We present an agent-based computer simulation that extends a previous experimental simulation of the cultural transmission of projectile-point technology in the prehistoric Great Basin, with participants replaced with computer-generated agents. As in the experiment, individual learning is found to generate low correlations between artifact attributes, whereas indirectly biased cultural transmission (copying the point design of the most successful hunter) generates high correlations between artifact attributes. These results support the hypothesis that low attribute correlations in prehistoric California resulted from individual learning, and high attribute correlations in prehistoric Nevada resulted from indirectly biased cultural transmission. However, alternative modes of cultural transmission, including conformist transmission and random copying, generated similarly high attribute correlations as indirect bias, suggesting that it may be difficult to infer which transmission rule generated this archaeological pattern. On the other hand, indirect bias out-performed all other cultural-transmission rules, lending plausibility to the original hypothesis. Importantly, this advantage depends on the assumption of a multimodal adaptive landscape in which there are multiple locally optimal artifact designs. Indeed, in unimodal fitness environments no cultural transmission rule outperformed individual learning, highlighting how the shape of the adaptive landscape within which cultural evolution occurs can strongly influence the dynamics of cultural transmission. Generally, experimental and computer simulations can be useful in answering questions that are difficult to address with archaeological data, such as identifying the consequences of different modes of cultural transmission or exploring the effect of different selective environments.