We analyze deformation of the Fars Arc in the eastern Zagros, Iran, including earthquake slip vectors, GPS velocities, paleomagnetism data, and fold orientations, to understand how this fold‐and‐thrust belt works and so better understand the generic issue of fold‐and‐thrust belt curvature. The Fars Arc is curved, convex southward. GPS‐derived rotation rates are ≤0.5° Myr−1: Rotation is clockwise west of 53°E and counterclockwise to the east. These rotation senses are opposite to previous predictions of passive “bookshelf” models for strike‐slip faults during north‐south convergence. West of 53°E, average GPS vectors, thrust earthquake slip vectors, strain axes derived from GPS data, and orthogonal directions to fold trends are all aligned, toward ~218°. East of this meridian, the average GPS vector is toward 208°, but the averages of the other data sets are distinctly different, all toward ~190°. We propose that fault blocks in eastern Fars, each ~20–40 km long, rotate predominantly counterclockwise, whereas in western Fars, the regional clockwise rotation takes place mainly on the array of active right‐lateral faults in this area. Thus, localized block faulting and rotations accumulate to produce the overall strain and regional curvature. Active folds of different orientations in eastern Fars intersect to produce domal interference patterns, without involving separate deformation phases, indicating that fold interference patterns should not be interpreted in terms of changing stress orientations unless there is clear evidence. Fars Arc curvature is best explained by deformation being restricted at tectonic boundaries at its eastern and western margins, with significant gravitational spreading.