A neurofuzzy controller has been used as a nonlinear compensator for a flexible four-link manipulator. Two classes of neurofuzzy models, the Takagi-Sugeno fuzzy model and the rectangular local linear model network have been applied as a feedforward controller to compensate the nonlinearities. The first model incorporates expert-based fuzzy rules into the controller, whereas the second model structure automatically partitions the input space. An adaptation algorithm is developed to train the controller in order to stabilize the whole system. Two control problems have been considered, namely, joint and tip position control schemes. The output signal redefinition strategy is adapted to stabilize the tip position control scheme. A tradeoff between the tracking accuracy and manipulator link vibration can be achieved. Experimental studies have been carried out on a planar flexible four-link manipulator testbed.