Two approaches to approximate feedback linearization are applied to the nonlinear control of a fluidized bed reactor. The first approach, third order approximate input-state linearization, results in improved performance for setpoint tracking over a narrow region around the operating point. On the basis of an analysis of the assumptions of approximate input-state linearization, an approximate input-output linearization scheme is proposed. Implementation of a control scheme for the fluidized bed reactor based upon this approach resulted in excellent setpoint tracking over a 30 degrees C temperature range in the reactor. The issues of nonlinear model selection and computational issues are discussed. Finally, a novel internal model control based nonlinear anti-windup scheme is implemented to ensure that improved performance is obtained without violating hard actuator constraints.