Our understanding of neuroplasticity following stroke is predominantly based on neuroimaging measures that cannot address the subsecond neurodynamics of impaired language processing. We combined for the first time behavioral and electrophysiological measures and structural- connectivity estimates to characterize neuroplasticity underlying successful compensation of language abilities after left-hemispheric stroke. We recorded the electroencephalogram from patients with stroke lesions to the left temporal lobe and matched controls during context-driven word retrieval. Participants heard context sentences that either constrained the final word ("He locked the door with the") or not ("She walked in here with the"). The last word was shown as a picture to be named. We conducted individual-participant analyses and focused on oscillatory power as a subsecond indicator of a brain region's functional neurophysiological computations. All participants named pictures faster following constrained than unconstrained sentences, except for two patients, who had extensive damage to the left temporal lobe. Left-lateralized alpha-beta oscillatory power decreased in controls pre-picture presentation for constrained relative to unconstrained contexts. In patients, the alpha-beta power decreases were observed with the same time course as in controls but were lateralized to the intact right hemisphere. The right lateralization depended on the probability of white-matter connections between the bilateral temporal lobes. The two patients who performed poorly behaviorally showed no alpha-beta power decreases. Our findings suggest that incorporating well-understood direct measures of neural activity into investigations of neuroplasticity can provide important neural markers to help predict language recovery, assess the progress of neurorehabilitation, and delineate targets for therapeutic neuromodulation.