The concept of empirical generalisation has recently received close attention in the causal modelling debate. Notably, in Woodward’s account, empirical generalisations replace the traditional concept of law. This paper investigates the role of empirical generalisations in quantitative social science and argues that invariance and counterfactual-based approaches are problematic in observational contexts. An alternative account of empirical generalisation based on a non-counterfactual definition of invariance is offered within a structural modelling framework.