ABSTRACT
The manipulation of the independent variable distinguishes a true experimental design from pre-experimental designs, and from non-experimental designs. Ideally, the experimental and control groups should be identical except for the experimental treatment. Any other difference between the conditions leaves room for an alternative explanation for the results. One should be careful not to compromise random assignment, even though an experimental procedure may take longer or be more cumbersome than a control procedure. For example, an experimental condition may require the presence of a confederate. An especially serious kind of bias can result when subjects assign themselves to particular experimental conditions. Most experimenters are aware of the problems of different refusal rates in different experimental conditions. There are two basic types of experimental designs, single-factor, and multi-factor or factorial designs. Mortality for non-experimental reasons should be taken care of by random assignment to conditions.
