Good, Bad, and Beyond: Perils of Partialing Positive and Negative Couple Variables Within Individuals

Partialing correlated predictors to test independent effects is an essential tool in couple research. In actor-partner models, partners' parallel scores are partialed in tests of unique associations with outcomes. Correlated aspects of couple functioning are also often partialed within individuals to examine separate effects. Partialed versions of measures are typically interpreted as assessing the same construct as original unadjusted variables, but in fact their meaning can change. Extending a prior report on changes in construct validity resulting from partialing partners' parallel scores, the present analyses examined effects of partialing measures of positive and negative relationship variables within individuals, 4 specifically perceived support from the partner and relationship conflict. In 300 middle-aged and older couples, we utilized participants' interpersonal circumplex ratings of their partner's typical behavior during marital interactions to compare interpersonal correlates of unadjusted and withinperson partialed forms of the Quality of Relationships Inventory Support and Conflict scales. Compared to unadjusted scores, partialed support scores (i.e., adjusted for conflict) were substantially less closely associated with ratings of partner's warmth, and were associated with a less submissive (i.e., less agreeable, cooperative) form of warmth. Compared to unadjusted scores, partialed conflict scores (i.e., adjusted for support) were substantially less closely correlated with ratings of the partner's hostility, and were associated with a more controlling (i.e., critical, coercive) form of hostility. Results were nearly identical for wives and husbands. Partialing correlated scores within individuals can alter the construct validity of relationship measures, suggesting the need for careful reporting and interpretation in couple research.