Proceedings of the Annual Meeting of the Cognitive Science Society

Why do languages change? One possibility is they evolve in response to two competing pressures: (1) to be easily learned,and (2) to be effective for communication. In a number of domains, variation in the worlds natural languages appears tobe accounted for by different but near-optimal tradeoffs between these pressures. Models of these evolutionary processeshave used transmission chain paradigms in which errors of learning by one agent become the input for the subsequentgeneration. However, a critical feature of human language is that children do not learn in isolation. Rather, they learn incommunicative interactions with caregivers who draw inferences from their errorful productions to their intended interests.In a set of iterated reproduction experiments, we show that this supportive context can have a powerful stabilizing role inthe development of artificial patterned systems, allowing them to achieve higher levels of complexity than they would byvertical transmission alone while retaining equivalent transmission accuracies.