Experiment/Experimental Methods

An experiment is what John Hocking, Don Stacks, and Steven McDermott have labeled the sine qua non of all research. The experiment is a carefully controlled research study that allows the researcher to state within certain limits that a relationship exists between two elements and what causes the relationship to change. The experiment, as John Pavlik has lamented, is underrepresented in public relations, partly because of a limited theoretical base and the applied nature of public relations. Another reason for its lack of use may be the 24/7 nature of public relations practice.

The experiment is a very powerful research method that is extremely limited in its ability to generalize to other situations. The limitation is based on the fact that the experiment requires as much control as can be placed on the actual implementation of the study. When an experiment is conducted, the researcher attempts to keep all other influences constant, varying only what is manipulated (the “independent” variable) and observing its effect on the “dependent” variable. The dependent variable, then, is dependent for its result on the manipulation of the independent variable. Because most experiments attempt to test a theory that makes predictions based on the relationship(s) between the independent variable and the dependent variable, an experiment predicts—or hypothesizes—the expected relationships. On the basis of a theory, the experiment is conducted, data are collected, and the results are submitted to a statistical analysis in a way that allows the researcher to state with a certain degree of confidence that the experiment's results were based on hypothesized relationships and not on measurement or other types of error. Error is what the experiment attempts to reduce when the researcher establishes control.

There are a number of different experiments, each based on the degree of control placed on the [Page 311]study. The most controlled experiment is the laboratory experiment, where the researcher attempts to keep all possible sources of error—“spurious variables,” as Donald Campbell and Julian Stanley (1963) call them—out of the study. Relaxing control somewhat results in an experiment that simulates what happens when other variables are allowed to influence the hypothesized relationships. Relaxing the study even more results in a field experiment, where the study is conducted in a natural environment, but still with certain degrees of control.