A Consideration of Some Problems of Content Identification in Scaling

It is possible that something other than the intended attitudinal content may be measured in cumulative scaling. This paper includes an empirical demonstration of how cumulative content can be “built” into a scale, and how the simple Guttman scale is in this case improved by the H-technique. The authors demonstrate that even in a high reproducibility scale the factor loadings of questions differ somewhat withthe cutting points used, and that the factor loadings on items may vary substantially with the frequency of positive response to the question when cutting points are held constant