This paper examined the psychometric properties of the Interpersonal Relationship Inventory with a sample of homeless men who were first-time shelter users. The measure, based on social exchange theory and equity theory, has two subscales, one for support and the other for conflict. As yet, few measurements have been developed to assess conflict from within social networks. As part of a larger longitudinal study a sample of 166 first-time shelter users was administered the inventory. A student sample also filled out the inventory for comparison purposes. Internal consistency for both subscales was good, .90 for Support and .83 for Conflict. Weak interitem correlations were prevalent among many items in the Conflict subscale. A nearly zero correlation was found between scores on the Support and Conflict subscales, suggesting independence. An exploratory factor analysis using varimax rotation confirmed a dual factor structure. Analyses of variance and Scheffé contrast tests detected no statistically significant differences between age groups on either the Support or the Conflict subscales. A multiple regression analysis indicated that, when controlling for age, scores on the Support scale discriminated the homeless and student samples, while scores on the Conflict subscale do not. The Support subscale indicates excellent psychometric qualities, while the Conflict subscale should be refined.