A note on the model reduction problem