The subdivision of the embryo into physically distinct regions is one of the most fundamental processes in development. General hypotheses for tissue separation based on differential adhesion or tension have been proposed in the past, but with little experimental support. During the last decade, the field has experienced a strong revival, largely driven by renewed interest in biophysical modeling of development. Here, I will discuss the various models of boundary formation and summarize recent studies that have shifted our understanding of the process from the simple juxtaposition of global tissue properties to the characterization of local cellular reactions. Current evidence favors a model whereby separation is controlled by cell surface cues, which, upon cell-cell contact, generate acute changes in cytoskeletal and adhesive properties to inhibit cell mixing, and whereby the integration of multiple local cues may dictate both the global morphogenetic properties of a tissue and its separation from adjacent cell populations.