This study investigated the extent to which researchers and clinicians can obtain valid retrospective self-reports of couples’ conflict interactions outside a laboratory setting. A distinction was made between relationship attribute variance, regarding a shared perspective of both partners, and informant-specific variance, regarding the unique vantage point of each partner. By examining convergent and divergent associations for each type of variance, this study clarified the risk that responses might be influenced by informant-specific biases related to levels of relationship satisfaction. This study also investigated potential moderators of validity. Participants included both members of 269 married and cohabiting couples (538 individuals) who completed online questionnaires. Results were analyzed using a correlated trait–correlated method minus one model. The total true variance included large components of both shared relationship attribute variance and informant-specific variance. Although the shared component was moderately correlated with relationship satisfaction, the informant-specific component was mostly distinct from satisfaction, suggesting minimal bias. Convergent correlations between partners were strong and mostly unrelated to potential moderating variables, albeit slightly smaller than reported in studies conducted in laboratory settings. The results generally support the validity for retrospective self-reports of conflict interactions, especially when reports are obtained from both members of a couple.