Connectivity of Implicit Statistical Learning Networks is Associated with Language Impairment Severity in Post-Stroke Aphasia

Rehabilitation of language for individuals with post-stroke aphasia (PSA) requires relearning to use cognitive-linguistic skills, but little is known about how these individuals relearn, or how brain reorganization may support different types of learning. One type of learning critical to language acquisition is implicit-statistical learning (ISL), which supports robust, accurate performance with minimal conscious effort. ISL relies on brain subnetworks that include language areas as well as domain general regions (i.e., salience and cognitive control) with the latter considered less likely to be damaged in individuals with PSA. Using three brain subnetworks associated with ISL, the present study used graph theory metrics to (1) compare their functional connectivity in PSA (n = 30) relative to neurologically healthy controls (n=18), and (2) examine their relationship with aphasia severity in the PSA group. Resting-state functional connectivity was weaker for the language and salience subnetworks in PSA than controls, but not for the cognitive control subnetwork. Additionally, edges within all three networks correlated with aphasia severity, particularly between regions of the right frontal cortex, such that stronger connectivity in these ISL subnetworks were associated with milder aphasia. While it is unsurprising that the damaged left-sided brain regions in the PSA group resulted in connectivity that differed from controls, these differences were also observed in the salience subnetwork. Further, the correspondence to aphasia severity was evident in all subnetworks, highlighting that domain general ISL networks may contribute to variability in language outcomes.