The dynamic properties of joints are extremely difficult to model accurately using a purely analytical approach. However, these properties can be extracted from experimental data. In this paper we present a method for establishing a theoretical model of a joint from the substructures and assembly frequency–response function (FRF) data. The identification process considers not only translational, but also rotational degrees of freedom (RDOFs). The validity of the proposed method is demonstrated numerically and experimentally. A combined numerical–experimental approach was used to identify the mass, stiffness and damping effects of a real bolted joint. Using the least-squares method, data from the wide frequency range were used. A substructure synthesis method with the joint effects included was used to check the extracted values.