Chapter 9: Experimental Designs: Within-Subjects Design

Within-subjects experimental design represents a research methodology in which all participants experience every treatment condition, allowing researchers to measure the same individuals across multiple levels of an independent variable. This approach offers substantial statistical advantages by eliminating between-group differences in participant characteristics, thereby reducing error variance and increasing the power to detect genuine treatment effects. The design requires fewer total participants than between-subjects alternatives while providing more direct comparisons within each individual. However, within-subjects designs introduce distinct threats to internal validity that researchers must carefully manage. Time-related confounds such as history effects, maturation, instrumentation changes, and statistical regression can compromise result validity when conditions are administered sequentially. Order effects—including practice effects that improve performance across successive conditions, fatigue effects that degrade performance, and carryover effects where experiences in one condition influence subsequent conditions—present particular challenges requiring systematic control strategies. Counterbalancing techniques address these threats by varying the sequence in which participants encounter treatment conditions. Complete counterbalancing presents all possible orderings, partial counterbalancing uses selected orderings, and Latin square designs provide efficient systematic arrangements that balance order while reducing the total number of required sequences. Matched-subjects designs offer a hybrid approach that combines elements of within-subjects and between-subjects strategies by grouping participants based on relevant characteristics before assigning them to conditions. Statistical analysis of within-subjects data requires specialized tests appropriate for repeated measurements: repeated-measures analysis of variance accommodates multiple treatment conditions, repeated-measures t-tests compare two conditions, and non-parametric alternatives such as sign tests and Wilcoxon signed-ranks tests apply when data violate normality assumptions. Understanding when to implement within-subjects designs and how to properly control methodological threats remains essential for conducting rigorous experimental research.