This paper presents a flight control system design method with a neural-fuzzy gain-scheduling algirithm with learning capability for a high performance fighter aircraft undergoing a high angle of attack velocity vector roll manoeuvre. This manoeuvre is similar to the Herbst manoeuvre, hence also called a Herbst-like manoeuvre. Two linear H infinity robust controllers are designed by using the mu-synthesis method at two trim conditions that cover the whole dynamic range of this manoeuvre. Then a scheduling technique is proposed, using the neural-fuzzy concept, together with a dynamic backpropagation algorithm for its training. Simulation results, using a recently developed non-linear six-degree-of-freedom aircraft simulation package, demonstrate that the neural-fuzzy scheduler can be trained by using the dynamic backpropagation algorithm to achieve better closed-loop tracking performance, and the proposed neural-fuzzy scheduled flight control system gives satisfactory performance in executing this Merbst-like manoeuvre.