Context effects have traditionally been explored as a reliability concern and detected through the use of univariate, test/retest techniques. We argue that context effects may actually represent a threat to validity and, as a consequence, traditional univariate measurement techniques
will fail to detect their full impact. Using an experimental design which manipulates the context of questions involving ethics, we demonstrate that priming influences the relationship of variables to each other and has an effect on the dimensional reduction of the responses for the two experimental
groups. We conclude that the problem posed by context effects is not just one of reliability, but also one of validity. Implications for future basic research on attitude formation are drawn.