More than 2 million people in the U.S. have aphasia, a language disorder most often caused by stroke that reduces participation in preferred activities, functional independence, and health-related quality of life. Language therapy for aphasia is efficacious, but outcomes vary across patients, presenting challenges for treatment-planning and prognostication for recovery. Recent evidence suggests brain network properties derived from functional connectivity data and quantified via graph theory may help explain this variability and predict treatment outcomes. However, only a few studies have used graph theory to investigate aphasia and the relationships between graph metrics, stroke- related brain damage, and patients’ response to specific types of intervention remain unclear. This study seeks to address these knowledge gaps by leveraging two potentially informative graph metrics, modularity and global efficiency, which characterize the brain’s segregation into functionally distinct subsystems and its capacity to integrate information among separate regions, respectively. To advance knowledge of the relationship between brain damage and neural function in aphasia, this study will determine the association between lesion size and modularity and global efficiency in Veterans with chronic aphasia (Aim 1). Additionally, to inform predictive models of recovery, the study will determine if pre-treatment modularity and/or global efficiency are associated with outcomes from semantic feature analysis (SFA), a popular intervention for naming impairments (Aim 2a), and whether they provide unique predictive information relative to other neural and behavioral predictors (e.g., lesion size, pre-treatment aphasia severity, demographics) (Aim 2b). This study will include 10 Veterans with chronic aphasia due to left-hemisphere stroke, all of whom will undergo neuroimaging and treatment in a larger randomized clinical trial of SFA therapy. Specifically, participants will complete a language evaluation, structural MRI, and resting-state fMRI (RSfMRI) prior to receiving 60 hours of SFA therapy over 15 days. Treatment outcomes will be derived from pre- and post-treatment naming assessments of trained items. Lesion volume will be calculated from lesion maps drawn on participants’ structural scans. Functional connectivity-based brain graphs (i.e., network representations) consisting of nodes (i.e., 264 brain regions, per a parcellation scheme from Power et al., 2011) and edges (i.e., pairwise correlations in the BOLD signal over time between nodes) will be constructed from participants’ RSfMRI scans, and the modularity and global efficiently of each participant’s graph will subsequently be computed using the Brain Connectivity Toolbox. Aim 1 will be addressed by correlating lesion volume with modularity and global efficiency. Aim 2 will be addressed by regressing treatment outcomes on modularity and global efficiency (Aim 2a), as well as other predictive variables (Aim ...