ABSTRACT
In his book on measurement, W. S. Torgerson states that one limitation of Guttman's method of scaling resides in its incapacity to test statistically whether the cumulative scale model is a good fit. While there is truth to Torgerson's statement, it is perhaps unnecessarily categorical and may act to divert attention from scale analysis before its potentialities as an ordering device have been properly considered by the investigator. The intraclass correlation coefficient is a function of the variance ratio, and, consequently, to test its significance it is necessary to satisfy the conditions of the F-distribution. On the other hand, it is a conventional measure of correlation, and if such a measure be required, then its use would be dictated by that need, even though its statistical significance could not be exactly determined.
