Humphry (this issue) deserves credit for drawing attention to the long-neglected fact that differences in item discrimination parameters are often due to empirical factors and not the product of random error components. In doing so, Humphry offers a psychometrically elegant, coherent, and practically important new model that is more flexible while making greater demands on theoretical considerations as to why discrimination parameters differ. However, the author thinks that while the presented model may solve certain psychometrical problems, it simultaneously raises new questions. The author wishes to mention two of them: one related to the problem of data-model fit and the other more closely related to its purpose as a model of empirical psychological phenomena. The author hopes that Humphry's approach will, not only start a discussion about empirical factors leading to varying discrimination parameters, but will also commence a discussion on some of the fundamental measurement problems in psychology and related disciplines.