Marital separation can be a time of extreme stress and can significantly disrupt support networks and support processes when they are needed most. Reliance on self-report measures of social support has limited our understanding of how support takes place following transitions. The purpose of this study was to test mechanisms of support erosion in a longitudinal multimethod study of 138 single mothers and their confidants. Previous work is extended by employing observational measures of confidant support in a process model including 1) maternal characteristics, 2) support person characteristics, and 3) relationship characteristics. An hypothesized mechanism of erosion was supported in which maternal distress was positively associated with levels of confidant negativity, which in turn was related to lower levels of support. A path via intimacy was partially supported, in that chronic distress was related to lower levels of intimacy; however, contrary to expectation, intimacy was related to lower levels of observed support. Lower levels of observed report were found for intimate partners than for friends or family member confidants.