The peri-sylvian cortico-striatal language system, has been demonstrated by numerous techniques (including lesion, PET, fMRI) to play a fundamental role in sentence comprehension.  We have explored this system in thematic role assignment (determining who did what to whom) in sentence processing, and developed a neural network model that explains a significant component of this human behaviour, based on neurophysiological functions that are not specific to language (Dominey et al. 2006, 2009).  This modelling work has led to a number of predictions concerning a shared system for linguistic and non-linguistic sequence processing (Hoen et al. 2006).  A major challenge was to determine if our language processing model could accommodate the “real world”.  We thus embedded the neural network in a robotic system that could see, hear, and learn to describe physical events using the language learning network (Dominey & Boucher 2005).  More recently we have extended this neuro-robotic approach to using high degree-of-freedom humanoid robots (Dominey et al. 2009b).  We currently focus on neuro-robotic approaches to embodied representations of meaning (Madden et al. 2009).  The talk will provide an overview of this research.