The neurofunctional correlates of sentence processing: focus on difficulties of morphosyntactic processing and thematic role assignment in aphasia

Left hemisphere damage is a frequent cause of aphasia. Analyses of deviant linguistic behaviors provide valuable information about the functional architecture of language. Correlating specific language difficulties with damage to the brain helps shed light on the relationships between language and the neural substrate. The aim of this Ph.D. thesis is to contribute to the understanding of the neural correlates of sentence comprehension, based on behavioral and neuroimaging evidence from aphasia. A substantial amount of research based on lesion-symptom mapping has been devoted to this issue, but several issues remain to be clarified. To consider just an example, lesion-symptom mapping studies have systematically linked the posterior regions of the left hemisphere to sentence comprehension. Surprisingly, however, the same studies failed to provide similarly strong evidence for prefrontal regions, contradicting the results of previous neuropsychological investigations that clearly supported the critical role of these regions in sentence processing. To date, there are enough controversial issues on sentence processing as to warrant reconsideration of available evidence. The present project focused on the neural correlates of the mechanisms involved in thematic role assignment and in the processing of morphosyntactic features. This is because both sets of mechanisms are critical for sentence interpretation both in comprehension and in production. The first step of the project consisted of a systematic literature review and meta-analysis of lesion-symptom investigations of sentence processing (study 1 – Chapter 1). The literature search yielded 43 studies eligible for review, of which 27 were used in the meta-analysis. The main goal was to identify the correlates of thematic role assignment and of morphosyntactic processing. Thematic role assignment errors correlated mainly with damage in the left temporo-parietal regions, and morphosyntactic errors mainly with damage in the prefrontal regions. However, careful consideration of the reviewed and meta-analyzed studies shows that conclusions are biased under several aspects. Data on thematic deficits are based almost exclusively on sentence comprehension, and data on morphosyntactic deficits on sentence production. Furthermore, even the very few studies that evaluated both impairments did so in distinct linguistic contexts, or in different response modalities. In addition, studies that focused on one set of mechanisms did not consider the possibility that performance on their dimension of interest was influenced by damage to the other. For example, studies focusing on thematic comprehension administered thematic foils, but not morphosyntactic foils. Therefore, the neurofunctional correlates emerging from the meta-analysis and the review may offer a biased and/or partial view. As a first attempt at overcoming these limitations, a lesion study on native speakers of Italian with post-left stroke aphasia was conducted (study 2 – Chapter 2) to clarify the neural substrates of morphosyntactic and thematic processes in comprehension. Experimental stimuli consisted of simple declarative, semantically reversible sentences presented in the active or passive voice. In an auditory sentence comprehension task, participants were asked to match a sentence spoken by the computer to the corresponding picture, that had to be distinguished from a thematic, a morphosyntactic or a lexical-semantic foil. Thirty-three left brain-damaged individuals (out of an initial sample of 70) were selected because they fared normally on lexical-semantic foils, but poorly on morphosyntactic (n=15) and/or thematic (n=18) contrasts. Voxel-based Lesion Symptom Mapping (VLSM) analyses retrieved non-overlapping substrates. Morphosyntactic difficulties were uninfluenced by sentence voice and correlated with left inferior and middle frontal damage, whereas thematic role reversals were more frequent on passives and correlated with damage to the superior and middle temporal gyrus and to the superior occipitolateral cortex. Both correlations persisted after covarying for phonological short-term memory. When response accuracy to passive vs active sentences in the presence of thematic foils was considered, portions of the angular and supramarginal gyrus were retrieved. They could provide the neural substrate for thematic reanalysis, that is critical for comprehending sentences with noncanonical word order. However interesting and strong, these results were obtained by considering just one sentence type (declaratives) and by relying on basic neuroimaging data. To go beyond these limitations, the final step of the project relied on more comprehensive behavioral analyses and more advanced neuroimaging techniques (study 3 – Chapter 3). The SCOPRO (Sentence Comprehension and PROduction) language battery was developed, that focuses on thematic and morphosyntactic processes and allows assessing these processes in a variety of reversible sentences in both comprehension and production. SCOPRO was administered to 50 neurotypical subjects (to assess applicability and establish cutoff levels) and 27 aphasic participants (native Italian speakers with left post-stroke aphasia). Of the latter, 21 were included in an MRI-based lesion-symptom mapping study. Results obtained in comprehension tasks were correlated with neuroimaging data (structural T1 and DWI). Lesion maps, disconnectome maps, tract disconnection probability and personalized deterministic tractography data demonstrated the involvement of grey and white matter. Thematic role reversals correlated to cortical damage in the left angular gyrus. They also correlated to cortical damage in the left supramarginal gyrus when controlling for single-word processing in a voxel-based disconnectome-symptom mapping analysis. Thematic errors were associated also with underlying white matter damage. Correlating the probability of tract disconnections and personalized deterministic tractography with thematic role performance involved the left arcuate fasciculus. The posterior segment was associated with thematic role reversals, even after controlling for morphosyntactic and single-word processing. The anterior segment was linked to accuracy on thematic roles when single-word processing was used as a covariate. The long segment also correlated with the level of thematic role performance, but the correlation was no longer present when morphosyntactic performance was used as a covariate. SCOPRO can be used not only to assess language processes in a broad sense (e.g., morphosyntactic vs thematic), but also to look into more detailed issues. Contrasting accuracy on declarative and comparative sentences is an interesting case in point. Both sentence types express reversible relations, but only declaratives require thematic role mapping. Hence, contrasting results between the two could help distinguish the correlates of role mapping from those of reversibility per se. The supramarginal gyrus was damaged in participants who fared poorly in both declaratives and comparatives but, interestingly, the aphasics with selective thematic difficulties had suffered damage to the posterior division of the middle temporal gyrus and to the angular gyrus, whereas those with selective difficulties on comparatives presented with lesions in the parietal and central opercular cortex. Clearly, these results are preliminary and require further investigation. It is unanimously accepted that sentence processing involves a large-scale network including frontal, temporal and parietal cortices and the underlying white matter pathways. The main contribution of the present project is that it allows articulating more detailed hypotheses on the role played by some components of the network during sentence comprehension. Results tie left frontal regions to morphosyntactic processing, posterior temporal regions to the retrieval of verb argument structure, and a posterior-superior parietal area to thematic reanalysis. Preliminary observations also suggest that different neural substrates could be involved in processing reversibility as such and when more specifically implemented in thematic roles. Further studies exploiting detailed behavioral tools like the SCOPRO battery and sophisticated neuroimaging techniques in larger samples will lead to a better understanding of language functions and their processing in the brain.