A large scale model analysis, using embedded strain gauges, of the strain distribution in the cement mantle surrounding a femoral prosthesis is underway. In order to predict, and so avoid, positions of locally high strain gradients in this model, a finite element and experimental analysis of a similar problem was undertaken. For this purpose, a loose fitting rectangular steel insert inside a surrounding rectangular epoxy sheath was used to model an extreme case of the torsional and bending components of hip joint load. The axial component of joint load was modelled using an axisymmetric finite element model of a tapered shaft. The finite element results were used to determine suitable positions for embedding gauges in the experimental model. Results showed that the finite element analysis failed to adequately model the close sliding fit between the steel insert and epoxy. Altering the experimental model to artificially replicate the finite element contact conditions produced good correlation in bending, with experimental strains agreeing with simple bending theory to within 6%. Satisfactory correlation under torsional loading was not obtained, but strain magnitudes were low. Predicted positions for embedding gauges give conservative results, lessening the possibility of strain gradient induced error in the large scale model test of the cement mantle and prosthesis.