In studying the causes of offending, the aim should be to develop an explicit scientific theory that yields quantitative predictions for testing against empirical data. There are fundamental differences between theoretical constructs and empirical variables, and statements about causal relationships are at the core of any theory. Further, theory testing requires operational definitions of theoretical constructs so that they can be measured by empirical variables. This involves the study of change within subjects in longitudinal analysis. In such analysis, each person acts as his or her own control and extraneous independent influences on offending are held constant. Key theoretical constructs in the development of a causal theory of offending are examined, as well as empirical variables, problems in existing research, and the Cambridge study of delinquent development. 38 references and 1 figure