Within-Subjects Design

Advantages and Disadvantages

In experimental settings, subjects often differ substantially from one another for reasons unrelated to the studied factor(s). In between-subjects designs, differences among subjects are uncontrolled and are treated as random error. By contrast, in within-sub-jects designs, individual differences can be separated from the error term, as the same subjects are observed under multiple conditions, and thus, each subject can serve as his or her own control. Consequently, a major advantage of within-subjects designs is their ability to control for individual differences or heterogeneity among subjects. Moreover, one can obtain substantially greater information by collecting multiple scores from each subject. A practical implication of both the removal of individual differences from the error term and the collection of a larger number of observations from the same subjects is that statistical power can be increased to detect treatment effects in within-subjects designs.

Along with these important advantages, within-subjects designs also bring complexities and challenges to data analysis. In some experiments, treatment conditions are administered simultaneously in such a way that their order is not relevant. More commonly, however, treatments are administered one after the other and thus the order of treatments is a matter for concern because subjects might systematically change depending on the order. Order effects or practice effects exist if treatments administered earlier to a subject continue to have an effect that carries over to influence the subject's behavior during subsequent [Page 1639]treatment conditions. For example, there might exist a general improvement on related tests (e.g., learning effect) or a decrease in performance (e.g., fatigue effect). A strategy commonly implemented to prevent order effects from contaminating the treatment effects is counterbalancing, which is a technique of ordering sequences of conditions so that each treatment is administered first, second, third, and so on, an equal number of times across subjects. This type of design is known as a crossover or counterbalanced design. The logic behind the crossover design is that any main effect of order will be controlled for and, as a result, the crossover within-subjects design possesses stronger internal validity than a design without counterbalancing.

...