Independent Variable (In Nonexperimental Research)

A defining characteristic of every true experiment is the independent variable (IV), an adjustable or alterable feature of the experiment controlled by the researcher. Such variables are termed independent because they can be made independent of their natural sources of spatial and temporal covariation. The researcher decides whether or not the IV is administered to a particular participant, or group of participants, and the variable's magnitude or strength. Controlled variations in the IV are termed experimental treatments or manipulations. Variations in the presence or amount of an IV are termed the levels of the treatment. Treatment levels can be quantitative (e.g., 5, 10, 35 mg of a drug) or qualitative (participant assigned to Workgroup A, B, or C). IVs may be combined in FACTORIAL DESIGNS, and their interactive effects assessed. The IV is critical in the LABORATORY EXPERIMENT because it forms the basis for the inference of causation. Participants assigned randomly to different conditions of an experiment are assumed initially equivalent. Posttreatment differences on the DEPENDENT VARIABLE between participants receiving different levels of the IV arecausally attributed to the IV.

Crano and Brewer (2002) distinguish three general forms of IVs: social, environmental, and instructional treatments. Social treatments depend on the actions of people in the experiment, usually actors employed by the experimenter, whose behavior is scripted or controlled. In Asch's (1951) classic conformity studies, naïve participants made a series of comparative judgments in concert with others. In fact, these others were not naïve, but were accomplices who erred consistently on prespecified trials. In one condition, one accomplice whose erroneous report was different from that of his peers shattered the (incorrect) unanimity of the accomplice majority. Differences between unanimous and nonunanimous groups were interpreted as having been caused by controlled variations in the uniformity of accomplices' responses.

[Page 484]Environmental treatments involve the systematic manipulation of some feature(s) of the physical setting. In an attitude change experiment, for example, all participants may be exposed to the same persuasive communication, but some may receive the message under highly distracting circumstances, with noise and commotion purposely created for the experiment. Differences in susceptibility to the message between participants exposed under normal versus distracting conditions are interpreted causally, if participants were randomly assigned to the distraction or normal communication contexts, and the experimenter controlled the presence or absence of the distraction.

Instructional manipulations depend on differences in instructions provided to participants. For example, Zanna and Cooper (1974) gave participants a pill and suggested to some that they would feel tense and nervous, whereas others were informed that the pill would help them relax. In fact, the pill contained an inert substance that had no pharmacological effects. Differences in participants' subsequent judgments were attributed to variations in the instructional manipulation, the IV.

For strict experimentalists, factors that differentiate participants (e.g., sex, religion, IQ, personality factors), and other variables not under the control of the researcher (e.g., homicide rates in Los Angeles), are not considered independent and thus are not interpreted causally. However, in some research traditions, variables not under experimental control sometimes are suggested as causes. For example, a strong negative correlation between marriage rates in a society at Time 1, and suicides at Time 2, might be interpreted as marriage causing a negative impact on suicides. Marriage, it might be suggested, causes contentment and thus affects the likelihood of suicide. The opposite conclusion is untenable: Suicide rates at Time 2 could not have affected earlier marriage rates. However, such conclusions are usually (and appropriately) presented tentatively, insofar as a third, unmeasured variable might be the true cause of the apparent relationship. In the present example, economic conditions might be the true causal operator, affecting marriage rates at Time 1, and suicides at Time 2. Owing to the possibility of such third-variable causes, causal inferences based on correlational results are best offered tentatively. In the ideal experimental context, such extraneous influences are controlled, and hence their effects cannot be used to explain differences between groups receiving different levels of the IVs.

...