Removing the influence of feature repetitions on the congruency sequence effect: why regressing out confounds from a nested design will often fall short

This article illustrates a shortcoming of using regression to control for confounds in nested designs. As an example, we consider the congruency sequence effect, which is the observation that the congruency effect in distractor interference (e.g., Stroop) tasks is smaller following incongruent as compared with congruent trials. The congruency sequence effect is often interpreted as indexing conflict adaptation: a relative increase of attention to the target following incongruent trials. However, feature repetitions across consecutive trials can complicate this interpretation. To control for this confound, the standard procedure is to delete all trials with a stimulus or response repetition and analyze the remaining trials. Notebaert and Verguts (2007) present an alternative method that allows researchers to use all trials. Specifically, they employ multiple regression to model conflict adaptation independent of feature repetitions. We show here that this approach fails to account for certain feature repetition effects. Furthermore, modeling these additional effects is typically not possible because of an upper bound on the number of degrees of freedom in the experiment. These findings have important implications for future investigations of conflict adaptation and, more broadly, for all researchers who attempt to regress out confounds in nested designs