A Cautionary Note on the Interpretation of Relationship Effects in the Social Relations Model

Abstract

Kenny and La Voie's Social Relations Model (1984) is a promising approach to the analysis of social interaction data which simultaneously and independently assesses individual differences and relationship-specific effects among interacting subjects. Unfortunately, some authors have used one-replication studies (which confound error and relationship-specific variance) and have interpreted the confounded relationshiplerror variance component as a relationship effect. We reanalyzed behaviorally coded data from four experiential groups. Side by side one-replication and two-replication Round Robin Analyses of Variance of two behaviors demonstrated the danger of interpreting the eonfounded relationship/error component of one-replication analyses as a relationship effect. With questioning behavior, one-replication analyses suggested a relationship-specific effect, but a two-replication analysis indicated that this finding was error variance. For self-disclosing behavior, the two-replication analysis confirmed the suggested relationship effect from the one-replication analyses. Researchers interested in relationship-specific effects using the Social Relations Model should design their studies to allow for two or more replications and report relationship variance terms unconfounded with error.