A modified pole placement controller on the basis of a dynamic neural network is developed for controlling nonlinear multi-inputs and single output (MISO) processes. Using process data, this dynamic network composed of several local single-input and single-output (SISO) neural network models is trained to model general nonlinear MISO processes. At each time step each local neural model is linearized to produce the linear transfer functions. Hence, the original MISO control problem is decomposed into several SISO linear control problems. As a consequence, the well-established pole placement design technique for linear systems can be implemented directly. Like the indirect self-tuning control design, the control synthesis algorithm is adaptive when the linear model is extracted from the nonlinear neural model at each sampling instant. In the end, a pH neutralization system is used to verify the control performance of the proposed method.