In this paper we illustrate a new method for measuring family relationship through the structural equation modeling (SEM), considering both family members’ points of view and dyadic relationships. Starting from the model developed by Cole and colleagues (Cole & Jordan, 1989; Cole & Mcpherson, 1993), we proposed and tested a new measurement model in which the family dimension is operation- alized as a second-order latent variable. We applied the measurement model to companionship — a fundamental feature of close relationships which had previously been investigated exclusively as an in- dividual construct. The different dimensions of relational construct predict the adjustment of the family members differently. The first aim of the study is to test three different measurement models to identify the individual, dyadic, and family dimensions of companionship, and the second whether companion- ship as a multidimensional construct is related to depression in family members. The participants are 107 family triads who were asked to fill out a questionnaire containing Companionship subscale of Network of Relationships Inventory (NRI; Furman & Buhrmester, 1985) and Depression Scale (Radloff, 1977). Results reveal that the best model for measuring family companionship is that investigated in our study which hypothesizes family dimension as a second-order factor, and that the family dimension of companionship is related to depression in family members.