Coordinate-based meta-analyses and connectivity profiles of an implicit language learning brain network in healthy adults
Amy Ramage1, Kaila Cote1; 1University of New Hampshire
Baddeley (1993) proposed that rehabilitation should center on modifying a behavior through experience-based learning. However, approaches to understanding recovery of language in aphasia focus on characteristics of the brain and the presenting language deficits. Previous work has identified favorable brain-based biomarkers of recovery, but not how rehabilitation efforts enhance brain changes to result in favorable recovery of language. That is, the learning that is required for successful rehabilitative outcomes following brain injury is not often considered relative to how the damaged brain learns. Thus, we embark on a program of research that begins with identifying language learning paradigms and their corresponding brain networks. While language rehabilitation is not synonymous with learning a grammar, we focus first on learning known to be involved in language acquisition – implicit or statistical learning, to identify the network of brain regions actively involved in implicit language learning in healthy adults. Multiple coordinate-based meta-analyses (CBMA) were conducted to identify common and distinct brain activity across studies. Inclusion criteria for studies were: study of healthy adults 18+ years, use of implicit language learning tasks, reporting of coordinates for whole brain analyses, and reporting of experimental contrasts indicating rule learning. Papers were coded for meta-data and experimental contrasts were categorized as grammatical (rule-following) or non-grammatical (rule-violating). Activation likelihood estimation (ALE) was used to compute the activation probabilities across all voxels, generating regions of interest (ROI) that were common across studies. Connectivity amongst the ROIs was characterized by using them as seeds in task-independent and task-dependent functional connectivity analyses. Hierarchical clustering of the connectivity profiles grouped brain regions into subnetworks associated with grammatical/non-grammatical processes. Functional decoding further characterized the mental operations associated with those sub-networks. The CBMAs were conducted on 25 functional magnetic resonance imaging (fMRI) studies that used artificial or natural language learning tasks, with adjacent/non-adjacent or pairwise dependency learning, hierarchical, chunk-based, or phrase structure rule learning. ALE results identified bilateral ROIs in frontal, insular, and parietal cortex. However, regional activity did not differ when participants were distinguishing rule-following versus rule-violating stimuli. There was more left-lateralized activation as well as left superior temporal gyrus activation when the stimuli were rule-following (grammatical). Connectivity profiles indicated strong connections amongst the ROIs, as well as inclusion of subcortical structures (e.g., thalamus, pallidum, putamen), but the general pattern was consistent with structures of the frontoparietal network. These data support a left-dominant cognitive control network as a scaffold for grammar rule identification, maintenance, and rule application in healthy adults. This suggests that cognitive control is necessary to track regularities across stimuli and imperative for rule identification and application of grammar, as has been predicted by others reporting frontoparietal activity during language learning tasks. The frontoparietal brain network is domain general and largely overlaps with the cognitive control (e.g., multi-demand) networks, the integrity of which is known to be a positive prognostic sign for language recovery in individuals with aphasia. Future study will determine whether connectivity of the additional subcortical structures, and other temporal structures, are also relevant to successful language learning.
Topic Areas: Disorders: Acquired, Control, Selection, and Executive Processes