A fish swims by stimulating its muscles and causing its body to "wiggle", which in turn generates the thrust required for propulsion. The relationship between the forces generated by the fish muscles and the observed pattern of movement is governed by the mechanics of the internal structure ofthe fish, and the fluid mechanics of the surrounding water. The mathematical modell ing of how fish swim involves coupling the external "biofluiddynamics" to the body's internal solid mechanics. The best-known theory for the hydrodynamics of fish swimming is Lighthill's elongated body theory (Lighthill, 1975). In Lighthill's theory the curvature of the fish is assumed small and the effect on the fish of the vortex wake is neglected. Cheng et al. (1991) did not make these simplifications in developing their vortex lattice panel method, but the fish was assumed to be infinitely thin and its undulations of small amplitude. Lighthill's "recoil correction" is the addition of a solid-body motion to ensure that an imposed "swimming description" satisfies the conservation of momentum and angular momentum. A real fish is expected to minimize such sideways translation and rotation to avoid wasteful vortex shedding. Cheng and Blickhan (1994) found that the panel method model required a smaller recoil than did Lighthill's model. Our approach is to extend Cheng's model to large amplitude. Thus we include the effect of the wake on the fish, and the self-induced deformation of the wake itself. In studying the internal mechanics of the body we model the fish as an active bending beam. Using the equations of motion of cross-sectional slices of the body we can form a set of coupled differential equations for the bending moment distribution. At large amplitude the bending moment equations involve the tangential forces acting on the body (which may be neglected in the small amplitude version). Consequently we include the boundary layer along the fish in order to estimate the viscous drag directly. The panel method has been used successfully for the fluid mechanical calculations associated with large-amplitude fish swimming. We are able to use its results as input to calculate the bending moment distribution. The boundary layer calculations are based on a crude model; solutions to the large amplitude bending moment equations should also be considered in this light.