The aim of the present work is to investigate the application of the self-tuning control method to water level control in the steam generators of the Pressurized Water Reactor (PWR) nuclear powr plant. A well-known generalized self-tuning control algorithm is reviewed and modified to eliminate a steady-state offset error and to facilitate tuning the control system. Then a new algorithm incoporates pole-assignment into the minimum variance strategy with a particular form of cost function. The pole-assignment algorithm enables the stability characteristics of the closed-loop system to be readily specified. For the computer simulation, the control system design is based on a fourth-order linear ARMAX model (autoregressive moving average model with auxilary input). In order to ensure that the low-order model describes the complex real dynamics well enough for control purposes, control parameters are updated on-line with the recursive estimation sequences of the least-squares method. Weighting polynomials in the pole-assignment controller are also adjusted on-line to keep closed-loop poles at desired locations and to satisfy the requirements of no offset errors and disturbance rejection. The present pole-assignment controller is quite robust and can easily handle unstable non-minimum phase system. Computer simulation and experimental results confirm the robustness and stability of the present self-tuning control algorithm.