The effect of task structure on diffusion dynamics: Implications for diffusion curve and network-based analyses

Abstract

Theoretical analyses within the broad field of social learning research give mixed conclusions on whether the shape of a diffusion curve can be used to infer that a learned trait increases through social or asocial learning. Here we explore how factors such as task structure (e.g., multiple-step tasks), task abandonment, subgoal learning, and neophobia affect the shape of the diffusion curve for both asocially learned and socially learned behavior. We demonstrate that, whereas social learning increases the likelihood of S-shaped curves, sigmoidal patterns can be generated by entirely asocial processes, and cannot be reliably interpreted as indicators of social learning. Our findings reinforce the view that diffusion curve analysis is not a reliable way of detecting social transmission. We also draw attention to the fact that task structure can similarly confound interpretation of network-based diffusion analyses, and suggest resolutions to this problem. Supplemental materials for this article may be downloaded from http://lb.psychonomic-journals.org/content/supplemental.