Wireless Sensor Network (WSN) software is typically organized as a set of modules and depending on the application only necessary modules are selected, compiled and installed on a sensor node. This software flexibility is currently not matched by available hardware platforms. A sensor node always provides the same single microcontroller regardless of application requirements; efficiency is limited by the need to support the worst-case processing demands of the system. We argue that the central processing unit can be replaced by a number of less capable, interconnected and more specialized microcontrollers focused on specific tasks. Thus, unutilised processing hardware can easily be removed from the platform. This paper presents the results of an early experimental study into the benefits of using a multiprocessor sensor node architecture. The study shows that the desired hardware configuration flexibility can be achieved with relatively low overheads while supporting established sensor node programming concepts.