanisms is an opportunity to understand the origins of variation not just in the selective milieu but also in the variability of the developmental process, the substrate for morphological change. Ultimately, evolutionary developmental biology (EvoDevo) expects to articulate how the diversity of organic form results from adaptive variation in development. This ambition demands a shift in the way biologists describe causality, and the central problem of EvoDevo is to understand how the architecture of development confers evolvability. We argue in this essay that it makes little sense to think of this question in terms of individual gene function or isolated morphometrics, but rather in terms of higher-order modules such as gene networks and homologous characters. We outline the conceptual challenges raised by this shift in perspective, then present a selection of case studies we believe to be paradigmatic for how biologists think about modularity in development and evolution. J. Exp. Zool. (Mol. Dev. Evol.) 285:307-325, 1999.