ABSTRACT
James Beshers goes on to advocate the inclusion of “all relevant variables” in explanatory surveys. However, when Beshers says that an experiment “can only uncover causal relationships if the experimental design encompasses all relevant variables” [sic] and that “randomization does not eliminate correlated biases” [sic], he is simply wrong. As Hyman points out, “The analysis of complex relationships provides a solution to the problem social scientistsshall call spuriousness and permits the analyst to infer that the original relationship involves cause and effect”. Perhaps the confusion that Beshers finds may come from scientists different conceptions of causal inference. In general, to determine whether two variables are causally related, one must show that the observed relationship is not, or could not reasonably be, the result of “extraneous” variables. These extraneous variables can, in turn, be divided into two classes: those variables that can be identified by the researcher and those that cannot.
