Pairs of endothelial cells on adhesive micropatterns rotate persistently, but pairs of fibroblasts do not; coherent rotation is present in normal mammary acini and kidney cells but absent in cancerous cells. Why? To answer this question, we develop a computational model of pairs of mammalian cells on adhesive micropatterns using a phase field method and study the conditions under which persistent rotational motion (PRM) emerges. Our model couples the shape of the cell, the cell's internal chemical polarity, and interactions between cells such as volume exclusion and adhesion. We show that PRM can emerge from this minimal model and that the cell-cell interface may be influenced by the nucleus. We study the effect of various cell polarity mechanisms on rotational motion, including contact inhibition of locomotion, neighbor alignment, and velocity alignment, where cells align their polarity to their velocity. These polarity mechanisms strongly regulate PRM: Small differences in polarity mechanisms can create significant differences in collective rotation. We argue that the existence or absence of rotation under confinement may lead to insight into the cell's methods for coordinating collective cell motility. C ollective cell migration is a crucial aspect of wound healing, growth and development of organs and tissues, and cancer invasion (1-3). Cells may move in cohesive groups ranging from small clusters of invading cancerous cells to ducts and branches during morphogenesis to monolayers of epithelial or endothelial cells. Two hallmarks of collective migration are strong cell-cell adhesion and multicellular polarity-an organization of the cellular orientation beyond the single-cell level (1). Cell-cell interactions can lead to collective behavior not evident in any single cell, including chemotaxis in clusters of cells that singly do not chemotax (4). Collective behavior may arise from cell-cell interactions altering the polarity of individual cells (5, 6). Many theories have been proposed for how this multicellular order appears, either in specific biological contexts (7-11) or in simpler, more generic models (12-16). Some authors argue that these dynamics are relatively universal and can be understood with minimal knowledge of the signaling pathways involved (2, 17).Collective rotation is commonly observed in collectively migrating cells, especially in confinement. Persistent rotations have been observed in the slime mold Dictyostelium discoideum (18), canine kidney epithelial cells on adhesive micropatterns (19), and small numbers of endothelial cells on micropatterns (20,21). Transient swirling patterns are also seen in epithelial monolayers (22). Recent work has also observed that the growth of spherical acini of human mammary epithelial cells in 3D matrix involves a coherent rotation persisting from a single cell to several cells; this rotation is not present in randomly motile cancerous cells (23). Similarly, cancerous cells on adhesive micropatterns do not develop coherent rotation (19). In a recent review ...