In a world that is becoming more multilingual, bilingual aphasia is a clinical problem with a major clinical impact. However, at present we lack causal explanations of the many features of recovery patterns and there is no consensus about the language in which the patient should receive speech therapy. Further advance requires an understanding of the dynamics of recovery. In a novel longitudinal, single-case study, we combine fMRI and dynamic causal modeling to examine the effects of specific language treatment for picture naming on the representation and control of language areas during the course of recovery. Improved performance in the treated language was associated with increased activation in language areas. Consistent with theoretical expectations, causal modeling indicated increased connectedness of the control and language networks for the treated language. This functional approach holds great promise for investigating recovery patterns and the effects of specific language treatment in bilingual aphasic patients.