This paper illustrates the chaining of neural network bused self-adapting modules for the visual servoing of a robot manipulator. Arm movements are controlled using visual features. Camera orientation is also controlled to increase the feedback control precision. In this case it proves to be efficient to chain two NeuroModules in sequence. We propose an algorithm to adapt online and simultaneously two serially chained NeuroModules. The approach is applied to the control of a 2 d.o.f. planar arm using a single 2 d.o.f. pan-tilt camera.