Distinct neural correlates of morphosyntactic and thematic comprehension processes in aphasia

Abstract

Functional neuroimaging studies in neurotypical subjects correlate sentence comprehension to a left fronto-temporo-parietal network. Recent voxel-based lesion-symptom mapping (VLSM) studies of aphasia confirm the link between sentence comprehension and a left posterior region including the angular gyrus, the supra-marginal gyrus and the postero-superior division of the temporal lobe but support left pre-frontal involvement inconsistently. However, these studies focus on thematic role assignment without considering morphosyntactic processes. Hence, available VLSM evidence could provide a partial view of the neurofunctional substrate of sentence comprehension. In the present VLSM study, both morphosyntactic and thematic processes were evaluated systematically and in the same sentence types in each participant, to provide a more detailed picture of the sentence comprehension network. Participants (33 patients with post-stroke aphasia and 90 healthy controls) completed a sentence-picture matching task in which active and passive, declarative reversible sentences were paired with morphosyntactic, thematic and lexical-semantic alternatives. Phonological short-term memory tasks were also administered. Aphasic participants were selected from an initial pool of 70 because they scored below norm on thematic foils (n = 18) or on thematic and morphological foils (n = 15), but within the norm on lexical-semantic foils. The neurofunctional correlates of morphosyntactic and thematic processes were starkly distinguishable. Pre-frontal areas including the inferior and middle frontal gyrus were involved directly in processing local morphosyntactic features and only indirectly in thematic processes. When these areas were damaged, morphosyntactic errors always co-occurred with thematic errors, probably because morphosyntactic damage disrupts the assignment of grammatical roles and ultimately that of thematic roles. Morphosyntactic errors were not influenced by word order canonicity. In contrast, selective thematic role reversals were linked to temporal and parietal damage and were significantly influenced by word order, occurring on passive more than on active sentences. An area including the angular and supra-marginal gyrus was critical for processing non-canonical word order. In sentence comprehension, pre-frontal regions are critical for processing local morphosyntactic features (at least in simple declarative sentences). Temporal and parietal regions are critical for thematic processes. Postero-superior temporal areas are involved in retrieving verb argument structure. Parietal areas are critical for assigning morphosyntactically analysed constituents to the appropriate thematic role, thus serving a crucial function in thematic re-analysis. Each area plays a prevailing but not exclusive role in these processes, interacting with other areas in the network and possibly providing both the language-specific and the domain-general resources needed at various stages of sentence comprehension.

Keywords: lesion-symptom mapping; morphosyntactic processing; sentence comprehension; thematic re-analysis; thematic role assignment.

Graphical Abstract

Figure 1

Lesion overlap. Lesion overlaps are shown on sagittal, axial and coronal slices, with MNI152 coordinates (in mm) below each slice. Colours indicate the number of patients contributing to the overlap, as indicated in the colour bars. (A) Lesions in the entire group (33 patients) covered a large portion of the left hemisphere, with pre-frontal and parieto-temporal maxima; the insula, the parietal operculum, the supra-marginal gyrus and a small portion of the angular gyrus were also damaged in all patients. (B) In patients with morphosyntactic errors (n = 15), lesions overlapped mostly in the inferior and middle frontal gyri, with extensive white matter involvement. Pre-central and post-central cortices were also affected. (C) In patients with selective role assignment errors (n = 18), lesions overlapped mostly in parieto-temporal regions, including the angular gyrus, and extending to the lateral occipital cortex.

Figure 2

VLSM Brunner–Munzel tests: correlation with severity of morphosyntactic processing and thematic role assignment deficits in the experimental sample (n = 33). Voxels whose damage was significantly associated to poorer performance on morphosyntactic foils (A) or on thematic role assignment foils (B). The colour bars reflect the Z-scores of the significant clusters (P < 0.01, FDR corrected). Left hemispheric clusters are shown on 3D lateral views (top row) and on axial (middle row) and coronal slices (bottom row). MNI152 coordinates (in mm) along the z-axis (axial) and y-axis (coronal) are shown below each slice.

Figure 3

VLSM Liebermeister tests. Voxels whose damage is significantly associated with morphosyntactic difficulties (orange colour scale; n = 15), and voxels whose damage is significantly and selectively associated with thematic role difficulties (blue colour scale; n = 18). The colour bars reflect the Z-scores of the significant clusters (P < 0.05, FDR corrected). Clusters on the left hemisphere are shown, with MNI152 coordinates (in mm) shown below each slice (x on sagittal, z on axial, and y on coronal views).

Figure 4

VLSM: Brunner–Munzel tests with short-term memory as a covariate in the experimental sample (n = 33). Voxels whose damage was significantly associated to poorer performance on morphosyntactic foils (A) or on thematic role assignment foils (B), with short-term memory as a covariate. The colour bars reflect the Z-scores of the significant clusters (P < 0.01, FDR corrected). Left hemispheric clusters are shown on 3D lateral views (top row) and on axial (middle row) and coronal slices (bottom row). MNI152 coordinates (in mm) along the z-axis (axial) and y-axis (coronal) are shown below each slice.

Figure 5

VLSM Brunner–Munzel tests: correlation with thematic role assignment errors on passive sentences, with errors on active sentences as covariate in the experimental sample (n = 33). Voxels whose damage is significantly associated to poorer performance on thematic role foils in passive sentences, with thematic errors in active sentences as a covariate. The results of the VLSM analysis on thematic errors with short-term memory as a covariate were used as an inclusive mask. The colour bar reflects the Z-scores of the significant clusters (P < 0.05, FDR corrected). Clusters on the left hemisphere are shown, with MNI152 coordinates (in mm) reported below each slice (x on sagittal, z on axial, and y on coronal views).