ABSTRACT
The notion of test-retest or any other "reliability" measure involving a time sequence is antithetical to social science since it must make the incorrect assumption that human thought and behavior is static and, therefore, that any change in response is a reflection of either instrument error or deception. Most of the methods for estimating the unreliability that comes from disagreement between items were developed in the area of psychological testing. At a more abstract level of methodology, mathematical models of change and error provide sharp tests of different theories of unreliability. Whether the scales are arbitrarily weighted combinations of items or rational procedures like Guttman scales, the main purpose is the same. In order to make the logic of Guttman scaling clear, it may be helpful to think of the items as observers. One reason for the vogue of Guttman scaling is its simplicity.
