Brain and Language 201 (2020) 104721

Contents lists available at ScienceDirect

Brain and Language

journal homepage: www.elsevier.com/locate/b&l

The left inferior longitudinal fasciculus supports orthographic processing: T

Evidence from a lesion-behavior mapping analysis
Ke Wanga,1, Xiaonan Lia,b,1, Ruiwang Huangc, Junhua Dingd, Luping Songe, , Zaizhu Hana,
⁎ ⁎

a
State Key Laboratory of Cognitive Neuroscience and Learning & IDG/McGovern Institute for Brain Research, Beijing Normal University, Beijing 100875, China
b
CAS Key Laboratory of Behavioral Science, Institute of Psychology, Chinese Academy of Sciences, Beijing 100101, China
c
School of Psychology, South China Normal University, Guangzhou 510631, China
d
Department of Neurosurgery, Baylor College of Medicine, Houston 77030, USA
e
Shenzhen University General Hospital, Department of Rehabilitation Medicine, Shenzhen 518055, China

ARTICLE INFO ABSTRACT

Keywords: Orthographic processing is a critical stage in visual word recognition. However, the white-matter pathways that
Orthographic processing support this processing are unclear, as prior findings might have been confounded by impure behavioral mea-
Connectome sures, potential structural reorganization of the brain, and limited sample sizes. To address this issue, we in-
Lesion-behavior mapping vestigated the correlations between the integrity of 20 major tracts in the whole brain and the pure orthographic
White-matter tract
index across 67 patients with short-term brain damage. The integrity of the tracts was measured by the lesion
volume percentage and the mean fractional anisotropy value. The orthographic index was calculated as the
residual of the orthographic tasks after regressing out corresponding nonorthographic tasks and the orthographic
factor from the principal component analysis (PCA) on the basis of four orthographic tasks. We found significant
correlations associated with the left inferior longitudinal fasciculus (ILF), even after controlling for the influence
of potential confounding variables. These observations strengthen evidence for the vital role of the left ILF in
orthographic processing.

1. Introduction processing, remains elusive. Relevant findings are mainly from in-
vestigations on word reading. The white-matter tracts connecting to the
Reading text is an important skill for literate people. One of the VWFA in the ventral pathway were postulated to be associated with
critical processes in reading is orthographic processing, i.e., accessing orthographic processing (Yeatman, Rauschecker, & Wandell, 2013).
the written form of a word in the orthographic input lexicon (Grainger The inferior longitudinal fasciculus (ILF) connects the occipital and
& Jacobs, 1996). This type of processing is supported not only by cer- temporal regions and passes within the VWFA (Catani, Jones, Donato, &
tain gray-matter cerebral regions but also by the white-matter tracts Ffytche, 2003; Yeatman et al., 2013). Damage to the left ILF led to
connecting those regions (Feldman, Lee, Yeatman, & Yeom, 2012; Lebel reading disorders in a dyslexic patient (Epelbaum et al., 2008) and in
et al., 2013; Saur et al., 2008). Rich evidence has shown that there is an patients with glioma (Sarubbo et al., 2015; Zemmoura, Herbet, Moritz-
orthographic processing-related area in the left ventral occipito- Gasser, & Duffau, 2015). The degree of the integrity of the left tract was
temporal cortex (vOT), named the visual word form area (VWFA), correlated with reading skills in children with typical and poor reading
which has a more strongly selective response to written letter strings abilities as well as abilities in adults (Horowitz-Kraus, Wang, Plante, &
and words than to other categories, such as line drawings and objects. Holland, 2014; Nikki Arrington, Kulesz, Juranek, Cirino, & Fletcher,
Damage to this region has a causal link to pure alexia (Binder, Medler, 2017; Yeatman, Dougherty, Ben-Shachar, & Wandell, 2012). The left
Westbury, Liebenthal, & Buchanan, 2006; Cohen et al., 2000; Dehaene inferior frontal-occipital fasciculus (IFOF) has been identified as an-
& Cohen, 2011; Grainger, Dufau, Montant, Ziegler, & Fagot, 2012; other tract that connects to the VWFA from the occipital lobes to the
Hirshorn et al., 2016; Pflugshaupt et al., 2009; Purcell et al., 2017). inferior frontal gyrus (Catani, Howard, Pajevic, & Jones, 2002; Forkel
However, although numerous researches have been investigated it, et al., 2014; Martino, Brogna, Robles, Vergani, & Duffau, 2010). The
the specific anatomical connectivity contributing to orthographic white matter integrity of the left IFOF has been reported to be related to

⁎
Corresponding authors.
E-mail addresses: zzhhan@bnu.edu.cn (Z. Han), songluping882002@aliyun.com (L. Song).
1
K. Wang and X. Li are co-first authors.

https://doi.org/10.1016/j.bandl.2019.104721
Received 29 August 2018; Received in revised form 5 October 2019; Accepted 10 November 2019
Available online 20 December 2019
0093-934X/ © 2019 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/BY-NC-ND/4.0/).
K. Wang, et al. Brain and Language 201 (2020) 104721

the orthographic component in adults and children (Vanderauwera had suffered from a stroke, whereas the remaining subjects (n = 10)
et al., 2018; Vandermosten et al., 2012). Meanwhile, other tracts as- had suffered a traumatic brain injury. The mean years of age and formal
sociated with word reading, such as the arcuate fasciculus (Gullick & education were 46 (SD = 13; range = 19 to 70 years) and 13 (SD = 3;
Booth, 2015; Horowitz-Kraus et al., 2014; Yeatman et al., 2012, 2011), range: 4 to 22 years), respectively. The mean score on the Mini-Mental
right ILF (Horowitz-Kraus et al., 2014), right uncinated fasciculus (UF) State Examination (MMSE, Folstein, Folstein, & Mchugh, 1975) was 22
(Nikki Arrington et al., 2017), and corpus callosum (Andrews et al., (SD = 8; range = 0 to 30). Neuropsychological tests of Chinese aphasia
2010), were also identified. (Gao et al., 1993) revealed that 8 patients had no aphasia, and the
Although the abovementioned studies provide important insights remaining patients suffered from motor (11%), sensory (13%), anomia
into the neuroanatomical connectivity of orthographic processing, the (17%), conduction (7%), global/mixed (26%), or subcortical (7%)
findings should be interpreted with caution for the following reasons. aphasia or dysgraphia (1%).
(1) Reading involves (but is not limited to) primary visual form per-
ception, semantic processing, and phonological encoding, in addition to 2.1.2. Healthy participants
orthographic processing (Coltheart, Rastle, Perry, Langdon, & Ziegler, Fifty-one healthy subjects (27 males) who had no neurological or
2001; Plaut, Seidenberg, & Mcclelland, 1996; Schlaggar & McCandliss, psychiatric illness were selected. All were right-handed. The mean age
2007). Therefore, the tracts that were observed to be involved in word and years of formal education of the subjects were 50 years (SD = 11;
reading might engage in other reading processes rather than ortho- range: 26–72 years) and 13 years (SD = 4; range: 6–22 years), respec-
graphic processing. Hence, pure orthographic-processing measures tively. The mean MMSE score was 28 (SD = 1; range: 24–30).
need to be adopted to identify orthographic fiber pathways. (2) Find- The difference in age (t(116) = 1.93, p > 0.05) and education level
ings from studies with small sample sizes (Epelbaum et al., 2008) need (t(116) = −0.27, p > 0.79) between patients and healthy subjects was
to be confirmed by studies with larger subject samples. (3) The cerebral insignificant, but the differences in the gender ratio (χ2 = 7.97, d f = 1,
connectivity in those with long-lasting tumors or developmental dys- p < 0.005) and MMSE scores (t(116) = 5.80, p < 10−7) were sig-
lexia might undergo structural/functional reorganization (Kinno, Ohta, nificant.
Muragaki, Maruyama, & Sakai, 2014; Lazar, Alexander, Thottakara, Most participants in this study (51 patients, all healthy controls)
Badie, & Field, 2006; MacSweeney, Brammer, Waters, & Goswami, were derived from our previous subject cohort (Bi et al., 2015; Han
2009; Park & Friston, 2013). Some studies found that the white matter et al., 2013). The differences in the subjects among these studies were
integrity near brain tumors was disrupted (Yen et al., 2009), and the due to differences in their available behavioral or imaging data.
white matter network began to recover when the tumors were resected
(Lazar et al., 2006). Moreover, abnormal functional connectivity in the 2.2. Behavioral data collection
language network was also observed in patients with glioma (Briganti
et al., 2012; Kinno et al., 2014). In addition, the individuals with de- No task only measures orthographic processing without the in-
velopmental dyslexia showed structural and functional alterations in volvement of other processes. In practice, a pure orthographic index is
the occipitotemporal area, parietotemporal area, inferior frontal gyri, obtained by at least two common methods on the basis of multiple
and fusiform gyri (MacSweeney et al., 2009; Raschle, Chang, & Gaab, tasks: multivariable regression analysis and principal component ana-
2011; Raschle, Zuk, & Gaab, 2012; Richlan, Kronbichler, & Wimmer, lysis (PCA). In the former method, the orthographic index is defined as
2009). Thus, studies of patients with relatively short-term brain injury the residual of the performance on orthographic tasks regressing out
are in demand. performance on nonorthographic tasks. In the latter method, the index
The purpose of the present study was to determine which white- is defined as the common element (i.e., orthographic component) from
matter fiber tracts in the entire brain are responsible for orthographic distinct orthographic tasks. These two methods were both adopted in
processing in visual word recognition. We investigated the correlational the present study. To obtain reliable orthographic indices, we designed
relationship between the integrity of major tracts in the whole brain two sets of tasks for the former method and two additional orthographic
and the pure orthographic-processing index in a group of 67 adults with tasks for the latter method (see Table 1).
acquired short-term brain injury. The effects of each observed ortho-
graphic-relevant tract were further validated by statistically factoring 2.2.1. First set of tasks used in the regression analysis method
out the influence of some potential confounding variables (e.g., type of The visual lexical decision task (n = 10) was adopted as an ortho-
brain injury). The flowchart of this study is shown in Fig. 1. graphic task. The stimuli consisted of ten Chinese text samples (5
words, 5 nonwords). Each word or nonword was visually presented on a
2. Methods screen, and participants were required to decide whether it was a real
word. This task involves (but is not limited to) primary visual form
2.1. Participants perception, orthographic processing, judgment/decision making, and
response execution. Two nonorthographic control tasks were used. The
Patients with brain damage and healthy subjects participated in the first task was an auditory lexical decision task (n = 10), which was
present study. Behavioral and neuroimaging data were collected for identical to the visual lexical decision task except that the ten written
each subject using an identical procedure. All subjects were native words were replaced with ten other auditory words presented via
Chinese speakers and provided written informed consent. This study headphones. This task could be used to rule out the nonorthographic
was approved by the Institutional Review Board of the State Key components except for primary visual form perception processing.
Laboratory of Cognitive Neuroscience and Learning, Beijing Normal Therefore, the second control task was designated to exclude this re-
University. dundant component: we adopted the visual form perception test (test 3,
n = 30) of the Birmingham Object Recognition Battery (Riddoch &
2.1.1. Patients Humphreys, 1993). Subjects were required to decide whether the two
Sixty-seven patients (52 males) were recruited from the China circles on the screen were of the same size.
Rehabilitation Research Centre according to the following inclusion
criteria. The patients had suffered from their first brain injury with 2.2.2. Second set of tasks used in the regression analysis method
more than a 1-month onset (mean = 3.28 months; SD = 2.44; range: The next orthographic task was word associative matching (n = 70).
1–10 months), had no other neurological or psychiatric diseases, were Each item showed three object words positioned within a triangle.
able to follow task instructions and were right-handed (Edinburgh Participants were instructed to judge which of the two bottom words
Handedness Inventory; Oldfield, 1971). Most of the subjects (n = 57) (e.g., chopstick, ruler) was semantically closer to the top word (e.g.,

2
K. Wang, et al. Brain and Language 201 (2020) 104721

Fig. 1. Flowchart for the analyses of this study. (1–2) Computing the percentage of voxels with the lesion or the mean FA value of each tract for each patient by
inputting the tract mask (c, JHU white-matter tractography atlas, Hua et al., 2008) into the lesion map (a) or the FA map (b) of each patient, respectively. (3–4)
Calculating the orthographic indices by extracting the orthographic factor from the principal component analysis (PCA) on the basis of orthographic tasks or
regressing out the accuracy scores of nonorthographic control tasks from the accuracy scores of orthographic tasks. (5–8) Obtaining orthography-relevant tracts (d, e,
f, g) in separate analyses by correlating the PCA scores or the composite scores with the lesion percentages or the mean FA values of each tract. (9) Extracting the
common tracts of the lesion analysis and the mean FA analysis with orthographic composite scores and PCA scores (h). (10) Validating the observed orthographic-
related tract by additionally controlling for potential confounding variables (i). (11) Extracting the white-matter fiber map connecting with the word form area
(VWFA, Bolger et al., 2005) (l) by performing DTI deterministic tractography with the VWFA as the seed (j) on the neuroimaging data (k) for healthy participants.
(12) Masking the common voxels between the VWFA-connected map and the orthography-relevant tract (m). (13) Splitting the mask into anterior (n) and posterior
(o) branches along the VWFA and separately evaluating their roles in orthographic processing. Step 13 was repeated except that the split mask was replaced with that
of the whole tract instead of that for part of the tract (i.e., the VWFA-connected fibers).

spoon). This task involves (but is not limited to) primary form per- two separate blocks, once with the target object word (e.g., tiger) and
ception, orthographic processing, semantic comprehension, judgment/ once with a semantically related object word from the same category
decision making, and response execution. A control task, picture asso- (e.g., leopard). The task was to decide whether the object and the word
ciative matching (n = 70), was adopted to exclude the nonorthographic were identical (see Han et al., 2016 for details).
processes. This task was identical to the word task except that the All tasks were run with the DMDX program (Forster & Forster,
written words were replaced with corresponding object pictures (see 2003). The participants were tested individually in a sound-attenuated
Han et al., 2013 for details). room. Each task was administered in a separate session. Each session
lasted less than two hours. The presentation order of the items in the
task was pseudorandom and was identical across subjects. Participant
2.2.3. Additional orthographic tasks used in the PCA method responses except for those in the oral word reading task were made by
The first orthographic task used in PCA was oral word reading pressing the “YES”/“NO” button or the corresponding stimuli on a
(n = 140). Participants were instructed to read visually presented touch screen. A pause was allowed, if required.
words aloud. The second task was word-picture verification (n = 70). For
each item, a photographed object (e.g., tiger) was visually presented in

Table 1
Behavioral performance scores of participants.
Healthy subjects (n = 51) Patients (n = 67)

Mean raw accuracy % (SD) Mean raw accuracy % (SD) Mean standardized t score (SD)

Orthographic tasks
Visual lexical decision (n = 10) 96.54 (7.27) 91.94 (13.05)* −0.62 (1.73)
Word associative matching (n = 70) 96.44 (2.70) 89.02 (10.07)*** −2.71 (3.76)
Oral word reading (n = 140) 99.14 (1.06) 86.39 (23.74)*** −11.91 (22.07)
Word picture verification (n = 70) 96.70 (2.62) 88.44 (12.10)*** −3.45 (4.79)

Non-orthographic control tasks

Auditory lexical decision (n = 10) 93.68 (9.18) 91.34 (9.68) −0.33 (0.99)
Visual form perception (n = 25) 91.76 (5.98) 83.28 (10.89)*** −1.27 (1.81)
Picture associative matching (n = 70) 94.45 (4.02) 86.16 (10.59)*** −1.97 (2.59)

Group difference of raw accuracy: *P < 0.05, ***P < 0.001.

3
K. Wang, et al. Brain and Language 201 (2020) 104721

2.3. Behavioral data preprocessing 2.4. Imaging data collection and preprocessing

2.3.1. Correcting the raw scores 2.4.1. Data collection

The first complete response or button-press response for each item All patients and healthy controls were scanned using a 1.5T GE Signa
was scored for the oral word reading task or other tasks, respectively. EXCITE system at the China Rehabilitation Research Centre. We acquired
Moreover, an item was scored as incorrect if no response was made three types of images: high-resolution 3D T1-weighted images (3D images)
within 60 s after the stimulus was presented. Thus, each subject had a (using two scanning sequences), T2-weighted fluid attenuated inversion
raw accuracy score for each task. Given that the patient group showed recovery (FLAIR) images (1 sequence), and diffusion-weighted images
considerable variation across demographic properties (age, gender, (DWIs) (two sequences with 15 directions and two sequences with 17
education), their raw accuracy scores might not meaningfully reflect directions). The 3D images were T1-weighted three-dimensional magne-
the degree of deficits. To eliminate the influence of these demographic tization-prepared rapid gradient echo (MPRAGE) images on the sagittal
variables, the raw accuracy score of each patient on each task was plane with the following parameters: matrix size = 512 × 512 mm2, voxel
corrected to a standardized t-score by comparing their performance to size = 0.49 × 0.49 × 0.70 mm3, repetition time (TR) = 12.26 s, echo
the actual performance distribution of the 51 healthy subjects. time (TE) = 4.2 ms, inversion time (TI) = 400 ms, field of view
Specifically, we first established a regression model on the basis of the (FOV) = 250 × 250 mm2, flip angle = 15°, slice number = 248 slices,
demographic properties of the healthy controls, in which the accuracy slice thickness = 1.4 mm, and slice spacing = 0.70 mm. The FLAIR T2
score was treated as the dependent variable and the demographic images were fluid-attenuated inversion recovery T2-weighted images on
variables (age, gender and education level) were treated as predictors. the axial plane with the following parameters: matrix
Then a predicted accuracy score for each patient was acquired by in- size = 512 × 512 mm2, voxel size = 0.49 × 0.49 × 5 mm3,
troducing his or her demographic information into the regression TR = 8002 ms, TE = 127.57 ms, TI = 2 s, FOV = 250 × 250 mm2, flip
model, and a discrepancy value was generated (Discrepancypatient) (i.e., angle = 90°, slice number = 28 slices, slice thickness = 5 mm, and slice
observed accuracy - predicted accuracy). A corrected standard error of spacing = 0 mm. DWIs had two separate sequences with different diffu-
estimate for each patient (SEpatient) was obtained using the following sion weighting direction sets so that 32 directions were covered in total.
formula: The first acquisition had the following parameters: 15 diffusion weighting
directions, matrix size = 128 × 128 mm2, voxel size = 1.95 × 1.95
1 1 2 × 2.6 mm3, TR = 13000 ms, TE = 69.3 ms, TI = 0 s,
SEpatient = SY . X 1 + + r iiZio2 + r ijZio Zjo
N N 1 N 1 FOV = 250 × 250 mm2, flip angle = 90°, slice number = 53 slices, slice
where SY.X and N are the standard error and sample size, respectively, thickness = 2.6 mm, and slice spacing = 2.6 mm. The other acquisition
for the control group; rii and rij are the main diagonal and off-diagonal had the same parameters except that it included 17 different directions.
elements of the inverted correlation matrix, respectively, for the k The first two volumes were b0 volumes, and the b-value of other volumes
predictor variables (k = 3, i.e., age, education, gender); and z0 (z10, … , was 1000 s/mm2 in each sequence. All the sequences except for FLAIR T2
zk0) are the z-scores of the patient’s accuracies on the predictor vari- images were scanned twice to improve the quality of images.
ables. The patient’s t-score was then computed: t-
scorepatient = Discrepancypatient/SEpatient (Crawford & Garthwaite, 2006; 2.4.2. Data preprocessing
Han et al., 2013). In our dataset, each voxel in the image of each patient’s brain ob-
tained a lesion value (categorical variable) from the lesion map and a
fractional anisotropy (FA) value (continuous variable) from the nor-
2.3.2. Computing orthographic composite scores using the regression malized FA map. For the lesion map of a patient, the two sequences of
analysis method the 3D imaging data in the same native space were first coregistered
For each of the two sets of tasks used in the regression analysis using the tri-linear interpolation method applied in SPM5 (http://www.
method (see above), we calculated the orthographic scores of each fil.ion.ucl.ac.uk/spm/software/spm5); then, the FLAIR T2 images were
patient: the residual of the corrected t scores of the orthographic task coregistered and resliced in the native space of the averaged 3D images
after regressing out the corrected t scores of its corresponding control using tri-linear interpolation method in SPM5. Each patient’s lesion
tasks. Thus, a patient could obtain two residuals corresponding to two contour was drawn slice by slice on 3D images by two trained persons,
orthographic tasks (i.e., visual lexical decision and word associative visually referring to FLAIR T2 images. Patients’ structural images were
matching). Finally, the mean value of the two residuals of each patient registered into Talairach space via the ‘3D Volume Tools’ in
was treated as his or her orthographic composite score. BrainVoyager QX v2.0, in which we manually marked the anterior to
posterior commissure plane and the borders of the cerebrum. The affine
2.3.3. Extracting orthographic PCA scores using the PCA method transformation matrix between native and Talairach spaces was ex-
Considering the visual lexical decision, one task used to calculate tracted with the ANT software package. The lesion images were
the orthographic composite score, had a very small number of items transformed into Talairach space using this matrix with the
(n = 10), which might lead to the composite score not stably reflecting ‘WarpImageMultiTransform’ program. The lesion image was finally
the orthographic processing ability of the subjects. Another ortho- transformed into the MNI space using the affine transformation matrix
graphic index was measured through PCA analysis. Two additional between MNI and Talairach spaces using a similar method. Given that
orthographic tasks that contained more items (oral word reading: the registration procedure was based on anatomical landmarks without
n = 140; visual word-picture verification: n = 70) were introduced, evaluating detailed local information in the brain, the processes was not
and PCA was conducted on four orthographic tasks (visual lexical de- affected by the lesions. For the FA map of a patient, we first merged
cision, word associative matching, oral word reading, and word-picture each of the 15-direction and 17-direction paired DWI sequences into
verification). The Kaiser-Meyer-Olkin measure of sampling adequacy one single 4D nifti-1 format file and merged diffusion-weighted gra-
(0.77) and Bartlett’s test of sphericity (χ2 = 148, d f = 6, p < 10−29) dient tables of the two sequences. We then executed the following steps
showed that the behavioral dataset was suitable for PCA. One compo- using a pipeline tool, PANDA (Cui, Zhong, Xu, Gong, & He, 2013): BET:
nent with eigenvalues > 1 was extracted. The component accounted skull removal; Eddycorrect: correction of eddy current distortion;
for 72% of the variance of the model, with the four tasks having high DTIFIT: build diffusion tensor model. The FA maps of each patient in
loading values (0.74–0.92). We considered this component to be the native space were further registered to the FMRIB FA template in MNI
orthographic processing component and derived the orthographic score space using ANTs (version 1.9). The normalization included two steps:
for each patient on the basis of this component. linear affine and nonlinear transform registration. In linear affine

4
K. Wang, et al. Brain and Language 201 (2020) 104721

transform, we used “ANTs” program to obtain one affine transform.txt 2.5.2. Fractional anisotropy-behavior correlation
file for each participant and then executed the WarpImageMultiTrans- For each of the 20 tracts, the mean FA value (averaging the FA
form program to obtain the FA map in MNI space. In non-linear trans- values of all the voxels in the tract) was correlated with the ortho-
form, a shell script ‘buildtemplate’ was used to obtain a more fine- graphic composite scores and PCA scores across patients (FDR corrected
grained normalized FA map of each patient in MNI space. q < 0.05).
Tracts showing significant correlations in both the lesion and FA
analyses with both the orthographic composite scores and PCA scores
2.5. Identifying orthographic tracts were considered to be orthographically relevant.

To determine the major white-matter tracts contributing to ortho-

graphic processing, we separately correlated the degree of the integrity 2.6. Further confirming the observed effects of orthographic tracts
of the major tracts (measured by lesion volume percentage and FA
values) with the degree of impairments in orthographic processing Although the above analyses considered the influence of some po-
(measured by orthographic composite scores and PCA scores), factoring tentially confounding factors (i.e., age, gender, educational level, total
out the overall lesion volume (i.e., the number of voxels with lesion in lesion volume and the duration of brain damage), to further examine
the whole brain) and the duration of brain damage (i.e., months after whether the observed orthographic-related white-matter tract results
the onset of the disease). Note that all the following tract integrity- were driven by other nuisance variables, we carried out multiple re-
behavior correlation analyses treated total lesion volume and the post- gression analysis using the “Enter” regression method. In the regression
onset time of the disease as covariates. For the sake of brevity, these model, the dependent variable was the orthographic composite index or
control variables are not mentioned in the following sections unless the the PCA index, and the predictors included the integrity of the left ILF
analyses included additional control variables. (the mean FA values or lesion percentages of the left ILF), as well as the
To extract tract integrity values, we created a mask of each tract. five potential confounding variables: (1) total lesion volume, (2)
The tracts were extracted from the subtemplate with 25% probability duration of brain damage, (3) effects of orthography-related gray
levels according to the JHU white-matter tractography atlas (http://fsl. matter (lesion percentages in the orthography-related gray-matter re-
fmrib.ox.ac.uk/fsl/fslwiki/Atlases/; Hua et al., 2008). The subset con- gion), (4) type of brain injury (a dichotomous variable: stroke or
tained 20 main white-matter tracts in the whole brain, and each tract trauma) and (5) type of aphasia (a categorical variable: 1–8, corre-
was masked (see Table 2). sponding to the eight types of aphasia). For the third control variable,
the orthographic region was derived from the widely accepted visual
word form processing area (i.e., VWFA; Dehaene, Le Clec’H, Poline, Le
2.5.1. Lesion-behavior correlation Bihan, & Cohen, 2002). We first created a sphere with a 12-mm radius
Three tracts (the left cingulum gyrus, the left cingulum hippo- centered in the peak point of the VWFA (Talairach coordinates: −45,
campus, and the right cingulum hippocampus) had lesions in fewer −57, −12; from a meta-analysis in Bolger, Perfetti, & Schneider,
than five patients (Table 2) and were excluded from our lesion analysis. 2005). Then, the sphere was overlaid on a gray-matter mask (SPM5
For each of the remaining 17 tracts, the lesion percentage (number of template, probability greater than 0.40). The common voxels consisted
voxels with lesion divided by the total number of voxels in the tract) composed the orthographic-relevant gray matter region.
was separately correlated with the orthographic composite scores and
PCA scores across patients. False discovery rate (FDR) correction
(q < 0.05) was implemented to adjust for multiple comparisons.

Table 2
Correlation coefficients between white matter tract integrity (lesion percentages and mean fractional anisotropy values) and orthographic index (composite scores
and PCA scores) across 67 patients, with total lesion volume and duration of brain damage as covariates.
Tract Total Volume (mm3) Number of patients with Composite scores PCA scores
lesion
Lesion analysis FA analysis Lesion analysis FA analysis

1. L anterior thalamic radiation (ATR) 8128 36 −0.25 0.31 −0.33* 0.54*

2. R anterior thalamic radiation (ATR) 7576 31 0.20 −0.23 0.32* −0.19
3. L corticospinal tract (CST) 5464 35 −0.09 0.11 −0.05 0.26
4. R corticospinal tract (CST) 4760 28 0.29 −0.33 0.40* −0.32*
5. L cingulum gyrus (CCG) 1552 3 – 0.11 – 0.24
6. R cingulum gyrus (CCG) 608 6 0.11 0.09 0.13 0.23
7. L cingulum hippocampus (CH) 248 0 – 0.30 – 0.34*
8. R cingulum hippocampus (CH) 544 1 – −0.07 – −0.06
9. Forceps major (F_MA) 5744 19 −0.13 −0.01 −0.09 0.07
10.Forceps minor (F_MI) 19,712 37 −0.25 0.001 −0.14 0.14
11. L inferior fronto-occipital fasciculus (IFOF) 5048 36 −0.31 0.22 −0.32* 0.42*
12. R inferior fronto-occipital fasciculus (IFOF) 6304 30 0.26 −0.22 0.37* −0.19
13. L inferior longitudinal fasciculus (ILF) 5400 23 −0.50* 0.35* −0.37* 0.43*
14. R inferior longitudinal fasciculus (ILF) 3152 9 0.12 −0.21 0.20 −0.10
15. L superior longitudinal fasciculus (SLF) 9472 36 −0.19 0.37* −0.24 0.52*
16. R superior longitudinal fasciculus (SLF) 7456 21 0.41* −0.20 0.36* −0.22
17. L uncinate fasciculus (UF) 744 20 −0.30 0.14 −0.35* 0.36*
18. R uncinate fasciculus (UF) 448 15 0.25 −0.20 0.28* −0.18
19. L superior longitudinal fasciculus (temporal part) 96 5 −0.20 0.30 −0.05 0.38*
(SLFT)
20. R Superior longitudinal fasciculus (temporal part) 72 7 0.24 −0.005 0.21 −0.1
(SLFT)

L = left, R = right; * false discovery rate (FDR) corrected q < 0.05; – could not carry out this analysis due to few patients; Lesion analysis = lesion percentage-
behavior analysis; FA analysis = fractional anisotropy-behavior analysis.

5
K. Wang, et al. Brain and Language 201 (2020) 104721

Fig. 2. Raw imaging schematic for participants. The value of each voxel is the number of patients showing a lesion in that voxel (A), the mean FA value (B, C), or the
t-value of the independent sample t-test (D). FA = fractional anisotropy.

3. Results the patients had lesions in five other tracts (right ATR, right CST, right
IFOF, left ILF, and left UF).
3.1. Behavioral performance of the participants
3.2.1. Lesion-behavior correlation analysis
The behavioral performance of the subjects is reported in Table 1. After controlling for the total lesion volume and the onset of brain
The raw accuracy scores of the patients were significantly lower than damage, the orthographic composite scores of the patients were sig-
those of the healthy controls in both the orthographic and nonortho- nificantly negatively correlated with the lesion percentages of only one
graphic tasks (t(116) = 2.26–5.31, p-values < 0.03), but not in the au- left-hemispheric tract: the left ILF (partial r(63) = −0.50, FDR corrected
ditory lexical decision task (t(116) = 1.33, p = 0.74). This pattern was q < 0.05). However, PCA scores showed a significant negative corre-
also reflected in the standardized t scores of patients, in which the most lation with the lesion percentages in four left-hemispheric tracts: left
severe impairments occurred in oral word reading (standardized t ATR, IFOF, ILF and UF (partial r(63) = −0.37 to −0.32, FDR corrected
score = −11.91), followed by word associative matching and word- q < 0.05).
picture verification (standardized t score = −3.45 to −0.62). Minimal
deficits were observed in auditory lexical decision (standardized t 3.2.2. Fractional anisotropy-behavior correlation analysis
score = −0.33). The orthographic composite scores of the patients were significantly
positively correlated with the mean FA values of two left-lateralized
3.2. Orthographic-relevant tracts tracts: the left ILF and SLF (partial r(63) = 0.35–0.37, FDR corrected
q < 0.05). However, PCA scores were significantly positively corre-
The lesion and mean FA maps of the subjects are illustrated in Fig. 2. lated with the mean FA values of seven left-hemispheric tracts: left ATR,
The lesions of patients were widely distributed, covering most white- IFOF, ILF, SLF, UF, cingulum hippocampus and temporal part of SLF
matter and gray-matter areas, with most patients having lesions in the (partial r(63) = 0.34–0.54, FDR corrected q < 0.05).
insula and its surrounding white-matter tissues (Fig. 2A). Both patients The two abovementioned analyses using the two orthographic in-
and healthy controls showed the basic white-matter connectivity ske- dices consistently revealed that only the pathology of the left ILF caused
leton (Fig. 2B–C), and patients had lower FA values than healthy con- impairments in orthographic processing (Fig. 3). Therefore, this tract
trols (Fig. 2D). was considered to be an orthographically-relevant fiber bundle and was
Table 2 displays the features of the white-matter tracts and the further examined in the following analyses.
correlations between their integrity and orthographic processing per-
formances in the patients. Lesions were found in all white-matter tracts 3.3. Further confirmation of the observed orthographically-relevant effects
except for three (left cingulum gyrus, left cingulum hippocampus, and in the left ILF
right cingulum hippocampus) in at least five patients for. Five tracts
[left anterior thalamic radiation (ATR), left corticospinal tract (CST), To verify whether the orthographic effect of the left ILF could be
the left IFOF, left superior longitudinal fasciculus (SLF), and the forceps explained by other potential confounding factors, we applied regression
minor] were damaged in more than 50% of the patients, and 30–50% of analysis to each orthographic measure. We simultaneously used the

6
K. Wang, et al. Brain and Language 201 (2020) 104721

Fig. 3. Correlation between the integrity values of the left inferior longitudinal fasciculus and orthographic indices in 67 patients. *P < 0.05, ** P < 0.01,
***P < 0.001. The left ILF was mapped onto the cortical surface using BrainNet Viewer (Xia, Wang, & He, 2013).

Table 3
Results of the regression analyses (simultaneous entering method) for orthographic index (composite scores and PCA scores).
Composite scores PCA scores

Predictors Beta t p VIF Beta t p VIF

Model 1
Lesion percentage of left ILF −0.48 −4.17 0.000*** 1.18 −0.34 −2.90 0.005** 1.18
Total lesion volume 0.07 0.56 0.58 1.19 −0.13 −1.10 0.28 1.19
Brain damage duration −0.004 −0.04 0.97 1.07 −0.14 −1.26 0.21 1.07
Orthographic-relevant gray matter −0.21 −1.84 0.07 1.18 −0.14 −1.23 0.22 1.18
Types of brain injury 0.12 1.06 0.30 1.09 0.22 1.94 0.06 1.09
Types of aphasia −0.05 −0.39 0.70 1.19 0.03 0.27 0.79 1.19

Model 2
Mean FA of left ILF 0.34 2.37 0.02* 1.54 0.39 2.89 0.005** 1.54
Total lesion volume 0.09 0.67 0.50 1.43 −0.04 −0.32 0.75 1.43
Brain damage duration −0.02 −0.18 0.86 1.07 −0.15 −1.29 0.20 1.07
Orthographic-relevant gray matter −0.28 −2.24 0.03* 1.15 −0.17 −1.48 0.14 1.15
Types of brain injury −0.02 −0.13 0.90 1.20 0.08 0.71 0.48 1.20
Types of aphasia −0.07 −0.57 0.57 1.22 −0.005 −0.04 0.97 1.22

*P < 0.05; ** P < 0.01; *** P < 0.001 (two-tailed test); VIF = variance inflation factor.

7
K. Wang, et al. Brain and Language 201 (2020) 104721

“Enter” method to introduce the integrity of the left ILF (the mean FA random patients for both composite residuals (actual mean
values or lesion percentages of the left ILF) and five potential con- score = −1.39, p = 0.001) and PCA residuals (actual mean
founding variables as predictors and the orthographic composite index score = −2.06, p = 0.0002). These four patients with lesions in the
or the PCA index as the dependent variable in the regression model. The posterior segment had a very low possibility of having normal ortho-
results presented in Table 3 highlighted that the left ILF was the only graphic processing capabilities.
significant predictor of the dependent variable across the three re- To rule out the possibility that the observed effects in the two seg-
gression models (composite scores-lesion percentage: β = −0.48, ments were due to the collinearity of the lesions between these seg-
p < 10−4; PCA scores-lesion percentage: β = −0.34, p < 0.006; PCA ments, we carried out two further analyses. For the first analysis of the
scores-mean FA: β = 0.39, p < 0.006). The integrity of the left ILF as anterior segment, when we computed the correlations with this seg-
measured by either the mean FA values (β = 0.34, p < 0.03) or or- ment, the integrity values of the posterior segment, the total lesion
thography-related gray matter (β = −0.28, p < 0.03) both had sig- volume and the duration of brain damage were included as covariates.
nificant explanatory power for the dependent variable when the com- Such analyses revealed that the integrity of the anterior segment re-
posite index was used. These results demonstrated that the integrity of mained significantly correlated with orthographic indices (composite
the left ILF made the largest contribution to the regression equations, scores-lesion percentage: partial r(62) = −0.56, p < 10−6; composite
when holding all other predictor variables constant. In addition, the scores-mean FA: partial r(62) = 0.37, p < 0.003; PCA scores-lesion
values of the variance inflation factor (VIF), which was the most percentage: partial r(62) = −0.42, p < 0.0006; PCA scores-mean FA:
commonly used regression diagnostic for multicollinearity were much partial r(62) = 0.43, p < 0.0005). For the second analysis of the ante-
lower than 10 (range: 1.07–1.54), indicating that there was almost no rior segment, the correlations were only calculated in the 19 patients
collinearity, between the independent variables. In other words, the who had lesions in the anterior segment and no lesions in the posterior
analyses revealed that the left ILF plays a role in visual orthographic segment. This analysis revealed that the correlations were still sig-
processing. nificant for the composite scores (lesion percentage: partial
r(15) = −0.57, p < 0.02; mean FA: partial r(15) = 0.70, p < 0.002)
3.4. Roles of the anterior and posterior segments of the left ILF in and PCA scores in FA analysis (lesion percentage: partial r(15) = −0.20,
orthographic processing p = 0.44; mean FA: partial r(15) = 0.60, p < 0.02). For the first ana-
lysis of the posterior segment, the same permutation method as that
As the left ILF was found to be an orthographic tract, a relevant described above was applied, except that the residual scores of each
question is whether its anterior and posterior segments are both in- patient were calculated by regressing out the lesion volume of the
volved in orthographic processing. Zemmoura et al. (2015) observed anterior segment, the total lesion volume and the duration of brain
that the left ILF could be divided into two distinct segments along the damage from the composite scores and PCA scores. This analysis re-
VWFA, and the disruption of the posterior (but not the anterior) seg- vealed that the actual average scores of the four patients with lesions in
ment caused reading disorders in two of 7 patients with low-grade the posterior segment were still significantly smaller than the shuffled
glioma. To confirm this finding, we performed the following two ana- average scores of the composite residuals (actual mean score = −1.38,
lyses. The results are shown in Figs. 4 and 5. p = 0.001) and PCA residuals (actual mean score = −2.06,
p = 0.0002). For the second analysis of the posterior segment, because
3.4.1. Analysis of the entire left ILF only one patient had a lesion only in the posterior segment, we per-
This analysis aimed to determine whether the anterior segment formed the permutation test by randomly extracting one residual value
and/or posterior segment of the entire left ILF (i.e., the tract mask that from the 67 patients 10,000 times and generating a null distribution of
we used in the above analyses) plays a critical role in orthographic the residuals instead of mean scores. The true residual values of com-
processing. We first divided the left ILF into anterior and posterior posite scores (−3.17, p = 0.0001) and PCA scores (−3.00, p = 0.0003)
segments along the y-axis of the VWFA (MNI coordinates: −54, −57, were both significantly smaller than those from the shuffled distribu-
−12; Bolger et al., 2005). The two segments contained 593 voxels and tion.
82 voxels, respectively. The number of patients with lesions in only the
anterior branch, only in the posterior branch and in both branches were 3.4.2. Analysis of the connection of the left ILF to the VWFA
19, 1 and 3, respectively. Since only 4 patients had lesions in the pos- Not all the fibers in the left ILF tract connect to the VWFA. To
terior segment, a behavior-lesion correlation analysis was not appro- further examine whether the fibers that connect to the VWFA play a
priate for this branch. Therefore, a permutation test analysis was used crucial role in orthographic processing, we obtained a mask of the fibers
for this segment, while behavior-lesion correlation analysis was used connecting to the VWFA from the left ILF based on the 51 healthy
for the anterior segment. For the anterior segment, behavior-lesion subjects via deterministic fiber tracking. The mask was obtained by the
correlations were significant after partialling out the total lesion vo- following steps (see similar methods in Bi et al., 2015; Fang et al.,
lume and the duration of brain damage (composite scores-lesion per- 2015). (1) We created a 15-mm-radius sphere whose central point was
centage: partial r(63) = −0.47, p < 10−4; composite scores-mean FA: localized in the VWFA seed (MNI coordinates: −45, −57, −12; Bolger
partial r(63) = 0.36, p < 0.004; PCA scores-lesion percentage: partial et al., 2005). (2) Fiber tracking was performed starting with each voxel
r(63) = −0.34, p < 0.007; PCA scores-mean FA: partial r(63) = 0.44, in the sphere in the native space of each subject and terminating when
p < 0.0003). For the posterior segment, we first computed the re- the angle between two consecutive orientations was larger than 45 or
siduals of orthographic composite scores and PCA scores after regres- when the FA value was smaller than 0.20. To avoid bias from the initial
sing out the total lesion volume and the duration of brain damage for seed position in a given voxel, 100 seed points were randomly selected
each of the 67 patients. Then, we calculated the actual average scores of within each voxel. (3) The voxels of the fibers of each subject were
the residuals for the 4 patients with lesions in the posterior segment. masked, resulting in a binary map. This binary map was further
Finally, a permutation test was performed to examine the possible transformed into the MNI space. (4) The binary maps for all subjects
distribution of the scores of the 4 patients within all 67 patients. Spe- were overlaid to generate a count map. A group-level threshold was set
cifically, we randomly extracted the residual values of these four pa- at a voxel value seen in > 25% of the subjects (i.e., 12 subjects). Thus,
tients from those of the 67 patients, and averaged the four values. Thus, we obtained a whole-brain tracking map from the VWFA. (5) The
we obtained a mean value. This extraction process repeated 10,000 tracking map was overlaid on the left ILF, and common voxels were
times and generated a distribution of 10,000 random mean scores. The masked. This mask corresponded to the white matter in the left ILF,
actual average scores of the four patients with lesions in the posterior which was connected to the VWFA. In this new mask, we carried out an
segment were significantly smaller than the shuffled average scores of 4 analysis identical to that for the entire left ILF, except that the entire left

8
K. Wang, et al. Brain and Language 201 (2020) 104721

Fig. 4. Correlation scatter diagrams between the orthographic indices and the integrity values of the anterior branches of the left ILF. LV = lesion volume,
FA = fractional anisotropy; VWFA = visual word form area; *P < 0.05, ** P < 0.01, ***P < 0.001.

ILF mask was replaced with this newly obtained mask. The result pat- segment: 297 voxels; posterior segment: 78 voxels). The numbers of
tern of this analysis was highly similar to that of the analysis of the patients with lesions only in the anterior branch, only in the posterior
entire ILF tract. This new mask contained 375 voxels in total (anterior branch and in both branches were 12, 2 and 2, respectively. The

9
K. Wang, et al. Brain and Language 201 (2020) 104721

Fig. 5. Distribution histograms of the mean orthographic scores from one sample (the patients who had lesions in the left ILF) and 10,000 samples (each included the
ranking + 1
patients which were randomly from all the 67 patients). p = 10000 + 1 .

orthographic indices were significantly correlated with the integrity p = 0.002; PCA scores = −2.07, p = 0.0002). Even for the two patients
values of the anterior segment, even after controlling for the total lesion with a lesion only in the posterior segment, the actual average scores
volume and the duration of brain damage (composite scores-lesion were still significantly smaller than the random average scores of the
percentage: partial r(63) = −0.29, p < 0.03; composite scores-mean two patients after excluding the influence of the total lesion volume and
FA: r(63) = 0.33, p < 0.007; PCA scores-lesion percentage: partial the duration of brain damage (composite scores = −2.37, p = 0.0008;
r(63) = −0.08, p = 0.52; PCA scores-mean FA: r(63) = 0.44, PCA scores = −2.70, p = 0.0008).
p < 0.0003) and the influence of the other branch (composite scores- Generally, we partly replicated the findings of Zemmoura et al.
lesion percentage: partial r(62) = −0.39, p < 0.002; composite scores- (2015), showing that the posterior segment of the left ILF contributed to
mean FA: r(62) = 0.35, p < 0.006; PCA scores-lesion percentage: par- orthographic recognition. Unexpectedly, we observed that the anterior
tial r(62) = −0.15, p = 0.23; PCA scores-mean FA: r(62) = 0.44, segment of this tract was also involved in orthographic processing.
p < 0.0004). Even for the 12 patients who had a lesion only in the
anterior branch, the FA analysis results were still significant after ex-
4. Discussion
cluding the influence of the total lesion volume and the duration of
brain damage (composite scores-lesion percentage: partial
In 67 patients with brain damage, we observed that the integrity of
r(8) = −0.43, p = 0.21; composite scores-mean FA: partial r(8) = 0.79,
the left ILF measured by the percentage of damaged voxels and the
p < 0.007; PCA scores-lesion percentage: partial r(8) = −0.05,
mean FA values could successfully predict orthographic processing
p = 0.89; PCA scores-mean FA: partial r(8) = 0.67, p < 0.04). The
ability as measured by orthographic component scores and PCA scores.
permutation test, which followed the same procedure as the above-
This relationship persisted even when we ruled out a wide range of
mentioned statistical analyses was adopted to explore the effects of the
potentially confounding variables (e.g., influence of orthographic-re-
posterior segment. The actual average scores of orthographic indices,
levant gray-matter cortex, type of brain damage and aphasia type). We
regressing out total lesion volume, the duration of the brain damage
further found that both the anterior and posterior parts of the left ILF
and the lesion volume of anterior segment, were still significantly
contribute to orthographic processing. Note that anatomical con-
smaller than the shuffled average scores (composite scores = −1.42,
nectivity related to orthographic processing has been investigated in

10
K. Wang, et al. Brain and Language 201 (2020) 104721

individuals with long-term tumors and developmental dyslexia, and achieved via a network containing orthographic, phonological, and
studies in stroke patients who have a relatively short-term onset dura- semantic information layers that communicate with each other bidir-
tion are still necessary. Previous clinical observations showed that the ectionally. The higher order linguistic representations (e.g., semantic,
removal of some eloquent areas in patients with low-grade gliomas did phonological information) modulate early orthographic processing via
not cause the patients to lose corresponding functions (Duffau et al., top-down control. Successful word form recognition in the VWFA is a
2003). The functional recovery of patients with low-grade gliomas was consequence of the interaction between bottom-up sensory inputs and
better than that of stroke patients (Duffau et al., 2003; Varona, top-down linguistic information. Specifically, the posterior branch of
Bermejo, Guerra, & Molina, 2004). Structurally, the white matter tracts the left ILF transfers primary visual lower level information regarding
underwent a series of alterations near the tumors (Lazar et al., 2006; the morphological characteristics of a word to the VWFA, which can
Yen et al., 2009). Functionally, syntax-related and attentional networks lead to an ambiguous representation of that word in the VWFA. This
have been reported to be changed in patients with tumors (Charras ambiguous activation is disambiguated by top-down predictions re-
et al., 2015; Kinno et al., 2014). Moreover, Duffau and his colleagues presenting associations between visual stimuli and meanings. In other
summarized four main neural reorganizational patterns in tumor pa- words, bottom-up and top-down information for a visual word inter-
tients, including the redistribution of eloquent areas near the tumor and actively and jointly contribute to word recognition in the VWFA.
a distributed network of functional areas in the hemisphere containing Therefore, damage to either the anterior or posterior branch of the left
the tumor (Desmurget, Bonnetblanc, & Duffau, 2007). With regard to ILF would lead to deficits in orthographic processing. Consistent with
developmental dyslexia, some evidence has shown that structural and our findings, previous studies have shown that orthographic recogni-
functional reorganization occur even before literacy acquisition tion is modulated by prefrontal and sensorimotor regions in an inter-
(Raschle et al., 2011, 2012). In summary, compared with individuals active manner (Whaley, Kadipasaoglu, Cox, & Tandon, 2016). In ad-
suffering from brain tumors or developmental dyslexia, stroke patients dition, the activation of the VWFA is modulated by the top-down
had more similar brain structures and functions to those of normal in- processing of symbols (Reinke, Fernandes, Schwindt, O’Craven, &
dividuals. Grady, 2008; Song, Tian, & Liu, 2012; Starrfelt & Gerlach, 2007). Re-
cently, Hirshorn et al. (2016) revealed that the VWFA contributes to
4.1. The role of the left ILF in orthographic processing multiple stages of orthographic representations: an early stage for gist-
level visual representations that are sensitive to orthographic statistics
The left ILF connects the occipital lobe with the anterior part of the and a later stage for more precise representations that are sufficient for
temporal lobe (ATL), running laterally and inferiorly to the lateral wall the individuation of orthographic word forms. The early stage might
of the temporal horn (Catani & Thiebaut de Schotten, 2008; Catani correspond to the bottom-up processing of orthographic recognition,
et al., 2002, 2003; Latini, 2015). Our results showing a vital role of left and the later stage might correspond to the interaction between bottom-
ILF in orthographic processing are consistent with previous studies up and top-down processing.
(Epelbaum et al., 2008; Gullick & Booth, 2015; Hayashi, Okita,
Kinoshita, Miyashita, & Nakada, 2014; Sarubbo et al., 2015; Yeatman 4.2. Other orthographic tracts reported in the literature
et al., 2012) and add important evidence that clarifies the interpreta-
tion of those studies. Notably, Zemmoura et al. (2015) showed that In addition to the left ILF, we also found that the left ATR, IFOF, UF,
some fibers in the left ILF are connected to the VWFA and that damage SLF and temporal part of SLF showed FA or lesion volumes that were
to posterior (but not anterior) fibers caused dyslexia. Our study partly significantly correlated with the orthographic composite scores or PCA
confirms this finding of Zemmoura et al. (2015). Indeed, the specific scores. We briefly discuss their potential role in processing orthographic
role of the posterior part of this tract may be inferred from the cortical information.
areas to which it connects. Several studies have revealed that the oc- The IFOF connects the frontal lobe with the occipital lobe, the su-
cipital cortices are involved in primary form perception across subject perior parietal lobe and the posterior temporo-basal area (Martino
samples (Heeley & Buchanan-Smith, 1996; Hegdé & Van Essen, 2000; et al., 2010; Sarubbo et al., 2015), and it has been reported to be an
Ito & Komatsu, 2004; Kapadia, Westheimer, & Gilbert, 2000) and ex- important subcortical component of the face processing network and
perimental techniques (Hammond & Andrews, 1978; Ito & Komatsu, the ventral pathway of reading as well as the semantic system
2004). The VWFA has been identified as a key region for visual word (Almairac, Herbet, Moritz-Gasser, de Champfleur, & Duffau, 2015;
form processing (Baeck, Kravitz, Baker, & Op de Beeck, 2015; Fischer- Duffau et al., 2005; Han et al., 2013; Harvey & Schnur, 2015; Motomura
Baum, Bruggemann, Gallego, Li, & Tamez, 2017; Glezer, Jiang, & et al., 2014; Thomas et al., 2009). In addition, the left IFOF has been
Riesenhuber, 2009; Zhao et al., 2016). Studies have demonstrated the found to participate in orthographic processing in adults and children,
preference for letters and words in this region over other types of visual as the FA values of this tract were significantly correlated with per-
objects (Baker et al., 2007; Dehaene & Cohen, 2011; Szwed et al., formance in a visual flashed-word identification task and the ortho-
2011), and this preference remained invariant to changes in visual graphic component extracted from a series of phoneme detection tasks
scripts, fonts, and location in the visual field (Bolger et al., 2005; Cohen and reading tasks (Vanderauwera et al., 2018; Vandermosten et al.,
et al., 2000; Dehaene et al., 2001). Therefore, we speculate that the 2012). Evidence for the role of this tract in orthographic processing was
posterior portion of the left ILF supports bottom-up processing in which partly in line with our finding that the integrity of the left IFOF was
primary visual stimuli (e.g., angle, shape) are transferred and combined significantly correlated with the PCA scores, but not the composite
with higher level orthographic symbols. scores, of four orthographic tasks (visual lexical decision, word asso-
An unexpected finding was the involvement of the anterior segment ciative matching, oral word reading, and word-picture verification). We
of the left ILF in orthographic processing. This pathway connects the suspected that the PCA component extracted from orthographic related
VWFA to the anterior temporal lobe (ATL). The ATL has been found to tasks and the visual flashed-word identification task may include a
be a critical area for semantic processing (Damasio, Tranel, Grabowski, lexical/semantic component in addition to an orthographic component.
Adolphs, & Damasio, 2004; Lambon Ralph, 2014; Patterson, Nestor, & Moreover, it is now well known that the left IFOF is crucial to lexical-
Rogers, 2007). An interesting question is why dysfunction in the con- semantic processing, with serious evidence in intraoperative electrical
nection between semantic and orthographic processes leads to disorders stimulation works, as well as in brain-behavior mapping studies (Duffau
in orthographic processing. Theories related to reading might provide et al., 2005; Harvey & Schnur, 2015). Similar to the left IFOF, the in-
some insights into this issue. The interactive theory (Carreiras, tegrity of the left ATR connecting the thalamus with the inferior frontal
Armstrong, Perea, & Frost, 2014; Price & Devlin, 2011) based on con- cortex and the left UF connecting the anterior temporal lobe with the
nectionist models (Coltheart, 2005), argues that word reading is orbitofrontal cortex were more sensitive to the orthographic PCA scores

11
K. Wang, et al. Brain and Language 201 (2020) 104721

than composite scores. These tracts have also been proven to be sub- Acknowledgements
cortical components of the semantic system (De Zubicaray, Rose, &
McMahon, 2011; Duffau, Gatignol, Moritz-Gasser, & Mandonnet, 2009; We would like to thank BNU-CNLab members for data collection
Han et al., 2013). For instance, Han et al. (2013) identified three major and imaging preprocessing. We are also grateful to all research parti-
tracts underlying the semantic system: the left IFOF, ATR, and UF. cipants. This work was supported by the National Key R&D Program of
Furthermore, the left UF has been reported to be involved in semantic China (2018YFC1315200; 2017YFF0207400), the National Natural
control in word comprehension and contributes to name exception Science Foundation of China (31872785; 81171019; 81972144;
words (Cummine et al., 2015; Harvey, Wei, Ellmore, Hamilton, & 81871338) and the Beijing Natural Science Foundation (7182088).
Schnur, 2013). In summary, we speculate that the significant correla-
tions between the PCA scores extracted from orthographic tasks and the References
integrity of these tracts mentioned above may indicate, to some extent,
their role in lexical-semantic processing. Almairac, F., Herbet, G., Moritz-Gasser, S., de Champfleur, N. M., & Duffau, H. (2015).
The AF/SLF connects Wernicke’s area in the posterior superior The left inferior fronto-occipital fasciculus subserves language semantics: A multi-
level lesion study. Brain Structure and Function, 220(4), 1983–1995.
temporal gyrus to Broca’s area in the inferior frontal regions (see review Andrews, J. S., Ben-Shachar, M., Yeatman, J. D., Flom, L. L., Luna, B., & Feldman, H. M.
in Dick & Tremblay, 2012). This fiber bundle has also been reported to (2010). Reading performance correlates with white-matter properties in preterm and
play a role in word reading (Bakhtiari, Boliek, & Cummine, 2014; term children. Developmental Medicine & Child Neurology, 52(6), e94–e100.
Baeck, A., Kravitz, D., Baker, C., & Op de Beeck, H. P. (2015). Influence of lexical status
Gullick & Booth, 2015). In the theoretical framework of a dual-route and orthographic similarity on the multi-voxel response of the visual word form area.
model of word reading (Coltheart et al., 2001; Price, 2012), this tract NeuroImage, 111, 321–328. https://doi.org/10.1016/j.neuroimage.2015.01.060.
would present the dorsal pathway, which is responsible for phonolo- Baker, C. I., Liu, J., Wald, L. L., Kwong, K. K., Benner, T., & Kanwisher, N. (2007). Visual
word processing and experiential origins of functional selectivity in human extra-
gical (Han et al., 2016; Mandonnet, Nouet, Gatignol, Capelle, & Duffau,
striate cortex. Proceedings of the National Academy of Sciences of the United States of
2007) and syntactic processes (Friederici, 2009; Wilson et al., 2011). America, 104(21), 9087–9092.
Hence, a defect in this pathway would not cause the deterioration of Bakhtiari, R., Boliek, C., & Cummine, J. (2014). Investigating the contribution of ventral-
lexical and dorsal-sublexical pathways during reading in bilinguals. Frontiers in
orthographic processing in reading. In our results, although the FA
Human Neuroscience, 8(July), 507. https://doi.org/10.3389/fnhum.2014.00507.
values of the SLF were significantly correlated with the behavioral in- Bi, Y., Han, Z., Zhong, S., Ma, Y., Gong, G., Huang, R., ... Caramazza, A. (2015). The white
dices, the lesion volume of the tract failed to reach a significant level. matter structural network underlying human tool use and tool understanding. Journal
of Neuroscience, 35(17), 6822–6835.
Binder, J. R., Medler, D. A., Westbury, C. F., Liebenthal, E., & Buchanan, L. (2006).
4.3. Limitations of the present study Tuning of the human left fusiform gyrus to sublexical orthographic structure.
NeuroImage, 33(2), 739–748. https://doi.org/10.1016/j.neuroimage.2006.06.053.
Bolger, D. J., Perfetti, C. A., & Schneider, W. (2005). Cross-cultural effect on the brain
The present study had at least the following caveats. (1) Our lexical revisited: Universal structures plus writing system variation. Human Brain Mapping,
decision tasks included only a few items, which might result in low 25(1), 92–104. https://doi.org/10.1002/hbm.20124.
statistical power. (2) The distribution of the number of patients with Briganti, C., Sestieri, C., Mattei, P. A., Esposito, R., Galzio, R. J., Tartaro, A., ... Caulo, M.
(2012). Reorganization of functional connectivity of the language network in patients
lesions was not equivalent across tracts, which did not give an oppor- with brain gliomas. American Journal of Neuroradiology, 33(10), 1983–1990.
tunity to reveal the effects of tracts that are lesioned in only a few Carreiras, M., Armstrong, B. C., Perea, M., & Frost, R. (2014). The what, when, where, and
patients. (3) The present study only investigated orthographic proces- how of visual word recognition. Trends in Cognitive Sciences, 18(2), 90–98. https://
doi.org/10.1016/j.tics.2013.11.005.
sing at the whole-word level in the Chinese language via one input Catani, M., Howard, R. J., Pajevic, S., & Jones, D. K. (2002). Virtual in vivo interactive
modality, and the orthographic information was retrieved from long- dissection of white matter fasciculi in the human brain. NeuroImage, 17(1), 77–94.
term memory. Previous studies have revealed dissociations of ortho- https://doi.org/10.1006/nimg.2002.1136.
Catani, M., Jones, D. K., Donato, R., & Ffytche, D. H. (2003). Occipito-temporal con-
graphic processing between the whole-word level and other levels, the nections in the human brain. Brain, 126(9), 2093–2107. https://doi.org/10.1093/
reading input modality and the spelling output modality, Chinese and brain/awg203.
other languages, and long-term memory and working memory (Rapp, Catani, M., & Thiebaut de Schotten, M. (2008). A diffusion tensor imaging tractography
atlas for virtual in vivo dissections. Cortex, 44(8), 1105–1132. https://doi.org/10.
Purcell, Hillis, Capasso, & Miceli, 2015). Thus, it is unknown whether 1016/j.cortex.2008.05.004.
our findings could be generalized to other aspects of orthographic Charras, P., Herbet, G., Deverdun, J., De Champfleur, N. M., Duffau, H., Bartolomeo, P., &
processing. (4) The definition of the VWFA in the current study was Bonnetblanc, F. (2015). Functional reorganization of the attentional networks in low-
grade glioma patients: A longitudinal study. Cortex, 63, 27–41.
based on a meta-analysis of the VWFA (Bolger et al., 2005), the peak of
Cohen, L., Dehaene, S., Naccache, L., Lehéricy, S., Dehaene-Lambertz, G., Hénaff, M. A., &
which show variability across individuals (Glezer & Riesenhuber, Michel, F. (2000). The visual word form area: Spatial and temporal characterization
2013). Finally, (5) our DWI sequences had no isovoxels or gaps between of an initial stage of reading in normal subjects and posterior split-brain patients.
slices, which was not optimal for tractography. Brain, 123, 291–307. https://doi.org/10.1093/brain/123.2.291.
Coltheart, M. (2005). Modeling reading: The dual-route approach. The science of reading:
A handbook (pp. 6–23).
Coltheart, M., Rastle, K., Perry, C., Langdon, R., & Ziegler, J. (2001). DRC: A dual route
4.4. Conclusion
cascaded model of visual word recognition and reading aloud. Psychological Review,
108(1), 204.
By mapping measures of pure orthographic processing deficits with Crawford, J. R., & Garthwaite, P. H. (2006). Comparing patients’ predicted test scores
the integrity of major white-matter tracts in the entire brains of 67 from a regression equation with their obtained scores: A significance test and point
estimate of abnormality with accompanying confidence limits. Neuropsychology,
patients with short-term brain damage, we revealed that the left ILF is a 20(3), 259.
critical tract for orthographic processing. The posterior segment of this Cui, Z., Zhong, S., Xu, P., Gong, G., & He, Y. (2013). PANDA: A pipeline toolbox for
tract is responsible for the feedforward sensory input of visual words, analyzing brain diffusion images. Frontiers in Human Neuroscience, 7, 42.
Cummine, J., Dai, W., Borowsky, R., Gould, L., Rollans, C., & Boliek, C. (2015).
and the anterior segment is responsible for delivering feedback re- Investigating the ventral-lexical, dorsal-sublexical model of basic reading processes
garding associative predictions between visual word stimuli and using diffusion tensor imaging. Brain Structure and Function, 220(1), 445–455.
meanings. The interaction of the information from these two segments Damasio, H., Tranel, D., Grabowski, T., Adolphs, R., & Damasio, A. (2004). Neural sys-
tems behind word and concept retrieval. Cognition, 92(1–2), 179–229.
leads to the successful orthographic recognition of visual words. These De Zubicaray, G. I., Rose, S. E., & McMahon, K. L. (2011). The structure and connectivity
results reconstructed anatomical substrates of orthographic recognition of semantic memory in the healthy older adult brain. NeuroImage, 54(2), 1488–1494.
and underscored the causal role of the left ILF in this processing. https://doi.org/10.1016/j.neuroimage.2010.08.058.
Dehaene, S., Le Clec’H, G., Poline, J. B., Le Bihan, D., & Cohen, L. (2002). The visual word
form area: A prelexical representation of visual words in the fusiform gyrus.
Declaration of Competing Interest NeuroReport, 13(3), 321–325. https://doi.org/10.1097/00001756-200203040-
00015.
Dehaene, S., & Cohen, L. (2011). The unique role of the visual word form area in reading.
The authors declare no competing financial interests.

12
K. Wang, et al. Brain and Language 201 (2020) 104721

Trends in Cognitive Sciences, 15(6), 254–262. https://doi.org/10.1016/j.tics.2011.04. pnas.1604126113.

003. Horowitz-Kraus, T., Wang, Y., Plante, E., & Holland, S. K. (2014). Involvement of the right
Dehaene, S., Naccache, L., Cohen, L., Bihan, D. L., Mangin, J. F., Poline, J. B., & Rivière, hemisphere in reading comprehension: A DTI study. Brain Research, 1582, 34–44.
D. (2001). Cerebral mechanisms of word masking and unconscious repetition https://doi.org/10.1016/j.brainres.2014.05.034.
priming. Nature Neuroscience, 4(7), 752–758. https://doi.org/10.1038/89551. Hua, K., Zhang, J., Wakana, S., Jiang, H., Li, X., Reich, D. S., ... Mori, S. (2008). Tract
Desmurget, M., Bonnetblanc, F., & Duffau, H. (2007). Contrasting acute and slow-growing probability maps in stereotaxic spaces: Analyses of white matter anatomy and tract-
lesions: A new door to brain plasticity. Brain, 130(4), 898–914. specific quantification. NeuroImage, 39(1), 336–347. https://doi.org/10.1016/j.
Dick, A. S., & Tremblay, P. (2012). Beyond the arcuate fasciculus: Consensus and con- neuroimage.2007.07.053.
troversy in the connectional anatomy of language. Brain, 135(12), 3529–3550. Ito, M., & Komatsu, H. (2004). Representation of angles embedded within contour stimuli
https://doi.org/10.1093/brain/aws222. in area V2 of macaque monkeys. Journal of Neuroscience, 24(13), 3313–3324.
Duffau, H., Capelle, L., Denvil, D., Sichez, N., Gatignol, P., Taillandier, L., ... Muller, J.-C. Kapadia, M. K., Westheimer, G., & Gilbert, C. D. (2000). Spatial distribution of contextual
(2003). Usefulness of intraoperative electrical subcortical mapping during surgery for interactions in primary visual cortex and in visual perception. Journal of
low-grade gliomas located within eloquent brain regions: Functional results in a Neurophysiology, 84(4), 2048–2062.
consecutive series of 103 patients. Journal of Neurosurgery, 98(4), 764–778. Kinno, R., Ohta, S., Muragaki, Y., Maruyama, T., & Sakai, K. L. (2014). Differential re-
Duffau, H., Gatignol, P., Mandonnet, E., Peruzzi, P., Tzourio-Mazoyer, N., & Capelle, L. organization of three syntax-related networks induced by a left frontal glioma. Brain,
(2005). New insights into the anatomo-functional connectivity of the semantic 137(4), 1193–1212. https://doi.org/10.1093/brain/awu013.
system: A study using cortico-subcortical electrostimulations. Brain, 128(4), Lambon Ralph, M. A. (2014). Neurocognitive insights on conceptual knowledge and its
797–810. https://doi.org/10.1093/brain/awh423. breakdown. Philosophical Transactions of the Royal Society of London. Series B,
Duffau, H., Gatignol, P., Moritz-Gasser, S., & Mandonnet, E. (2009). Is the left uncinate Biological Sciences, 369(December 2013), 20120392. https://doi.org/10.1098/rstb.
fasciculus essential for language? Journal of Neurology, 256(3), 382. 2012.0392.
Epelbaum, S., Pinel, P., Gaillard, R., Delmaire, C., Perrin, M., Dupont, S., ... Cohen, L. Latini, F. (2015). New insights in the limbic modulation of visual inputs: The role of the
(2008). Pure alexia as a disconnection syndrome: New diffusion imaging evidence for inferior longitudinal fasciculus and the Li-Am bundle. Neurosurgical Review, 38(1),
an old concept. Cortex, 44(8), 962–974. https://doi.org/10.1016/j.cortex.2008.05. 179–190. https://doi.org/10.1007/s10143-014-0583-1.
003. Lazar, M., Alexander, A. L., Thottakara, P. J., Badie, B., & Field, A. S. (2006). White matte
Fang, Y., Han, Z., Zhong, S., Gong, G., Song, L., Liu, F., ... Bi, Y. (2015). The semantic reorganization after surgical resection of brain tumors and vascular malformations.
anatomical network: Evidence from healthy and brain-damaged patient populations. American Journal of Neuroradiology, 27(1), 1258–1271 https://doi.org/27/6/1258
Human Brain Mapping, 36(9), 3499–3515. https://doi.org/10.1002/hbm.22858. [PII].
Feldman, H. M., Lee, E. S., Yeatman, J. D., & Yeom, K. W. (2012). Language and reading Lebel, C., Shaywitz, B., Holahan, J., Shaywitz, S., Marchione, K., & Beaulieu, C. (2013).
skills in school-aged children and adolescents born preterm are associated with white Diffusion tensor imaging correlates of reading ability in dysfluent and non-impaired
matter properties on diffusion tensor imaging. Neuropsychologia, 50(14), 3348–3362. readers. Brain & Language, 125(2), 215–222.
Fischer-Baum, S., Bruggemann, D., Gallego, I. F., Li, D. S., & Tamez, E. R. (2017). MacSweeney, M., Brammer, M. J., Waters, D., & Goswami, U. (2009). Enhanced activa-
Decoding levels of representation in reading: A representational similarity approach. tion of the left inferior frontal gyrus in deaf and dyslexic adults during rhyming.
Cortex, 90, 88. Brain, 132(7), 1928–1940.
Folstein, M. F., Folstein, S. E., & Mchugh, P. R. (1975). “Mini-mental state”. A practical Mandonnet, E., Nouet, A., Gatignol, P., Capelle, L., & Duffau, H. (2007). Does the left
method for grading the cognitive state of patients for the clinician. Journal of inferior longitudinal fasciculus play a role in language? A brain stimulation study.
Psychiatric Research, 12(3), 189–198. Brain, 130(3), 623–629. https://doi.org/10.1093/brain/awl361.
Forkel, S. J., de Schotten, M. T., Kawadler, J. M., Dell’Acqua, F., Danek, A., & Catani, M. Martino, J., Brogna, C., Robles, S. G., Vergani, F., & Duffau, H. (2010). Anatomic dis-
(2014). The anatomy of fronto-occipital connections from early blunt dissections to section of the inferior fronto-occipital fasciculus revisited in the lights of brain sti-
contemporary tractography. Cortex, 56, 73–84. mulation data. Cortex, 46(5), 691–699.
Forster, K. I., & Forster, J. C. (2003). DMDX: A Windows display program with milli- Motomura, K., Fujii, M., Maesawa, S., Kuramitsu, S., Natsume, A., & Wakabayashi, T.
second accuracy. Behavior Research Methods, Instruments, & Computers, 35(1), (2014). Association of dorsal inferior frontooccipital fasciculus fibers in the deep
116–124. parietal lobe with both reading and writing processes: A brain mapping study: Case
Friederici, A. D. (2009). Pathways to language: Fiber tracts in the human brain. Trends in report. Journal of Neurosurgery, 121(1), 142–148.
Cognitive Sciences, 13(4), 175–181. https://doi.org/10.1016/j.tics.2009.01.001. Nikki Arrington, C., Kulesz, P. A., Juranek, J., Cirino, P. T., & Fletcher, J. M. (2017).
Gao, S. R., Wang, Y., Shi, S., Liu, J., Lin, G., & Rao, B. (1993). Aphasia. Beijing: Beijing White matter microstructure integrity in relation to reading proficiency. Brain and
Medicine University and China Xiehe Medicine University Joint Press. Language, 174(August), 103–111. https://doi.org/10.1016/j.bandl.2017.08.002.
Glezer, L. S., Jiang, X., & Riesenhuber, M. (2009). Evidence for highly selective neuronal Oldfield, R. C. (1971). The assessment and analysis of handedness: The Edinburgh in-
tuning to whole words in the “Visual Word Form Area”. Neuron, 62(2), 199. ventory. Neuropsychologia, 9(1), 97–113.
Glezer, L. S., & Riesenhuber, M. (2013). Individual variability in location impacts or- Park, H.-J., & Friston, K. (2013). Structural and functional brain networks: From con-
thographic selectivity in the “visual word form area”. Journal of Neuroscience, 33(27), nections to cognition. Science, 342(6158), 1238411.
11221–11226. Patterson, K., Nestor, P. J., & Rogers, T. T. (2007). Where do you know what you know?
Grainger, J., Dufau, S., Montant, M., Ziegler, J. C., & Fagot, J. (2012). Orthographic The representation of semantic knowledge in the human brain. Nature Reviews
processing in baboons (Papio papio). Science, 336(6078), 245–248. Neuroscience, 8(12), 976.
Grainger, J., & Jacobs, A. M. (1996). Orthographic processing in visual word recognition: Pflugshaupt, T., Gutbrod, K., Wurtz, P., Von Wartburg, R., Nyffeler, T., De Haan, B., ...
A multiple read-out model. Psychological Review, 103(3), 518–565. Mueri, R. M. (2009). About the role of visual field defects in pure alexia. Brain,
Gullick, M. M., & Booth, J. R. (2015). The direct segment of the arcuate fasciculus is 132(7), 1907–1917.
predictive of longitudinal reading change. Developmental Cognitive Neuroscience, 13, Plaut, D. C., Seidenberg, M. S., & Mcclelland, J. L. (1996). Understanding normal and
68–74. https://doi.org/10.1016/j.dcn.2015.05.002. impaired word reading: Computational principles in quasi-regular domains.
Hammond, P., & Andrews, D. P. (1978). Collinearity tolerance of cells in areas 17 and 18 Psychological Review, 103(1), 56–115.
of the cat’s visual cortex: Relative sensitivity to straight lines and chevrons. Price, C. J. (2012). A review and synthesis of the first 20years of PET and fMRI studies of
Experimental Brain Research, 31(3), 329–339. heard speech, spoken language and reading. NeuroImage, 62(2), 816–847. https://
Han, Z., Ma, Y., Gong, G., He, Y., Caramazza, A., & Bi, Y. (2013). White matter structural doi.org/10.1016/j.neuroimage.2012.04.062.
connectivity underlying semantic processing: Evidence from brain damaged patients. Price, C. J., & Devlin, J. T. (2011). The Interactive Account of ventral occipitotemporal
Brain, 136(10), 2952–2965. https://doi.org/10.1093/brain/awt205. contributions to reading. Trends in Cognitive Sciences, 15(6), 246–253. https://doi.
Han, Z., Ma, Y., Gong, G., Huang, R., Song, L., & Bi, Y. (2016). White matter pathway org/10.1016/j.tics.2011.04.001.
supporting phonological encoding in speech production: A multi-modal imaging Purcell, J., Sebastian, R., Leigh, R., Jarso, S., Davis, C., Posner, J., ... Hillis, A. E. (2017).
study of brain damage patients. Brain Structure and Function, 221(1), 577–589. Recovery of orthographic processing after stroke: A longitudinal fMRI study. Cortex,
https://doi.org/10.1007/s00429-014-0926-2. 92, 103.
Harvey, D. Y., & Schnur, T. T. (2015). Distinct loci of lexical and semantic access deficits Rapp, B., Purcell, J., Hillis, A. E., Capasso, R., & Miceli, G. (2015). Neural bases of or-
in aphasia: Evidence from voxel-based lesion-symptom mapping and diffusion tensor thographic long-term memory and working memory in dysgraphia. Brain, 139(2),
imaging. Cortex, 67, 37–58. 588–604. https://doi.org/10.1093/brain/awv348.
Harvey, D. Y., Wei, T., Ellmore, T. M., Hamilton, A. C., & Schnur, T. T. (2013). Raschle, N. M., Chang, M., & Gaab, N. (2011). Structural brain alterations associated with
Neuropsychological evidence for the functional role of the uncinate fasciculus in dyslexia predate reading onset. Neuroimage, 57(3), 742–749.
semantic control. Neuropsychologia, 51(5), 789–801. Raschle, N. M., Zuk, J., & Gaab, N. (2012). Functional characteristics of developmental
Hayashi, Y., Okita, H., Kinoshita, M., Miyashita, K., & Nakada, M. (2014). Functional dyslexia in left-hemispheric posterior brain regions predate reading onset. Proceedings
recovery from pure dyslexia with preservation of subcortical association fiber net- of the National Academy of Sciences, 109(6), 2156–2161.
works. Interdisciplinary Neurosurgery: Advanced Techniques and Case Management, 1(3), Reinke, K., Fernandes, M., Schwindt, G., O’Craven, K., & Grady, C. L. (2008). Functional
59–62. https://doi.org/10.1016/j.inat.2014.06.004. specificity of the visual word form area: General activation for words and symbols but
Heeley, D. W., & Buchanan-Smith, H. M. (1996). Mechanisms specialized for the per- specific network activation for words. Brain and Language, 104(2), 180–189. https://
ception of image geometry. Vision Research, 36(22), 3607–3627. https://doi.org/10. doi.org/10.1016/j.bandl.2007.04.006.
1016/0042-6989(96)00077-6. Richlan, F., Kronbichler, M., & Wimmer, H. (2009). Functional abnormalities in the
Hegdé, J., & Van Essen, D. C. (2000). Selectivity for complex shapes in primate visual area dyslexic brain: A quantitative meta-analysis of neuroimaging studies. Human Brain
V2. Journal of Neuroscience, 20(5) RC61–RC61. Mapping, 30(10), 3299–3308.
Hirshorn, E. A., Li, Y., Ward, M. J., Richardson, R. M., Fiez, J. A., & Ghuman, A. S. (2016). Riddoch, M. J., & Humphreys, G. W. (1993). Birmingham object recognition battery.
Decoding and disrupting left midfusiform gyrus activity during word reading. Lawrence Erlbaum Associates.
Proceedings of the National Academy of Sciences, 201604126. https://doi.org/10.1073/ Sarubbo, S., De Benedictis, A., Merler, S., Mandonnet, E., Balbi, S., Granieri, E., & Duffau,

13
K. Wang, et al. Brain and Language 201 (2020) 104721

H. (2015). Towards a functional atlas of human white matter: Functional Atlas of Whaley, M. L., Kadipasaoglu, C. M., Cox, S. J., & Tandon, N. (2016). Modulation of or-
White Matter. Human Brain Mapping, 36(8), 3117–3136 http://doi.wiley.com/ thographic decoding by frontal cortex. Journal of Neuroscience the Official Journal of
10.1002/hbm.v36.8 http://doi.wiley.com/10.1002/hbm.22832. the Society for Neuroscience, 36(4), 1173.
Saur, D., Kreher, B. W., Schnell, S., Kümmerer, D., Kellmeyer, P., Vry, M. S., ... Abel, S. Wilson, S. M., Galantucci, S., Tartaglia, M. C., Rising, K., Patterson, D. K., Henry, M. L., ...
(2008). Ventral and dorsal pathways for language. Proceedings of the National Gorno-Tempini, M. L. (2011). Syntactic processing depends on dorsal language tracts.
Academy of Sciences of the United States of America, 105(46), 18035. Neuron, 72(2), 397–403. https://doi.org/10.1016/j.neuron.2011.09.014.
Schlaggar, B. L., & McCandliss, B. D. (2007). Development of neural systems for reading. Xia, M., Wang, J., & He, Y. (2013). BrainNet Viewer: A network visualization tool for
Annual Review of Neuroscience, 30, 475–503. https://doi.org/10.1146/annurev.neuro. human brain connectomics. PloS One, 8(7), e68910.
28.061604.135645. Yeatman, J. D., Dougherty, R. F., Ben-Shachar, M., & Wandell, B. A. (2012). Development
Song, Y., Tian, M., & Liu, J. (2012). Top-down processing of symbolic meanings mod- of white matter and reading skills. Proceedings of the National Academy of Sciences of
ulates the visual word form area. Journal of Neuroscience, 32(35), 12277–12283. the United States of America, 109(44), E3045–53. https://doi.org/10.1073/pnas.
Starrfelt, R., & Gerlach, C. (2007). The Visual What For Area: Words and pictures in the 1206792109.
left fusiform gyrus. NeuroImage, 35(1), 334–342. https://doi.org/10.1016/j. Yeatman, J. D., Dougherty, R. F., Rykhlevskaia, E., Sherbondy, A. J., Deutsch, G. K.,
neuroimage.2006.12.003. Wandell, B. A., & Ben-Shachar, M. (2011). Anatomical properties of the arcuate
Szwed, M., Dehaene, S., Kleinschmidt, A., Eger, E., Valabrègue, R., Amadon, A., & Cohen, fasciculus predict phonological and reading skills in children. Journal of Cognitive
L. (2011). Specialization for written words over objects in the visual cortex. Neuroscience, 23(11), 3304–3317. https://doi.org/10.1162/jocn_a_00061.
Neuroimage, 56(1), 330–344. Yeatman, J. D., Rauschecker, A. M., & Wandell, B. A. (2013). Anatomy of the visual word
Thomas, C., Avidan, G., Humphreys, K., Jung, K., Gao, F., & Behrmann, M. (2009). form area: Adjacent cortical circuits and long-range white matter connections. Brain
Reduced structural connectivity in ventral visual cortex in congenital prosopagnosia. and Language, 125(2), 146–155. https://doi.org/10.1016/j.bandl.2012.04.010.
Nature Neuroscience, 12(1), 29. Yen, P. S., Teo, B. T., Chiu, C. H., Chen, S. C., Chiu, T. L., & Su, C. F. (2009). White matter
Vanderauwera, J., De Vos, A., Forkel, S. J., Catani, M., Wouters, J., Vandermosten, M., & tract involvement in brain tumors: A diffusion tensor imaging analysis. Surgical
Ghesquière, P. (2018). Neural organization of ventral white matter tracts parallels the Neurology, 72(5), 464–469.
initial steps of reading development: A DTI tractography study. Brain and Language, Zemmoura, I., Herbet, G., Moritz-Gasser, S., & Duffau, H. (2015). New insights into the
183(May), 32–40. https://doi.org/10.1016/j.bandl.2018.05.007. neural network mediating reading processes provided by cortico-subcortical elec-
Vandermosten, M., Boets, B., Poelmans, H., Sunaert, S., Wouters, J., & Ghesquière, P. trical mapping. Human Brain Mapping, 36(6), 2215–2230. https://doi.org/10.1002/
(2012). A tractography study in dyslexia: Neuroanatomic correlates of orthographic, hbm.22766.
phonological and speech processing. Brain, 135(3), 935–948. https://doi.org/10. Zhao, L., Chen, C., Shao, L., Wang, Y., Xiao, X., Chen, C., ... Xue, G. (2016). Orthographic
1093/brain/awr363. and phonological representations in the fusiform cortex. Cerebral Cortex, 1–14.
Varona, J. F., Bermejo, F., Guerra, J. M., & Molina, J. A. (2004). Long-term prognosis of https://doi.org/10.1093/cercor/bhw300.
ischemic stroke in young adults. Journal of Neurology, 251(12), 1507–1514.

14