Abstract: Causality is central to our understanding of the world and central to scientific explanation. In recent years two approaches to causality have come to prominence and have had a major impact on the social sciences: these are the counterfactual or potential outcomes model of causality and the approach that understands causality in terms of a causal structure represented by a graph. I present both of these and explain how they can be used to identify causal relationships in situations when we do not have access to experimental data. I discuss the principles underlying the most widely used strategies for estimating causal effects in these situations. Finally, I discuss questions of external validity, and, in particular, the conditions under which sociologists' causal estimates can be of more than historical interest.