We used the diffusion chain method in order to investigate whether irrelevant actions would be transmitted along chains of 3- and 5-year-old children. Four chains of eight children witnessed a trained “expert” child perform a sequence of actions in order to retrieve a reward from either a transparent or an opaque puzzle box. The action sequence involved both goal relevant and goal irrelevant actions. In the transparent box chains the participants could potentially determine which of the actions were irrelevant as the causal effects were clearly visible. Results indicated that irrelevant actions transmitted down chains of 3-year-old children irrespective of box transparency. In contrast, irrelevant actions dropped out of the transparent box chain extremely quickly at 5 years, but were maintained within the opaque chain. These findings highlight the power of the diffusion chain method as a tool for exploring cultural learning.