The third limitation of the paper concerns the use of confirmatory factor analysis, which is based on a domain sampling paradigm with LISREL. The concern is that the study constructs can in some circumstances be criticized as more formative than reflective by nature; as a result the indicator variables can have the weakness of measuring different aspects of the latent variables (see Gefen, Straub, and Boudreau 2000) instead of the same aspect. When using a domain sampling paradigm the indicator variables should be highly correlated and thus exchangeable.