Vocabulary relearning in aphasia is supported by hippocampal memory and cortical language systems

Speech and language therapy can be an effective tool in improving language in post-stroke aphasia. Despite an increasing literature on the efficacy of language therapies, there is a dearth of evidence about the neurocognitive mechanisms that underpin language re-learning, including the mechanisms implicated in neurotypical learning. Neurotypical word acquisition fits within the idea of Complementary Learning Systems, whereby an episodic hippocampal system supports initial rapid and sparse learning, whilst longer-term consolidation and extraction of statistical regularities across items is underpinned by neocortical systems. Therapy may drive these neurotypical learning mechanisms, and efficacy outcome may depend on whether there is available spared tissue across these dual systems to support learning. Here, for the first time, we utilised a reverse translation approach to explore these learning mechanisms in post-stroke aphasia, spanning a continuum of consolidation success. After three weeks of daily anomia treatment, 16 patients completed a functional magnetic resonance imaging protocol; a picture naming task which probed (i) premorbid vocabulary retained despite aphasia, (ii) newly re-learned treated items and (iii) untreated/unknown and therefore unconsolidated items. The treatment was successful, significantly improving patients naming accuracy and reaction time post-treatment. Consistent with the Complementary Learning Systems hypothesis, patients overall naming of treated items, like that of controls when learning new vocabulary, was associated with increased activation of both episodic and language regions. Patients with relatively preserved left hemisphere language regions, aligned with the control data in that hippocampal activity during naming of treated items was associated with lower accuracy and slower responses - demonstrating the shifting division of labour from hippocampally-dependent new learning towards cortical support for the efficiently-named consolidated items. In contrast, patients with greater damage to the left inferior frontal gyrus displayed the opposite pattern (greater hippocampal activity when naming treated items was associated with quicker responses), implying that their therapy-driven learning was still wholly hippocampally reliant. Open accessFor the purpose of open access, the UKRI-funded authors have applied a Creative Commons Attribution (CC-BY) licence to any Author Accepted Manuscript version arising from this submission.