This paper reviews an evolving family of surface modeling primitives for use in computer vision, graphics, CAGD, and data analysis. Dynamic surfaces offer significantly greater power to these applications domains than do conventional, geometric surface modeling primitives. The benefits stem from computational physics: Simulated forces influence the shapes and motions of dynamic surface according to the basic mechanical principles of nonrigid bodies with given mass and damping distributions and deformation energies. Dynamic surfaces are applicable both to static surface reconstruction problems and to more general surface estimation from structured or unstructured time-varying data. In particular, efficient, recursive solutions to time dependent problems result from a Kalman estimation approach in which dynamic surfaces serve as nonstationary system models.