ABSTRACT
Given the indirectness of measurement procedures in general, and the fact that in most sociological research the operational indicators are linked with the theoretical variables by rather poorly understood processes, it is clear that our analytic techniques must allow for substantial measurement errors. A number of papers have considered various approaches to measurement error by utilizing explicit causal models linking unmeasured variables to their measured indicators. In this chapter, the author combines selected features of the arguments developed by Costner and Heise while emphasizing the flexibility of the general approach that is common to both papers. With a single measure of each variable, one can remain blissfully unaware of the possibility of measurement error, but in no sense will this make his inferences more valid. Where there is only a single indicator of each variable, the method of instrumental variables may be used to take out random measurement errors.
