Right hemisphere language network plasticity in aphasia

Abstract The role of the right hemisphere in aphasia recovery has been controversial since the 19th century. Imaging studies have sometimes found increased activation in right hemisphere regions homotopic to canonical left hemisphere language regions, but these results have been questioned due to small sample sizes, unreliable imaging tasks, and task performance confounds that affect right hemisphere activation levels even in neurologically healthy adults. Several principles of right hemisphere language recruitment in aphasia have been proposed based on these studies: that the right hemisphere is recruited primarily by individuals with severe left hemisphere damage, that transcallosal disinhibition results in recruitment of right hemisphere regions homotopic to the lesion, and that increased right hemisphere activation diminishes to baseline levels over time. It is debated whether engagement of language homotopes reflects upregulation of weakly active right hemisphere processors in a bihemispheric language network, versus recruitment of new processors into the network. We address these issues in 76 chronic left hemisphere stroke survivors with ongoing or prior aphasia and 69 neurologically healthy older adults using a semantic decision fMRI paradigm that elicits reliable and strongly left-lateralized individual-participant language activation and adapts to require effortful performance irrespective of participant ability levels. Right hemisphere activation was greater in stroke survivors than controls, and related to younger age, left-handedness, and higher education. Right hemisphere activation magnitude was modest compared to left hemisphere activation. In contrast to prior proposals, right hemisphere activation was unrelated to lesion size and greater with longer time-since-stroke. Right ventral inferior frontal and mid-anterior temporal regions were weakly engaged in language processing in controls and co-activated with their homotopic left hemisphere counterparts. Lesions to those left hemisphere counterparts resulted in greater homotopic activation that contributed to naming and word reading outcomes. In contrast, the right dorsal inferior frontal cortex was not engaged during language processing in controls and did not co-activate with its left hemisphere counterpart. It exhibited the largest group-level difference in activation for stroke survivors relative to controls due to complex lesion-activation interactions, but the activation was unrelated to the aphasia outcomes tested here. In sum, right hemisphere language homotopes are recruited in some chronic left hemisphere stroke survivors, consistent with both upregulation of weakly active processors in a bihemispheric language network and new recruitment of the dorsal inferior frontal gyrus into the network. These findings clarify the mechanisms of, and constraints on, right hemisphere language network plasticity after left hemisphere stroke.