JSLHR

Research Article

Resting-State Functional Magnetic

Resonance Imaging Connectivity Between
Semantic and Phonological Regions
of Interest May Inform Language
Targets in Aphasia
Amy E. Ramage,a Semra Aytur,b and Kirrie J. Ballardc

Purpose: Brain imaging has provided puzzle pieces in Results: Network coherence was observed in healthy
the understanding of language. In neurologically healthy controls and participants with stroke. The left–right
populations, the structure of certain brain regions is premotor cortex connection was stronger in healthy
associated with particular language functions (e.g., controls, as reported by New et al. (2015) in the same
semantics, phonology). In studies on focal brain damage, data set. FC of (a) connections between temporal
certain brain regions or connections are considered regions, in the left hemisphere and bilaterally, predicted
sufficient or necessary for a given language function. lexical–semantic processing for auditory comprehension
However, few of these account for the effects of lesioned and (b) ipsilateral connections between temporal and
tissue on the “functional” dynamics of the brain for frontal regions in both hemispheres predicted access to
language processing. Here, functional connectivity (FC) semantic–phonological representations and processing
among semantic–phonological regions of interest (ROIs) for verbal production.
is assessed to fill a gap in our understanding about the Conclusions: Network connectivity of brain regions associated
neural substrates of impaired language and whether with semantic–phonological processing is predictive of
connectivity strength can predict language performance language performance in poststroke aphasia. The most
on a clinical tool in individuals with aphasia. predictive connections involved right-hemisphere ROIs—
Method: Clinical assessment of language, using the particularly those for which structural adaptions are known
Western Aphasia Battery–Revised, and resting-state to associate with recovered word retrieval performance.
functional magnetic resonance imaging data were obtained Predictions may be made, based on these findings, about
for 30 individuals with chronic aphasia secondary to left- which connections have potential as targets for neuroplastic
hemisphere stroke and 18 age-matched healthy controls. functional changes with intervention in aphasia.
FC between bilateral ROIs was contrasted by group and Supplemental Material: https://doi.org/10.23641/asha.
used to predict Western Aphasia Battery–Revised scores. 12735785

S
everal modern brain imaging methodologies have
been used to investigate aphasia, each providing its
piece in understanding language processing. The
a tradition of attaching cognitive function to particular brain
Department of Communication Sciences and Disorders, University
regions began long before imaging was available as a tool,
of New Hampshire, Durham
b
Department of Health Policy and Management, University of when Paul Broca and Carl Wernicke made their postmortem
New Hampshire, Durham discoveries. Broca provided the first postmortem evidence
c
Faculty of Medicine and Health and the Brain and Mind Centre, of language function localization, that is, that the ability to
The University of Sydney, New South Wales, Australia articulate language is in the posterior two thirds of the in-
Correspondence to Amy E. Ramage: Amy.ramage@unh.edu ferior frontal gyrus (IFG; Grodzinsky & Amunts, 2006).
Editor-in-Chief: Stephen M. Camarata About 10 years later, Wernicke found that damage to
Editor: Christos Salis the superior temporal gyrus is associated with language
Received August 5, 2019
Revision received March 16, 2020
Accepted June 1, 2020 Disclosures: The authors have declared that no competing interests existed at the time
https://doi.org/10.1044/2020_JSLHR-19-00117 of publication.

Journal of Speech, Language, and Hearing Research • Vol. 63 • 3051–3067 • September 2020 • Copyright © 2020 The Authors 3051
This work is licensed under a Creative Commons Attribution 4.0 International License.
Downloaded from: https://pubs.asha.org 157.51.53.89 on 02/25/2024, Terms of Use: https://pubs.asha.org/pubs/rights_and_permissions
comprehension. This established brain–behavior relation- symptom mapping, which associates lesions in particular
ships or functional specialization of brain regions (Gesch- brain areas (defined by voxels) with language symptoms,
wind, 1970). With the advent of neuroimaging, the and therefore, the results are relevant to lesion locations
mapping of language functions to brain regions has be- present in the participant sample and do not assess the
come more complex (Holland & Ramage, 2000), facilitat- dynamic functional brain activity.
ing a narrowing of the focus to specific language functions. As such, there remains a gap in knowledge about
Current neuroimaging data hinder translation of re- whether or how well previous findings based on structural,
search to practice given the multitude of approaches taken anatomical, or lesion-based studies are suited for select-
to identify the brain regions that are necessary or suffi- ing brain regions that may be informative in “functional”
cient for language processing (Poldrack, 2000; Price & imaging studies to understand language breakdown in indi-
Mummery, 1999; Price et al., 1999). These methods often viduals with aphasia. That is, lesion studies indicate the
yield different results regarding which regions and func- areas of the brain that are “necessary” to perform a language
tions are most important for studies of language or aphasia, process, and activation studies in neurotypical individuals
largely because the methodology or population of study indicate brain regions that are “sufficient” to perform a
differs across studies. For example, some studies identify language process, but neither indicates how those regions
regions that are necessary for processing of word meanings interact to sufficiently approximate language in aphasia.
or semantics (e.g., Halai et al., 2017; Jackson et al., 2016) It is likely that, rather than only investigating structural
or for phonology (e.g., Okada et al., 2018) in neurologically characteristics of lesions within certain brain regions, in-
healthy populations using task-based functional magnetic vestigations of structural or functional connections among
resonance imaging (fMRI). Others study individuals with regions (i.e., connectivity) may facilitate prediction of as-
focal brain damage to indicate which brain “regions” (Price, pects of language performance in individuals with aphasia
2012) or structural “connections” (Fridriksson et al., 2018; (Pustina et al., 2017).
Ivanova et al., 2016) are necessary or sufficient for a given Voxel-based and connectivity-based lesion–symptom
language function. However, few investigate the causal rela- mapping (e.g., den Ouden et al., 2019; Fridriksson et al.,
tionship between a brain lesion to an impaired function to 2018) and metrics of white matter integrity (e.g., Ivanova
indicate the “necessity” of a region for a specific function et al., 2016) have been used to correlate or predict language
(Poldrack, 2000). performance in individuals with aphasia. Language perfor-
mance in these studies has been characterized using aphasia
batteries (e.g., Western Aphasia Battery–Revised [WAB-R];
Lesion Studies Kertesz, 2006) as well as with assessments of semantic
While isolating language processes for study may associations (e.g., with the Pyramids and Palm Trees Test;
not be representative of language as the dynamic process Howard & Patterson, 1992) and other word retrieval mea-
that it is, this gold standard approach has yielded immense sures (e.g., the Philadelphia Naming Test [Roach et al.,
understanding about language and the brain. This approach 1996] that includes items that differ by word class) that
is exemplified by Price and colleagues, who have developed are used in clinical practice. These studies have improved
a cumulative model of language processes and their asso- understanding of the structural connections most likely
ciated brain regions (Price, 2012). They have identified involved in language processing and have even documented
lesions that are necessary to disrupt specific language func- changes in white matter integrity as a function of treatment
tions in persons with aphasia (PWAs) secondary to left- (McKinnon et al., 2017). They have also informed the con-
hemisphere stroke (Hope et al., 2017; Price, 2012; Seghier tribution of the right hemisphere in recovery in aphasia, with
et al., 2010) establishing specific language process–brain structural adaptation that is uniquely associated with re-
lesion correspondence (Seghier et al., 2015) and improving covery of specific language performance (Hope et al., 2017).
understanding of recovery from aphasia. However, most of However, a gap remains in understanding how lesions
the predictions have been tested as they relate to language affecting white matter integrity or connectivity change the
performance during task-based fMRI in neurologically functional dynamics of the brain for language performance
healthy adults (e.g., Hope et al., 2014; Ludersdorfer et al., in aphasia. In fact, recent predictive models of language
2019). Only a few have been tested using fMRI in individ- performance indicate that structural connectivity metrics
uals with aphasia, as described below. are highly correlated with lesion volume and thus do not
Similarly, the work of Schwartz and colleagues has add much to the functional predictions about language per-
focused on isolating components of language processing to formance (Hope et al., 2017).
associate them with lesion characteristics. This work has
also carefully characterized language using single-word
processing assessments to specify semantics (Schnur et al., Functional Connectivity Studies
2009; Schwartz et al., 2009), phonology (Schwartz et al., There have been several informative studies of the
2012; Thompson-Schill et al., 2010), and sentence compre- functional dynamics of brain function in aphasia with fMRI
hension and production (Thothathiri, Gagliardi, et al., using language tasks (e.g., Kiran & Thompson, 2019) or
2012; Thothathiri, Kimberg, et al., 2012; Thothathiri et al., by modeling the causal dynamics of task-based fMRI (e.g.,
2015, 2010). These studies have utilized voxel-based lesion Meier et al., 2016). However, as noted by Klingbeil et al.

3052 Journal of Speech, Language, and Hearing Research • Vol. 63 • 3051–3067 • September 2020

Downloaded from: https://pubs.asha.org 157.51.53.89 on 02/25/2024, Terms of Use: https://pubs.asha.org/pubs/rights_and_permissions

(2017), performing language tasks in an MRI scanner is it provides a unique opportunity to link ROIs to seman-
awkward at best, and performance under such conditions tic and phonological processing data typically collected
is dubious given the known intra- and interindividual vari- in clinical settings. Thus, we determine how well FC
ability that is characteristic of individuals with aphasia. among these ROIs can predict overall aphasia severity
There continues to be a need to identify relatively coherent (the WAB-R Aphasia Quotient), as well as subtest scores
brain states that are impervious to differing task demands for Spontaneous Speech (with subcomponents Content
and that may be investigated (a) in recovery from aphasia and Fluency), Auditory Comprehension, Repetition, and
and (b) as targets to assess neuroplastic change associated Naming/Word Finding.
with a treatment.
Resting-state fMRI (rsfMRI) provides information
about brain networks that demonstrate intrinsic coherent Method
activity. Networks demonstrating intrinsic coherent activ- Participants
ity at rest are impervious to task-related fluctuations and
are considered to be more permanent, trait-like functional An existing data set was used for the study, which
signatures (Hjelmervik et al., 2014). In fact, resting-state included 31 right-handed, chronic left-hemisphere stroke
intrinsic networks are thought to provide the scaffolding participants with aphasia (New et al., 2015). Speech-language
for activity engaged during task performance (X. J. Chai, assessments and rsfMRI connectivity analyses were con-
Berken, et al., 2016; Smith et al., 2009). For example, ducted. Fifteen of the participants had comorbid apraxia
Jackson et al. (2016) investigated a network of regions func- of speech (AOS; see New et al., 2015, for more details).
tionally connected to the ventral anterior temporal gyrus in Eighteen healthy volunteers (HC) without neurological or
rsfMRI to establish a “semantic network” based on the pre- psychiatric disorders also participated. All participants were
mise that this brain region is a critical hub of multimodal 18–75 years of age, right-handed, self-reported (by PWA
semantic conceptual representations. They then compared or caregiver) proficient English speakers prior to the stroke,
that network at rest with patterns of activation during a and had no contraindications for undergoing an MRI. Ad-
triad semantic judgment task (e.g., matching a target word ditionally, eligible participants had no history of uncorrected
—“hen”—to either a strongly associated word—“cage”—or hearing, vision, or other sensory impairments; cognitive
conceptually similar word—“robin”—relative to unrelated impairments; premorbid speech, language, or reading im-
foils), further specifying the components of the network to pairments; or substance abuse. All control participants
semantic processing. Thus, rsfMRI can index intrinsic con- scored within normal limits ( > 28) on the Mini-Mental
nectivity networks that may or may not be anatomically State Examination (Folstein et al., 1975). All procedures
connected but have relevance to the networks that are conformed to the Declaration of Helsinki (World Medical
engaged during language performance. In concert with the Association, 2013) and were approved by the Sydney Local
exquisite detail provided by studies of poststroke lesions Health District Human Research Ethics Committee, Con-
and structural connectivity (Fridriksson et al., 2018; Halai cord Repatriation General Hospital in Sydney and the
et al., 2018; Hope et al., 2017; Price, 2012; Seghier et al., University of Sydney, Australia. The examiner, a qualified
2015) and the process-based focus of task-based fMRI speech-language pathologist, walked through the partici-
studies (Hope et al., 2014), rsfMRI is a complementary pant information statement and consent form with the par-
tool providing a simple yet robust indicator of network ticipant and caregiver and asked comprehension questions
structure that may be useful in studying the relationships to ensure understanding of study details prior to obtaining
of these networks to performance. written consent. All participants provided written informed
Thus, this study utilizes rsfMRI to determine whether consent. In addition, because the assessments included ver-
a set of regions found previously to be “necessary” for bal expression, participants were excluded if they had AOS
semantic or phonological processing (Halai et al., 2017; judged as severe, as described below.
Seghier et al., 2010) or for speech motor planning (New
et al., 2015; Price, 2012) demonstrate coherent activity Speech and Language Evaluation
(a network) at rest in healthy controls (HC) and PWAs. The Because the data used here are from an existing
a priori hypothesis underlying the study is that coherent data set, we did not have the ability to select our own test
activity between regions of interest (ROIs) present in the batteries. The existing data set included stroke participants’
control group represents a resting-state intrinsic network performance on a battery of assessments used to diagnose
that serves as the scaffolding upon which language perfor- and determine presence and severity of dysarthria, apraxia,
mance depends (Fox & Raichle, 2007; Smith et al., 2009). as well as severity and type of aphasia. The WAB-R was
Group differences in functional connectivity (FC) strength administered as a measure of aphasia severity and served
among this network between PWAs and control participants as the primary measure of language in the study. While
may have relevance for language performance; therefore, the WAB-R does not isolate semantics or phonology, ac-
this study assesses the “predictive value” of FC among cess to both is required to a varying extent for all subtests.
these regions for language performance in aphasia using Additional measures were administered to rule out
a commonly used clinical tool—the WAB-R. While this apha- auditory discrimination deficits (Psycholinguistic Assess-
sia battery does not isolate language processes specifically, ments of Language Processing in Aphasia: Same–Different

Ramage et al.: Semantic–Phonological rsfMRI Connectivity in Aphasia 3053

Downloaded from: https://pubs.asha.org 157.51.53.89 on 02/25/2024, Terms of Use: https://pubs.asha.org/pubs/rights_and_permissions
Discrimination Using Word Minimal Pairs subtest; Kay Auditory Comprehension, r(30) = −.20, p = .33. No such
et al., 1992) and nonverbal cognitive deficit (Raven’s Pro- correlations existed between WAB-R scores and age, sex, or
gressive Colored Matrices; Raven, 1978). The Motor Speech months postonset of stroke (see Supplemental Material S2
Examination (Duffy, 2013) and Apraxia Battery for Adults– for all correlations). Henceforth, connectivity among the re-
Second Edition (ABA-2; Dabul, 2000) Increasing Word gions will be evaluated in the stroke patients only and con-
Length subtest characterized motor speech. For published trolled for lesion volume and AOS severity.
tests that involve perceptual judgment and scoring (e.g.,
WAB-R Fluency subtest), research assistants were trained Image Acquisition
by author K. J. B., and discrepancies in scoring were dis-
cussed and determined by consensus. Reliability for the T1-weighted structural and resting-state echo-planar
ABA-2 Increasing Word Length (interrater reliability) was imaging fMRI data were acquired on a Philips 3T TX MRI
98%. Additionally, connected speech samples were elicited scanner. Blood oxygen-level–dependent contrast (gradient-
either through the Story Retell Procedure (McNeil et al., echo echo-planar imaging pulse sequence, time to repeat =
1997) or unconstrained conversation with the examiner 2.2 s, time to echo = 30 ms, flip angle = 90°, in-plane reso-
(Ballard et al., 2016). All testing was completed by research lution = 3.1 × 3.1 mm2, 36 axial slices [3.1 mm thickness]
assistants or author K. J. B. A research assistant edited (to covering the entire brain) was utilized to acquire 216 resting-
remove silences and off-task behaviors) and compiled the state echo-planar images.
recorded tasks into a single approximately 20-min video,
with the responses to each task labeled in sequence. These Image Preprocessing
videos were uploaded to a secure server that was accessed Structural scans were normalized to standard Montreal
by the three expert judges (Ballard et al., 2016). The videos Neurological Institute space in SPM8 (http://www.fil.ion.
contained no identifying information other than the audio- ucl.ac.uk/spm) using the “unified segmentation” algorithm
visual image. These judges were blinded to any other de- (Ashburner & Friston, 2005). For the stroke participants,
mographic information, did not administer any of the testing, an extra empirically derived tissue class (“lesion”) was added
and were blinded to scores of any formal tests (e.g., WAB-R, to the segmentation priors to be represented in a tissue
ABA-2 subtests, Psycholinguistic Assessments of Lan- class separate from gray/white/cerebrospinal fluid (Seghier
guage Processing in Aphasia subtest) or neuroimaging et al., 2008). All segmentation output images were smoothed
findings at the time of making their ratings. They judged with an isotropic kernel of 8 mm at full-width at half-
participant speech samples for presence/severity of apraxia, maximum. After smoothing, the value of each voxel in
dysarthria, nonverbal oral apraxia, and phonologically the image presented the probability that the tissue belongs
based sound errors on a 7-point Likert scale (1 = normal, to a single class and not to one of the others. The lesion
7 = severe; see Ballard et al., 2016; New et al., 2015, for tissue class image for each subject was also used to deter-
more details). mine lesion volumes using the automated lesion identifi-
Participants’ demographic variables and individual cation algorithm (ALI toolbox) implemented in SPM8
testing performance are reported in Table 1. All participants (Seghier et al., 2008). This method is used in many analyses
spoke English proficiently, though there was one partici- of stroke data, with established accuracy (Crinion et al.,
pant whose first language was Hindi (DIS048) and one 2007). Though error in delineating the lesion is minimized
whose first language was Cantonese (DIS006). Neither were with this and similar procedures (Pustina et al., 2016), there
outliers for any of the language or neuroimaging mea- is potential for mischaracterization in periventricular le-
sures and therefore were included in the study. One PWA sions or when atrophy is significant. Lesion volumes (cm3)
(DIS002) was removed as he had extreme outlier values for were calculated for each participant and included in the
both WAB-R Aphasia Quotient and Auditory Comprehen- analyses described below.
sion subtests. Figure 1 demonstrates the extent and over- Echo-planar images were corrected for head move-
lap of lesions in the sample. Language performance on the ment by applying affine registration using a two-pass
WAB-R for each participant is presented in Table 2. The procedure in SPM8. Mean echo-planar images for each
PWA and HC groups did not differ for age or education subject were created and spatially normalized to the Mon-
level, but there were more females in the HC group than in treal Neurological Institute template as above and smoothed
the PWA group (see Table 2). using a 5-mm full-width at half-maximum Gaussian ker-
PWAs with larger lesion volumes performed more nel. Variance associated with physiological noise and
poorly on WAB-R Spontaneous Speech, r(30) = −.67, p < motion was removed to reduce false correlations (Bandettini
.0001; Fluency, r(30) = −.67, p < .0001; Content, r(30) = −.66, & Bullmore, 2008; Jakobs et al., 2012). Data were then
p < .0001; and Naming and Word Finding, r(30) = −.58, p = bandpass filtered, preserving frequencies between 0.01
.002, subtests. More severe AOS was associated with poorer and 0.08 Hz (Fox & Raichle, 2007; zu Eulenburg et al.,
performance on all WAB-R subtests—Spontaneous Speech, 2012).
r(30) = −.55, p = .001; Fluency, r(30) = −.67, p < .0001;
Content, r(30) = −.46, p = .018; Repetition, r(30) = −.65, ROI Selection
p < .0001; and Naming and Word Finding, r(30) = −.60, p = The brain regions used as ROIs in our analysis are
.001—except Aphasia Quotient, r(30) = −.14, p = 0.49, and those proposed in the literature to be necessary for semantic

3054 Journal of Speech, Language, and Hearing Research • Vol. 63 • 3051–3067 • September 2020

Downloaded from: https://pubs.asha.org 157.51.53.89 on 02/25/2024, Terms of Use: https://pubs.asha.org/pubs/rights_and_permissions

Table 1. Demographic data and behavioral test scores for each participant with stroke.

Western Aphasia Battery–Revised

LV AOS Aphasia
ID Age Gender Education MPO (cm3) RCPM PALPA AOS severity type AQ SS AC Rep Nam/WF

DIS001 70 M 13 52 80.7 19 67 N 1 AN 86 19 7.4 7.6 9

DIS003 72 M 11 156a 34.3 26 63 Y 2.5 AN 81.3 14 8.65 8.8 9.2
DIS004 66 M 17 32 163 34 70 Y 3.5 AN 83.7 15 9.85 8.9 8.1
DIS005 45 F 13 44 66.18 32 72 N 1 AN 93 19 9 8.6 9.9
DIS006b 54 M 16 36 59.2 31 68 Y 3.5 AN 75.1 15 8.15 8.8 5.6
DIS007AC+ 71 M 13 17 32.4 28 56 N 1.5 CO 73.7 15 9.75 6.6 5.5
DIS008AC 58 M 15 10 27.1 34 66 N 1 WE 68.3 14 6.45 4.6 9.1
DIS009 71 M 11 16 1.5 32 70 N 1 AN 91.6 17 9.8 9.7 9.3
DIS010 67 M 13 58 70.4 35 63 N 1 AN 86.4 15 9.95 8.85 9.4
DIS011 77 M 15 81 140.3 30 70 Y 6.5 CO 60.5 10 9.55 3.2 7.5
DIS012 69 M 19 27 55.3 34 66 Y 3 AN 80.8 14 9.4 8.1 8.9
DIS014 48 M 15 13 62.4 36 – Y 4.5 BR 41.6 5 7.4 2.6 5.8
DIS015AC+ 66 M 11 84 171.1 32 60 Y 3 CO 75.3 13 9.95 6.4 8.3
DIS017 76 M 15 120a 158.6 22 63 Y 6.5 BR 39.6 5 9.3 3.3 2.2
DIS018 66 M 16 21 21 27 71 N 1.5 NL 97.3 20 9.45 10 9.2
DIS022 64 F 17 121a 222.4 29 63 N 2 BR 54.5 12 4.85 5 5.4
DIS023 49 M 17 14 49.5 32 69 N 1 AN 72.5 12 9.65 9 5.6
DIS024 59 M 11 69 132.9 27 66 N 1 AN 80.7 14 8.65 8.9 8.8
DIS025 55 F 19 92 17.7 35 71 N 1 NL 98.7 20 9.95 9.4 10
DIS026 71 M 13 11 217.4 25 72 N 1 TS 66.6 9 6.4 9.8 8.1
DIS027 73 M 17 26 298.5 31 72 N 1 BR 50 9 6.8 4.2 5
DIS028 50 M 15 9 62.2 – 68 Y 5 CO 60.2 13 8.4 0 8.7
DIS029 75 M 15 36 117.2 15 – Y 6 GL 17.8 5 2.9a 0.7 0.3
DIS030 61 M 13 3 1.1 35 67 N 1 AN 88.9 18 9.45 7.3 9.7
DIS031 63 M 19 23 56.2 30 61 Y 2.5 BR 62.8 9 8.3 6.6 7.5
DIS047 45 F 17 37 161.5 3 67 N 1 BR 36.9 7 6.25 1.8 3.4
DIS048b 40 M 16 13 143.2 21 66 Y 5 BR 23.9 2 7.45 1.7 0.8
DIS050 51 M 10 6 37.4 34 68 Y 4 CO 69.5 14 9.65 3.4 7.7
DIS051 57 M 11 1 129.8 28 68 Y 4.5 TM 64.8 11 6.8 8.1 6.5
DIS052 74 F 21 5 1.3 35 68 N 1 NL 96 19 10 9.6 9.4

Note. En dash (–) indicates missing data. MPO = months postonset of stroke; LV = lesion volume; RCPM = Ravens Colored Progressive
Matrices; PALPA = Psycholinguistic Assessments of Language Processing in Aphasia Same–Different Discrimination Using Word Minimal
Pairs subtest; AOS = presence of apraxia of speech; AOS severity = rating of severity ranging from 1 to 7; AQ = Aphasia Quotient standard
score (range: 0–100); SS = Spontaneous Speech standard score (range: 0–20); AC = Auditory Comprehension standard score (range: 0–10);
Rep = Repetition standard score (range: 0–10); Nam/WF = Naming and Word Finding standard score (range: 0–10); M = male; N = no; Y = yes;
F = female; AN = Anomic; CO = Conduction; WE = Wernicke; BR = Broca; NL = within normal limits; TS = Transcortical Sensory; GL = Global;
TM = Transcortical Motor.
a
Outlier. bParticipant’s primary language was not English.

or phonological processing and for motor programming. and written word-to-picture matching, and type–token ratio
Please see Table 3 for a list of ROIs and their coordinates, (Halai et al., 2017). Another study that contributed seman-
reasons for their selection as ROIs, and hypotheses about tic ROIs investigated the correlations between gray matter
the WAB-R scores each may predict. Please also see Sup- density and change in word retrieval performance of PWAs
plemental Material S1 for a map of ROI locations. The on object or action picture-naming tasks over time (recovery;
bilateral homologs of all of the ROIs were also included Hope et al., 2017). Finally, Seghier et al. (2010) isolated
in the study, given the known changes in laterality of lan- activation of the left ventral angular gyrus in processing
guage following stroke (Saur et al., 2006) and language- of semantic matching, relative to perceptual matching, in
relevant function in the right hemisphere (L. R. Chai, healthy participants.
Mattar, et al., 2016). Semantic ROIs were regions in which ROIs attributed to phonological aspects of word re-
the structure (gray matter volume or density) was found trieval or production included the left planum polare and
previously to associate with tasks that isolated specific pro- left supramarginal gyrus (phonological factor from Halai
cesses associated with access to word meaning. Several of et al., 2017), and left posterior IFG (Halai et al., 2017;
the ROIs were borrowed from the work of Halai et al. (2017). Okada et al., 2018). The latter region was found to activate
These investigators administered a large battery of language in an fMRI study of repetition suppression and the phono-
measures to a cohort of PWAs (n = 31) and factor analyzed logical planning involved in speech production (Okada et al.,
the data, deriving factors identified as semantic, speech flu- 2018). ROIs associated with speech motor programming
ency, or phonology. Tasks loading on the semantic factor included the left premotor cortex (New et al., 2015), found
included confrontation naming, synonym judgment, spoken previously in these data to relate to AOS severity in this

Ramage et al.: Semantic–Phonological rsfMRI Connectivity in Aphasia 3055

Downloaded from: https://pubs.asha.org 157.51.53.89 on 02/25/2024, Terms of Use: https://pubs.asha.org/pubs/rights_and_permissions
 Figure 1. Lesion overlap map for all stroke participants demonstrating the extent of the lesion and the peak point
of overlap among patients (in white). The peak overlap (23–24 persons with aphasia had lesioned tissue) was in
the left posterior insula (x = −24, y = −30, z = 27) and the body of the caudate (x = −21, y = −8, z = 26). Coordinates
are in Montreal Neurological Institute space. Images generated in Mango (http://ric.uthscsa.edu/mango/mango.html;
Lancaster et al., 2010).

sample, and the right precentral gyrus, noted by Hope et al. between FC among these regions and WAB-R perfor-
(2017) as the premotor cortex and found to negatively corre- mance (see Table 3). For example, damage to the left fu-
late with recovery of picture naming. siform gyrus was associated with poorer performance on
Having chosen these ROIs based on findings about their tasks requiring recognition of speech and access to seman-
associations with fairly specific language task performance tic representations (spoken word-to-picture matching, audi-
allowed for a priori hypotheses regarding relationships tory input and nonverbal response), word retrieval for
single-word production (confrontation naming), and flu-
ent expression (type–token ratio; Halai et al., 2017). As
Table 2. Means (medians) ± standard deviations for the group such, the left fusiform is hypothesized to be connected
demographics for participants with aphasia (PWA) and control with other ROIs involved in semantic processing for tasks
participants and the Western Aphasia Battery–Revised Aphasia with similar semantic access demands—that is, Auditory
Quotient (WAB-R AQ) and subtest scaled scores.
Comprehension, Spontaneous Speech, and Naming and
Word Finding.
Variable PWA Control

n 30 18
Age 62 (64) ± 10 63 ± 9 Data Analysis
Gender (female) 3 10 The G*Power 3.1.9.4 software package was used for
Education in year 15 ± 3 16 ± 4
power calculations (Faul et al., 2009, 2007). The sample
Months postonset 41 (26) ± 40
Lesion volume (cm3) 93 (64) ± 74 size of n = 48 provided 77% power to detect a large effect
AOS severity 3 (2) ± 2 size and 49% power to detect a medium effect size (Faul
PALPA Word Minimal Pairs 0.8 ± 0.2 et al., 2009). Group differences for demographic variables
WAB-R AQ 66.8 (71.6) ± 25 and WAB-R scores were calculated in SPSS 24 (IBM Cor-
Spontaneous Speech 12.8 (14) ± 5
Content 7 (8) ± 3 poration, 2015). The distribution for each of the language
Fluency 5.8 (5) ± 3 variables was visually inspected on Q-Q plots and assessed
Auditory Comprehension 8.3 (9) ± 2 using one-sample Kolmogorov–Smirnov tests, assessing the
Repetition 6.4 (7) ± 3 null hypothesis that the WAB-R variables are normally
Naming and Word Finding 7.1 (8) ± 3
distributed. The null hypothesis was not rejected for the
Note. The groups did not differ significantly for age, U = 225.5, WAB-R Aphasia Quotient, Spontaneous Speech, and Flu-
p = .84, or education, U = 285.5, p = .22. There were more females ency but was rejected for the Naming and Word Finding
in the control group than in the PWA group, χ2(1, N = 48) = 10, (skewness = −1.13, kurtosis = 0.31, D(31) = .19, p = .005),
p = .002. AOS = apraxia of speech; PALPA Word Minimal Pairs =
Psycholinguistic Assessments of Language Processing in Aphasia Auditory Comprehension (skewness = −1.56, kurtosis =
Same–Different Discrimination Using Word Minimal Pairs subtest. 2.39, D(31) = .18, p = .01), and Repetition (skewness = −0.58,
kurtosis = −1.10, D(31) = .18, p = .016) subtests, indicating

3056 Journal of Speech, Language, and Hearing Research • Vol. 63 • 3051–3067 • September 2020

Downloaded from: https://pubs.asha.org 157.51.53.89 on 02/25/2024, Terms of Use: https://pubs.asha.org/pubs/rights_and_permissions

 Table 3. Coordinate locations for the regions of interest based on findings of either semantic, phonological, or motor aspects of word retrieval or production, with input and response
modality noted (input: auditory, visual; response: verbal, nonverbal).

Response
MNI coordinates Region Reason for selection of ROI Input modality modality Predicted WAB-R scores
x y z

54 −18 −22 Anterior middle Gray matter density positively correlated with Visual Verbal Spontaneous Speech
temporal gyrus recovery of word retrieval for confrontation Naming and Word Finding
naming (Hope et al., 2017)
62 −40 −20 Superior inferior Gray matter density peaks that negatively Visual Verbal Spontaneous Speech
temporal gyrus correlated with recovery of word retrieval
46 −38 −28 Inferior inferior for confrontation naming (Hope et al., 2017) Naming and Word Finding
temporal gyrus
−38 −6 −28 Fusiform gyrus A peak area of gray and white matter damage Visual or Auditory Verbal and Auditory Comprehension
associated with performance on confrontation Nonverbal Spontaneous Speech
naming, synonym judgment, spoken and Naming and Word
written word-to-picture matching, type–token Finding
ratio (semantic factor; Halai et al., 2017)
−48 −68 28 Angular gyrus A ventral portion of the angular gyrus involved in Visual Nonverbal Auditory Comprehension
semantic and phonological aspects of word
Ramage et al.: Semantic–Phonological rsfMRI Connectivity in Aphasia

processing on a semantic matching task,

particularly for visual inputs (Seghier et al.,
2010)
−48 −16 0 Planum polare Peaks of gray and white matter damage Auditory Verbal Repetition
−48 −50 14 Posterior supramarginal correlated with word and nonword Spontaneous Speech
gyrus repetition (phonology factor; Halai et al., Auditory Comprehension
2017)
58 −38 −8 Posterior middle Gray matter density peak positively correlated Visual Verbal Naming and Word Finding
temporal gyrus with recovery of word retrieval for confrontation
naming (Hope et al., 2017)
−58 1 23 Premotor cortex Connectivity with the right premotor cortex — Verbal All requiring verbal response
associated with severity of apraxia of speech
in this cohort of persons with aphasia (New
et al., 2015)
50 −6 40 Precentral gyrus Gray matter density peak negatively correlated Visual Verbal Naming and Word Finding
with recovery of word retrieval for confrontation
naming (Hope et al., 2017)
−56 17 15 Posterior inferior Functional activity of this region observed in Auditory Verbal Repetition
frontal gyrus repetition suppression and phonological Spontaneous Speech
planning involved in speech production Naming and Word Finding
(Okada et al., 2018)
−51 23 30 Middle frontal Region associated with word retrieval, particularly Visual or Auditory Verbal and Spontaneous Speech
gyrus when controlled for the influences of semantics Nonverbal Naming and Word Finding
and motor execution, across several neuroimaging
methods (Price, 2012)

Note. Em dash indicates no input. Regions labeled by coordinates provided by the authors and using the Montreal Neurological Institute’s (MNI) coordinate system in the Talairach
Daemon (Lancaster et al., 2000). ROI = region of interest; WAB-R = Western Aphasia Battery–Revised.
3057

Downloaded from: https://pubs.asha.org 157.51.53.89 on 02/25/2024, Terms of Use: https://pubs.asha.org/pubs/rights_and_permissions

that the latter were not normally distributed. To identify scores with small differences in the scaled score (< 1 point),
the influences of lesion volume or AOS severity on WAB-R while significant FC predictors were less precise, with
scaled scores, Spearman correlations were computed. larger confidence intervals for scaled scores. Thus, for each
WAB-R subtest, comparison of fit statistics (Akaike’s In-
FC formation Criterion and Bayesian Information Criterion;
Each ROI’s time course was extracted within 5 mm Müller et al., 2013) indicated whether the inclusion of the
of the respective peak coordinate (see Table 1). Inter-ROI FC measures improved model fit. The results reported be-
connectivity was assessed with linear Pearson correlations low represent the best fitting models for each score. Omni-
(transformed into Fisher’s Z values), and independent- bus test statistics and Wald χ2 values and significance for
samples t tests determined group differences (p < .05, false all predictors by each WAB-R standard score are reported
discovery rate–corrected for multiple comparisons). in Supplemental Material S4, best fitting model in Supple-
mental Material S5. Significant findings are noted in Table 4
and presented in Figure 4.
Generalized Linear Models
To identify potential predictors of WAB-R perfor-
mance, the associations between FC and WAB-R scores Prediction of WAB-R Aphasia Quotient.
were assessed with multivariate general linear models, with PWAs with smaller lesions, absence of or less severe
potential predictors being variables (including sex, lesion AOS, and weaker FC between the right posterior middle
volume, months postonset, and AOS severity) associated temporal gyrus (pMTG) and right IFG had higher Aphasia
with each score at p < .05 with Type I or Type III sums of Quotient scores (see Table 5).
squares. The identified predictor variables for each WAB-R
standard score were input to a generalized linear model. Spontaneous Speech Standard Score
Generalized linear models provide a flexible extension of Lesion volume was the only significant predictor for
ordinary least squares regression and can accommodate WAB-R Spontaneous Speech (and Fluency) scores. How-
dependent variables with either Gaussian or non-Gaussian ever, model fit for the Content score, a component of the
error distributions (McCullagh & Nelder, 1989). General- Spontaneous Speech standard score, indicated that older
ized linear models enable the analyst to specify a link func- PWAs with smaller lesions, no or mild AOS, and stronger
tion that relates the expected value of the response variable FC between the left fusiform and the right inferior inferior
to the linear predictors in the model. temporal gyrus (ITG) had higher Content scores.
Visual inspection of the variable distributions and
fit statistics indicated which model best fit the data (e.g., Auditory Comprehension Standard Score
normal, Tweedie, Inverse Gaussian) and which predictors Stronger FC between the left and right angular gyrus,
were significant ( p < .05, corrected for multiple compari- and the left angular gyrus and the left superior ITG pre-
sons with the false discovery rate). Given the relatively small dicted higher WAB-R Auditory Comprehension scores.
sample size, the robust estimator was used.
Repetition Standard Score
Results AOS severity was the only significant predictor of
Repetition performance, with less severe AOS predicting
FC higher Repetition scores.
Figure 3A presents the 57 ROI–ROI resting-state
connections that were coherent and significant in both the Naming and Word Finding
control and stroke participants. Of the 57, 36 were present Smaller lesion volumes, absent or minimal AOS, as
in both groups, 16 only in the controls, and five only in the well as stronger FC between the left IFG–left middle fron-
PWAs (see Figure 2; averages and standard deviations for tal gyrus (MFG) and the right pMTG–right IFG predicted
each are presented in Supplemental Material S3. The pres- higher WAB-R Naming and Word Finding scores.
ence of statistically significant connections in the PWA
and control groups suggests that these ROIs represent an
intrinsically coherent network that is present at rest in Discussion
healthy and lesioned brains. Controls had stronger FC This study demonstrates that selecting semantic and
than PWAs for the left and right premotor cortex connection, phonological ROIs a priori, using existing neuroimaging
t(46) = 3.5, p < .05, false discovery rate–corrected. findings from structural and functional studies of semantic
Figure 3B presents the connections that were identi- and phonological processing, elucidates a coherent resting-
fied as potential predictors of one or more of the WAB-R state network in HC and individuals with aphasia secondary
scores, while Figure 3C presents those connections found to left-hemisphere stroke. This aligns with previous reports
to be significant predictors. For all of the generalized linear of coherent frequency oscillations among the regions of the
models, the beta and Wald confidence intervals indicated language network at rest (e.g., Siegel et al., 2016). Con-
that participant characteristics like age, lesion volume, or trols had stronger connectivity than PWAs between the
AOS severity were more precise predictors of language left and right premotor cortex, associated with the presence

3058 Journal of Speech, Language, and Hearing Research • Vol. 63 • 3051–3067 • September 2020

Downloaded from: https://pubs.asha.org 157.51.53.89 on 02/25/2024, Terms of Use: https://pubs.asha.org/pubs/rights_and_permissions

Figure 2. The healthy control (HC) group (A) had a greater number of resting-state connections (indicated by black lines connecting regions of
interest labeled) than the persons with aphasia (PWA) group (B) but only differed significantly for strength of the left PM–right PM connection
(HC > PWA). The circles (left) allow for visualization of all connections (between hemispheres and within hemisphere) in two dimensions, while
the left and right lateral views (right) allow for visualization of within-hemisphere connections and their anatomic locations. Figures generated
in BrainNet Viewer (http://www.nitrc.org/projects/bnv/; Xia et al., 2013). PM = premotor cortex; MFG = middle frontal gyrus; IFG = inferior
frontal gyrus; PcG = precentral gyrus; PP = planum polare; Fusi = fusiform gyrus; aMTG = anterior middle temporal gyrus; sITG = superior aspect
of the inferior temporal gyrus; iITG = inferior aspect of the inferior temporal gyrus; pMTG = posterior middle temporal gyrus; AG = angular gyrus;
SMG = supramarginal gyrus.

and severity of AOS (New et al., 2015), but there were no participant characteristics known to be associated with
other significant group differences. aphasia severity (e.g., lesion volume, AOS severity), were
To validate this network of brain regions for its tested as predictors of performance on a commonly used
relevance to language, between-regions FC, along with clinical assessment in aphasia, the WAB-R. Smaller lesion

Figure 3. Of the connections present at rest in the study participants (A; red line indicates left PM–right PM group difference, t(46) = 3.5, p < .05
corrected for false discovery rate), 15 were potential predictors (green) of language performance on the Western Aphasia Battery–Revised (B).
Only six were significant predictors (red) of language performance in the generalized linear models (C). PM = premotor cortex.

Ramage et al.: Semantic–Phonological rsfMRI Connectivity in Aphasia 3059

Downloaded from: https://pubs.asha.org 157.51.53.89 on 02/25/2024, Terms of Use: https://pubs.asha.org/pubs/rights_and_permissions
 3060
Journal of Speech, Language, and Hearing Research • Vol. 63 • 3051–3067 • September 2020

Table 4. Significant predictors, all χ2(df = 2–4, N = 48) > 16, all ps < .05, FDR-corrected for multiple comparisons, for each of the Western Aphasia Battery–Revised standard scores
included PWA characteristics including lesion volume, apraxia of speech (AOS) severity, and age, as well as functional connectivity among regions of interest in the frontal, temporal,
and parietal lobes and involving unilateral (right [R], left [L]) and bilateral (B) connections.

Aphasia Spontaneous Auditory Naming and

Predictor Quotient Speech Content Fluency Comprehension Repetition Word Finding

Lesion volume Larger lesion = Larger lesion = Larger lesion = Larger lesion = Larger lesion =
lower score lower score lower score lower score lower score
AOS severity More severe More severe More severe More severe
AOS = lower AOS = lower AOS = lower AOS = lower
score score score score
Age Older = better
score
Connection location
Frontal L lIFG-lMFG Weaker FC =
lower score
Frontotemporal R rpMTG-rIFG Weaker FC = Weaker FC =
lower score lower score
Temporal B lFusi-riITG Stronger FC =
higher score
Temporoparietal L lAG-lsITG Stronger FC =
higher score
Parietal B lAG-rAG Stronger FC =
higher score

Note. FC = functional connectivity; l = left; r = right; IFG = inferior frontal gyrus; MFG = middle frontal gyrus; pMTG = posterior middle temporal gyrus; iITG = inferior inferior temporal
gyrus; Fusi = fusiform gyrus; AG = angular gyrus; sITG = superior inferior temporal gyrus.

Downloaded from: https://pubs.asha.org 157.51.53.89 on 02/25/2024, Terms of Use: https://pubs.asha.org/pubs/rights_and_permissions

Figure 4. Axial (left), left hemisphere (middle), and right hemisphere (right) depictions of functional connectivity that is predictive of language
performance on the Western Aphasia Battery–Revised in the stroke participants. Table 4 provides the generalized linear models and the
combinations of variables predicting performance, including each of the connections depicted here, all χ2(df = 2–4, N = 48) > 16, ps < .05
false discovery rate–corrected. Stronger connectivity strength between the left fusiform and the right inferior ITG (A) predicts better Content
score. Stronger connectivity strength between the bilateral angular gyrus and between the left angular gyrus and left superior ITG (B) predicts
better Auditory Comprehension score. Weaker connectivity strength between the left IFG and the left MFG (C) predicts better Naming and
Word Finding score. Weaker connectivity strength between the right pMTG and the right IFG (D) predicts better Aphasia Quotient and Naming
and Word Finding scores. Figures generated in BrainNet Viewer (http://www.nitrc.org/projects/bnv/; Xia et al., 2013). l = left; r = right; MFG =
middle frontal gyrus; IFG = inferior frontal gyrus; Fusi = fusiform gyrus; sITG = superior aspect of the inferior temporal gyrus; pMTG = posterior
middle temporal gyrus; AG = angular gyrus.

volume and presence of less severe AOS predicted better hemispheres that followed a pattern of stronger frontal
language performance on all the WAB-R subtests except and stronger temporoparietal connectivity corresponding
Auditory Comprehension. In fact, Spontaneous Speech to better performance.
(particularly the Fluency rating) and Repetition were best
predicted by these two variables—that is, none of the FC
variables improved model fit. However, the predictive
Stronger Connectivity Involving Frontal ROIs
models for the Aphasia Quotient, the Content component Predicts Better Word Retrieval
of Spontaneous Speech, and Naming and Word Finding Stronger connectivity involving frontal lobe ROIs
scores fit best with the inclusion of FC variables between (left IFG–left MFG, right pMTG–right IFG), along with
frontal, temporal, and parietal regions in the left and right smaller lesion volumes and less severe AOS, predicted better

Table 5. Summary of the predictors for each of the Western Aphasia Battery–Revised scores.

Western Aphasia Battery–Revised Scores

Predictors AQ SS AC Rep Naming

Lesion volume − − −
Apraxia of speech severity − − −
Connection Location Connection Prediction about the function
of the connection
Frontal L lIFG-lMFG Phonological planning + word −
retrieval
Frontal-temporal R rpMTG-rIFG Naming recovery − −
Temporal B lAG-rAG Semantic and phonological +
aspects of processing
visual inputs
Temporal L lAG-lsITG Naming recovery +

Note. Based on the predictions of region–behavior associations predicted in Price (2012) and the findings in this study, hypothesized
functions for the connections are proposed. AQ = Aphasia Quotient; SS = Spontaneous Speech; AC = Auditory Comprehension; Rep =
Repetition; + = positive predictor; − = negative predictor; L = left hemisphere; R = right hemisphere; B = bilateral; l = left; r = right; IFG =
inferior frontal gyrus; MFG = middle frontal gyrus; pMTG = posterior middle temporal gyrus; AG = angular gyrus; sITG = superior inferior
temporal gyrus.

Ramage et al.: Semantic–Phonological rsfMRI Connectivity in Aphasia 3061

Downloaded from: https://pubs.asha.org 157.51.53.89 on 02/25/2024, Terms of Use: https://pubs.asha.org/pubs/rights_and_permissions
Naming and Word Finding performance (see Table 4). pMTG–right IFG connection relates to language perfor-
According to the studies from which these regions were mance in HC would aid in interpretation of this finding.
borrowed, the left IFG (phonological planning for speech
production; Okada et al., 2018) and the left MFG (word re- Stronger Connectivity Among Temporoparietal
trieval; Price, 2012) were predicted to be involved in concert
for tasks requiring word retrieval and verbal expression—
Regions Predicts Better Access to Semantics
that is, Spontaneous Speech (Content) and Naming and in Visual or Auditory Modalities
Word Finding. This prediction was correct for Naming and The bilateral connection between the left fusiform
Word Finding, but not for Spontaneous Speech. However, and the right inferior ITG, two regions associated with ac-
similar connections involving the frontal lobes were “poten- cess to semantic information for word retrieval, is necessary
tial” but nonsignificant predictors for Spontaneous Speech for imparting relevant Content in Spontaneous Speech.
(e.g., right pMTG–right IFG; see Supplemental Material S4) Both ROIs were predicted to be involved in semantic access
as well as its subcomponents of Content (e.g., right superior for Spontaneous Speech and Naming and Word Finding,
ITG–right MFG) and Fluency (e.g., right superior ITG– with the common factor being word retrieval. Both regions
left MFG). Nonetheless, this pattern of bilateral temporal– may be considered part of the “ventral stream” or the
frontal connections highlights the importance of these “what” stream of Hickok and Poeppel (2007).
relatively long-distance connections for verbal production Additionally, stronger connectivity between the left
of retrieved words in isolation (confrontation naming) and angular gyrus–right angular gyrus and the left angular
in connected language (picture description). gyrus–left superior ITG was the only significant predictor
Weaker connectivity among frontal regions, par- of Auditory Comprehension scores, that is, no characteris-
ticularly the left IFG–left MFG connection, may reflect tics of aphasia or stroke predicted comprehension. These
an influence of the presence of AOS in 15 members of the regions were found in previous studies to be involved in
PWA group. AOS severity has a negative impact on Nam- semantic association tasks (i.e., angular gyrus) and with re-
ing and Word Finding performance (see Table 4), likely covery of word retrieval (i.e., superior ITG). The common
a result of the requirement of a verbal response. However, denominator for these two tasks is access to word meaning,
AOS severity was not correlated with left IFG–left MFG and thus, we propose that these connections are necessary
connectivity strength, ρ(30) = .18, p = .34. These ROIs for access to semantics for understanding auditory input.
were hypothesized to be involved in phonological process- These connections are also in agreement with the flow of
ing and word retrieval (see Table 3). Given the minimal language processing proposed in the Dual Stream Model
effect of AOS on left IFG–left MFG connection strength (Hickok & Poeppel, 2007), with the angular gyrus aligning
and the a priori predictions from the ROI selection (see with the “conceptual network” and in proximity to the
Table 3), the results suggest that this connection is likely sensorimotor interface (sylvian-parieto-temporal area) and
involved in retrieving phonological representations of target the superior ITG aligning with the lexical interface (linking
words and translating phonological information for pro- phonological to semantic representations).
gramming of motor execution.
The most robust finding in the study is the relation-
Right-Hemisphere Connectivity and Language
ship between right pMTG–right IFG connectivity and
Aphasia Quotient and Naming and Word Finding scores. in Aphasia
Greater gray matter density of the right pMTG was found The right pMTG–right IFG connection would likely
previously to be a positive predictor of naming recovery be structurally connected via a right-hemisphere homolog
in individuals with aphasia—that is, greater gray matter of the “dorsal stream” (Hickok & Poeppel, 2007). In the
density of the right pMTG was associated with better nam- left hemisphere, it is conceivable that this connection could
ing and word finding performance (Hope et al., 2017). Hope be strengthened (since weaker connectivity = more severe
and colleagues report that right-hemisphere regions un- aphasia), with improvements in access to phonological rep-
dergo systematic structural adaptations to support language resentations. This has been observed as improved naming
functions, even in the chronic phase of recovery, and that and word finding performance following constraint-induced
the degree of these changes relates to word retrieval. This language therapy, via enhanced integrity of the left infe-
study adds that weaker connectivity of the right pMTG– rior longitudinal fasciculus (McKinnon et al., 2017). While
right IFG also associates with better Aphasia Quotient McKinnon et al. found that group-based constraint-induced
scores. It is outside the scope of this study to identify whether language therapy improved naming performance and re-
or how right pMTG gray matter density may relate to right duced semantic error production (likely ventral stream
pMTG–right IFG connectivity, but we speculate that the changes), others have reported improvement in other language
structural adaptation of this structure may facilitate con- measures (e.g., repetition in Bi et al., 2011, or WAB-R
nections to other regions and that FC of right pMTG–right performance across subtests in Mozeiko et al., 2018). These
IFG may be a maladaptive functional adaptation to apha- data suggest that integrity and strength of frontotemporal
sia. Longitudinal study of this connection would best ad- connections is predictive of language performance and
dress the validity and utility of this prediction. As well, “necessary” for language in aphasia. That is, input and
having an understanding about whether and how the right output processing are integrated through this pathway, and

3062 Journal of Speech, Language, and Hearing Research • Vol. 63 • 3051–3067 • September 2020

Downloaded from: https://pubs.asha.org 157.51.53.89 on 02/25/2024, Terms of Use: https://pubs.asha.org/pubs/rights_and_permissions

therefore, treatments should affect change in it. However, though one other study has reported increased connectiv-
the right-hemisphere homolog of the dorsal stream has not ity among posterior regions of the default mode network
yet been investigated as a target for change secondary to following a naming treatment (Marcotte et al., 2013). We
language interventions. Given the findings here and those therefore are cautious of overinterpreting this finding, given
of Hope et al. (2017), it is conceivable that structural and that resting-state FC in stroke may be confounded by varia-
functional integrity of right-hemisphere frontotemporal tion in hemodynamic lag (Siegel et al., 2017) and that the
connectivity may be a marker of outcomes in intervention. bulk of lesioned tissue in this cohort is in the left frontal
The majority of studies in aphasia on left–right brain cortex; thus, replication in a larger study, potentially with
structure and function has operated under the theory that hemodynamic lag as a variable, is warranted. Nonetheless,
right-hemisphere involvement in language processing is the data here suggest that interhemispheric connectivity in-
compensatory or maladaptive (Saur et al., 2006). Laterality volving the left angular gyrus and superior ITG is signifi-
of language function is known to shift to the right in the cant in predicting performance on language tasks requiring
acute phase and reshift to the left hemisphere with recovery semantic and phonological access and maintenance of rep-
(cf. Hartwigsen & Saur, 2017) or to shift to the right with resentations for comprehension and production, requiring
certain types of treatments (e.g., Melodic Intonation Ther- synthesis of information processing across the language and
apy in Wan et al., 2014). This study is merely a snapshot other cognitive domains (L. R. Chai, Mattar, et al., 2016).
of resting-state FC in the chronic phase and cannot address
compensation, but the patterns of positive and negative
predictive value for the bilateral connections suggest that A Comment on Compensatory Connectivity
right-hemispheric connectivity is not simply compensatory The clinical significance of the right pMTG–right
or maladaptive. As mentioned above, aphasia severity is IFG connection, potentially as a maladaptive response to
predicted, in part, by connectivity strength between the right stroke in the left hemisphere, is intriguing. To be clear,
pMTG–right IFG, and the significant bilateral predictors the data here suggest that the stronger the connectivity
were positive (i.e., stronger connectivity = better perfor- between these right-hemisphere homologs of Broca’s and
mance). As such, strengthening these connections may be Wernicke’s areas, the poorer the language performance
helpful in (a) improving semantic access for processing overall (stronger connectivity = more severe aphasia). It
information regardless of the need for output (left angular has been suggested that right-hemisphere involvement in
gyrus–right angular gyrus predicting Auditory Comprehen- language processing in individuals with aphasia, as well
sion) and (b) formulating spontaneous speech (left fusiform– as in HC, may be indicative of “lexical learning” or reso-
right inferior ITG predicting Content in Spontaneous lution of lexical conflict in learning (Raboyeau et al., 2010).
Speech). Though, in most cases, the right-sided activity during lan-
guage performance, particularly in the right IFG, is sub-
optimal (Rosen et al., 2000). Study of the whole brain at
Importance of Bilateral Temporoparietal Connectivity rest indicates that, while the left-hemisphere ROIs in this
The bilateral connections that were predictive of lan- study may represent a “core” language network, right-
guage performance involved temporal or parietal regions hemisphere homologs are considered to be “peripheral”
(see Figure 4). This finding is akin to those of Braun et al. with less stable (or more flexible) connectivity at rest, sug-
(2001) and Silbert et al. (2014), who report that earlier gesting that it may have a less specialized functional role
stages of lexical processing for comprehension and produc- in language (L. R. Chai, Mattar, et al., 2016). For exam-
tion occur in posterior regions bilaterally, while the later ple, the right-hemisphere auditory cortex (primarily the
stages necessary for production are left-lateralized and more parietotemporal junction) is involved in error detection and
anterior. Similarly, a recent coordinate-based meta-analysis making feedback-feedforward predictions about verbal
of sentence comprehension and production points to bi- productions—that is, when productions match the intended
lateral posterior temporal cortex (temporo-occipital cortex) sounds (cf. Houde et al., 2002; Kort et al., 2014). Findings
in lexical–semantic processing (Walenski et al., 2019) for of this effect have been observed in electroencephalograpy or
sentence production. All three of these studies were con- magnetoencephalography (MEG) specific to voicing (e.g.,
ducted in neurologically healthy, relatively young adults Parkinson et al., 2014) and vowel production (e.g., Niziolek
during sentence comprehension and/or production tasks et al., 2013) and at discrete timescales. However, it also has
but suggest roles of bilateral posterior temporal regions in been documented as reduced right-hemisphere MEG signal
lexical–semantic processing for comprehension and pro- during whole-item versus fine-grained lexical decision anal-
duction of complex language. Specific to resting-state FC, ysis of phonological stimuli (pseudowords)—an effect oppo-
increased bilateral connectivity at rest is relevant to that site of that seen in the left hemisphere (Ylinen et al., 2015).
observed during language processing (Tzourio-Mazoyer Though these findings are not localized to the ROIs in this
et al., 2015) and is predictive of better overall outcomes study per se, they document a specialized role of the right
following stroke (Puig et al., 2018). We are not aware of temporal cortex in monitoring verbal productions, which
other studies that have investigated “resting-state” connec- has also been noted in real-time monitoring of language
tivity in stroke and aphasia indicating anterior–posterior production in individuals with aphasia (Sreedharan et al.,
interhemispheric differences relative to language performance, 2019). Because this connection was a negative predictor of

Ramage et al.: Semantic–Phonological rsfMRI Connectivity in Aphasia 3063

Downloaded from: https://pubs.asha.org 157.51.53.89 on 02/25/2024, Terms of Use: https://pubs.asha.org/pubs/rights_and_permissions
Naming and Word Finding here (weaker connection = bet- was predictive of aphasia severity and word retrieval for
ter performance), which involves verbal production, we verbal production.
speculate that the right pMTG–right IFG connection may The connections that were predictive of language
be involved in monitoring of verbal production in such a performance in aphasia have potential as targets for the
way that is maladaptive when uninhibited. However, the study of neuroplastic functional adaptations as evidence of
poor temporal resolution of fMRI relative to structural im- treatment efficacy. For example, one might predict that
aging or electroencephalograpy/MEG data limits our ability efficacy of treatments targeting semantic access for word
to address this speculation in this study. retrieval (e.g., semantic feature analysis; Boyle & Coehlo,
1995) should increase FC between the left angular gyrus–
left superior ITG and left fusiform–right inferior ITG and
Limitations weaker connectivity between the right pMTG–right IFG.
Though the intent of using the WAB-R as the language Or, one might predict that connectivity of the left IFG–left
measure allows for application to a well-known clinical MFG should strengthen given a phonologically based
measure for aphasia, it does not allow for specific defini- treatment (e.g., phonological components analysis; Leonard
tion of the language processes assessed in each subtest. The et al., 2008). Further study is required to determine the
loosely defined language processes do not allow for more utility of such predictions. Rather, this study demonstrates
direct predictions about the many language processes in- that differing types of information may be gained by care-
volved in each subtest and potential disagreement about ful consideration of the brain regions assessed and fine-
the hypotheses put forth in Table 4. Given that we had tuning the analytic approach to identify specific language
relatively low power to detect small-to-medium effect sizes, processes, even when using clinical assessment measures.
there may be a risk of Type II error. As such, replication These types of data may be invaluable in determining
of this study in a larger sample with more detailed charac- whether the changes that clinicians seek through intervention
terization of the semantic–phonological impairment, as well are (a) targeting the correct neural system to address a
as extension to understanding whether and how rsfMRI language process and (b) guiding outcomes to address treat-
relates to communication for life participation, is warranted. ment effects. Future study is needed to determine whether
In addition, the intent and design of the study centered the FC targets identified here can be useful in making these
on an ROI, seed-based analysis and did not consider the determinations.
potential for use of differing time windows in the approach
to account for variability in the hemodynamic response or Acknowledgments
a lag of the response as a result of stroke. Future studies
This work was supported by National Health and Medical
may consider alternative approaches (e.g., as in L. R. Chai,
Research Council Project Grant 632763 and Australian Research
Mattar, et al., 2016) that may not only control for hemo- Council Future Fellowship FT120100355 (Principal Investigator:
dynamic lag but also consider the stability and functional Kirrie J. Ballard). The authors acknowledge the contributions
specification of bilateral or right-hemisphere connections. of the members of the University of New Hampshire Cognition,
Finally, the cross-sectional design employed in this study pre- Brain and Language Team (Hannah Franz, Gwyneth Horne, Mara
cludes causal inferences, and a longitudinal study in future Callahan, Jessica Lee, and Elizabeth Kinney) and their help in
study would expand the utility of the predictions to recov- organizing and analyzing the data. They also thank Cathy Price
ery or treatment effects. and the Predicting Language Outcomes and Recovery After Stroke
laboratory for conducting the automated lesion identification for
this project.

Conclusions
References
The network of semantic and phonological brain
regions chosen as ROIs for their roles in semantic and Ashburner, J., & Friston, K. J. (2005). Unified segmentation. Neuro-
Image, 26(3), 839–851. https://doi.org/10.1016/j.neuroimage.
phonological processing show patterns of FC that predict
2005.02.018
language performance in poststroke aphasia. Connections Ballard, K. J., Azizi, L., Duffy, J. R., McNeil, M. R., Halaki, M.,
between regions that relate to recovery of specific language O’Dwyer, N., Layfield, C., Scholl, D. I., Vogel, A. P., & Robin,
functions (e.g., naming; Hope et al., 2017) were most pre- D. A. (2016). A predictive model for diagnosing stroke-related
dictive of language performance. For example, at least one apraxia of speech. Neuropsychologia, 81, 129–139. https://doi.
of the ROIs drawn from the Hope study was present in org/10.1016/j.neuropsychologia.2015.12.010
three of the five connections found to predict language per- Bandettini, P. A., & Bullmore, E. (2008). Endogenous oscillations
formance. Several of these connections involve the right and networks in functional magnetic resonance imaging. Hu-
hemisphere, and bilateral connections were limited to the man Brain Mapping, 29(7), 737–739. https://doi.org/10.1002/
hbm.20607
temporal cortex, supporting the notion that semantic pro- Bi, Y., Wei, T., Wu, C., Han, Z., Jiang, T., & Caramazza, A.
cessing and lexical access involves a bilateral posterior sub- (2011). The role of the left anterior temporal lobe in language
network in individuals with aphasia. Other connections processing revisited: Evidence from an individual with ATL
involved frontal regions and were largely left lateralized, resection. Cortex, 47(5), 575–587. https://doi.org/10.1016/
with the exception of the right pMTG–right IFG, which j.cortex.2009.12.002

3064 Journal of Speech, Language, and Hearing Research • Vol. 63 • 3051–3067 • September 2020

Downloaded from: https://pubs.asha.org 157.51.53.89 on 02/25/2024, Terms of Use: https://pubs.asha.org/pubs/rights_and_permissions

Boyle, M., & Coehlo, C. A. (1995). Application of semantic fea- speech fluency, phonology and semantics. Cortex, 86, 275–289.
ture analysis as a treatment for aphasic dysnomia. American https://doi.org/10.1016/j.cortex.2016.04.016
Journal of Speech Language Pathology. 4(4), 94–98. https://doi. Halai, A. D., Woollams, A. M., & Lambon Ralph, M. A. (2018). Pre-
org/10.1044/1058-0360.0404.94 dicting the pattern and severity of chronic post-stroke language
Braun, A. R., Guillemin, A., Hosey, L., & Varga, M. (2001). The deficits from functionally-partitioned structural lesions. Neuro-
neural organization of discourse: An H2 15O-PET study of Image: Clinical, 19, 1–13. https://doi.org/10.1016/j.nicl.2018.03.011
narrative production in English and American Sign Language. Hartwigsen, G., & Saur, D. (2017). Neuroimaging of stroke recov-
Brain: A Journal of Neurology, 124(Pt 10), 2028–2044. https:// ery from aphasia—Insights into plasticity of the human lan-
doi.org/10.1093/brain/124.10.2028 guage network. NeuroImage, 190, 14–31. https://doi.org/10.1016/
Chai, L. R., Mattar, M. G., Blank, I. A., Fedorenko, E., & Bassett, j.neuroimage.2017.11.056
D. S. (2016). Functional network dynamics of the language Hickok, G., & Poeppel, D. (2007). The cortical organization of
system. Cerebral Cortex, 26(11), 4148–4159. https://doi.org/ speech processing. Nature Reviews. Neuroscience, 8(5), 393–402.
10.1093/cercor/bhw238 https://doi.org/10.1038/nrn2113
Chai, X. J., Berken, J. A., Barbeau, E. B., Soles, J., Callahan, M., Hjelmervik, H., Hausmann, M., Osnes, B., Westerhausen, R., &
Chen, J.-K., & Klein, D. (2016). Intrinsic functional connectiv- Specht, K. (2014). Resting states are resting traits—An fMRI
ity in the adult brain and success in second-language learning. study of sex differences and menstrual cycle effects in resting
Journal of Neuroscience, 36(3), 755–761. https://doi.org/10.1523/ state cognitive control networks. PLOS ONE, 9(7), e103492.
JNEUROSCI.2234-15.2016 https://doi.org/10.1371/journal.pone.0103492
Crinion, J., Ashburner, J., Leff, A., Brett, M., Price, C., & Friston, K. Holland, A. L., & Ramage, A. E. (2000). Thinking outside the
(2007). Spatial normalization of lesioned brains: Performance (black) box. Brain and Language, 71(1), 93–95. https://doi.org/
evaluation and impact on fMRI analyses. NeuroImage, 37(3), 10.1006/brln.1999.2222
866–875. https://doi.org/10.1016/j.neuroimage.2007.04.065 Hope, T. M. H., Leff, A. P., Prejawa, S., Bruce, R., Haigh, Z.,
Dabul, B. (2000). Apraxia Battery for Adults: Examiner’s manual. Lim, L., Ramsden, S., Oberhuber, M., Ludersdorfer, P., Crinion,
Pro-Ed. J., Seghier, M. L., & Price, C. J. (2017). Right hemisphere
den Ouden, D.-B., Malyutina, S., Basilakos, A., Bonilha, L., structural adaptation and changing language skills years af-
Gleichgerrcht, E., Yourganov, G., Hillis, A. E., Hickok, G., ter left hemisphere stroke. Brain, 140(6), 1718–1728. https://
Rorden, C., & Fridriksson, J. (2019). Cortical and structural- doi.org/10.1093/brain/awx086
connectivity damage correlated with impaired syntactic pro- Hope, T. M. H., Prejawa, S., Parker Jones, Ō., Oberhuber, M.,
cessing in aphasia. Human Brain Mapping, 40(7), 2153–2173. Seghier, M. L., Green, D. W., & Price, C. J. (2014). Dissecting
https://doi.org/10.1002/hbm.24514 the functional anatomy of auditory word repetition. Frontiers
Duffy, J. R. (2013). Motor speech disorders: Substrates, differential in Human Neuroscience, 8, 246. https://doi.org/10.3389/fnhum.
diagnosis, and management. Elsevier Health Sciences. 2014.00246
Faul, F., Erdfelder, E., Buchner, A., & Lang, A.-G. (2009). Statisti- Houde, J. F., Nagarajan, S. S., Sekihara, K., & Merzenich, M. M.
cal power analyses using G*Power 3.1: Tests for correlation (2002). Modulation of the auditory cortex during speech: An
and regression analyses. Behavior Research Methods, 41(4), MEG study. Journal of Cognitive Neuroscience, 14(8), 1125–1138.
1149–1160. https://doi.org/10.3758/BRM.41.4.1149 https://doi.org/10.1162/089892902760807140
Faul, F., Erdfelder, E., Lang, A.-G., & Buchner, A. (2007). G*Power 3: Howard, D., & Patterson, K. E. (1992). The Pyramids and Palm
A flexible statistical power analysis program for the social, be- Trees Test: A test of semantic access from words and pictures.
havioral, and biomedical sciences. Behavior Research Methods, Thames Valley Test.
39(2), 175–191. https://doi.org/10.3758/BF03193146 IBM Corporation. (2015). IBM SPSS Statistics for Mac (Version 24.0)
Folstein, M. F., Folstein, S. E., & McHugh, P. R. (1975). “Mini- [Computer software].
mental state.” A practical method for grading the cognitive state Ivanova, M. V., Isaev, D. Y., Dragoy, O. V., Akinina, Y. S.,
of patients for the clinician. Journal of Psychiatric Research, Petrushevskiy, A. G., Fedina, O. N., Shklovsky, V. M., &
12(3), 189–198. https://doi.org/10.1016/0022-3956(75)90026-6 Dronkers, N. F. (2016). Diffusion-tensor imaging of major
Fox, M. D., & Raichle, M. E. (2007). Spontaneous fluctuations white matter tracts and their role in language processing in
in brain activity observed with functional magnetic resonance aphasia. Cortex, 85, 165–181. https://doi.org/10.1016/j.cortex.
imaging. Nature Reviews Neuroscience, 8(9), 700–711. 2016.04.019
Fridriksson, J., den Ouden, D.-B., Hillis, A. E., Hickok, G., Rorden, Jackson, R. L., Hoffman, P., Pobric, G., & Lambon Ralph, M. A.
C., Basilakos, A., Yourganov, G., & Bonilha, L. (2018). Anatomy (2016). The semantic network at work and rest: Differential
of aphasia revisited. Brain, 141(3), 848–862. https://doi.org/ connectivity of anterior temporal lobe subregions. Journal
10.1093/brain/awx363 of Neuroscience, 36(5), 1490–1501. https://doi.org/10.1523/
Geschwind, N. (1970). The organization of language and the brain. JNEUROSCI.2999-15.2016
Science, 170(3961), 940–944. Jakobs, O., Langner, R., Caspers, S., Roski, C., Cieslik, E. C.,
Grodzinsky, Y., & Amunts, K. (2006). Comments regarding the Zilles, K., Laird, A. R., Fox, P. T., & Eickhoff, S. B. (2012).
seat of the faculty of spoken language, followed by an obser- Across-study and within-subject functional connectivity of a
vation of aphemia (loss of speech) (1861). In Y. Grodzinsky right temporo-parietal junction subregion involved in stimu-
& K. Amunts (Eds.), Broca’s region (Chap. 18). Oxford Uni- lus–context integration. NeuroImage, 60(4), 2389–2398. https://
versity Press. https://www.oxfordscholarship.com/view/10.1093/ doi.org/10.1016/j.neuroimage.2012.02.037
acprof:oso/9780195177640.001.0001/acprof-9780195177640- Kay, J., Lesser, R., & Coltheart, M. (1992). PALPA: Psycholin-
chapter-18 guistic Assessments of Language Processing in Aphasia. Psy-
Halai, A. D., Woollams, A. M., & Lambon Ralph, M. A. (2017). chology Press.
Using principal component analysis to capture individual Kertesz, A. (2006). Western Aphasia BatteryTM—Revised (WAB-RTM).
differences within a unified neuropsychological model of chronic Pearson. https://www.pearsonclinical.ca/en/products/product-
post-stroke aphasia: Revealing the unique neural correlates of master/item-176.html

Ramage et al.: Semantic–Phonological rsfMRI Connectivity in Aphasia 3065

Downloaded from: https://pubs.asha.org 157.51.53.89 on 02/25/2024, Terms of Use: https://pubs.asha.org/pubs/rights_and_permissions
Kiran, S., & Thompson, C. K. (2019). Neuroplasticity of language & Ballard, K. J. (2015). Altered resting-state network connec-
networks in aphasia: Advances, updates, and future challenges. tivity in stroke patients with and without apraxia of speech.
Frontiers in Neurology, 10, 1–15. https://doi.org/10.3389/fneur. NeuroImage: Clinical, 8, 429–439. https://doi.org/10.1016/j.nicl.
2019.00295 2015.03.013
Klingbeil, J., Wawrzyniak, M., Stockert, A., & Saur, D. (2017). Niziolek, C. A., Nagarajan, S. S., & Houde, J. F. (2013). What
Resting-state functional connectivity: An emerging method for does motor efference copy represent? Evidence from speech
the study of language networks in post-stroke aphasia. Brain production. Journal of Neuroscience, 33(41), 16110–16116.
and Cognition, 131, 22–33. https://doi.org/10.1016/j.bandc.2017. https://doi.org/10.1523/JNEUROSCI.2137-13.2013
08.005 Okada, K., Matchin, W., & Hickok, G. (2018). Phonological fea-
Kort, N. S., Nagarajan, S. S., & Houde, J. F. (2014). A bilateral ture repetition suppression in the left inferior frontal gyrus.
cortical network responds to pitch perturbations in speech Journal of Cognitive Neuroscience, 30(10), 1549–1557. https://
feedback. NeuroImage, 86, 525–535. https://doi.org/10.1016/ doi.org/10.1162/jocn_a_01287
j.neuroimage.2013.09.042 Parkinson, A. L., Behroozmand, R., Ibrahim, N., Korzyukov, O.,
Lancaster, J. L., Cykowski, M. D., McKay, D. R., Kochunov, P. V., Larson, C. R., & Robin, D. A. (2014). Effective connectivity
Fox, P. T., Rogers, W., Toga, A. W., Zilles, K., Amunts, K., & associated with auditory error detection in musicians with ab-
Mazziotta, J. (2010). Anatomical global spatial normalization. solute pitch. Frontiers in Neuroscience, 8, 46. https://doi.org/
Neuroinformatics, 8(3), 171–182. https://doi.org/10.1007/s12021- 10.3389/fnins.2014.00046
010-9074-x Poldrack, R. A. (2000). Imaging brain plasticity: Conceptual and
Lancaster, J. L., Woldorff, M. G., Parsons, L. M., Liotti, M., methodological issues—A theoretical review. NeuroImage, 12(1),
Freitas, C. S., Rainey, L., Kochunov, P. V., Nickerson, D., 1–13. https://doi.org/10.1006/nimg.2000.0596
Mikiten, S. A., & Fox, P. T. (2000). Automated Talairach Atlas Price, C. J. (2012). A review and synthesis of the first 20 years of
labels for functional brain mapping. Human Brain Mapping, PET and fMRI studies of heard speech, spoken language and
10(3), 120–131. https://doi.org/10.1002/1097-0193(200007) reading. NeuroImage, 62(2), 816–847. https://doi.org/10.1016/
10:3<120::AID-HBM30>3.0.CO;2-8 j.neuroimage.2012.04.062
Leonard, C., Rochon, E., & Laird, L. (2008). Treating naming im- Price, C. J., & Mummery, C. J. (1999). Delineating necessary and
pairments in aphasia: Findings from a phonological compo- sufficient neural system with functional imaging studies of
nents analysis treatment. Aphasiology, 22(9), 923–947. https:// neuropsychological patients. Journal of Cognitive Neuroscience,
doi.org/10.1080/02687030701831474 11(4), 371–382. https://doi.org/10.1162/089892999563481
Ludersdorfer, P., Price, C. J., Kawabata Duncan, K. J., DeDuck, K., Price, C. J., Mummery, C. J., Moore, C. J., Frakowiak, R. S., &
Neufeld, N. H., & Seghier, M. L. (2019). Dissociating the func- Friston, K. J. (1999). Delineating necessary and sufficient neural
tions of superior and inferior parts of the left ventral occipito- systems with functional imaging studies of neuropsychological
temporal cortex during visual word and object processing. patients. Journal of Cognitive Neuroscience, 11(4), 371–382.
NeuroImage, 199, 325–335. https://doi.org/10.1016/j.neuroimage. https://doi.org/10.1162/089892999563481
2019.06.003 Puig, J., Blasco, G., Alberich-Bayarri, A., Schlaug, G., Deco, G.,
Marcotte, K., Perlbarg, V., Marrelec, G., Benali, H., & Ansaldo, Biarnes, C., Navas-Martí, M., Rivero, M., Gich, J., Figueras, J.,
A. I. (2013). Default-mode network functional connectivity Torres, C., Daunis-I-Estadella, P., Oramas-Requejo, C. L.,
in aphasia: Therapy-induced neuroplasticity. Brain and Lan- Serena, J., Stinear, C. M., Kuceyeski, A., Soriano-Mas, C.,
guage, 124(1), 45–55. https://doi.org/10.1016/j.bandl.2012. Thomalla, G., Essig, M., . . . Pedraza, S. (2018). Resting-state
11.004 functional connectivity magnetic resonance imaging and out-
McCullagh, P., & Nelder, J. (1989). Generalized linear models come after acute stroke. Stroke, 49(10), 2353–2360. https://doi.
(2nd ed.). Chapman & Hall/CRC. org/10.1161/STROKEAHA.118.021319
McKinnon, E. T., Fridriksson, J., Glenn, G. R., Jensen, J. H., Helpern, Pustina, D., Coslett, H. B., Turkeltaub, P. E., Tustison, N., Schwartz,
J. A., Basilakos, A., Rorden, C., Shih, A. Y., Spampinato, M. V., M. F., & Avants, B. (2016). Automated segmentation of chronic
& Bonilha, L. (2017). Structural plasticity of the ventral stream stroke lesions using LINDA: Lesion identification with neigh-
and aphasia recovery. Annals of Neurology, 82(1), 147–151. borhood data analysis. Human Brain Mapping, 37(4), 1405–1421.
https://doi.org/10.1002/ana.24983 https://doi.org/10.1002/hbm.23110
McNeil, M. R., Robin, D. A., & Schmidt, R. A. (1997). Apraxia Pustina, D., Coslett, H. B., Ungar, L., Faseyitan, O. K., Medaglia,
of speech: Definition, differentiation, and treatment. Clinical J. D., Avants, B., & Schwartz, M. F. (2017). Enhanced estima-
Management of Sensorimotor Speech Disorders, 1, 311–344. tions of post-stroke aphasia severity using stacked multimodal
Meier, E. L., Kapse, K. J., & Kiran, S. (2016). The relationship predictions. Human Brain Mapping, 38(11), 5603–5615. https://
between frontotemporal effective connectivity during picture doi.org/10.1002/hbm.23752
naming, behavior, and preserved cortical tissue in chronic Raboyeau, G., Marcotte, K., Adrover-Roig, D., & Ansaldo, A. I.
aphasia. Frontiers in Human Neuroscience, 10(109), 1–23. (2010). Brain activation and lexical learning: The impact of
https://doi.org/10.3389/fnhum.2016.00109 learning phase and word type. NeuroImage, 49(3), 2850–2861.
Mozeiko, J., Myers, E. B., & Coelho, C. A. (2018). Treatment https://doi.org/10.1016/j.neuroimage.2009.10.007
response to a double administration of constraint-induced lan- Raven, J. C. (1978). Manual for Raven’s progressive matrices and
guage therapy in chronic aphasia. Journal of Speech, Language vocabulary scales. Western Psychological Services, H. K.
& Hearing Research, 61(7), 1664–1690. https://doi.org/10.1044/ Lewis. https://search.library.wisc.edu/catalog/999608145802121
2018_JSLHR-L-16-0102 Roach, A., Schwartz, M. F., Martin, N., Grewal, R. S., & Brecher,
Müller, S., Scealy, J. L., & Welsh, A. H. (2013). Model selection A. (1996). The Philadelphia Naming Test: Scoring and ratio-
in linear mixed models. Statistical Science, 28(2), 135–167. nale. Clinical Aphasiology, 24, 121–134. https://doi.org/10.1037/
https://doi.org/10.1214/12-STS410 t56477-000
New, A. B., Robin, D. A., Parkinson, A. L., Duffy, J. R., McNeil, Rosen, H. J., Petersen, S. E., Linenweber, M. R., Snyder, A. Z.,
M. R., Piguet, O., Hornberger, M., Price, C. J., Eickhoff, S. B., White, D. A., Chapman, L., Dromerick, A. W., Fiez, J. A.,

3066 Journal of Speech, Language, and Hearing Research • Vol. 63 • 3051–3067 • September 2020

Downloaded from: https://pubs.asha.org 157.51.53.89 on 02/25/2024, Terms of Use: https://pubs.asha.org/pubs/rights_and_permissions

 & Corbetta, M. D. (2000). Neural correlates of recovery from Sreedharan, S., Chandran, A., Yanamala, V. R., Sylaja, P., Kesavadas,
aphasia after damage to left inferior frontal cortex. Neurol- C., & Sitaram, R. (2019). Self-regulation of language areas using
ogy, 55(12), 1883–1894. https://doi.org/10.1212/WNL.55.12. real-time functional MRI in stroke patients with expressive
1883 aphasia. Brain Imaging and Behavior. https://doi.org/10.1007/
Saur, D., Lange, R., Baumgaertner, A., Schraknepper, V., Willmes, s11682-019-00106-7
K., Rijntjes, M., & Weiller, C. (2006). Dynamics of language Thompson-Schill, S., Barde, L. H. F., Chrysikou, E. G., Thompson-
reorganization after stroke. Brain: A Journal of Neurology, Schill, S. L., & Schwartz, M. F. (2010). Reduced short-term
129(Pt 6), 1371–1384. https://doi.org/10.1093/brain/awl090 memory span in aphasia and susceptibility to interference: Contri-
Schnur, T. T., Schwartz, M. F., Kimberg, D. Y., Hirshorn, E., bution of material-specific maintenance deficits. Neuropsycholo-
Coslett, H. B., & Thompson-Schill, S. L. (2009). Localizing gia, 48(4), 909–920. https://doi.org/10.1016/j.neuropsychologia.
interference during naming: Convergent neuroimaging and 2009.11.010
neuropsychological evidence for the function of Broca’s area. Thothathiri, M., Gagliardi, M., & Schwartz, M. F. (2012). Subdi-
Proceedings of the National Academy of Sciences, 106(1), vision of frontal cortex mechanisms for language production
322–327. https://doi.org/10.1073/pnas.0805874106 in aphasia. Neuropsychologia, 50(14), 3284–3294. https://doi.
Schwartz, M. F., Faseyitan, O., Kim, J., & Coslett, H. B. (2012). org/10.1016/j.neuropsychologia.2012.09.021
The dorsal stream contribution to phonological retrieval in Thothathiri, M., Kimberg, D. Y., & Schwartz, M. F. (2012). The
object naming. Brain: A Journal of Neurology, 135(Pt 12), neural basis of reversible sentence comprehension: Evidence
3799–3814. https://doi.org/10.1093/brain/aws300 from voxel-based lesion symptom mapping in aphasia. Jour-
Schwartz, M. F., Kimberg, D. Y., Walker, G. M., Faseyitan, O., nal of Cognitive Neuroscience, 24(1), 212–222. https://doi.org/
Brecher, A., Dell, G. S., & Coslett, H. B. (2009). Anterior tem- 10.1162/jocn_a_00118
poral involvement in semantic word retrieval: Voxel-based Thothathiri, M., Rattinger, M., & Trivedi, B. (2015). Cognitive
lesion–symptom mapping evidence from aphasia. Brain: A control during sentence generation. Cognitive Neuroscience,
Journal of Neurology, 132(Pt 12), 3411–3427. https://doi.org/ 8(1), 1–11. https://doi.org/10.1080/17588928.2015.1090421
10.1093/brain/awp284 Thothathiri, M., Schwartz, M. F., & Thompson-Schill, S. L.
Seghier, M. L., Fagan, E., & Price, C. J. (2010). Functional sub- (2010). Selection for position: The role of left ventrolateral
divisions in the left angular gyrus where the semantic system prefrontal cortex in sequencing language. Brain and Lan-
meets and diverges from the default network. Journal of Neuro- guage, 113(1), 28–38. https://doi.org/10.1016/j.bandl.2010.
science, 30(50), 16809–16817. https://doi.org/10.1523/JNEUROSCI. 01.002
3377-10.2010 Tzourio-Mazoyer, N., Joliot, M., Marie, D., & Mazoyer, B. (2015).
Seghier, M. L., Patel, E., Prejawa, S., Ramsden, S., Selmer, A., Variation in homotopic areas’ activity and inter-hemispheric
Lim, L., Browne, R., Rae, J., Haigh, Z., Ezekiel, D., Hope, intrinsic connectivity with type of language lateralization: An
T. M. H., Leff, A. P., & Price, C. J. (2015). The PLORAS fMRI study of covert sentence generation in 297 healthy volun-
database: A data repository for predicting language outcome teers. Brain Structure & Function, 221(5), 2735–2753. https://
and recovery after stroke. NeuroImage, 124(Part B), 1208–1212. doi.org/10.1007/s00429-015-1068-x
https://doi.org/10.1016/j.neuroimage.2015.03.083 Walenski, M., Europa, E., Caplan, D., & Thompson, C. K. (2019).
Seghier, M. L., Ramlackhansingh, A., Crinion, J., Leff, A. P., & Neural networks for sentence comprehension and production:
Price, C. J. (2008). Lesion identification using unified segmen- An ALE-based meta-analysis of neuroimaging studies. Human
tation-normalisation models and fuzzy clustering. NeuroImage, Brain Mapping, 40(8), 2275–2304. https://doi.org/10.1002/
41(4), 1253–1266. https://doi.org/10.1016/j.neuroimage.2008. hbm.24523
03.028 Wan, C. Y., Zheng, X., Marchina, S., Norton, A., & Schlaug, G.
Siegel, J. S., Ramsey, L. E., Snyder, A. Z., Metcalf, N. V., Chacko, (2014). Intensive therapy induces contralateral white matter
R. V., Weinberger, K., Baldassarre, A., Hacker, C. D., Shulman, changes in chronic stroke patients with Broca’s aphasia. Brain
G. L., & Corbetta, M. (2016). Disruptions of network connec- and Language, 136, 1–7. https://doi.org/10.1016/j.bandl.2014.
tivity predict impairment in multiple behavioral domains after 03.011
stroke. Proceedings of the National Academy of Sciences, 113(30), World Medical Association. (2013). World Medical Association
E4367–E4376. https://doi.org/10.1073/pnas.1521083113 Declaration of Helsinki: Ethical principles for medical research
Siegel, J. S., Shulman, G. L., & Corbetta, M. (2017). Measuring involving human subjects. JAMA, 310(20), 2191–2194. https://
functional connectivity in stroke: Approaches and consider- doi.org/10.1001/jama.2013.281053.
ations. Journal of Cerebral Blood Flow & Metabolism, 37(8), Xia, M., Wang, J., & He, Y. (2013). BrainNet Viewer: A network
2665–2678. https://doi.org/10.1177/0271678X17709198 visualization tool for human brain connectomics. PLOS ONE,
Silbert, L. J., Honey, C. J., Simony, E., Poeppel, D., & Hasson, 8(7), 1–15. https://doi.org/10.1371/journal.pone.0068910
U. (2014). Coupled neural systems underlie the production and Ylinen, S., Nora, A., Leminen, A., Hakala, T., Huotilainen, M.,
comprehension of naturalistic narrative speech. Proceedings Shtyrov, Y., Mäkelä, J. P., & Service, E. (2015). Two distinct
of the National Academy of Sciences, 111(43), E4687–E4696. auditory–motor circuits for monitoring speech production as
https://doi.org/10.1073/pnas.1323812111 revealed by content-specific suppression of auditory cortex.
Smith, S. M., Fox, P. T., Miller, K. L., Glahn, D. C., Fox, P. M., Cerebral Cortex, 25(6), 1576–1586. https://doi.org/10.1093/
Mackay, C. E., Filippini, N., Watkins, K. E., Toro, R., Laird, cercor/bht351
A. R., & Beckmann, C. F. (2009). Correspondence of the zu Eulenburg, P., Caspers, S., Roski, C., & Eickhoff, S. B. (2012).
brain’s functional architecture during activation and rest. Meta-analytical definition and functional connectivity of the
Proceedings of the National Academy of Sciences, 106(31), human vestibular cortex. Neuroimage, 60(1), 162–169. https://
13040–13045. https://doi.org/10.1073/pnas.0905267106 doi.org/10.1016/j.neuroimage.2011.12.032

Ramage et al.: Semantic–Phonological rsfMRI Connectivity in Aphasia 3067

Downloaded from: https://pubs.asha.org 157.51.53.89 on 02/25/2024, Terms of Use: https://pubs.asha.org/pubs/rights_and_permissions