Background. Recent studies of social learning have revealed that adult humans are “over-imitators” who frequently reproduce a model's causally irrelevant tool actions to the detriment of task efficiency. At present our knowledge of adult over-imitation is limited to the fact that adults do over-imitate, we know very little about the causes of this behavior. The current study aimed to provide novel insights into adult over-imitation by extending a paradigm recently used with human children to explore social aspects of over-imitation. In the child study observers saw two models demonstrate a tool-use task using the same inefficient approach, or two models demonstrate different approaches to the task (one inefficient and one efficient). The manipulation of social influence came in the testing phase where the observer completed the task in the presence of either an inefficient model or an efficient model. Methodology/Principal Findings. We adapted the paradigm used in the child study to provide the first systematic exploration of factors which may lead to adult over-imitation including: 1) the presence of the model(s) during testing, 2) the presence of a competing efficient task demonstration, 3) the presence of a majority displaying the inefficient approach, and 4) the ‘removal’ of the experimental context during task completion. We show that the adult participants only over-imitated in conditions where the inefficient strategy was the majority approach witnessed. This tendency towards over-imitation was almost entirely eliminated when the participants interacted with the task when they believed the experiment to be complete. Conclusions. Our results suggest that adult over-imitation is best explained as a result of an evolved ‘conformist bias’ argued to be crucial to the transmission of human cultural behavior and one which may be unique in the animal kingdom.