Generalization to untrained items is one measure of therapeutic success in aphasia treatment research. Protocols that capitalize on theorized relationships within the semantic system are more likely to promote generalization. Previous work based on the complexity account of treatment efficacy (CATE; Thompson et al., 2003) and leveraging the unique relationship between abstract and concrete words has shown that when abstract words are trained, concrete words also improve, but not vice versa (e.g., Kiran et al., 2009). Additionally, previous work based on the revised hierarchical model (RHM; Kroll & Stewart, 1994) has shown that training in the non-dominant language promotes cross-language generalization in bilingual persons with aphasia (e.g., Edmonds & Kiran, 2006). This work replicates previous work showing the generalization benefit of training abstract words in monolingual aphasia, successfully extends this abstract word training protocol to bilingual aphasia, and shows the additive benefit of training in the non-dominant language in bilingual aphasia.