ABSTRACT
The principal difficulty in applying tests of significance to sociological research centers about the concept of experimental control. In principle, then, tests of significance have a place in non-experimental research. Designing non-experimental studies so that tests of significance can be used validly is at best difficult. The problems of interpreting significance tests fall into three groups—problems of meaning, of random process, and of selection. Insofar as the level of significance is interpreted in this essentially negative way, there is no problem of meaning. But one frequently finds the level of significance given an erroneous positive interpretation—as a virtual “seal of approval” or a measure of substantive importance. The level of significance is only one link in a chain of methodological evidence that the results are substantially as claimed; to offer it as the only piece of evidence is misleading.
