Proceedings of the Annual Meeting of the Cognitive Science Society

Classical models of the emergence of compositionality in communication focused on the compositional nature of the en-vironment (Cangelosi, 2001; Cornish et al., 2008). Here we advance a model in which compositional structure emergesfrom integrating environments properties with agents actions. We take as a starting point Cangelosis (2001) model, wherea population of agents searched for edible mushrooms. Given opportunity to communicate, they evolved a system inwhich combinations of signs were sensorily grounded in combinations of mushroom properties. We modify this modelby grounding the communication also in agents’ actions. With this, we are able to evolve communication systems con-taining meaningful compositions of mushroom properties and agent actions. We investigate how such compositions canfacilitate a) learning the communication protocol, b) learning the adequate behavior policy. This kind of sensory-motorcompositionality seems better suited for coordinating navigation in dynamic environments.