Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Lesion-related psycholinguistic effects in post-stroke agraphia based on a graded measure of spelling errors
Poster Session F, Friday, October 2, 2:45 - 4:45 pm, Wangari Maathai
Zoe White1, Ryan Staples2,3, Peter Turkeltaub1,2,3,4, Andrew DeMarco1,2,3; 11Department of Rehabilitation Medicine, Georgetown University Medical Center, 2Department of Neurology, Georgetown University Medical Center, 3Center for Brain Plasticity and Recovery, Georgetown University Medical Center, 4MedStar National Rehabilitation Hospital
Acquired agraphia reflects disruption of highly distributed and commingled phonological, orthographic, and semantic networks. Exaggerated psycholinguistic effects during spelling provides insight into damage to distinct network subcomponents. While psycholinguistic variables are known to influence spelling performance, their brain substrates remain poorly characterized in post-stroke agraphia. One reason for this may be that lesion studies of agraphia have relied on binary accuracy, which may fail to capture graded differences in spelling impairment. Here, we extend our prior work examining whole-word accuracy, relying on a measure of partial credit for errors (Levenstein Distance; LD). We hypothesized that LD would enhance sensitivity to lesion-related effects in post-stroke agraphia, providing additional insight into the brain networks for spelling. We studied 114 chronic left-hemisphere stroke survivors and 71 healthy controls. Participants completed a typing-to-dictation task consisting of 32 auditorily presented real words balanced for frequency, imageability, and regularity. Mixed-effects models of accuracy included fixed effects of group, psycholinguistic variables, and their interactions, while controlling for education, with random intercepts of participant and item. Multivariate lesion-symptom mapping identified lesion correlates of overall spelling performance and psycholinguistic effects (e.g., low-frequency words covarying for high-frequency performance), controlling for age and education (voxelwise p<.005, clusterwise p<.05, 5,000 permutations). Parallel analyses were performed for whole-word scoring and partial-credit (LD) scoring. Patients showed reduced behavioral accuracy relative to controls' scores both via whole-word (59% vs 90%) and by LD (0.26 vs 0.04 letters). Whole-word analyses found main effects of education (p = .01) and group (p<.001) but no psycholinguistic main effects or interactions. LD analyses also found a main effect of group (p<.001) plus both a frequency×group (p=0.03) and imageability×group interaction (p<.001). VLSM identified one cluster for overall whole-word and LD accuracy, both encompassing mid/posterior superior temporal gyrus (STG), including supramarginal gyrus (SMG, and underlying white matter (MNI = -53,-34,12, volume= 47.7cc and MNI=-52,-31,9, volume=41.3cc). LD analyses found a cluster for frequency (low controlling for high) encompassing STG/SMG (MNI=-53,-30,9, volume=35.9cc); for imageability (low controlling for high) encompassing supramarginal and angular gyrus (MNI=-50,-50,32, volume=15.8cc); and regularity (irregular controlling for regular) encompassing SMG (MNI=-41,-49,30, volume=6.8cc) and extending posteroventrally through subcortical white matter. Results support our hypothesis that scoring spelling errors for partial credit enhances sensitivity to psycholinguistic-based deficits in agrapha. LD behavioral models revealed frequency and imageability effects not evident using whole-word scoring. Moreover, the LD lesion models found results for each psycholinguistic variable, none of which were present in whole-word accuracy analyses. Temporoparietal lesions may have caused an exaggerated frequency effect due to damaging phonological processing and increasing reliance on frequency-sensitive lexical orthographic representations. Parietal lobe lesions may have caused an exaggerated imageability effect by also causing increased reliance on semantics, which benefits highly imageable words. Finally, lesions associated with an exaggerated regularity effect may reflect disruption of pathways to ventral occipitotemporal cortex areas specialized for wordform processing. Next steps involve examination of error types and self-corrections, and applying connectomic analysis to LD.
Topic Areas: Disorders: Acquired, Writing and Spelling