We compare in detail central collisions Xe(50 AMeV) + Sn, recently measured by the INDRA collaboration, with the Quantum Molecular Dynamics (QMD) model in order to identify the reaction mechanism which leads to multifragmentation. We find that QMD describes most of the data quite well, in the projectile/target region as well as in the midrapidity zone where also statistical models can be and have been employed. The agreement between QMD and data allows us to use this dynamical model to investigate the reaction in detail. We arrive at the following observation: (a) the in-medium nucleon-nucleon cross section is not significantly different from the free cross section, (b) even the central collisions have a binary character, (c) most of the fragments are produced in the central collisions, (d) the simulations as well as the data show a strong attractive in-plane flow resembling deep inelastic collisions, and (e) at midrapidity the results from QMD and those from statistical model calculations agree for almost all of the observables with the exception of d 2σ dZ dE . This renders it difficult to extract the reaction mechanism from midrapidity fragments only. According to the simulations the reaction shows a very early formation of fragments, even in central collisions, which pass through the reaction zone without being destroyed. The final transverse momentum of the fragments is very close to the initial one and due to the Fermi motion. A heating up of the systems is not observed and hence a thermal origin of the spectra cannot be confirmed. The disagreement between the simulations and the data for some observables is presumably due to a force range which is too large as compared to reality but necessary to keep the nuclei stable in these semiclassical approach.