Statistical adjustments (i.e., partialing) for between-partner correlations on the same variable and for correlations among multiple predictors within individuals are common in relationship research. Although useful, partialing can alter the construct validity or meaning of measured variables in ways that are typically not considered. In this study of 300 middle-aged and older couples, unadjusted, partner-partialed, partialed warmth and hostility component, and common fate model scores for observer-rated behavior during marital conflict discussions were compared using interpersonal circumplex-based spouse ratings of targets' behavior during those discussions, as well as self-reports of marital quality (i.e., overall marital adjustment, support from spouse, conflict). Unadjusted scores for observer-rated affiliation-and its two components, warmth, and hostility-had expected associations with interpersonal circumplex-based spouse ratings and self-reports of marital quality. Compared to unadjusted scores, partner-partialing, analogous to the actor-partner interdependence model, resulted in significantly weaker associations of behavioral scores with the expected interpersonal content of spouse ratings, and weaker associations with reported marital quality. Partialing of warmth and hostility within individuals also resulted in weaker associations with spouse ratings and marital quality, and some shifts in the theme of spouse ratings. In contrast, common fate model scores had expected associations with these criteria that equaled or exceeded the magnitude for unadjusted scores. Thus, common forms of partialing in relationship research can weaken the construct validity of behavioral observation variables, and common fate model scores represent a viable alternative in some instances. Implications for the design, reporting, and evaluation of couple research are discussed. (PsycInfo Database Record (c) 2026 APA, all rights reserved).
Smith et al. (Mon,) studied this question.