Experiment

In social sciences, an experiment is a research strategy used by a social scientist to establish causal relationships between one or more independent variables and one or more dependent variables. An independent variable is a variable that is manipulated by a researcher; its causal impact on the dependent variable is investigated by the researcher. An independent variable is alternatively called the treatment variable or factor. The dependent variable is the observed phenomenon or measurement that has been affected by the manipulation of the independent variable. Dependent variables are alternatively referred to as outcome or criterion variables.

In establishing the causal relationship between independent variables and dependent variables, the researcher should design an experiment that includes the following elements:

• Manipulation of the amount (as in the case of quantitative independent variables) or the level of the independent variable (as in the case of qualitative independent variables)
• Control of nuisance (or confounding) variables using random selection and random assignment of subjects into treatment conditions
• Careful recording or observation of the change in the dependent variable

The first and the second requirements are achieved in either laboratory studies or field experiments (see FIELD EXPERIMENT); they distinguish the experimental research strategy from other research strategies such as quasi-experimental studies, surveys, or naturalistic studies. For these reasons, experiments that possess the above three characteristics are sometimes called true experiments (Campbell & Stanley, 1966).

An example of a true experiment is reported in Cordova and Lepper’s (1996) study of elementary school children’s learning of arithmetical order-of-operations rules. All learning activities took place in a computerized environment. First, children were randomly assigned, within gender, to either a control condition or one of four experimental conditions. In the [Page 350]control condition, learning materials were presented abstractly. In the four experimental conditions, identical materials were contextualized in a meaningful and visually enhanced format. For half of the students in the four experimental conditions, the learning context was individually personalized; the other half were presented with a generic format. Furthermore, in each experimental condition, half of the group was given choices regarding instructional aspects of the learning context while the other half was not. Thus, the four experimental conditions were (a) learning context personalized with choices, (b) learning context personalized without choices, (c) learning context not personalized but with choices, and (d) learning context not personalized and also without choices. In sum, three independent variables were manipulated in the study: contextualization, personalization, and choice. The dependent variables were several, including students’ intrinsic motivation for learning, depth of engagement during learning, the amount of learning acquired in a fixed period of time, students’ perceived self-competence, and levels of aspiration. The causal relationship between three independent variables and multiple dependent variables was established through (a) random assignment of subjects into control and experimental conditions; (b) control of miscellaneous variables such as grade level of subjects, gender, and learning materials; and (c) statistical theory of inference making.

Indeed, inferential statistical theories enable social and behavioral scientists to make sense of empirical data collected in the so-called true experiments, in which causal relationships are established not by ruling out all possible alternative explanations or controlling all possible confounding variables, but by probabilities. The logic behind inferential statistical theories is testing statistical hypotheses, in the light of data, followed by making inferences about the underlying populations in the tradition of inductive logical reasoning (Maxwell & Delaney, 1990). The principles and statistical theories that have guided experimental research studies in social and behavioral sciences were, to a great extent, formulated by Karl Pearson (1857–1936), Sir Ronald A. Fisher (1890–1962), and Jerzy Neyman (1894–1981).

...