Movement compositionality can be realized by a hierarchic architecture of syn re chains

A hierarchic structure of syn re chains has been proposed as the neural substrate for compositionality; individual chains represent aspects of a complex object, and the dynamic binding of several chains integrates the entire object. In this study we demonstrate that such a structure is capable of modelling compositionality in a generic fashion, i.e. not just for reactive computations such as perceiving stimuli, but also for active computations such as the generation of complex behavior. As an example case which is the subject of many experimental investigations, we address the perception and generation of movement. A small number of syn re chains is su cient to reliably discriminate between di erent interstimulus delays and to generate arbitrary planar point-to-point trajectories on the basis of simple motion primitives. Moreover, when the network is constructed in a closed-loop fashion, autonomous scribbling-like trajectories are generated.