A RETRAN model was developed for determining the stability of boiling water reactors operating with high power/flow ratios. This model was applied to Peach Bottom 2 and compared with the stability tests conducted at the end of fuel cycle 2. Because of the deterministic nature of the RETRAN model, it appears logical to select those transients produced by periodic setpoint switching rather than those where pseudorandom switching was employed. Comparisons were made in both the time domain and the frequency domain between the computed equilibrium cycle and a measured equilibrium cycle obtained as an average over a large number of cycles. The variables of interest are the system pressure measured in the steam dome and the average neutron flux measured by an average power range monitor. Comparisons in the time domain were adversely affected by an apparent, strong nonlinearity in the pressure controller and/or turbine control valve response. Since this effect could neither be identified nor appropriately simulated, a second model was developed, for which the measured system pressure is specified as time-dependent volume condition. The equivalence of these two models with regard to the pressure-to-flux transfer function was established. This provides the justification for using any arbitrary, simple pressure perturbation of the second model to determine this transfer function, which is commonly used to provide a measure of stability. Sensitivity studies revealed three groups of parameters with a strong influence on stability as well as the nature of this influence. These parameters are related to: heat conduction from the fuel pellet and through the helium-filled gap, the axial power profile, and the moderator feedback of reactivity. The choice of two-phase flow model also was shown to be very important; a case using the homogeneous equilibrium model option resulted in diverging oscillations, while corresponding calculations with the algebraic slip model exhibited considerable damping similar to that observed in the measured response.