We estimate the quantum state of a system between two measurements via a maximum likelihood strategy. This incorporates all the information provided by preselection and postselection in a legitimate quantum state. Moreover, it allows to include any other constraint derived from practical considerations.