ABSTRACT
Hanan Selvin has confused statistical inference with causal inference. All his statements to the effect that sociologists need not employ significance tests in survey analysis are based upon this confusion. Selvin claims to find support for his views in an article by Herman Wold. Selvin argues that statistical inference may be substituted for causal inference in experimental research, but may not be substituted in survey research. In fact, statistical inference may not be substituted in either case. Survey research can only uncover causal relationships if the survey design encompasses all pertinent variables. Experimental research can only uncover causal relationships if the experimental design encompasses all pertinent variables. Research technique is no substitute for theory; experimental research is no exception.
