This article replicates and extends an agent-based model of cultural transmission (Acerbi andamp; Parisi, 2006). The original model uses artificial neural networks to inquire about the role of noise and selective cultural reproduction in imitation learning dynamics, both for static and dynamic environments. The replication tests the robustness of the original results, whereas the extension focuses on implementing an alternative type of learning: Direct instruction. The results of the extension suggest this type of learning could negatively affect the emergence of adaptive behavioral traits at the population level. Because of its reliance on explicit one-way communication and its reduced chance to question the traits transmitted, direct instruction might increase the time taken to find effective behavioral variants, in comparison with imitation. Yet, if the limit that defines inadequate behavior is chosen loosely enough, a sufficient amount of behavioral variations could be introduced in the behavioral pool so to ensure the development of highly adaptive variations. The text uses the implementation of direct instruction to discuss the role of extension in scientific endeavor, especially in interdisciplinary areas of research, such as the science of cultural evolution or agent-based computational social science.
