There were many failures before humans successfully learned to fly. After watching birds flap their wings, bold and adventurous individuals built huge winglike structures, leaped off cliffs, flapped their wings vigorously, and broke their necks. There are principles of flight to be learned from watching the birds all right, but the wrong analogy had been drawn. In similar fashion, our empirical approach to social behavior is based on an analogy. The natural sciences are incredibly appealing, the physical and biological sciences have been generating counterintuitive and esthetically elegant ideas for centuries. No wonder scholars thought: “Why not a science of...

A key feature of the experimental approach is that the subjects (whether they be individuals, groups, organizations, nations, or what have you) are randomly assigned to the conditions under study. If the assignment process is not random, then the investigator must be fully satisfied that it has no bearing on the likely outcome. In the social sciences, however, we are continuously dealing with situations in which the subjects have not been randomly assigned to the different conditions; rather, some selective process is operating which itself may be influencing the outcomes observed in the conditions under study. The ensuing simulation of...

Sociological research, in one form or another, iscomparativeresearch. If high values of X are thought to cause certain levels of Y to occur, then the truth is ascertained by comparing the Y outcome when X is at a medium or low level. Should one wish to determine if race or ethnic origin affects the judicial process, for example, it is not enough simply to examine the outcome for one particular group. Rather, the research must involve comparisons between groups who are matched as far as possible on all other attributes (witness the research on black-white differences reported in...

Some social researchers employ a probabilistic approach to causality; others use a deterministic approach. In either case, rarely is symmetric (or bidirectional) causality distinguished from asymmetric (or undirectional) causes. This is a different matter from the distinction between recursive and nonrecursive relationships, the former referring to models where “all the causal linkages run ‘one way,’ that is, if no two variables are reciprocally related in such a way that each affects arid depends on the other, and no variable ‘feeds back’ upon itself through any indirect concatenation of causal linkages, however circuitous” (Duncan 1975: 25). The symmetric/asymmetric distinction is, however,...

Virtually all social researchers want data with variation in the characteristics of interest to them.¹ The notion is that existing theories will then be validated, and new ones developed, by determining what variables help account for the variation found in the data set. These two steps in the research process—the search for data sets that incorporate variation and, in turn, the effort to account for the variation that is found—are so ingrained and widespread that we probably take them for granted. It turns out that the logic underlying these steps is more complicated and shakier than practitioners seem...

It is widely believed that the application of control variables is both desirable and possible. In one form or another, the control approach is viewed as a way of determining the influence some specific characteristic has on a dependent variable net of the influence due to other characteristics. A variety of statistical procedures is available to control or otherwise take into account the influence of variables.¹ This routine everyday procedure must be one of the most widespread practices found in contemporary social research. It is very much an accepted part of the logic of social science, even for those who...

This is the last chapter in which the primary goal is a critical evaluation of current research; sooner or later, we must start to consider alternative procedures.¹ The heart of my concern is again with the present-day pursuit of the undoable. This entails certaintheoreticalfacets to the control issue, but first we will examine the assumption that there is only one true or correct interpretation of empirical results. A study’s goals and the theory driving the research (the two arenotthe same) affect the interpretation of results in ways that are not appreciated at the moment. As matters...

And now what? As has been said before, the first step toward reaching a solution is to recognize that there is indeed a problem, that something is wrong. I have argued that several of the most widely accepted procedures in contemporary social research are often based on models and assumptions that are patently inappropriate and that, in such circumstances, their application is counterproductive. There is no point in arguing the matter further; the first part of this volume has either convinced the reader or not. Assuming there is agreement, then at the very least we must recognize that “business as...

Neither researchers nor theorists can continue to neglect the distinction between symmetric and asymmetric forms of causality. To my knowledge, most standard statistical research tools are applied with symmetrical causal relationships in mind. The researcher seeks to determine whether a change in one variable is associated with a change in another variable, possibly net of the influences of various other attributes. In any case, this is implicitly done in the form of a symmetrical test or description. The question of whether an increase in X yields an increase (or a decline) in Y is not distinguished from the question of...

Social researchers operate as if they have actually created an experiment in which the possible influence of certain variables is “controlled.” It is as if the researcher has used some sort of random assignment procedure or, if nonrandom, then at least one that is not selective on attributes that affect the processes under consideration. Our statistical analyses are in effect saying:here is what the influence of each variable is, had such results been found under conditions of random assignment.In point of fact, our social science knowledge is often telling us that there are no grounds for assuming that...

At the outset of a paper criticizing the use of cross-sectional methods (in terms not unrelated to the points made in chapter 9 of this volume), Carlsson describes the unavoidable linkage between the nature of what one wants to understand and the appropriate tools for reaching that understanding.

Research is a game against nature in which nature counters with a strategy of concealment, another name for the laws or dynamic principles holding in the area under study. Obviously, the effectiveness of a given strategy of discovery will depend on nature’s strategy of concealment, and conversely, the effectiveness of the laws...