Slide Slam H4
Lesion Correlates of Auditory Sentence Comprehension Deficits in Post-Stroke Aphasia
Daniel Mirman1, Erica Adezati1, Melissa Thye1, Jerzy P. Szaflarski2; 1Department of Psychology, University of Edinburgh, Edinburgh, UK, 2Department of Neurology and the University of Alabama at Birmingham Epilepsy Center, University of Alabama at Birmingham, Birmingham, AL, USA
Auditory sentence comprehension requires coordination of multiple levels of processing: auditory-phonological, lexical-semantic comprehension, syntactic and discourse processing, as well as executive functions such as verbal working memory (WM) and cognitive control. The phonological and lexical aspects appear to be supported by a “ventral stream” of regions in the lateral temporal lobe (e.g., Hickok & Poeppel, 2007), but the critical regions are less clear at the sentence level. Left inferior frontal cortex, inferior parietal cortex superior, and posterior superior temporal cortex are the most consistently implicated regions for sentence-level comprehension deficits (for a review see Wilson, 2017). Most prior studies on this topic have either used composite measures of comprehension or specifically focused on syntactically complex (or non-canonical) sentence structures. In the present study, sentence comprehension was measured using the Token Test (De Renzi & Vignolo, 1962), which does not require complex syntactic or semantic processing, but it does require combinatorial processing, sequencing, verbal WM, and possibly semantic control (repetition of items from semantic categories tends to produce competition that requires control resources to resolve, as seen in blocked cyclic naming and in “access” deficits more generally; e.g., (Mirman & Britt, 2014)). Prospectively collected MRI and psycholinguistic data from 50 participants with aphasia secondary to a single left hemisphere stroke were analysed (see also Thye et al., 2021). Lesion-symptom mapping (LSM) analyses were conducted in R (version 3.5.1) using the LESYMAP package (version 0.0.0.9220). All analyses only considered voxels with at least 10% (n=5) lesion involvement and were corrected for overall lesion size. Mass-univariate analyses were also corrected for multiple comparisons using continuous permutation-based FWER with p<0.05 of observing more than v=100 false positive voxels (Mirman et al., 2018). Multivariate LSM was conducted using SCCAN with 4-fold cross-validation to optimise sparseness (Pustina et al., 2018). The mass-univariate VLSM identified 4,142 suprathreshold voxels, primarily in the posterior superior temporal lobe (STG, MTG, and Heschl’s gyrus) and a smaller cluster in IFG pars triangularis. Multivariate SCCAN LSM also identified a relatively sparse solution (optimal sparseness = 0.173, CV correlation = 0.55, p<0.0001) with suprathreshold voxels in posterior superior temporal cortex, IFG pars triangularis, and frontal white matter (superior and anterior corona radiata). These results converge with prior work that identified posterior superior temporal and inferior frontal regions as critical for sentence comprehension. We did not observe involvement of anterior temporal regions, perhaps because the semantic demands were quite limited. Nor did we observe involvement of inferior parietal regions, perhaps because the demands for hierarchical syntactic processing were limited (Matchin & Hickok, 2020). Rather, the critical regions are ones associated with speech perception, verbal WM, and semantic control, suggesting that these are the primary drivers of Token Test performance. The Token Test is also used in clinical contexts, so in addition to providing further insight into the neural correlates of sentence comprehension deficits, these results are relevant for interpreting Token Test results in clinical settings.