The origins of human language is a fascinating question which so far has defied a solid scientific explanation. After being dormant for almost a century, this research topic has in recent years been taken up with new vigour by the many sciences interested in language, from linguistics to anthropology and from neurobiology to genetics. Modeling plays an important role in all sciences, and it could contribute also here, particularly because we obviously do not have any data of the very beginnings of language.
This talk will show that the state of the art in models of language evolution has dramatically advanced over the past decade. Language evolution models typically introduce a population of agents that model the cognitive capacities of individual language users and their interaction patterns. The agents are then set up to play language games with each other with the goal of seeing whether a language-like communication system indeed emerges from scratch given a particular constellation of cognitive and social capacities. Language evolution models can not only be tested with computer simulations. The agents can be implemented as autonomous physical robots with all the necessary visual perception, motor control, and social embodied cognition to play grounded language games.
The outcome of this experimental research has already been remarkable. We now know how to build robots that establish both a lexicon and an inventory of perceptually grounded concepts about the world, and recent experiments have shown how human-language-like case grammars could emerge in situated embodied language games. In contrast to the hypothesis that language originated in a unique genetic event that gave rise to a highly specialised language organ in the brain, these models show that advanced cognition and social interaction patterns must have been the crucial factors.