This paper summarizes similitude requirements for liquid-metal fast breeder reactor natural convection shutdown heat removal test facilities. It is generally not possible under natural circulation flow to reproduce both hydrodynamic and thermodynamic similarity when the thermal driving head is determined by structure-to-fluid heat transfer. This paper explores the source of the scaling difficulties and examines a number of possible solutions. Design and operating restrictions imposed on the test facility and the validity of extrapolating test facility data to the full-scale system are discussed.