Several methods have been developed to test interfacial adhesion in composite materials such as pull-out, microbond and push-in/push-out tests. Some of them can only be applied to single fibre matrix composites and others are difficult to perform on brittle fibres due to premature fracture of the fibre. Push-in tests, consisting on pushing the fibre with a micro or nanoindenter on a bulk specimen, constitute a powerful technique that can be applied directly on composite laminates. However, the interfacial adhesion values obtained from different tests (microbond, push in) often differ and even are subjected to a large scatter. This might be due to the fact that the existing analytical solutions that are typically used to interpret the experimental data take into account the constrain effect of the surrounding fibres on a simplified manner. To study this, we have carried out a careful micromechanical modelling of the push-in test, coupled with experimental adhesion testing in a glass fibre reinforced epoxy matrix composite. The model takes into account the interfacial fracture process by means of interface cohesive elements at the fibre–matrix interface and focuses on the study of the constrain effects due to the local configuration of the surrounding fibres.