This paper is intended as a stimulus to wider discussion on the nature of cause–effect models, and their role in educational research. It is clearly not intended to be definitive. Nor is it based on new evidence. It considers in turn models based on no causation at all, models based on causation alongside other explanatory processes, models based solely on causation, and variations of the latter. Among these are purportedly weaker forms of causation, including Granger-causation, social determinants, and so-called fuzzy generalisation. Cause–effect models have been based variously on temporal and spatial conjunction, the impact of interventions, and theoretical explanations. The paper concludes that cause–effect provides a powerful, persuasive and near-universal explanation for social and psychological processes, despite our inability to detect it directly. Because of this inability, cause–effect models are more impressive where they have more of the characteristics just noted. This conclusion, if accepted, has implications for the nature of educational research.