Toward a functionalist account of so-called ‘extraction constraints’