restricted access The Interplay of Theory and Measurement
In lieu of an abstract, here is a brief excerpt of the content:

The Interplay of Theory and Measurement


This Viewpoint concerns all of the sociological sciences that deal with empirical phenomena, including sociology, demography, political science, social anthropology, communication, and socio-environmental science. The reason for the focus on these disciplines rather than on demography alone is that the theoretical base of sociology underpins all of the others. Put differently, each of the others takes its fundamental theory from sociology.

The discussion that follows consists of five parts. The first part details what I think a theory is. The second presents the gist of a well known article that I wrote with two co-authors on status attainment (Sewell, Haller, and Portes 1969). Written some forty years ago, it serves as an example of the mix between theory and measurement, referred to herein as a mini-theory. Having been republished several times over the years, it is still more or less fresh. The third part summarizes the general theory that underlies the mini-theory's concepts. The fourth discusses the general theory in more detail and provides a brief history of the concepts and measurements involved in the mini-theory. The fifth and final part consists of concluding remarks.


If memory serves me right, Immanuel Kant once stated a key dictum that easily applies to scientific reasoning: 'No concept without percepts, no percepts without concepts'. This implies a sort of progression in which concepts yield new percepts and percepts yield new concepts, oscillating back and forth in a never ending series of interchanges, some of which bring on shifts of theoretic paradigms. This is precisely how scientific theory advances. As applied to such research, Kant's dictum may be restated: No science without theory, no theory without observation—or even better, no theory without measurement.

Theory itself is purely conceptual. That is, a theory is made up of concepts. It is not made up of observations no matter how relevant they may be for a given theory. More specifically, theory is a set of inter-related concepts. A theory in an empirical science is made up of a set of inter-related concepts purporting to explain the behavior of a domain of observed phenomena. It consists of all the interdependent sets of well reasoned concepts and their conceptual linkages. In the empirical sciences theory is tied to empirical observations. This is true of all such sciences.

In the physical sciences the tie between a theoretic concept and the observed phenomenon to which it refers is taken for granted. That is, physical scientists have become so adept at measuring the referents to their conceptual variables that concept and measurement are isomorphic.

Not so, unfortunately, in the sociological sciences. In these fields the connection between concept and observation is often widely separated, with so-called 'proxies' and 'indicators' being quite common. Despite this, among sociological scientists much effort has gone into determining the extent to which the operational definition of a conceptual variable is a good measure of it. Techniques for testing the validity of an operational definition are well known. Similarly, considerable effort has gone into determining the extent to which the measurement of an operational form of a conceptual variable can be trusted. Like those of validity, techniques for testing the reliability of the operational form of a conceptual variable are well known.

Even so, though well known, such devices are not often used by sociologists, demographers, political scientists, etc. In part this is due to the widespread dependence on so-called 'secondary data', data sets designed for one set of purposes that are used for another. Many of these data sets (maybe most of them) are designed to serve the purposes of governments. Often their validity is assumed, not tested. And tests of instrument reliability seem rare indeed.

Still, it's easy to see why secondary data are so frequently used. Researchers often aim at generalizing to specifiable universes, such as a nation. The cost of doing this, even in medium-sized nations, is so large that only a government can cover it. But there is also a cost to using them. The measurement devices that can be constructed from them may not be very precise...