In their seminal theory on modularity in design, Baldwin and Clark (2000, Design Rules: The Power of Modularity. MIT Press) focus on the macro-level: They show that modularity limits the degree to which changes propagate throughout a complex system, as constituted by how interdependencies among all parts pattern into a nested design structure with shared parts at the top transmitting functionalities via central parts down to the bottom. In this paper, we shift the focus to the microlevel structures of interdependencies that reflect the local decisions of designers with limited rationality. Such decisions relate to five fundamental types of interdependence structures among three parts: the design motifs. Following Baldwin and Clark (2000, Design Rules: The Power of Modularity. MIT Press), we assume that value-seeking designers weigh costs and benefits when creating motifs. We develop a new design motif theory to understand the sources of macro-level modularity. We empirically explore this theory using a sample of more than 20,000 design structures of a complex software. We observe a stable, recurring design motif signature: high-value design motifs occur more frequently than low -alue ones because of each motif’s idiosyncratic microstructure. The common resource motif occurs most frequently because of its low costs and benefits for reliability in design. The sequential loop motif holds a critical role: it offers unique benefits for innovating system functionalities, but when placed in the center, it inflates the propagation of changes. This explains why central rather than shared parts are a potential source of innovation at the expense of modularity. These findings contribute to the research and practice of modularity in design.