Abstract In shoulder arthroplasty, there is no consensus about the ideal mismatch between a prosthetic humeral head and a glenoïd component. Thus, investigations into mismatch effects from a biomechanical point of view can be useful. The aim of this in vitro study was to help us understand mismatch influence on bone strains, translational forces in the joint and implant/bone displacements in implanted scapulae. Five fresh cadaveric scapulae were implanted with a cemented keeled polyethylene implant. The lower part of the scapulae was embedded and the loadings were carried out using five metallic spheres simulating mismatches of 0, 2, 4, 5 and 6 mm. Loadings included a constant compressive preload of 392 N and an anterior, posterior, inferior and superior translation of 2.5 mm. We measured the transversal force necessary to produce the imposed translation, the strains at six locations around the peripheral cortex of the glenoïd using strain gages and the relative implant/bone displacements using CCD cameras. Generally, the increase of mismatch reduced the translational forces, the strains around the glenoïd and, except for the anterior loading, the relative implant/bone displacements. Few and even no significant differences were observed when the mismatch varied from 0 to 2 mm; the number of significant differences increased when the mismatch varied from 0 to 4 mm and from 0 to 5 mm; the results obtained for a 0–6 mm variation in mismatch were comparable to those obtained for a 0–5 mm variation. This study underlines that the mismatch has a significant effect on bone strains, relative implant/bone displacements and induced translational forces when a compressive preload and imposed translations were applied on implanted scapulae.