Abstract

Cognitive neuroscience exploring the architecture of semantics has shown that coherent supramodal concepts are computed in the anterior temporal lobes (ATL), but it is unknown how/where modular information implemented by posterior cortices (word/object/face forms) is conveyed to the ATL hub. We investigated the semantic module-hub network in healthy adults (n = 19) and in semantic dementia patients (n = 28) by combining semantic assessments of verbal and nonverbal stimuli and MRI-based fiber tracking using seeds in three module-related cortices implementing (i) written word forms (visual word form area), (ii) abstract lexical representations (posterior–superior temporal cortices), and (iii) face/object representations (face form area). Fiber tracking revealed three key tracts linking the ATL with the three module-related cortices. Correlation analyses between tract parameters and semantic scores indicated that the three tracts subserve semantics, transferring modular verbal or nonverbal object/face information to the left and right ATL, respectively. The module-hub tracts were functionally and microstructurally damaged in semantic dementia, whereas damage to non-module-specific ATL tracts (inferior longitudinal fasciculus, uncinate fasciculus) had more limited impact on semantic failure. These findings identify major components of the white matter module-hub network of semantics, and they corroborate/materialize claims of cognitive models positing direct links between modular and semantic representations. In combination with modular accounts of cognition, they also suggest that the currently prevailing “hub-and-spokes” model of semantics could be extended by incorporating an intermediate module level containing invariant representations, in addition to “spokes,” which subserve the processing of a near-unlimited number of sensorimotor and speech-sound features.

INTRODUCTION

A core issue of cognitive neuroscience is to understand how and where the meaning of words, faces, and objects of the surrounding world is extracted and combined into coherent concepts. Much work has been done to identify the cortical regions underpinning semantics, and most authors currently advocate that the implementation of abstract concepts is subtended by a supramodal hub situated in the anterior temporal lobes (ATL; Lambon Ralph, Jefferies, Patterson, & Rogers, 2017; Rice, Lambon Ralph, & Hoffman, 2015; Patterson, Nestor, & Rogers, 2007). However, it is largely unknown how modular information, such as, for example, word forms or face and object representations implemented by remote brain modules, is conveyed to the anterior temporal hub. The currently prevailing “hub-and-spokes” model of semantics, which is grounded on connectionist accounts, states that the so-called “spokes,” located mostly in posterior brain regions, represent sensory/speech-sound/motor features that aliment the ATL hub (Lambon Ralph et al., 2017) via white matter connections (e.g., Rice, Hoffman, & Lambon Ralph, 2015; Guo et al., 2013; Binney, Parker, & Lambon Ralph, 2012). However, sensory–motor features are “complex” and variable (Lambon Ralph et al., 2017), and the “hub-and-spokes” model hardly accounts for an important line of cognitive research having demonstrated that highly variable features of, for example, speech sounds are concatenated into invariant modular language representations (word forms) stored in the mental lexicon (Pinker & Jackendoff, 2005; Tyler, Marslen-Wilson, & Stamatakis, 2005; Indefrey & Levelt, 2004; Ullman, 2001; Levelt, Roelofs, & Meyer, 1999; Caramazza & Miozzo, 1997; Fodor, 1983). Such invariant modular representations (modules) might provide robust and exploitable information for the supramodal semantic hub, which attributes meaning to word forms. The same holds for object representations that need to be extracted from highly variable signals of the surrounding world and which are stored in an object repertory/lexicon (sometimes called “structural description system”; Gerlach, Law, Gade, & Paulson, 1999; Lloyd-Jones & Humphreys, 1997). An important issue is therefore to explore whether there are, in addition to connections with “spokes,” white matter connections that directly link the ATL hub with cortical regions involved in cognitive modules.

Semantic dementia (SD), affecting selectively semantic processing in its early stage, has initially allowed for the identification of the crucial role of the ATL (Patterson et al., 2007), and it might be an appropriate model for investigating the unsolved module-hub network issue. SD is a neurodegenerative condition characterized by multimodal semantic deficits surfacing with single-word comprehension deficits, anomia, and face and object recognition difficulties (Gorno-Tempini et al., 2011). It is currently integrated into the spectrum of primary progressive aphasias as the “semantic variant of primary progressive aphasia” (Gorno-Tempini et al., 2011). However, the terminology “primary progressive aphasia” can be misleading given the multimodal semantic failure, and we will therefore use the more integrative disease label, that is, SD. SD is associated with bilateral damage to the ATL, which is left-predominant in early stages, including both cortical and subcortical structures of this region (Agosta-Cabronero et al., 2011; Mummery et al., 2000). Multiple anatomo-functional studies on SD have stressed the link between ATL atrophy and breakdown of semantics (Iaccarino et al., 2015; Mesulam et al., 2013; Mion et al., 2010; Snowden, Thompson, & Neary, 2004; Bozeat, Lambon Ralph, Patterson, Garrard, & Hodges, 2000), indicating that the ATL constitute a bilaterally distributed hub underlying semantic computations, which might bind modality-specific or modular information into coherent supramodal concepts (Lambon Ralph et al., 2017; Binney et al., 2012; Patterson et al., 2007; Lambon Ralph, Graham, Patterson, & Hodges, 1999). This view was also strengthened by investigations using inhibitory transcranial stimulation over the ATL inducing multimodal semantic impairment in healthy adults (Pobric, Jefferies, & Lambon Ralph, 2007, 2010). Accordingly, a meta-analysis of fMRI and PET investigations in healthy adults has shown that semantic tasks independently from the modality of stimuli input (verbal or nonverbal) generated bilateral activation of the ATL cortices (Rice, Lambon Ralph, et al., 2015). However, such research did not explore how modular information is conveyed to the semantic hub and whether subcortical fiber connections linking specific brain modules to the ATL are involved in semantic processes. Furthermore, it is unclear whether such connections might play a core role regarding the distinct processing of verbal versus nonverbal/object information given that several authors, criticizing the bilateral multimodal view, have attributed this distinction directly to left versus right ATL roles (Mesulam et al., 2013; Gainotti, 2013; Snowden et al., 2004).

The existence of multiple connections to the ATL dedicated to semantics is suggested by numerous findings indicating a semantic network including multiple cortical areas. For example, Damasio, Tranel, Grabowski, Adolphs, and Damasio (2004) have shown that semantic tasks such as picture naming depend, in addition to the ATL, on remote cortical regions such as left posterior inferior cortices of the temporal lobe. Likewise, Binder, Desai, Graves, and Conant (2009), Vandenberghe, Price, Wise, Josephs, and Frackowiak (1996), and Vandenberghe et al. (2013) have identified various cortical regions, such as the left ventral occipital-temporal junction, the left superior temporal gyrus, the left middle temporal gyrus, and the left inferior frontal gyrus, contributing to verbal and nonverbal semantics. Several of these non-ATL regions are involved in modular representations of words or faces or visual objects. In particular, the mid-posterior region of the right fusiform gyrus is dedicated to the representation/processing of faces, and the cortical region has been referred to as the “fusiform face area” (FFA; Kanwisher, McDermott, & Chun, 1997). The FFA and its vicinity have subsequently been shown to be also critical for the processing of well-known visual objects other than faces (Xu, 2005; Gauthier et al., 2000; Tarr & Gauthier, 2000; Gerlach et al., 1999). Regarding language, the left mid/posterior region of the fusiform gyrus has been reported to selectively underpin the representations of written words, and the cortical region has been referred to as the “visual word form area” (VWFA; McCandliss, Cohen, & Dehaene, 2003; Cohen et al., 2002). In addition, several authors have demonstrated that posterior parts of the left superior/middle temporal cortices subtend more generally lexical representations of both phonological and orthographic word forms (Race & Hillis, 2015; Graves, Grabowski, Mehta, & Gupta, 2008; Graves, Grabowski, Mehta, & Gordon, 2007; Kotz, Cappa, von Cramon, & Friederici, 2002). However, the anatomo-functional connectivity between such modular cortical regions and the ATL has not been explored given that previous research was mostly anatomically driven. This research has identified ATL connections to remote regions, which are not involved in modular language/object/face representations but in the processing of sensory–motor features such as the inferior longitudinal fasciculus (ILF) linking the ATL and the occipital lobe (Catani, Jones, Donato, & Ffytche, 2003). Likewise, several authors have shown connections with a variety of posterior regions, some of which probably implement sensory–motor features referred to as the “spokes” by the “hub-and-spokes” model (e.g., Guo et al., 2013; Binney et al., 2012). In the same vein, more anterior frontal regions such as the lateral/medial inferior frontal cortex, which connects with the ATL via the uncinate fasciculus (Catani, Howard, Pajevic, & Jones, 2002), is not involved in modular representations but rather in executive control functions that might impact on semantic processing. The roles of such connections and tracts in semantics remain unclear given contradicting findings. Most imaging studies in healthy adults, using various tasks tapping semantics and concomitantly also other capacities such as executive functioning and lexical competency, have suggested a semantic function of the so-called “ventral stream” tracts, including the ILF and the uncinate fasciculus (Friederici & Gierhan, 2013; Saur et al., 2008, 2010; Parker et al., 2005; Hickok & Poeppel, 2004). On the other side, intraoperative brain stimulation studies, directly mapping function and structure, have not confirmed the semantic involvement of these tracts (e.g., Duffau, 2008; Mandonnet, Nouet, Gatignol, Capelle, & Duffau, 2007).

The aim of this study was to investigate the white matter network of semantics, presumably linking the ATL and remote cortices implementing cognitive modules. It furthermore aimed at clarifying the semantic role of the ILF and the uncinate fasciculus. To this end, the study combined anatomical MRI (T1-weighted images), diffusion MRI, and semantic assessments of verbal and nonverbal/visual stimuli in a large cohort of SD patients and healthy controls. We more specifically explored with probabilistic fiber tracking in healthy controls whether the ATL are connected to three regions implementing cognitive modules: the FFA (right hemisphere), the VWFA (left hemisphere), and lexical-related posterior parts of the left superior/middle temporal cortices. We then studied whether these anatomical connections are altered in SD and whether they genuinely subtend semantics by conducting correlation analyses between tract parameters and scores of semantic tasks assessing verbal (words) and nonverbal object/face stimuli. Finally, we evaluated via correlation analyses the putative semantic roles of the ILF and the uncinate fasciculus. Post hoc correlations were also conducted for the semantic tasks and control tracts, as well as for the tracts of interest and a control task not involving semantic processing. One should note that this study focuses on module-hub connections that convey modular information, presumably automatically, to the semantic hub system to yield activation of concepts. It does not explore concept formation, which occurs during the time course of human development, or controlled semantic access, which requires a more “executive” component searching and retrieving information in the semantic system and which is referred to as “semantic control” implemented mainly by prefrontal cortices (Lambon Ralph et al., 2017).

METHODS

Participants

Twenty-eight SD patients and 19 healthy controls were recruited from a French multicenter investigation on primary progressive aphasias (PHRC-CAPP) including the SD variant. Diagnosis was established by expert neurologists following the international diagnostic criteria of Gorno-Tempini et al. (2011): progressive alteration of single-word comprehension, word finding, and object knowledge, without agrammatism or motor speech disorders. All SD patients also responded to the imaging-supported criteria for SD diagnosis showing atrophy and hypometabolism of the ATL on MRI and FDG-PET (Gorno-Tempini et al., 2011). They were at an early stage of the disease as reflected by aphasia severity scores ≥ 3 (normal = 5) in the Boston Diagnostic Aphasia Examination (Mazaux & Orgogozo, 1982). Patients did not present with any neurological/psychiatric disease other than SD. The 19 healthy controls were matched with the patients for age, handedness, level of education, and sex. All participants were native French speakers. The local ethics committee approved the study, and informed written consent was obtained from the participants. Demographical information is summarized in Table 1.

Table 1. 
Demographical Data of SD Patients and Healthy Controls (Means ± SD)
 SD PatientsHealthy Controls
Number of participants 28 19 
Sex (women/men) 13/15 13/6 
Age (years) 67.6 ± 7.3 64.9 ± 6.2 
Handedness (right/left) 27/1 19/0 
Years of education 13.2 ± 5.3 14.1 ± 4.8 
Age at symptom onset (years) 63.7 ± 7.3 — 
Disease duration since diagnosis (years) 1.2 ± 1.04 — 
Symptom duration (years) 3.9 ± 2.5 — 
 SD PatientsHealthy Controls
Number of participants 28 19 
Sex (women/men) 13/15 13/6 
Age (years) 67.6 ± 7.3 64.9 ± 6.2 
Handedness (right/left) 27/1 19/0 
Years of education 13.2 ± 5.3 14.1 ± 4.8 
Age at symptom onset (years) 63.7 ± 7.3 — 
Disease duration since diagnosis (years) 1.2 ± 1.04 — 
Symptom duration (years) 3.9 ± 2.5 — 

Cognitive Assessment

Standard Cognitive/Language Assessment

The general cognitive assessment included the Mini-Mental State Examination (Folstein, Folstein, & McHugh, 1975), the Mattis Dementia Rating Scale (Mattis, 1976), and the Frontal Assessment Battery (Dubois, Slachevsky, Litvan, & Pillon, 2000). The language assessment was based on the Boston Diagnostic Aphasia Evaluation (Mazaux & Orgogozo, 1982), including an evaluation of the severity of aphasia, a sentence repetition task, and a single-word comprehension task. We also applied a picture naming test (DO80; Deloche, Metz-Lutz, & Kremin, 1997) and a verbal fluency test comprising letter and category fluency (Cardebat, Doyon, Puel, Goulet, & Joanette, 1990).

The Raven's Progressive Matrices, which taps high-level processes of analogical/rule reasoning without involving semantic representations/processes (Raven, Raven, & Court, 1998), was used as a control task for supplementary post hoc analyses.

Semantic Tasks

Semantic performance was assessed by three tasks tapping both the processing of verbal (words) and nonverbal stimuli (visual objects, famous faces): (1) A familiarity judgment and semantic matching task manipulating famous and nonfamous people in a verbal version (names) and a nonverbal/visual version (black and white photographs; SemPer battery; Laisney, Eustache, & Desgranges, 2009). The task was subdivided into three parts: (a) “familiarity judgment,” (b) matching with the professional domain, and (c) matching with subcategories of the professional domain. (2) A semantic matching task including a verbal version (written words) and a nonverbal/visual version (line drawings; Pyramids and Palm Trees Test; Howard & Patterson, 1992). (3) A second semantic matching task with verbal stimuli (written words) and nonverbal/visual stimuli (line drawings; BECS-GRECO [Batterie d'Evaluation des Connaissances Sémantiques du Groupe de Réflexion sur les Evaluations COgnitives]; Merck et al., 2011). This latter task also included a supplementary condition of semantic matching in the nonverbal modality, referred to as “identity matching,” manipulating semantic/functional “identity” and physical but nonsemantic similarity. One should keep in mind that the use of picture/photograph stimuli to assess nonverbal aspects of semantics could induce a potential bias given that inner verbalization of such stimuli is possible, which is a near-universal and inevitable limitation of experimental tasks using visual stimuli to tap nonverbal processes.

In each participant, we identified a composite semantic score for both verbal and nonverbal modalities by calculating mean values derived from the percentage of correct responses of the three tasks and their respective verbal or nonverbal modalities. The three tasks are summarized in Table 2.

Table 2. 
Semantic Tasks Tapping Verbal and Nonverbal/Visual Modalities
Verbal ModalityNonverbal Modality
SemPer Battery 
Familiarity judgment: names Familiarity judgment: faces 
Semantic matching names—professional domain Semantic matching faces—professional domain 
Semantic matching names—subcategories of professional domain Semantic matching faces—subcategories of professional domain 
  
Pyramids and Palm Trees Test 
Semantic matching: words Semantic matching: pictures 
  
BECS-GRECO 
Semantic matching: words Semantic matching: pictures 
  “Identity” matching 
Verbal ModalityNonverbal Modality
SemPer Battery 
Familiarity judgment: names Familiarity judgment: faces 
Semantic matching names—professional domain Semantic matching faces—professional domain 
Semantic matching names—subcategories of professional domain Semantic matching faces—subcategories of professional domain 
  
Pyramids and Palm Trees Test 
Semantic matching: words Semantic matching: pictures 
  
BECS-GRECO 
Semantic matching: words Semantic matching: pictures 
  “Identity” matching 

Imaging

MRI acquisitions and cognitive assessments were conducted within a 1-month interval for all participants. All imaging centers involved in this investigation belong to the national network of the Centre d'Acquisition et de Traitement d'Images (CATI; cati-neuroimaging.com/; Operto et al., 2016). MRI sequences were harmonized by the CATI to minimize differences between imaging centers. The CATI performs onsite visits for the setup of imaging protocols and regular follow-up. This rigorous quality and harmonizing procedure provided the rationale for not conducting statistical corrections for the use of different scanners.

Participants underwent whole-brain MRI scanning on 3-T scanners at seven acquisition sites across France. Diffusion-weighted images (DWIs) were acquired using an EPI sequence (128 × 128 acquired matrix, voxel size: 2.0 × 2.0 × 2.0 mm3). Seventy separate images were extracted from each DWI scan: 10 T2-weighted images with no dedicated diffusion sensitization (b0 images) and 60 DWIs (b = 1500 sec/mm2). A fieldmap image was acquired to correct for geometrical distortions induced by the EPI sequence. Additionally, T1-weighted images were acquired with a 3-D gradient-echo sequence (240 × 256 acquired matrix, voxel size = 1.0 × 1.0 × 1.0 mm3, slice thickness = 1 mm, inversion time = 900 msec, repetition time = 2300 msec, echo time = 2.98 msec, flip angle = 9°).

Cortical ATL Atrophy

T1-weighted images were processed using the t1-freesurfer pipeline of the Clinica platform (www.clinica.run), which is a wrapper of different tools of the FreeSurfer image analysis software (surfer.nmr.mgh.harvard.edu; Fischl, 2012). The processing pipeline included nonuniformity and intensity correction, skull stripping, gray/white matter (GM/WM) segmentation, reconstruction of the cortical surface, cortical thickness estimation, and spatial normalization onto the FreeSurfer surface template (FsAverage). After segmentation, all data sets were checked visually for segmentation errors (errors of GM/WM and GM/cerebrospinal fluid [CSF] boundaries). The SD group was compared with the healthy controls applying a point-wise, vertex-to-vertex model based on the MATLAB SurfStat toolbox (www.math.mcgill.ca/keith/surfstat/) to analyze cortical thickness. The data were smoothed using a Gaussian kernel with a FWHM set to 20 mm.

Identification of Cortical ROIs for Fiber Tracking

We manually delineated the ATL and the cortical regions implementing modular representations (FFA, VWFA, lexical-related posterior parts of the left superior/middle temporal cortices). All cortical ROIs were drawn on the FsAverage template using the FreeSurfer toolbox Tksurfer (https://surfer.nmr.mgh.harvard.edu/fswiki/TkSurfer) and then mapped onto each individual participant. The regions were defined according to the results of the functional neuroimaging literature. For all regions, their spatial extent was slightly enlarged compared with the results of group-wise functional neuroimaging imaging studies to account for interindividual differences. We first delineated the ATL following the meta-analysis study of Rice, Lambon Ralph, et al. (2015). This region encompassed the most anterior regions of the temporal lobe and anterior portions of the superior temporal gyrus, the middle temporal gyrus, the inferior temporal gyrus, the fusiform gyrus, and the parahippocampal gyrus. We then identified the FFA and the VWFA as the right and left mid-posterior portion of the fusiform gyrus, respectively, based on previous fMRI studies (Cohen et al., 2000; Kanwisher et al., 1997). The lexical-related temporal area was defined as a region encompassing the posterior parts of the superior and middle temporal gyri in the left hemisphere according to previous imaging results (Race & Hillis, 2015; Graves et al., 2008; Kotz et al., 2002). The identified cortical regions are illustrated in Figure 1.

Figure 1. 

Cortical regions used for fiber tracking: (i) the ATL hub (yellow) and (ii) the three module-implementing areas: FFA (red), VWFA (blue), lexical-related area (LA) in posterior/superior temporal cortices (green).

Figure 1. 

Cortical regions used for fiber tracking: (i) the ATL hub (yellow) and (ii) the three module-implementing areas: FFA (red), VWFA (blue), lexical-related area (LA) in posterior/superior temporal cortices (green).

These ROIs were then mapped onto each individual participant. To that purpose, T1-weighted images were processed using FreeSurfer image analysis software (stable version 5.3; surfer.nmr.mgh.harvard.edu). The processing pipeline included nonuniformity and intensity correction, skull stripping, gray/white matter segmentation, reconstruction of the cortical surface, and segmentation of cortical structures. After segmentation, all data sets were checked visually for segmentation errors (errors of GM/WM and GM/CSF boundaries).

Finally, for each participant, the mean cortical thickness was computed within each of the identified cortical regions.

Fiber Tracking

First, we aligned for each participant all raw DWI volumes to the average b0 image (DWI volume with no diffusion sensitization) with 6 df to correct for head motion, and the diffusion weighted directions were appropriately updated as recommended by Leemans and Jones (2009). A registration with 12 df was used to correct for eddy current distortions. These registrations were done using the FSL flirt tool (www.fmrib.ox.ac.uk/fsl). To correct for EPI-induced susceptibility artifacts, the fieldmap image was used as proposed by Jezzard and Balaban (1995) with the FSL prelude/fugue tools. Finally, the DWIs were corrected for nonuniform intensity using ANTs N4 bias correction algorithm (Tustison & Avants, 2013). A single multiplicative bias field from the averaged b0 image was estimated, as suggested by Jeurissen, Tournier, Dhollander, Connelly, and Sijbers (2014). The DWI data set was up-sampled at 1 mm to improve the registration between the T1-weighted image and the DWI. A diffusion tensor model was fitted at each voxel to calculate fractional anisotropy (FA) and mean diffusivity (MD) maps.

We studied several white matter tracks connected to the ATL. The ILF and the uncinate fasciculus were explored in both hemispheres. Language-related tracts connecting the ATL to the VWFA and the ATL to the lexical area in posterior parts of the left superior/middle temporal cortex were explored in the left hemisphere. Tracts connecting the ATL to the FFA were explored in the right hemisphere.

The ILF and the uncinate fasciculus were defined using the JHU white matter tractography atlas (Mori, Wakana, Nagae-Poetscher, & van Zijl, 2005). Each tract of the JHU atlas is defined as the probabilistic average of 28 normal participants. Several thresholds of these probabilistic tracts are proposed (0%, 25%, 50%). After visual inspection, the 25% threshold was selected, which was neither too noisy (compared with the 0% threshold) nor too selective (compared with the 50% threshold). For each participant, the FA map of the participant was registered onto the FA map of the JHU atlas template with the ANTs SyN algorithm (Avants, Epstein, Grossman, & Gee, 2008). Then, the estimated nonlinear deformation was applied to the MD maps so that both the FA and MD maps of each participant were put into correspondence with the atlas.

Tracts connecting the ATL to module-implementing cortices (FFA, VWFA, lexical-related regions of the posterior temporal cortex) were identified by building a customized atlas of these tracts using fiber tracking. As fiber tracking might not be reliable in patients with severe atrophy, we performed tractography in healthy controls to build the customized atlas, which was subsequently mapped onto each SD patient. Fiber tracking was performed as follows with the MRtrix software package (Brain Research Institute; Tournier, Calamante, & Connelly, 2012; www.mrtrix.org). First, the fiber orientation distribution (FOD) was computed at highly anisotropic voxels (FA > 0.7) to determine the response function, which was used for constrained spherical deconvolution to accurately estimate the FOD (Tournier, Calamante, & Connelly, 2007). Then, fibers were generated with a probabilistic tracking algorithm that samples FOD at each step (Behrens et al., 2003). The algorithm generated 1,000,000 fibers of minimum length 20 mm. Default tracking parameters were used (step size of 0.5 mm, minimum radius of curvature of 1 mm and FOD cutoff of 0.1). All voxels in the 1-mm dilated white matter mask (obtained from the registered T1 image) were used as seeds. For each control participant, fibers passing through both the ATL and the FFA were selected, thus forming the “ATL–FFA tract.” Similarly, fibers passing through both the ATL and the VWFA formed the “ATL–VWFA tract.” Finally, fibers passing through both the ATL and the lexical-related posterior parts of the left superior/middle temporal cortices (lexicon area [LA]) formed the “ATL–LA tract.” A track density image (see Calamante, Tournier, Jackson, & Connelly, 2010) representing the density of fibers was generated for each tract. To normalize the track density images onto the JHU atlas, the nonlinear deformation described in the previous paragraph was applied to the track density images. The normalized track density images of all healthy controls were averaged, empirically thresholded at 10% and binarized to obtain the final white matter tract masks. Finally, for each participant, the average FA and MD value was computed within each of the studied tracts.

As a supplementary post hoc analysis, we also analyzed two control tracts to check for the predicted semantic roles of the three module-hub tracts: (i) the corticospinal tract, which, being a motor-related structure, should not have a role in semantic processing, and (ii) the right homologue of the ATL–LA tract, which should not have any role regarding the processing of the verbal modality of semantics. The corticospinal tract was defined with the same procedure used for the ILF and the uncinate fasciculus, based on the JHU white matter tractography atlas (Mori et al., 2005). The tracking of the right homologue of the ATL–LA tract was based on the same procedure used for the left hemisphere ATL–LA tract, except that we selected fibers linking the right ATL to the cortical region that is symmetric to the left hemisphere LA (posterior parts of the superior/middle temporal cortices).

Statistical Analysis

We used nonparametric tests given that semantic task scores and tract measures were not normally distributed. Regarding semantic performance, we tested the effects of group (SD patients, healthy controls) and task modality (verbal, nonverbal) using a two-way rank-based analysis of linear models according to the algorithm proposed by Hocking (1985) and implemented in R by Kloke and McKean (2012). Post hoc, we compared verbal and nonverbal task scores of SD patients and healthy controls using one-sided two-sample Wilcoxon rank-sum tests. Within each group (SD, controls), we also compared verbal and nonverbal performance using paired Wilcoxon signed rank. Regarding imaging data, we assessed differences in tract integrity (FA, MD) between SD patients and healthy controls using one-sided two-sampled Wilcoxon rank-sum tests. The relationship between semantic scores (verbal and nonverbal) and tract parameters was analyzed with Spearman's rank correlations. The correlation analyses were controlled for the effect of age and educational level. Correction for multiple comparisons was performed using the Bonferroni procedure, counting the number of comparison within each type of variable/test (group differences for semantic scores [n = 4], group differences for tract parameters [n = 16], correlations between tract parameters and semantic scores [n = 14]). Supplementary analyses (control tracts, control task, mean cortical thickness, partial correlations) were conducted as follows. As in the primary analyses, we assessed differences in tract parameters (FA, MD) for the control tracts (left and right corticospinal tract, right homologue of the ATL–LA tract) between SD patients and healthy controls using one-sided two-sampled Wilcoxon rank-sum tests. To ensure that the resulting p values were comparable to those of the primary analyses, we applied the same level of correction for multiple comparisons (n = 16). As in the primary analyses, the relationships between semantic scores and the parameters of control tracts as well as between the control task and the different tract parameters were analyzed with Spearman's rank correlations. Again, we applied the same level of correction for multiple comparisons as in the primary analyses (n = 14). The same rationale was applied for correlations between mean cortical thickness of the ATL and of the module-related cortical regions and semantic scores (verbal and nonverbal). Moreover, a Principal Component Analysis (PCA) between the different measures from MRI (mean cortical thickness, MD) and semantic scores (verbal, nonverbal) was performed. To understand the implication of gray matter loss for the relation between semantic scores (verbal, nonverbal) and white matter tracts, Spearman's rank partial correlations were conducted. In all analyses, the statistical significance threshold was set to p < .05. Statistical analyses were conducted using R (Version 3.1.1; R Core Team, 2013).

RESULTS

Standard Cognitive/Language Tests

Patients had a typical profile of SD showing impaired single-word comprehension, letter and category fluency, and picture naming contrasting with largely spared sentence repetition. Performance of SD patients and healthy controls are illustrated in Table 3.

Table 3. 
Results in the Standard Cognitive/Language Assessment of SD Patients and Healthy Controls (Means ± SD)
 SD PatientsHealthy ControlsNormal Threshold
General Cognitive Tests 
MMSE 24.96 ± 2.7 27.56 ± 1.8 ≥ 27 
FAB 14.54 ± 1.8 17.11 ± 0.8 ≥ 16 
MDRS 118.75 ± 11.9 140.00 ± 3.9 ≥ 137 
  
Language Tests 
Severity of aphasia (BDAE) 3.89 ± 0.57 5 ± 0 ≥ 4 
Single-word comprehension (BDAE) 61.32 ± 8.3 71.63 ± 0.6 ≥ 68 
Sentence repetition (BDAE) 14.96 ± 1.5 15.84 ± 0.7 ≥ 14 
DO80 43.51 ± 20 79.61 ± 0.6 ≥ 75 
DO80–semantic paraphasias 9.11 ± 6.4 0.33 ± 0.5 — 
Letter fluency (“P”/2 min) 11.57 ± 5.9 25.31 ± 7.1 ≥ 15 
Category fluency (“fruits”/2 min) 8.14 ± 4.88 21.36 ± 5.6 ≥ 15 
 SD PatientsHealthy ControlsNormal Threshold
General Cognitive Tests 
MMSE 24.96 ± 2.7 27.56 ± 1.8 ≥ 27 
FAB 14.54 ± 1.8 17.11 ± 0.8 ≥ 16 
MDRS 118.75 ± 11.9 140.00 ± 3.9 ≥ 137 
  
Language Tests 
Severity of aphasia (BDAE) 3.89 ± 0.57 5 ± 0 ≥ 4 
Single-word comprehension (BDAE) 61.32 ± 8.3 71.63 ± 0.6 ≥ 68 
Sentence repetition (BDAE) 14.96 ± 1.5 15.84 ± 0.7 ≥ 14 
DO80 43.51 ± 20 79.61 ± 0.6 ≥ 75 
DO80–semantic paraphasias 9.11 ± 6.4 0.33 ± 0.5 — 
Letter fluency (“P”/2 min) 11.57 ± 5.9 25.31 ± 7.1 ≥ 15 
Category fluency (“fruits”/2 min) 8.14 ± 4.88 21.36 ± 5.6 ≥ 15 

MMSE = Mini-Mental State Examination; FAB = Frontal Assessment Battery; MDRS = Mattis Dementia Rating Scale; BDAE = Boston Diagnostic Aphasia Evaluation; DO80 = picture naming test.

Semantic Tasks

There was a significant effect of Group (F = 171.4, p < .001), but not of Task Modality (F = 0.02, p = .90). The interaction between Group and Task Modality was not significant (F = 0.19, p = .66). Post hoc tests with one-sided two-sample Wilcoxon rank-sum tests showed that SD patients performed more poorly than healthy controls in both the verbal modality (patients: 69.36% correct ± 13.12, controls: 97.6% correct ± 2.13; W = 0, p < .001) and the nonverbal modality (patients: 72.81% correct ± 10.78, controls: 96.55% correct ± 3.39; W = 6, p < .001). Furthermore, there was no difference between scores of the verbal and nonverbal modality in SD patients (V = 138, p = .56) or in healthy controls (V = 118, p = .65). Results are illustrated in Figure 2.

Figure 2. 

Semantic performance for the verbal and nonverbal/visual modality in SD patients and healthy controls (mean values ± SD).

Figure 2. 

Semantic performance for the verbal and nonverbal/visual modality in SD patients and healthy controls (mean values ± SD).

Performance in the control task (Raven's Progressive Matrices) was similar in SD patients and healthy controls (patients: 85.03% correct ± 4.3, healthy controls: 88.13 correct ± 3.4; W = 232, p = .24).

Imaging

ATL Atrophy and Fiber Tracking in Healthy Controls

Cortical thickness of the left and right ATL was reduced in SD patients, slightly predominating in the left ATL and caudally extending beyond the left ATL region (see Figure 3). Individual findings of ATL atrophy also showed that seven SD patients had predominant right ATL atrophy and that the remaining 21 patients had predominant left ATL atrophy.

Figure 3. 

Brain regions with significantly reduced cortical thickness in SD patients as compared with healthy controls. Corrected p values for cluster correction are displayed with a blue color map.

Figure 3. 

Brain regions with significantly reduced cortical thickness in SD patients as compared with healthy controls. Corrected p values for cluster correction are displayed with a blue color map.

Fiber tracking identified a dense set of fibers connecting the ATL to the FFA and the VWFA, following the ventral part of the right and left temporal lobe, respectively, and to the lexical-related left posterior mid/superior temporal regions (LA) following more dorsal parts of the left temporal lobe. We will refer to these module-hub fiber bundles as the “ATL–FFA tract,” the “ATL–VWFA tract,” and the “ATL–LA tract.” The tracts are illustrated in Figure 4. We also identified the right homologue of the ATL–LA tract.

Figure 4. 

Tracts connecting the ATL to the FFA (red), to the VWFA (blue), and to the LA in posterior/superior temporal cortices (green).

Figure 4. 

Tracts connecting the ATL to the FFA (red), to the VWFA (blue), and to the LA in posterior/superior temporal cortices (green).

We then analyzed the percentages of overlap between the different tracts by calculating the intersection between them (pairwise analyses) and dividing the number of intersection voxels by the total number of voxels of each pair of tracts. The analyses showed that the left-hemisphere ATL–LA tract and ATL–VWFA tract demonstrate no overlap and that there was only poor overlap between these two tracts and the left ILF (respectively, 18.6% and 2.7%). Likewise, the right-hemisphere ATL–FFA tract has only poor overlap with the right ILF (16.0%). Finally, the uncinate fasciculus had no overlap with any of the other tracts.

Comparisons of Tract Integrity between SD Patients and Healthy Controls

There were significant tract alterations in SD patients for MD measures regarding the uncinate fasciculus (bilaterally), the ILF (bilaterally), the ATL–LA tract, the ATL–VWFA tract, and the ATL–FFA tract (Table 4). For FA measures, we found significant differences only for the uncinate fasciculus bilaterally and the ATL–VWFA tract. MD therefore appeared to be more sensitive than FA to white matter changes, which has already been reported in previous tractography studies on SD (Agosta et al., 2010). We therefore used MD measures for the subsequent correlation analyses with semantic scores. Regarding the two control tracts, which were predicted not to be involved in semantics (corticospinal tract) and more specifically not to contribute to verbal semantics (right homologue of the ATL–LA tract), we found that the former tract is not altered in our SD patients (bilaterally) and that the latter is altered for MD values. Results are summarized in Table 5.

Table 4. 
Comparisons of Tract Parameters (MD, FA) in SD Patients and Healthy Controls (Mean Values ± SD)
 ParameterTractSD PatientsControlsWilcoxon Rank-Sum Test
Left hemisphere FA Uncinate fasciculus 0.25 ± 0.04 0.33 ± 0.04 W = 61, p < .001 
ILF 0.32 ± 0.03 0.33 ± 0.04 W = 238, p = 1.000 
ATL–VWFA tract 0.20 ± 0.03 0.22 ± 0.02 W = 134, p = .029 
ATL–LA tract 0.31 ± 0.03 0.34 ± 0.03 W = 159, p = .150 
MD Uncinate fasciculus 1.20 ± 0.21 0.78 ± 0.03 W = 531, p < .001 
ILF 0.92 ± 0.15 0.73 ± 0.03 W = 516, p < .001 
ATL–VWFA tract 1.23 ± 0.31 0.75 ± 0.04 W = 532, p < .001 
ATL–LA tract 0.92 ± 0.13 0.74 ± 0.03 W = 504, p < .001 
Right hemisphere FA Uncinate fasciculus 0.31 ± 0.06 0.35 ± 0.05 W = 134, p = .029 
ILF 0.37 ± 0.05 0.37 ± 0.03 W = 249, p = 1.000 
ATL–FFA tract 0.28 ± 0.05 0.28 ± 0.03 W = 264, p = 1.000 
MD Uncinate fasciculus 0.97 ± 0.30 0.73 ± 0.30 W = 457, p < .001 
ILF 0.80 ± 0.13 0.69 ± 0.13 W = 417, p = .006 
ATL–FFA tract 0.81 ± 0.77 0.73 ± 0.77 W = 461, p < .001 
 ParameterTractSD PatientsControlsWilcoxon Rank-Sum Test
Left hemisphere FA Uncinate fasciculus 0.25 ± 0.04 0.33 ± 0.04 W = 61, p < .001 
ILF 0.32 ± 0.03 0.33 ± 0.04 W = 238, p = 1.000 
ATL–VWFA tract 0.20 ± 0.03 0.22 ± 0.02 W = 134, p = .029 
ATL–LA tract 0.31 ± 0.03 0.34 ± 0.03 W = 159, p = .150 
MD Uncinate fasciculus 1.20 ± 0.21 0.78 ± 0.03 W = 531, p < .001 
ILF 0.92 ± 0.15 0.73 ± 0.03 W = 516, p < .001 
ATL–VWFA tract 1.23 ± 0.31 0.75 ± 0.04 W = 532, p < .001 
ATL–LA tract 0.92 ± 0.13 0.74 ± 0.03 W = 504, p < .001 
Right hemisphere FA Uncinate fasciculus 0.31 ± 0.06 0.35 ± 0.05 W = 134, p = .029 
ILF 0.37 ± 0.05 0.37 ± 0.03 W = 249, p = 1.000 
ATL–FFA tract 0.28 ± 0.05 0.28 ± 0.03 W = 264, p = 1.000 
MD Uncinate fasciculus 0.97 ± 0.30 0.73 ± 0.30 W = 457, p < .001 
ILF 0.80 ± 0.13 0.69 ± 0.13 W = 417, p = .006 
ATL–FFA tract 0.81 ± 0.77 0.73 ± 0.77 W = 461, p < .001 

MD is expressed in mm2/sec. p Values are corrected for multiple comparisons (16 tests).

Table 5. 
Supplementary Analysis: Comparisons of Control Tract Parameters (MD, FA) in SD Patients and Healthy Controls (Mean Values ± SD)
 ParameterTractSD PatientsControlsWilcoxon Rank-Sum Test
Left hemisphere FA Control tract: corticospinal tract 0.51 ± 0.06 0.50 ± 0.03 W = 234, p = 1 
MD Control tract: corticospinal tract 0.71 ± 0.13 0.67 ± 0.04 W = 344, p = .745 
Right hemisphere FA Control tract: corticospinal tract 0.52 ± 0.06 0.52 ± 0.04 W = 239, p = 1 
Control tract: right homologue of ATL–LA tract 0.30 ± 0.04 0.33 ± 0.03 W = 250, p = 1 
MD Control tract: corticospinal tract 0.69 ± 0.15 0.64 ± 0.54 W = 340, p = .893 
Control tract: right homologue of ATL–LA tract 0.90 ± 0.14 0.72 ± 0.03 W = 443, p < .001 
 ParameterTractSD PatientsControlsWilcoxon Rank-Sum Test
Left hemisphere FA Control tract: corticospinal tract 0.51 ± 0.06 0.50 ± 0.03 W = 234, p = 1 
MD Control tract: corticospinal tract 0.71 ± 0.13 0.67 ± 0.04 W = 344, p = .745 
Right hemisphere FA Control tract: corticospinal tract 0.52 ± 0.06 0.52 ± 0.04 W = 239, p = 1 
Control tract: right homologue of ATL–LA tract 0.30 ± 0.04 0.33 ± 0.03 W = 250, p = 1 
MD Control tract: corticospinal tract 0.69 ± 0.15 0.64 ± 0.54 W = 340, p = .893 
Control tract: right homologue of ATL–LA tract 0.90 ± 0.14 0.72 ± 0.03 W = 443, p < .001 

MD is expressed in mm2/sec. To ensure that p values are comparable to those reported in Table 4, they were corrected for the same number of multiple comparisons (16 tests).

In addition, we compared in SD patients FA and MD measures between the left and right hemisphere tracts, which were bilaterally explored in the study (paired Wilcoxon signed rank tests). Alterations were more important in the left hemisphere compared with the right for the uncinate fasciculus (V = 50, p < .001 for FA; V = 339, p < .01 for MD) and the ILF (V = 10, p = < .001 for FA; V = 328, p < .01 for MD). The same holds for ATL–LA tract as compared with its right homologue (V = 395, p < .001 for FA; V = 279, p = .085 for MD) and the left hemisphere ATL–VWFA tract compared with its right homologue ATL–FFA tract (V = 2, p < .001 for FA; V = 364, p < .001 for MD).

Relationship between Tract Integrity and Semantic Performance

The results of the correlation analyses using composite semantic scores for the verbal and nonverbal task modalities and tract parameters (MD) are summarized in Table 6. Nonverbal semantic scores were significantly correlated with MD measures of all explored tracts in the right hemisphere: the uncinate fasciculus, the ILF, and the ATL–FFA tract. Verbal semantic scores were significantly correlated to MD measures of two left hemisphere tracts: the ATL–VWFA tract and the ATL–LA tract. No correlations were found between nonverbal semantic scores and the left hemisphere tracts or verbal semantic scores and the right hemisphere tracts. Results of supplementary analyses (control tracts and control task) are presented in Table 7. Regarding the two control tracts, there were significant correlations between the right corticospinal tract measures and nonverbal semantic scores, and between the right homologue of the ATL–LA tract and nonverbal semantic scores. Regarding the control task (Raven's Progressive Matrices), there were no significant correlations between task scores and the ATL–LA tract, the ATL–VWFA tract, the ATL–FFA tract, the ILF, or the uncinate fasciculus.

Table 6. 
Results of Correlation Analyses Using Tract Measures of MD and Composite Semantic Verbal and Nonverbal/Visual Scores
Semantic ModalityTractsLeft HemisphereRight Hemisphere
Rhop ValueRhop Value
Verbal Uncinate fasciculus −0.11 −0.45 .299 
ILF −0.54 .068 −0.40 .633 
ATL–VWFA tract −0.64 .007 — — 
ATL–FFA tract — — −0.55 .050 
ATL–LA tract −0.58 .029 — — 
Nonverbal Uncinate fasciculus 0.23 −0.67 .002 
ILF −0.18 −0.61 .012 
ATL–VWFA tract −0.24 — — 
ATL–FFA tract — — −0.75 < .001 
ATL–LA tract −0.30 — — 
Semantic ModalityTractsLeft HemisphereRight Hemisphere
Rhop ValueRhop Value
Verbal Uncinate fasciculus −0.11 −0.45 .299 
ILF −0.54 .068 −0.40 .633 
ATL–VWFA tract −0.64 .007 — — 
ATL–FFA tract — — −0.55 .050 
ATL–LA tract −0.58 .029 — — 
Nonverbal Uncinate fasciculus 0.23 −0.67 .002 
ILF −0.18 −0.61 .012 
ATL–VWFA tract −0.24 — — 
ATL–FFA tract — — −0.75 < .001 
ATL–LA tract −0.30 — — 

p Values are corrected for multiple comparisons (14 tests). Results in bold indicate significant correlations.

Table 7. 
Supplementary Analysis: Results of Correlation Analyses between Control Tract Measures of MD and Composite Semantic Verbal and Nonverbal/Visual Scores and between All Studied Tracts and the Control Task (Raven's Progressive Matrices)
Semantic ModalityTractLeft HemisphereRight Hemisphere
Rhop ValueRhop Value
Verbal Control tract: corticospinal tract −0.40 .593 −0.54 .059 
Control tract: right homologue of ATL–LA tract — — −0.52 .075 
Nonverbal Control tract: corticospinal tract −0.34 −0.55 .049 
Control tract: right homologue of ATL–LA tract — — −0.78 < .001 
Control task (Raven's matrices) Uncinate fasciculus 0.07 −0.41 .516 
ILF −0.12 −0.38 .724 
ATL–VWFA tract −0.10 — — 
ATL–FFA tract — — −0.41 .475 
ATL–LA tract −0.21 — — 
Control tract: right homologue of ATL–LA tract — — −0.44 .319 
Control tract: Corticospinal tract −0.20 −0.15 
Semantic ModalityTractLeft HemisphereRight Hemisphere
Rhop ValueRhop Value
Verbal Control tract: corticospinal tract −0.40 .593 −0.54 .059 
Control tract: right homologue of ATL–LA tract — — −0.52 .075 
Nonverbal Control tract: corticospinal tract −0.34 −0.55 .049 
Control tract: right homologue of ATL–LA tract — — −0.78 < .001 
Control task (Raven's matrices) Uncinate fasciculus 0.07 −0.41 .516 
ILF −0.12 −0.38 .724 
ATL–VWFA tract −0.10 — — 
ATL–FFA tract — — −0.41 .475 
ATL–LA tract −0.21 — — 
Control tract: right homologue of ATL–LA tract — — −0.44 .319 
Control tract: Corticospinal tract −0.20 −0.15 

To ensure that p values are comparable to those reported in Table 4, they were corrected for the same number of multiple comparisons (14 tests). Results in bold indicate significant correlations.

Relationship between Gray Matter Loss and Semantic Performance

The results of the correlation analyses using composite semantic scores for the verbal and nonverbal task modalities and mean cortical thickness are summarized in Table 8. Nonverbal semantic scores were significantly correlated with one region in the right hemisphere: the right ATL. No correlations were found between nonverbal semantic scores and regions in the left hemisphere or verbal semantic scores and any of the identified cortical regions.

Table 8. 
Results of Correlation Analyses between Mean Cortical Thickness and Composite Semantic Verbal and Nonverbal/Visual Scores
Semantic ModalityCortical RegionLeft HemisphereRight Hemisphere
Rhop ValueRhop Value
Verbal ATL 0.31 0.48 .16 
VWFA 0.23 — — 
FFA — — 0.15 
LA/right homologue of LA 0.53 .07 0.15 
Nonverbal ATL −0.06 0.744 < .001 
VWFA 0.12 — — 
FFA — — 0.50 .11 
LA/right homologue of LA 0.15 0.54 .051 
Semantic ModalityCortical RegionLeft HemisphereRight Hemisphere
Rhop ValueRhop Value
Verbal ATL 0.31 0.48 .16 
VWFA 0.23 — — 
FFA — — 0.15 
LA/right homologue of LA 0.53 .07 0.15 
Nonverbal ATL −0.06 0.744 < .001 
VWFA 0.12 — — 
FFA — — 0.50 .11 
LA/right homologue of LA 0.15 0.54 .051 

To ensure that p values are comparable to those reported in Table 4, they were corrected for the same number of multiple comparisons (14 tests). Results in bold indicate significant correlations.

Results of the PCA for MRI features (mean cortical thickness, MD) and semantic scores (verbal, nonverbal) are shown in the Supplementary Figure S1 (available at http://doi.org/10.5281/zenodo.3686064). The first principal component (horizontal axis), representing 39% of the total variance, seems to represent the level of neurodegeneration. Indeed, it opposes mean cortical thickness to white matter loss: the thinner the mean cortical thickness of a patient, the larger the MD (and thus the larger the white matter loss). The second principal component (vertical axis), representing 29% of the total variance, seems to differentiate MRI features from the left and the right hemisphere as well as verbal and nonverbal scores. It should be noted that verbal scores are associated with left hemisphere MRI features whereas nonverbal scores are associated with right hemisphere MRI features. Together, these two axes seem to indicate an effect of lateralization associating, on the one hand, the impairment of nonverbal semantics with neurodegeneration of the studied gray and white matter in the right hemisphere and, on the other hand, the impairment of verbal semantics with neurodegeneration of the gray and white matter in the left hemisphere.

Implication of the Gray Matter Loss for the Relation between Semantic Performance and White Matter Integrity

The results of the partial correlations analyses between composite semantic scores (verbal and nonverbal task modalities) and tract integrity, corrected for the mean cortical thickness of the region(s) connected to these tracts, are summarized in Table 9. Verbal semantic scores remained significantly correlated with the ATL–VWFA tract after inclusion of the ATL and/or the VWFA regions. Nonverbal semantic scores remained significantly correlated for two tracts in the right hemisphere: the ATL–FFA tract corrected for the FFA region and the right homologue of ATL–LA tract corrected for the right homologue of LA region.

Table 9. 
Results of Partial Correlation Analyses Using Tract Measures of MD and Composite Semantic Verbal and Nonverbal/Visual Scores Partialled Out for Cortical Thickness Measures
Semantic ModalityTractsCorrected ForLeft Hemisphere
Rhop Value
Verbal Uncinate fasciculus ATL 0.06 
ILF ATL −0.45 0.29 
ATL–VWFA tract ATL −0.63 0.009 
VWFA −0.61 0.015 
ATL + VWFA −0.63 0.013 
ATL–LA tract ATL −0.50 0.13 
LA −0.34 
ATL + LA −0.35 
  
 Right Hemisphere
Rhop Value
Nonverbal Uncinate fasciculus ATL −0.15 
ILF ATL −0.001 
ATL–FFA tract ATL −0.24 
FFA −0.64 0.006 
ATL + FFA −0.24 
Right homologue of ATL–LA tract ATL −0.42 0.47 
LA −0.66 0.003 
ATL + LA −0.43 0.52 
Semantic ModalityTractsCorrected ForLeft Hemisphere
Rhop Value
Verbal Uncinate fasciculus ATL 0.06 
ILF ATL −0.45 0.29 
ATL–VWFA tract ATL −0.63 0.009 
VWFA −0.61 0.015 
ATL + VWFA −0.63 0.013 
ATL–LA tract ATL −0.50 0.13 
LA −0.34 
ATL + LA −0.35 
  
 Right Hemisphere
Rhop Value
Nonverbal Uncinate fasciculus ATL −0.15 
ILF ATL −0.001 
ATL–FFA tract ATL −0.24 
FFA −0.64 0.006 
ATL + FFA −0.24 
Right homologue of ATL–LA tract ATL −0.42 0.47 
LA −0.66 0.003 
ATL + LA −0.43 0.52 

To ensure that p values are comparable to those reported in Table 4, they were corrected for the same number of multiple comparisons (14 tests). Results in bold indicate significant partial correlations.

DISCUSSION

The present investigation explored a large population of SD patients and healthy controls to identify the white matter network of semantics essential for conveying modular information from posterior cortices to the supramodal hub in the ATL. We first delineated three well-documented module-related cortices and performed probabilistic fiber tracking in healthy controls to identify the tracts connecting these cortical regions to the ATL. In a second step, we analyzed the alterations of these tracts in SD and conducted correlations analyses with tract parameters and semantic scores of verbal and nonverbal tasks to investigate the semantic roles of the different tracts. Finally, we evaluated the putative semantic functions of the ILF and the uncinate fasciculus, which have been shown to project to the ATL.

Our results reveal the existence of several tracts projecting from cortical modules to the semantic ATL hub: the ATL–FFA tract, the ATL–VWFA tract, and the ATL–LA tract. Our findings show that all tracts are microstructurally altered in SD and that they subserve specific semantic functions related to the transfer of modular verbal information to the left ATL (ATL–VWFA tract, ATL–LA tract) and of modular nonverbal/visual information to the right ATL (ATL–FFA tract). Furthermore, the semantic roles of the ILF and the uncinate fasciculus appear to be restricted to nonverbal operations in the right hemisphere. These findings have implications at multiples levels: Anatomically, they enrich the knowledge of the white matter network converging onto the ATL; neurocognitively, they elucidate major components of the semantic network conveying modular information to the supramodal hub of semantics; finally, regarding SD, they highlight wide-spread microstructural and functional damage to the semantic network extending beyond the ILF and the uncinate fasciculus.

Anatomical Connections between the ATL and Remote Cortices

Since the creation of fiber tracking techniques, numerous imaging studies have identified the major structural connections interlinking gray matter regions, but most of these studies were anatomically driven and did not build on cognitive/functional models. They have identified some tracts probably contributing to language or semantic processing, including the ILF, the superior longitudinal fasciculus comprising the arcuate, the uncinate fasciculus, and the inferior frontal-occipital fasciculus (Friederici & Gierhan, 2013; Saur et al., 2008, 2010; Duffau, 2008; Mandonnet et al., 2007). However, there is still a huge gap between the complexity and multitude of language/semantic processes and the relatively small number of explored tracts. Regarding semantics, the core computations are claimed to be implemented by the supramodal ATL hub, but structural connections between the ATL and remote module-implementing cortices have not been investigated. Only two ATL connections, the ILF and the uncinate fasciculus, have been highlighted, yet without considering whether they might connect the ATL to cortical input modules for semantics. The same holds for a less documented tract referred to as the “middle longitudinal fasciculus” (MLF; Makris et al., 2009, 2013, 2017).

We addressed these open issues in healthy adults by exploring ATL tracts specifically connecting cortices implementing modular representations, along with the two well-documented ATL connections. One novelty of our tracking study was the identification of tracts connecting the ATL to three modular cortical regions, which implement (i) written word forms: the ATL–VWFA tract, (ii) abstract lexical representations: the ATL–LA tract, and (iii) face/object representations: the ATL–FFA tract. Previous tracking studies did not provide evidence for these tracts probably because they did not explicitly search such connections given that they were hardly based on a cognitive module-driven approach. Furthermore, previous studies in healthy adults often attempted to replicate connectivity findings from animals/primates that do not have a human-like semantic processing system. There are, however, two exceptions suggesting connections between the ATL and module-implementing cortices. First, connections between the VWFA and the ATL have been identified by Bouhali et al. (2014), but this investigation did not check for their semantic role. Second, fiber tracking studies in the macaque monkey have identified the so-called “MLF,” which might eventually overlap with our ATL–LA tract. However, there are only few investigation in human adults confirming the existence of the MLF, and its involvement in semantics is still debated (Jouen et al., 2018; Makris et al., 2009, 2013, 2017). Furthermore, there is currently no published atlas for the MLF, which makes it difficult to know whether this tract corresponds to or partly overlaps with our ATL–LA tract.

In summary, the present tracking findings in healthy adults enrich and extend the results of previous studies by providing evidence for the existence of an anatomical ATL network, connecting the ATL to specific remote cortices involved in cognitive modules that require subsequent semantic processing. However, the study of healthy controls could not clarify whether these module-hub tracts are genuinely involved in semantics given the lack of variability of semantic task scores (near-ceiling effects), which impeded the rational for conducting correlation analyses between tract parameters and semantic scores. SD patients provide the opportunity to address this anatomo-functional issue.

Semantic Roles of ATL Tracts

A major issue was to check whether the ATL–FFA tract, the ATL–VWFA tract, and the ATL–LA tract play a genuine role in semantic processing. Our correlation analyses with SD patients bear on this issue, indicating that the three tracts are related to semantic processes and that the left hemisphere ATL–VWFA tract and ATL–LA tract convey modular language information to the semantic hub whereas the right hemisphere ATL–FFA tract transfers information of modular face and visual object representations. The specific semantic roles of these tracts are further strengthened by our results on a control task showing that scores of the Raven's Progressive Matrices, which taps high-level processes of analogical/rule reasoning without involving semantic representations/processes, do not correlate with the parameters of any of the three module-hub tracts. Furthermore, the right homologue of the ATL–LA tract, which was used as a control tract regarding the left hemisphere ATL–LA tract, does not correlate with scores of verbal semantics but only with those of nonverbal semantics. On the contrary, the right motor-related corticospinal tract, also used as a control tract, correlates with scores of nonverbal semantics. This is an unexpected result that might possibly suggest that patients with strong nonverbal semantic impairment have more extensive right hemisphere damage, including the right corticospinal tract, although the tract was not significantly damaged regarding its FA/MD parameters.

The observed connections between language and nonlanguage/visual modules and the semantic system are furthermore consistent with cognitive models of language and object processing. In language, the majority of psycholinguistic models assume that the mental lexicon and the semantic system are directly linked (Indefrey & Levelt, 2004; Levelt et al., 1999; Caramazza, 1997), which allows for the extraction of meaning from word forms during language perception and the access to semantically appropriate word forms during language production. These cognitive links between word forms and semantic representations are materialized by two tracts identified in our study: the ATL–LA tract and the ATL–VWFA tract connecting the ATL to posterior temporal cortices implementing the mental lexicon (e.g., Race & Hillis, 2015; Graves et al., 2007, 2008; Kotz et al., 2002) and to the mid-posterior region of the left fusiform gyrus implementing orthographical word forms (McCandliss et al., 2003; Cohen et al., 2002). A similar architecture holds for visual object recognition in cognitive models, which state the existence of a repertory of visual object/face forms, referred to as the “structural description system” (Gerlach et al., 1999; Lloyd-Jones & Humphreys, 1997). This modular repertory is assumed to be linked to the semantic system allowing for the extraction of object/face meanings. Such cognitively modeled links are, at least partly, embodied by the ATL–FFA tract, connecting the ATL to the mid-posterior region of the right fusiform gyrus implementing face and object representations (Xu, 2005; Tarr & Gauthier, 2000; Gerlach et al., 1999; Kanwisher et al., 1997). These findings represent a first step revealing the existence of anatomo-functional connections linking modules to the semantic hub system and, more generally, providing anatomical evidence corroborating major cognitive models and investigations of the modular organization of cognition (Pinker & Jackendoff, 2005; Tyler et al., 2005; Indefrey & Levelt, 2004; Ullman, 2001; Levelt et al., 1999; Caramazza & Miozzo, 1997; Fodor, 1983). Based on our findings and on cognitive models accounting for modules, unlike connectionist models, we propose to consider an extension of the “hub-and-spokes” model of semantics (Lambon Ralph et al., 2017) to a “hub-module-spokes” model where spokes represent/process motor–sensory features as stipulated by the authors of the model. Such motor–sensory features that are complex, highly variable, and near-unlimited in number are concatenated into invariant and robust modular representations (verbal—lexical representations and nonverbal—visual object representations), which are limited in number and which encode the limited number of verbal and nonverbal objects of the human environment. These modular representations might constitute the operational and exploitable basis for the central semantic hub system that forms, activates, and generalizes concepts, the number of which is also limited like real-world objects.

The identified tracts materializing module-hub connections also enrich previous anatomical investigations, which have explored multiple connections between the ATL and remote temporal, parietal, and frontal regions, yet without assessing their specific roles in semantics (Jouen et al., 2018; Pascual et al., 2015; Binney et al., 2012). Furthermore, our results are in line with several authors suggesting that semantic processing is subserved by a ventral stream pathway, including the ILF, the uncinate fasciculus (e.g., Saur et al., 2008, 2010), and the inferior fronto-occipital pathway (Duffau, 2008; Mandonnet et al., 2007; Duffau et al., 2005), while identifying the critical module-hub tracts within this ventral stream and assigning specific semantic roles to them.

Our findings open an avenue for future research that needs to replicate them in large settings of other patient populations using, in addition to correlation analyses, disruption of the module-hub tracts to further substantiate their causative role in semantics. Likewise functional connectivity studies based on fMRI resting state might confirm and enrich the existence of the semantic roles of the ATL–FFA, the ATL–VWFA, and the ATL–LA tract. Another open issue is the exploration of contralateral homologues of these tracts and of the three module-related cortices, which were investigated in this study. Our results on control tracts showed that parameters of the right homologue of the ATL–LA tract correlate with task scores of nonverbal semantics, which suggests that, in addition to the FFA, the right posterior mid/superior temporal cortex might play a role in the implementation of modular representations of visual objects and that the right homologue of the ATL–LA tract might convey modular information of visual object representations to the right ATL. Future research is required to address these issues with the aim of improving our understanding of the anatomical organization of cognitive modules and of the semantic module-hub network. Additional research is also needed to explore whether the bilateral ATL hub represents a unified system dedicated to supramodal semantics or whether there is a left versus right specialization or gradient. In particular, left versus right hemisphere module-hub tracts, such as those evidenced in this study, might provide input for verbal versus nonverbal/visual semantic computations in the left and right ATL, respectively. Providing evidence for this assumption would strengthen the claim of several authors positing that verbal semantics is primarily processed in the left ATL whereas visual nonverbal semantics might be computed in the right ATL (Gainotti, 2013; Mesulam et al., 2013; Snowden et al., 2004).

Regarding the semantic functions of the ILF and the uncinate fasciculus, our results suggest that their semantic contribution is restricted to their right hemisphere fiber bundles and that their semantic roles in conveying nonverbal information to the ATL hub are less salient than the role of the ATL–FFA tract. Accordingly, only the right hemisphere ILF and uncinate fasciculus were correlated with semantic performance restricted to nonverbal/visual task modalities. In the literature, the exact semantic roles of the ILF and the uncinate fasciculus are not fully elucidated because prior studies provided partially contradicting results regarding a potential semantic function of these tracts or did not contrast verbal and nonverbal modalities (Friederici & Gierhan, 2013; Duffau, 2008; Mandonnet et al., 2007). Further work needs also to explore whether fibers especially of the ILF are potentially intermingled with right hemisphere fibers of the ATL–FFA tract or with left hemisphere fibers of the ATL–VWFA tract or the ATL–LA tract. Such intermingled fiber trajectories might explain the inconsistent results of studies having investigated the semantic function of the ILF. Our results in healthy controls showed poor overlap of the ATL–VWFA tract and the ATL–LA tract with the left ILF and poor overlap between the ATL–FFA tract and the right ILF, indicating that the explored module-hub tracts are specific fiber bundles, yet probably sharing some fiber contingents with the ILF. Regarding the uncinate fasciculus, which links the ATL to prefrontal cortices, it remains speculative how it contributes to semantic processing given that prefrontal regions do not implement modular information. It might be that parts of the uncinate fasciculus constitute output channels conveying semantic information from the ATL to prefrontal cortices, which subserve semantic-dependent decision-making and reasoning (Shallice & Cipolotti, 2018). Conversely, some parts of the uncinate fasciculus might constitute input channels to the ATL allowing for executive-driven search, retrieval, and modulation of information in the semantic system (Harvey, Wei, Ellmore, Hamilton, & Schnur, 2013; Saur et al., 2010), which has been referred to as “semantic control” by Lambon Ralph et al. (2017).

ATL Tracts and SD

Our results showing damage to the ILF and the uncinate fasciculus are in line with several previous tracking studies in SD (Agosta et al., 2013, 2015; Galantucci et al., 2011). This study, in addition, demonstrates that tract damage in SD extends beyond these two well-documented connections toward semantic module-hub tracts including the ATL–VWFA tract, the ATL–LA tract, and the ATL–FFA tract. Furthermore, our results showed that tract alterations predominated in the left hemisphere, which is consistent with the finding that early stages of SD demonstrate predominant left ATL atrophy (e.g., Mummery et al., 2000), indicating that left-lateralized damage of ATL cortices and ATL tracts is a signature of early-stage SD. Such a wide-spread ATL network damage is in line with a recent imaging study in SD (Collins et al., 2017) exploring the functional connectivity between the ATL and multiple brain regions, yet without focusing on cortical modules. Hence, it appears that the degenerative process in SD, which primarily affects the ATL (Gorno-Tempini et al., 2004, 2011), spreads insidiously to more posterior regions through the progressive erosion of the entire white matter network of semantics. As a consequence, semantic breakdown in SD patients is caused by both, damage to the ATL and to tracts conveying modular information, which needs to be processed in the ATL cortices. It is, however, unclear whether the degenerative process selectively affects semantic-related networks or whether there is a general breakdown of various networks, shown to converge onto distinct subportions of the ATL (Papinutto et al., 2016).

Limitations

Our study has several limitations. First, as in any anatomo-cognitive correlation analysis, significant correlations do not necessarily imply involvement of a given anatomical structure/tract in a cognitive process. Another limitation is related to the use of white matter tractography from diffusion MRI. It has been highlighted in many studies that tractography is highly sensitive to the parameters of image acquisition and of the image processing procedures (Barrio-Arranz, de Luis-García, Tristán-Vega, Martín-Fernández, & Aja-Fernández, 2015; Tensaouti, Lahlou, Clarisse, Lotterie, & Berry, 2011). Although we used state-of-the-art acquisition and processing procedures, we cannot rule out the possibility that some artifactual fibers have been generated by tractography.

Finally, it remains unclear whether degenerative cortical diseases such as SD affect secondarily the white matter or whether there is also cortex-independent damage of white matter/tracts. On the one hand, the lack of a significant correlation between the left ATL and verbal semantic scores contrasts with significant correlations of the ATL–VWFA tract and the ATL–LA tract with verbal semantic performance. On the other hand, the significant correlation of the right ATL with nonverbal semantic scores goes along with significant correlations between right hemisphere tracts (ATL–FFA tract, ILF) and nonverbal semantic scores. Such diverging findings might be related to the severe atrophy in the left ATL as compared with the right ATL, reducing cortical thickness variability and thus the possibility to yield significant correlations with verbal semantic scores. Conversely, the results of our partial correlation analysis indicate that verbal semantics is related to the ATL–VWFA tract independently of the ATL or VWFA atrophy. A hypothesis could be that the verbal information is still conveyed to the left ATL by the ATL–VWFA tract, but that it is not sufficiently processed by this region because of the severe atrophy. Taken together, we are confident that significant correlations between tract parameters and verbal or nonverbal semantic scores of our study are an important indicator that the module-hub tracts, in addition to the ATL cortices, contribute to semantic processing. In the same vein, neuropathological investigations in the most frequent neurogenerative condition affecting the cortex, that is, Alzheimer's disease, have indicated that white matter/tract damage is triggered by cortical damage via Wallerian degeneration (McAleese et al., 2015, 2017), thus strongly suggesting that both gray and with matter structures are involved in a given cognitive process. We are, however, aware that additional studies are needed using focal brain lesions disrupting “module-hub tracts” to demonstrate their causal involved in semantic processing.

Conclusion

Our findings reveal crucial components of the white matter network of semantics by providing evidence for three tracts and their function in conveying modular verbal and nonverbal information to the supramodal semantic hub in the ATL. They enrich and extend the knowledge of the architecture of semantics and satisfy fundamental claims of cognitive processing models positing links between modular representations and the semantic system. Accordingly we propose to integrate modules into the currently prevailing connectionist-based “hub-and-spokes” model while extending it to a “hub-module-spokes” account of semantics. Our results also indicate that SD, like most degenerative conditions, is a network disease eroding in addition to the ATL almost the entire network of verbal and nonverbal semantics. Hence, both cortical and specific white matter imaging markers might be useful for monitoring disease evolution, indicating prognostic factors, and evaluating the potential efficacy of future therapy trials in SD. Our findings also open a route for future studies investigating whether the left versus right ATL cortices impact distinct module-related computations of semantics or whether a functional lateralization of the ATL system is primarily linked to the lateralized fiber connections with left and right hemisphere modules.

Acknowledgments

This research has received funding from the program Investissements d'avenir ANR-10-IAIHU-06 (Agence Nationale de la Recherche-10-IA Agence Institut Hospitalo-Universitaire-6) and from the French government under management of the Agence Nationale de la Recherche as part of the “Investissements d'avenir” program, reference ANR-19-P3IA-0001 (PRAIRIE 3IA Institute). O. C. was supported by a Contrat d'Interface Local from Assistance Publique-Hôpitaux de Paris (AP-HP).

Reprint requests should be sent to Marc Teichmann, Department of Neurology, National Reference Center for Rare Dementias, Hôpital de la Pitié Salpêtrière, 47-83, boulevard de l'Hôpital. 75013 Paris, France, or via e-mail: marc.teichmann@psl.aphp.fr.

REFERENCES

Acosta-Cabronero
,
J.
,
Patterson
,
K.
,
Fryer
,
T. D.
,
Hodges
,
J. R.
,
Pengas
,
G.
,
Williams
,
G. B.
, et al
(
2011
).
Atrophy, hypometabolism and white matter abnormalities in semantic dementia tell a coherent story
.
Brain
,
134
,
1869
1871
.
Agosta
,
F.
,
Ferraro
,
P. M.
,
Canu
,
E.
,
Copetti
,
M.
,
Galantucci
,
S.
,
Magnani
,
G.
, et al
(
2015
).
Differentiation between subtypes of primary progressive aphasia by using cortical thickness and diffusion-tensor MR imaging measures
.
Radiology
,
276
,
219
227
.
Agosta
,
F.
,
Galantucci
,
S.
,
Canu
,
E.
,
Cappa
,
S. F.
,
Magnani
,
G.
,
Franceschi
,
M.
, et al
(
2013
).
Disruption of structural connectivity along the dorsal and ventral language pathways in patients with nonfluent and semantic variant primary progressive aphasia: A DT MRI study and a literature review
.
Brain and Language
,
127
,
157
166
.
Agosta
,
F.
,
Henry
,
R. G.
,
Migliaccio
,
R.
,
Neuhaus
,
J.
,
Miller
,
B. L.
,
Dronkers
,
N. F.
, et al
(
2010
).
Language networks in semantic dementia
.
Brain
,
133
,
286
299
.
Avants
,
B. B.
,
Epstein
,
C. L.
,
Grossman
,
M.
, &
Gee
,
J. C.
(
2008
).
Symmetric diffeomorphic image registration with cross-correlation: Evaluating automated labeling of elderly and neurodegenerative brain
.
Medical Image Analysis
,
12
,
26
41
.
Barrio-Arranz
,
G.
,
de Luis-García
,
R.
,
Tristán-Vega
,
A.
,
Martín-Fernández
,
M.
, &
Aja-Fernández
,
S.
(
2015
).
Impact of MR acquisition parameters on DTI scalar indexes: A tractography based approach
.
PLoS One
,
10
,
e0137905
.
Behrens
,
T. E. J.
,
Woolrich
,
M. W.
,
Jenkinson
,
M.
,
Johansen-Berg
,
H.
,
Nunes
,
R. G.
,
Clare
,
S.
, et al
(
2003
).
Characterization and propagation of uncertainty in diffusion-weighted MR imaging
.
Magnetic Resonance in Medicine
,
50
,
1077
1088
.
Binder
,
J. R.
,
Desai
,
R. H.
,
Graves
,
W. W.
, &
Conant
,
L. L.
(
2009
).
Where is the semantic system? A critical review and meta-analysis of 120 functional neuroimaging studies
.
Cerebral Cortex
,
19
,
2767
2796
.
Binney
,
R. J.
,
Parker
,
G. J. M.
, &
Lambon Ralph
,
M. A.
(
2012
).
Convergent connectivity and graded specialization in the rostral human temporal lobe as revealed by diffusion-weighted imaging probabilistic tractography
.
Journal of Cognitive Neuroscience
,
24
,
1998
2014
.
Bouhali
,
F.
,
Thiebaut de Schotten
,
M.
,
Pinel
,
P.
,
Poupon
,
C.
,
Mangin
,
J.-F.
,
Dehaene
,
S.
, et al
(
2014
).
Anatomical connections of the visual word form area
.
Journal of Neuroscience
,
34
,
15402
15414
.
Bozeat
,
S.
,
Lambon Ralph
,
M. A.
,
Patterson
,
K.
,
Garrard
,
P.
, &
Hodges
,
J. R.
(
2000
).
Non-verbal semantic impairment in semantic dementia
.
Neuropsychologia
,
38
,
1207
1215
.
Calamante
,
F.
,
Tournier
,
J.-D.
,
Jackson
,
G. D.
, &
Connelly
,
A.
(
2010
).
Track-density imaging (TDI): Super-resolution white matter imaging using whole-brain track-density mapping
.
Neuroimage
,
53
,
1233
1243
.
Caramazza
,
A.
(
1997
).
How many levels of processing are there in lexical access?
Cognitive Neuropsychology
,
14
,
177
208
.
Caramazza
,
A.
, &
Miozzo
,
M.
(
1997
).
The relation between syntactic and phonological knowledge in lexical access: Evidence from the “tip-of-the-tongue” phenomenon
.
Cognition
,
64
,
309
343
.
Cardebat
,
D.
,
Doyon
,
B.
,
Puel
,
M.
,
Goulet
,
P.
, &
Joanette
,
Y.
(
1990
).
Formal and semantic lexical evocation in normal subjects: Performance and dynamics of production as a function of sex, age and educational level
.
Acta Neurologica Belgica
,
90
,
207
217
.
Catani
,
M.
,
Howard
,
R. J.
,
Pajevic
,
S.
, &
Jones
,
D. K.
(
2002
).
Virtual in vivo interactive dissection of white matter fasciculi in the human brain
.
Neuroimage
,
17
,
77
94
.
Catani
,
M.
,
Jones
,
D. K.
,
Donato
,
R.
, &
Ffytche
,
D. H.
(
2003
).
Occipito-temporal connections in the human brain
.
Brain
,
126
,
2093
2107
.
Cohen
,
L.
,
Dehaene
,
S.
,
Naccache
,
L.
,
Lehéricy
,
S.
,
Dehaene-Lambertz
,
G.
,
Hénaff
,
M.-A.
, et al
(
2000
).
The visual word form area: Spatial and temporal characterization of an initial stage of reading in normal subjects and posterior split-brain patients
.
Brain
,
123
,
291
307
.
Cohen
,
L.
,
Lehéricy
,
S.
,
Chochon
,
F.
,
Lemer
,
C.
,
Rivaud
,
S.
, &
Dehaene
,
S.
(
2002
).
Language-specific tuning of visual cortex? Functional properties of the visual word form area
.
Brain
,
125
,
1054
1069
.
Collins
,
J. A.
,
Montal
,
V.
,
Hochberg
,
D.
,
Quimby
,
M.
,
Mandelli
,
M. L.
,
Makris
,
N.
, et al
(
2017
).
Focal temporal pole atrophy and network degeneration in semantic variant primary progressive aphasia
.
Brain
,
140
,
457
471
.
Damasio
,
H.
,
Tranel
,
D.
,
Grabowski
,
T.
,
Adolphs
,
R.
, &
Damasio
,
A.
(
2004
).
Neural systems behind word and concept retrieval
.
Cognition
,
92
,
179
229
.
Deloche
,
G.
,
Metz-Lutz
,
M.
, &
Kremin
,
H.
(
1997
).
Test de dénomination orale de 80 images
. In
Editions du Centre de Psychologie Appliquée CDP
.
Paris
.
Dubois
,
B.
,
Slachevsky
,
A.
,
Litvan
,
I.
, &
Pillon
,
B.
(
2000
).
The FAB: A frontal assessment battery at bedside
.
Neurology
,
55
,
1621
1626
.
Duffau
,
H.
(
2008
).
The anatomo-functional connectivity of language revisited: New insights provided by electrostimulation and tractography
.
Neuropsychologia
,
46
,
927
934
.
Duffau
,
H.
,
Gatignol
,
P.
,
Mandonnet
,
E.
,
Peruzzi
,
P.
,
Tzourio-Mazoyer
,
N.
, &
Capelle
,
L.
(
2005
).
New insights into the anatomo-functional connectivity of the semantic system: A study using cortico-subcortical electrostimulations
.
Brain
,
128
,
797
810
.
Fischl
,
B.
(
2012
).
FreeSurfer
.
Neuroimage
,
62
,
774
781
.
Fodor
,
J. A.
(
1983
).
The modularity of mind
.
Cambridge, MA
:
MIT Press
.
Folstein
,
M. F.
,
Folstein
,
S. E.
, &
McHugh
,
P. R.
(
1975
).
“Mini-mental state”: A practical method for grading the cognitive state of patients for the clinician
.
Journal of Psychiatric Research
,
12
,
189
198
.
Friederici
,
A. D.
, &
Gierhan
,
S. M. E.
(
2013
).
The language network
.
Current Opinion in Neurobiology
,
23
,
250
254
.
Gainotti
,
G.
(
2013
).
Laterality effects in normal subjects' recognition of familiar faces, voices and names: Perceptual and representational components
.
Neuropsychologia
,
51
,
1151
1160
.
Galantucci
,
S.
,
Tartaglia
,
M. C.
,
Wilson
,
S. M.
,
Henry
,
M. L.
,
Filippi
,
M.
,
Agosta
,
F.
, et al
(
2011
).
White matter damage in primary progressive aphasias: A diffusion tensor tractography study
.
Brain
,
134
,
3011
3029
.
Gauthier
,
I.
,
Tarr
,
M. J.
,
Moylan
,
J.
,
Anderson
,
A. W.
,
Skudlarski
,
P.
, &
Gore
,
J. C.
(
2000
).
Does visual subordinate-level categorisation engage the functionally defined fusiform face area?
Cognitive Neuropsychology
,
17
,
143
163
.
Gerlach
,
C.
,
Law
,
I.
,
Gade
,
A.
, &
Paulson
,
O. B.
(
1999
).
Perceptual differentiation and category effects in normal object recognition: A PET study
.
Brain
,
122
,
2159
2170
.
Gorno-Tempini
,
M. L.
,
Dronkers
,
N. F.
,
Rankin
,
K. P.
,
Ogar
,
J. M.
,
Phengrasamy
,
L.
,
Rosen
,
H. J.
, et al
(
2004
).
Cognition and anatomy in three variants of primary progressive aphasia
.
Annals of Neurology
,
55
,
335
346
.
Gorno-Tempini
,
M. L.
,
Hillis
,
A. E.
,
Weintraub
,
S.
,
Kertesz
,
A.
,
Mendez
,
M.
,
Cappa
,
S. F.
, et al
(
2011
).
Classification of primary progressive aphasia and its variants
.
Neurology
,
76
,
1006
1014
.
Graves
,
W. W.
,
Grabowski
,
T. J.
,
Mehta
,
S.
, &
Gordon
,
J. K.
(
2007
).
A neural signature of phonological access: Distinguishing the effects of word frequency from familiarity and length in overt picture naming
.
Journal of Cognitive Neuroscience
,
19
,
617
631
.
Graves
,
W. W.
,
Grabowski
,
T. J.
,
Mehta
,
S.
, &
Gupta
,
P.
(
2008
).
The left posterior superior temporal gyrus participates specifically in accessing lexical phonology
.
Journal of Cognitive Neuroscience
,
20
,
1698
1710
.
Guo
,
C. C.
,
Gorno-Tempini
,
M. L.
,
Gesierich
,
B.
,
Henry
,
M.
,
Trujillo
,
A.
,
Shany-Ur
,
T.
, et al
(
2013
).
Anterior temporal lobe degeneration produces widespread network-driven dysfunction
.
Brain
,
136
,
2979
2991
.
Harvey
,
D. Y.
,
Wei
,
T.
,
Ellmore
,
T. M.
,
Hamilton
,
A. C.
, &
Schnur
,
T. T.
(
2013
).
Neuropsychological evidence for the functional role of the uncinate fasciculus in semantic control
.
Neuropsychologia
,
51
,
789
801
.
Hickok
,
G.
, &
Poeppel
,
D.
(
2004
).
Dorsal and ventral streams: A framework for understanding aspects of the functional anatomy of language
.
Cognition
,
92
,
67
99
.
Hocking
,
R. R.
(
1985
).
The analysis of linear models
.
Belmont, CA
:
Brooks/Cole Publishing Co
.
Howard
,
D.
, &
Patterson
,
K.
(
1992
).
Pyramids and palm trees: A test of semantic access from pictures and words
.
Bury St. Edmunds, UK
:
Thames Valley Test Company
.
Iaccarino
,
L.
,
Crespi
,
C.
,
Della Rosa
,
P. A.
,
Catricalà
,
E.
,
Guidi
,
L.
,
Marcone
,
A.
, et al
(
2015
).
The semantic variant of primary progressive aphasia: Clinical and neuroimaging evidence in single subjects
.
PLoS One
,
10
,
e0120197
.
Indefrey
,
P.
, &
Levelt
,
W. J. M.
(
2004
).
The spatial and temporal signatures of word production components
.
Cognition
,
92
,
101
144
.
Jeurissen
,
B.
,
Tournier
,
J.-D.
,
Dhollander
,
T.
,
Connelly
,
A.
, &
Sijbers
,
J.
(
2014
).
Multi-tissue constrained spherical deconvolution for improved analysis of multi-shell diffusion MRI data
.
Neuroimage
,
103
,
411
426
.
Jezzard
,
P.
, &
Balaban
,
R. S.
(
1995
).
Correction for geometric distortion in echo planar images from B0 field variations
.
Magnetic Resonance in Medicine
,
34
,
65
73
.
Jouen
,
A. L.
,
Ellmore
,
T. M.
,
Madden-Lombardi
,
C. J.
,
Pallier
,
C.
,
Dominey
,
P. F.
, &
Ventre-Dominey
,
J.
(
2018
).
Beyond the word and image: II—Structural and functional connectivity of a common semantic system
.
Neuroimage
,
166
,
185
197
.
Kanwisher
,
N.
,
McDermott
,
J.
, &
Chun
,
M. M.
(
1997
).
The fusiform face area: A module in human extrastriate cortex specialized for face perception
.
Journal of Neuroscience
,
17
,
4302
4311
.
Kloke
,
J. D.
, &
McKean
,
J. W.
(
2012
).
Rfit: Rank-based estimation for linear models
.
R Journal
,
4
,
57
64
.
Kotz
,
S. A.
,
Cappa
,
S. F.
,
von Cramon
,
D. Y.
, &
Friederici
,
A. D.
(
2002
).
Modulation of the lexical–semantic network by auditory semantic priming: An event-related functional MRI study
.
Neuroimage
,
17
,
1761
1772
.
Laisney
,
M.
,
Eustache
,
F.
, &
Desgranges
,
B.
(
2009
).
Assessment of semantic memory for famous people
.
Revue de Neuropsychologie
,
1
,
175
183
.
Lambon Ralph
,
M. A.
,
Graham
,
K. S.
,
Patterson
,
K.
, &
Hodges
,
J. R.
(
1999
).
Is a picture worth a thousand words? Evidence from concept definitions by patients with semantic dementia
.
Brain and Language
,
70
,
309
335
.
Lambon Ralph
,
M. A.
,
Jefferies
,
E.
,
Patterson
,
K.
, &
Rogers
,
T. T.
(
2017
).
The neural and computational bases of semantic cognition
.
Nature Reviews Neuroscience
,
18
,
42
55
.
Leemans
,
A.
, &
Jones
,
D. K.
(
2009
).
The B-matrix must be rotated when correcting for subject motion in DTI data
.
Magnetic Resonance in Medicine
,
61
,
1336
1349
.
Levelt
,
W. J. M.
,
Roelofs
,
A.
, &
Meyer
,
A. S.
(
1999
).
A theory of lexical access in speech production
.
Behavioral and Brain Sciences
,
22
,
1
38
.
Lloyd-Jones
,
T. J.
, &
Humphreys
,
G. W.
(
1997
).
Perceptual differentiation as a source of category effects in object processing: Evidence from naming and object decision
.
Memory & Cognition
,
25
,
18
35
.
Makris
,
N.
,
Papadimitriou
,
G. M.
,
Kaiser
,
J. R.
,
Sorg
,
S.
,
Kennedy
,
D. N.
, &
Pandya
,
D. N.
(
2009
).
Delineation of the middle longitudinal fascicle in humans: A quantitative, in vivo, DT-MRI study
.
Cerebral Cortex
,
19
,
777
785
.
Makris
,
N.
,
Preti
,
M. G.
,
Asami
,
T.
,
Pelavin
,
P.
,
Campbell
,
B.
,
Papadimitriou
,
G. M.
, et al
(
2013
).
Human middle longitudinal fascicle: Variations in patterns of anatomical connections
.
Brain Structure and Function
,
218
,
951
968
.
Makris
,
N.
,
Zhu
,
A.
,
Papadimitriou
,
G. M.
,
Mouradian
,
P.
,
Ng
,
I.
,
Scaccianoce
,
E.
, et al
(
2017
).
Mapping temporo-parietal and temporo-occipital cortico-cortical connections of the human middle longitudinal fascicle in subject-specific, probabilistic, and stereotaxic Talairach spaces
.
Brain Imaging and Behavior
,
11
,
1258
1277
.
Mandonnet
,
E.
,
Nouet
,
A.
,
Gatignol
,
P.
,
Capelle
,
L.
, &
Duffau
,
H.
(
2007
).
Does the left inferior longitudinal fasciculus play a role in language? A brain stimulation study
.
Brain
,
130
,
623
629
.
Mattis
,
S.
(
1976
).
Mental status examination for organic mental syndrome in the elderly patient
. In
L.
Bellak
&
T. B.
Karasu
(Eds.),
Geriatic psychiatry
(pp.
77
121
).
Orlando, FL
:
Grune & Stratton
.
Mazaux
,
J. M.
, &
Orgogozo
,
J. M.
(
1982
).
Boston diagnostic aphasia examination
.
Paris
:
Editions Scientifiques et Psychologiques
.
McAleese
,
K. E.
,
Firbank
,
M.
,
Dey
,
M.
,
Colloby
,
S. J.
,
Walker
,
L.
,
Johnson
,
M.
, et al
(
2015
).
Cortical tau load is associated with white matter hyperintensities
.
Acta Neuropathologica Communications
,
3
,
60
.
McAleese
,
K. E.
,
Walker
,
L.
,
Graham
,
S.
,
Moya
,
E. L. J.
,
Johnson
,
M.
,
Erskine
,
D.
, et al
(
2017
).
Parietal white matter lesions in Alzheimer's disease are associated with cortical neurodegenerative pathology, but not with small vessel disease
.
Acta Neuropathologica
,
134
,
459
473
.
McCandliss
,
B. D.
,
Cohen
,
L.
, &
Dehaene
,
S.
(
2003
).
The visual word form area: Expertise for reading in the fusiform gyrus
.
Trends in Cognitive Sciences
,
7
,
293
299
.
Merck
,
C.
,
Charnallet
,
A.
,
Auriacombe
,
S.
,
Belliard
,
S.
,
Hahn-Barma
,
V.
,
Kremin
,
H.
, et al
(
2011
).
La batterie d'évaluation des connaissances sémantiques du GRECO (BECS-GRECO): Validation et données normatives
.
Revue de Neuropsychologie
,
3
,
235
255
.
Mesulam
,
M.-M.
,
Wieneke
,
C.
,
Hurley
,
R.
,
Rademaker
,
A.
,
Thompson
,
C. K.
,
Weintraub
,
S.
, et al
(
2013
).
Words and objects at the tip of the left temporal lobe in primary progressive aphasia
.
Brain
,
136
,
601
618
.
Mion
,
M.
,
Patterson
,
K.
,
Acosta-Cabronero
,
J.
,
Pengas
,
G.
,
Izquierdo-Garcia
,
D.
,
Hong
,
Y. T.
, et al
(
2010
).
What the left and right anterior fusiform gyri tell us about semantic memory
.
Brain
,
133
,
3256
3268
.
Mori
,
S.
,
Wakana
,
S.
,
Nagae-Poetscher
,
L.
, &
van Zijl
,
P. C.
(
2005
).
MRI atlas of human white matter
.
Amsterdam
:
Elsevier
.
Mummery
,
C. J.
,
Patterson
,
K.
,
Price
,
C. J.
,
Ashburner
,
J.
,
Frackowiak
,
R. S. J.
, &
Hodges
,
J. R.
(
2000
).
A voxel-based morphometry study of semantic dementia: Relationship between temporal lobe atrophy and semantic memory
.
Annals of Neurology
,
47
,
36
45
.
Operto
,
G.
,
Chupin
,
M.
,
Batrancourt
,
B.
,
Habert
,
M.-O.
,
Colliot
,
O.
,
Benali
,
H.
, et al
(
2016
).
CATI: A large distributed infrastructure for the neuroimaging of cohorts
.
Neuroinformatics
,
14
,
253
264
.
Papinutto
,
N.
,
Galantucci
,
S.
,
Mandelli
,
M. L.
,
Gesierich
,
B.
,
Jovicich
,
J.
,
Caverzasi
,
E.
, et al
(
2016
).
Structural connectivity of the human anterior temporal lobe: A diffusion magnetic resonance imaging study
.
Human Brain Mapping
,
37
,
2210
2222
.
Parker
,
G. J. M.
,
Luzzi
,
S.
,
Alexander
,
D. C.
,
Wheeler-Kingshott
,
C. A. M.
,
Ciccarelli
,
O.
, &
Lambon Ralph
,
M. A.
(
2005
).
Lateralization of ventral and dorsal auditory-language pathways in the human brain
.
Neuroimage
,
24
,
656
666
.
Pascual
,
B.
,
Masdeu
,
J. C.
,
Hollenbeck
,
M.
,
Makris
,
N.
,
Insausti
,
R.
,
Ding
,
S.-L.
, et al
(
2015
).
Large-scale brain networks of the human left temporal pole: A functional connectivity MRI study
.
Cerebral Cortex
,
25
,
680
702
.
Patterson
,
K.
,
Nestor
,
P. J.
, &
Rogers
,
T. T.
(
2007
).
Where do you know what you know? The representation of semantic knowledge in the human brain
.
Nature Reviews Neuroscience
,
8
,
976
987
.
Pinker
,
S.
, &
Jackendoff
,
R.
(
2005
).
The faculty of language: What's special about it?
Cognition
,
95
,
201
236
.
Pobric
,
G.
,
Jefferies
,
E.
, &
Lambon Ralph
,
M. A.
(
2007
).
Anterior temporal lobes mediate semantic representation: Mimicking semantic dementia by using rTMS in normal participants
.
Proceedings of the National Academy of Sciences, U.S.A.
,
104
,
20137
20141
.
Pobric
,
G.
,
Jefferies
,
E.
, &
Lambon Ralph
,
M. A.
(
2010
).
Amodal semantic representations depend on both anterior temporal lobes: Evidence from repetitive transcranial magnetic stimulation
.
Neuropsychologia
,
48
,
1336
1342
.
Race
,
D. S.
, &
Hillis
,
A. B.
(
2015
).
Naming
. In
A. W.
Toga
(Ed.),
Brain mapping
(pp.
671
675
).
Waltham, MA
:
Academic Press
.
Raven
,
J.
,
Raven
,
J. C.
, &
Court
,
J. H.
(
1998
).
Raven manual: Standard progressive matrices
.
Oxford
:
Oxford Psychologists Press
.
R Core Team
. (
2013
).
R: A language and environment for statistical computing
.
Vienna, Austria
:
R Foundation for Statistical Computing
.
Rice
,
G. E.
,
Hoffman
,
P.
, &
Lambon Ralph
,
M. A.
(
2015
).
Graded specialization within and between the anterior temporal lobes
.
Annals of the New York Academy of Sciences
,
1359
,
84
97
.
Rice
,
G. E.
,
Lambon Ralph
,
M. A.
, &
Hoffman
,
P.
(
2015
).
The roles of left versus right anterior temporal lobes in conceptual knowledge: An ALE meta-analysis of 97 functional neuroimaging studies
.
Cerebral Cortex
,
25
,
4374
4391
.
Saur
,
D.
,
Kreher
,
B. W.
,
Schnell
,
S.
,
Kümmerer
,
D.
,
Kellmeyer
,
P.
,
Vry
,
M.-S.
, et al
(
2008
).
Ventral and dorsal pathways for language
.
Proceedings of the National Academy of Sciences, U.S.A.
,
105
,
18035
18040
.
Saur
,
D.
,
Schelter
,
B.
,
Schnell
,
S.
,
Kratochvil
,
D.
,
Küpper
,
H.
,
Kellmeyer
,
P.
, et al
(
2010
).
Combining functional and anatomical connectivity reveals brain networks for auditory language comprehension
.
Neuroimage
,
49
,
3187
3197
.
Shallice
,
T.
, &
Cipolotti
,
L.
(
2018
).
The prefrontal cortex and neurological impairments of active thought
.
Annual Review of Psychology
,
69
,
157
180
.
Snowden
,
J. S.
,
Thompson
,
J. C.
, &
Neary
,
D.
(
2004
).
Knowledge of famous faces and names in semantic dementia
.
Brain
,
127
,
860
872
.
Tarr
,
M. J.
, &
Gauthier
,
I.
(
2000
).
FFA: A flexible fusiform area for subordinate-level visual processing automatized by expertise
.
Nature Neuroscience
,
3
,
764
769
.
Tensaouti
,
F.
,
Lahlou
,
I.
,
Clarisse
,
P.
,
Lotterie
,
J. A.
, &
Berry
,
I.
(
2011
).
Quantitative and reproducibility study of four tractography algorithms used in clinical routine
.
Journal of Magnetic Resonance Imaging
,
34
,
165
172
.
Tournier
,
J.-D.
,
Calamante
,
F.
, &
Connelly
,
A.
(
2007
).
Robust determination of the fibre orientation distribution in diffusion MRI: Non-negativity constrained super-resolved spherical deconvolution
.
Neuroimage
,
35
,
1459
1472
.
Tournier
,
J.-D.
,
Calamante
,
F.
, &
Connelly
,
A.
(
2012
).
MRtrix: Diffusion tractography in crossing fiber regions
.
International Journal of Imaging Systems and Technology
,
22
,
53
66
.
Tustison
,
N. J.
, &
Avants
,
B. B.
(
2013
).
Explicit B-spline regularization in diffeomorphic image registration
.
Frontiers in Neuroinformatics
,
7
,
39
.
Tyler
,
L. K.
,
Marslen-Wilson
,
W. D.
, &
Stamatakis
,
E. A.
(
2005
).
Differentiating lexical form, meaning, and structure in the neural language system
.
Proceedings of the National Academy of Sciences, U.S.A.
,
102
,
8375
8380
.
Ullman
,
M. T.
(
2001
).
A neurocognitive perspective on language: The declarative/procedural model
.
Nature Reviews Neuroscience
,
2
,
717
726
.
Vandenberghe
,
R.
,
Price
,
C.
,
Wise
,
R.
,
Josephs
,
O.
, &
Frackowiak
,
R. S. J.
(
1996
).
Functional anatomy of a common semantic system for words and pictures
.
Nature
,
383
,
254
256
.
Vandenberghe
,
R.
,
Wang
,
Y.
,
Nelissen
,
N.
,
Vandenbulcke
,
M.
,
Dhollander
,
T.
,
Sunaert
,
S.
, et al
(
2013
).
The associative-semantic network for words and pictures: Effective connectivity and graph analysis
.
Brain and Language
,
127
,
264
272
.
Xu
,
Y.
(
2005
).
Revisiting the role of the fusiform face area in visual expertise
.
Cerebral Cortex
,
15
,
1234
1242
.

Author notes

*

These authors contributed equally to the study.

**

on behalf of the ‘CAPP study group