Introduction: Five classes of perceived family functioning have been identified among cancer patients in an Australian palliative care setting, spanning domains of cohesion, expressiveness, and conflict. Four classes have been identified among Japanese patients. Poor family functioning has been shown to relate to higher levels of psychosocial morbidity (Kissane et al., 1996). We used cluster analysis to determine the number of classes among American patients. Methods: Advanced cancer patients within the palliative care setting of a major cancer institute (N=1809) completed the Family Relationships Index (FRI; Moos & Moos, 1981) as eligibility screening for an RCT testing the efficacy of a family grief intervention. The average patient was middle-aged and Caucasian, with a primary diagnosis of GI cancer. K means and hierarchical agglomerative cluster analyses estimated the number of classes to extract across the three FRI subscales of cohesion, expressiveness, and conflict. The model-based cluster analysis bayesian information criterion (BIC) estimated an optimal number of classes. To examine clinical relevance of potential models, one-way ANOVAs tested relations between family functioning and indicators of psychosocial morbidity (e.g., depressive symptoms reported on the BDI-II). These indicators were completed by patients eligible for the RCT at a baseline measurement, prior to randomization. Results: The optimal statistical model contained seven classes of family functioning. However, poorer family functioning in a five-class model was significantly associated with heightened depressive symptoms (p=.016) whereas this was not the case for the seven-class model or a six-class model. Conclusions: Cultural differences among the American sample may account for the taxonomic differences in family functioning when compared to Australian and Japanese samples. In line with prior research, classifying patients by family functioning may identify those at risk for psychosocial morbidity.