The behaviour settings approach was introduced as a means to study the variability of human beings' behaviour outside the lab. More recently, it has been argued that it also provides a fruitful avenue for developing situated accounts of cognition. This article will provide a proof of concept for the latter suggestion, focusing on the science of self-control. Self-control is the ability of individuals to pursue goals they value in the face of conflicting motivations. The hypothesis we bring forward is that this ability should be understood as a set of skills by which individuals modulate their relation to their environment, more specifically the behaviour settings they inhabit. With this conception of self-control in hand, we will take a critical look at well-known experiments involving delayed gratification tasks and propose concrete suggestions on how to improve them. This will bring us to the conclusion that the behaviour settings framework might have a valuable role to play in developing a situated science of self-control. This article is part of the theme issue 'People, places, things and communities: expanding behaviour settings theory in the twenty-first century'.