The primary aim of this paper is to show how graphical models can be u
sed as a mathematical language for integrating statistical and subject
-matter information. In particular, the paper develops a principled, n
onparametric framework for causal inference, in which diagrams are que
ried to determine if the assumptions available are sufficient for iden
tifying causal effects from nonexperimental data. If so the diagrams c
an be queried to produce mathematical expressions for causal effects i
n terms of observed distributions; otherwise, the diagrams can be quer
ied to suggest additional observations or auxiliary experiments from w
hich the desired inferences can be obtained.