Three different understandings of causation, each importantly shaped by the
work of statisticians, are examined from the point of view of their value
to sociologists: causation as robust dependence, causation as consequential
manipulation, and causation as generative process. The last is favoured as
the basis for causal analysis in sociology. It allows the respective roles
of statistics and theory to be clarified and is appropriate to sociology a
s a largely non-experimental social science in which the concept of action
is central.