A method of identifying closed-loop systems is developed by using the orthogonal decomposition (ORT) method. The idea is to project the input and output data onto the space of exogenous inputs by using the LQ decomposition to obtain their deterministic components. The ORT-based method is then applied to deterministic components like the direct approach in order to derive state-space models of the plant. We also show that the present method is a subspace version of the two-stage method for transfer function estimation from closed loop data. Some simulation results are included to show the applicability of the present method. (C) 2007 Elsevier Ltd. All rights reserved.