NeuroImage 143 (2016) 15–25

Contents lists available at ScienceDirect

NeuroImage
journal homepage: www.elsevier.com/locate/neuroimage

Functional activity and white matter microstructure reveal

the independent effects of age of acquisition and proficiency
on second-language learning
Emily S. Nichols n, Marc F. Joanisse
The University of Western Ontario, Canada

art ic l e i nf o a b s t r a c t

Article history: Two key factors govern how bilingual speakers neurally maintain two languages: the speakers’ second
Received 22 May 2016 language age of acquisition (AoA) and their subsequent proficiency. However, the relative roles of these
Received in revised form two factors have been difficult to disentangle given that the two can be closely correlated, and most prior
23 August 2016
studies have examined the two factors in isolation. Here, we combine functional magnetic resonance
Accepted 24 August 2016
Available online 25 August 2016
imaging with diffusion tensor imaging to identify specific brain areas that are independently modulated
by AoA and proficiency in second language speakers. First-language Mandarin Chinese speakers who are
Keywords: second language speakers of English were scanned as they performed a picture-word matching task in
Bilingualism either language. In the same session we also acquired diffusion-weighted scans to assess white matter
fMRI
microstructure, along with behavioural measures of language proficiency prior to entering the scanner.
DTI
Results reveal gray- and white-matter networks involving both the left and right hemisphere that in-
Proficiency
Age of acquisition dependently vary as a function of a second-language speaker's AoA and proficiency, focused on the
superior temporal gyrus, middle and inferior frontal gyrus, parahippocampal gyrus, and the basal
ganglia. These results indicate that proficiency and AoA explain separate functional and structural net-
works in the bilingual brain, which we interpret as suggesting distinct types of plasticity for age-de-
pendent effects (i.e., AoA) versus experience and/or predisposition (i.e., proficiency).
& 2016 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).

1. Introduction than L1 learning. On this account L2 learning requires increased

Prior studies of bilingualism have identified ways in which the
neural representation of a second language (L2) differs from that of
an individual's first language (L1). Specifically, there are several
differences in activation between L2 and L1, both in terms of degree
and extent. L2s tend to not only show more activity within tradi-
tional left-hemisphere language areas, but also tend to activate more
regions outside the traditional language network (Chee et al., 2004;
Dehaene et al., 1997; Kim et al., 1997; Perani et al., 1998; Warten-
burger et al., 2003). Much of the current data come from studies
using fMRI to compare cortical activity, although an emerging lit-
erature also reveals differences between L2 and L1 in white matter
connectivity as examined with diffusion tensor imaging (DTI).
There are two predominant theories as to why neural signatures
of L2 differ from those of L1. The first is that these differences reflect
reduced neuroplasticity during L2 learning that occurs at a later age
n (simultaneous bilinguals) show higher white matter integrity in the
Correspondence to: The Brain and Mind Institute, Department of Psychology,
The University of Western Ontario, London, Ontario, Canada N6C 5B7.
E-mail address: enicho4@uwo.ca (E.S. Nichols).
neural resources due to maturational changes in neural plasticity
within regions and pathways supporting first language learning
(Abutalebi, 2008; Klein et al., 2013; Li et al., 2014; Mechelli et al.,
2004; Pakulak and Neville, 2011; Perani et al., 1996; Wartenburger
et al., 2003; Weber-Fox and Neville, 1996). Concordant with this
view, studies have found that age-of-acquisition (AoA) modulates
these effects; individuals who are early L2 learners show patterns of
brain activity that are more similar for L1 compared to late L2
learners (Perani and Abutalebi, 2005; Wartenburger et al., 2003).
Additionally, structural connectivity, as measured using DTI, appears
to vary as a function of AoA. A common measure of white matter
microstructure is fractional anisotropy (FA), which ranges from zero
to one and represents the cohesiveness of white matter tracts. High
FA suggests that water diffusion is restricted to a single direction and
thus the white matter tract is cohesive, while low FA suggests that
water diffusion is unrestricted and the tract is less cohesive. FA varies
with AoA such that children who learned two languages from birth
left inferior fronto-occipital fasciculus (IFOF), the tract connecting
anterior frontal regions with posterior temporal regions when

http://dx.doi.org/10.1016/j.neuroimage.2016.08.053
1053-8119/& 2016 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
16 E.S. Nichols, M.F. Joanisse / NeuroImage 143 (2016) 15–25
compared to children who learned their two languages sequentially
(Mohades et al., 2012). However, lower integrity was also found in
the tracts projecting from the anterior portion of the corpus callo-
sum to orbitofrontal cortex compared to late L2 learners. These re-
sults highlight how differences in brain connectivity may be related
to L2 AoA. Research suggests that there are separate L1 and L2
networks that are complementary in their importance as a function
of AoA (Mohades et al., 2012). The differing influence of AoA on
separate tracts may reflect their relative importance in L1 versus L2
processing.
The alternative theory is that neural differences in L1 versus L2
are instead driven by the fact that individuals' L2 is generally lower
in proficiency than their L1. On this account, second language
processing involves increased processing demands, and therefore
greater neural resources. While greater processing demands may
induce experience-based plasticity, these changes should be se-
parable from age-induced plasticity. Indeed, prior neuroimaging
studies have observed that higher-proficiency L2 users do show
patterns of neural activity that more closely resemble those of L1
users, compared to lower-proficiency L2 users, even when AoA is
controlled for (Chee et al., 1999; Newman et al., 2012; Pakulak and
Neville, 2011; Perani et al., 1998; Wartenburger et al., 2003).
While both AoA and proficiency explanations are appealing, it
is difficult to adjudicate among them given that the two are gen-
erally confounded: early L2 learners also tend to have higher
proficiency than late learners, making it difficult to tease apart the
relative contribution of both factors. Indeed, it could be argued
that the two are simply reflections of each other, and cannot be
disentangled. However, early AoA does not always mean high
proficiency, and high proficiency does not always mean early AoA.
Although it is well established that children acquire an L2 more
easily than adults, adults are still capable of becoming highly
proficient (Perani et al., 1998; Wartenburger et al., 2003), and
some children fail to fully acquire a second language in spite of
adequate opportunity to do so (Frost et al., 2013; Sahinkarakas,
2011). Factors such as motivation and environment also play a key
role in successful L2 acquisition (Bernaus and Gardner, 2008;
Gardner et al., 2004) unlike in L1, where failure to acquire lan-
guage is extremely rare outside situations involving severe sensory
deprivation (Fromkin et al., 1974). The present study addressed
individual differences in L2 learning success by examining neural
signatures of L2 processing in adult native speakers of Mandarin
who are second language English speakers. These speakers
showed significant variation in both AoA and proficiency, allowing
us to examine the two factors in parallel. Additionally, this ap-
proach allowed us to examine effects of L2 learning using a within-
subjects design, rather than creating (potentially arbitrary) groups
of high versus low proficiency and early versus late L2 learners.
We first used fMRI to isolate areas involved in L2 processing,
and examined how variation in patterns of activation can be ex-
plained either by AoA or proficiency. Next, we used DTI to identify
regions in which tract coherence correlated with either AoA or L2
proficiency, and which white matter tracts projected from these
regions. We identified networks of both functional and structural
organisation that were independently explained by proficiency
and AoA, suggesting the L2 speaker's brain organisation is not
wholly influenced by age-dependent learning, but is also suscep-
tible to L2 language ability levels.
2. Methods
2.1. Subjects
Twenty-two (17 female) neurologically healthy right-handed
native speakers of Mandarin were recruited via posters at the
Table 1
Subject demographics.
Measure Mean (SD) Min Max
Age (years) 22.18 (4.24) 18 35
Sex f 17
m 5
Years of schooling 14.8 (2.26) 12 20
Proficiency (% correct) L1 85.98 (6.34) 72.92 97.92
L2 68.28 (6.62) 54.17 77.08
Age of Acquisition 13.82 (7.12) 4 30
Duration of exposure to L2 (years) 8.36 (4.47) 0 15
Note: Proficiency score reflects percent correct on each language's proficiency test.
University of Western Ontario and from the London, Ontario
community. All subjects were English L2 speakers, aged 18–35
(M¼ 22.18, SD ¼4.24), and ranged in age of acquisition from 4–30
years (M¼13.82, SD ¼7.12). Two additional subjects were recruited
but excluded from analyses, one due to an incidental finding and
one due to chance-level performance on both the English and
Mandarin fMRI tasks. Additional subject demographics are pre-
sented in Table 1.
2.2. Behavioural materials
L1 Mandarin and L2 English proficiency levels were assessed
prior to scanning using a subset of 48 questions from the Hanyu
Shuiping Kaoshi (HSK Centre, Beijing, China) and the Test of Eng-
lish as a Second Language (ETS, Princeton, NJ) respectively, with
both tests consisting of three sections: Grammar, reading com-
prehension, and vocabulary. The three sections were combined to
give a final mark out of 48 for each language, representing overall
proficiency in these three domains. As the fMRI task was an au-
ditory lexical-semantic one, this general proficiency created from
not just lexical knowledge score helped avoid circularity in cor-
relating lexical knowledge with itself. Despite the written form of
the proficiency test and the auditory form of the fMRI task, lexical
knowledge is amodal (Coccia et al., 2004; Lambon Ralph and
Patterson, 2008; Patterson et al., 2007) and thus is not confounded
by the modality in which each task was administered.
AoA was defined as the age at which subjects first began
learning English. To verify right-handedness, subjects completed
an abridged version of the Edinburgh Handedness Inventory
(Oldfield, 1971). All behavioural measures were completed in
Mandarin aside from the English proficiency test. Letters of in-
formation, informed consent and task instructions were likewise
administered in Mandarin by a native speaker.
2.3. f MRI task
Subjects completed a picture-word matching task during
scanning. This task has been used extensively in the past to study
lexical knowledge (Breining et al., 2014; Dräger et al., 2004; We-
niger et al., 2000), and was chosen to examine lexical-semantic
processing. Because this task does not engage reading or syntax, it
provided an ideal task to investigate lexical-semantic processing.
Pictures appeared centred on a screen mounted at the head of
the scanner bore, which subjects viewed through a mirror placed
above the head coil. At the same time, a word was played binau-
rally through insert earphones (Sensimetrics Corporation, Malden,
MA). Both match and mismatch trials were presented in order to
maximise participants’ engagement during the task, and to en-
courage greater depth of processing than what might be expected
during passive listening or lexical decision. Subjects were required
to indicate as quickly as possible with a button press whether the
picture and word matched. Each picture was visible for 2.5 s.
 E.S. Nichols, M.F. Joanisse / NeuroImage 143 (2016) 15–25
Table 2
Contrasts of word measures between languages.
Measure Mean (SD) t p
English Mandarin
Frequency 213.83 (258.19) 129.69 (100.66)  1.92
Imageability 5.23 (0.81) 5.78 (.87)  2.89
Familiarity 5.48 (0.66) 5.78 (.62)  1.02
Picture/word match 6.03 (0.57) 6.21 (.57)  1.43
Stimulus presentation and response recording was controlled with
E-Prime software (Psychology Software Tools, Inc., Sharpsburg, PA)
and a Windows laptop.
The scanning session was divided into 8 runs of 20 trials each
for a total of 160 trials. A short break was provided between each
3.5-minute scanning run. Each image appeared twice during the
experiment, once in a matching pair and once in a semantically
unrelated mismatching pair. Subjects completed four English runs
alternating with four Mandarin runs, with starting language
counterbalanced. Each run began with an image reminding sub-
jects of which buttons to respond with, and subjects viewed a
fixation crosshair between trials. Subjects were also informed at
this time of the language in which the next run would be per-
formed. Each trial was 2.5 s in duration and presented at the onset
of an image acquisition, with inter-trial interval jittered between
2.5–12.5 s in 2.5 s increments, allowing us to more accurately de-
convolve the blood oxygen level dependent signal.
Stimulus word properties are described in Table 2. We selected
40 items in each language, which would each appear twice, once
in a matching pair and once in a mismatching pair. Items denoted
common single-word concepts in English and Mandarin, which
represented two-syllable words in both English and Mandarin, and
had frequencies greater than 40 per million in both languages
(English: CELEX Lexical Database, Baayen et al., 1995; Mandarin:
SUBTLEX-CH, Cai and Brysbaert, 2010). In a separate pilot study
involving different subjects we asked groups of native speakers of
English or Mandarin to rate the imageability and familiarity of the
stimulus words, as well as the correspondence of the pictures to
target words. As noted in Table 2, both groups showed equally
high ratings on all three sets of measures. Both groups of raters
showed highly similar ratings of familiarity and picture/word
correspondence, although imageability was higher for Mandarin
words than English words (t(39) ¼  2.89, p ¼.004), and frequency
approached significance. Although much effort was put into
matching languages on these measures, differences between lan-
guages should not affect interpretation of the results, as our focus
was on subject-wise variability in neural responses to L1 versus L2
rather than how neural responses to the two lists differ on a
whole-group basis.
2.4. fMRI acquisition and processing
Imaging was conducted on a Siemens Magnetom TIM Trio
whole-body 3 T scanner with a 32-channel head coil. T2*-weighted
functional scans were acquired in a transverse plane with 45 slices
per volume (TR¼ 2.5 s; TE¼38 ms; flip angle¼80°; FOV¼
192  192 mm; voxel size 3 mm3) using an iPAT parallel acquisition
sequence (generalised auto-calibrating partially parallel acquisition
[GRAPPA]; acceleration factor¼ 2), providing full coverage of the
cerebrum with partial loss of the cerebellum. A total of 576 func-
tional scans were acquired for each participant over 8 runs (3.5 min
per run). To ensure stability of the MR signal four functional volumes
that were run but then discarded prior to the start of image acqui-
sition. After the final functional run, a whole-head high-resolution
3D anatomical scan was acquired within the sagittal plane, using a
17
3D pulse sequence weighted for T1 contrast (MPRAGE; TR¼ 2.3 s;
TE¼2.98 ms; FOV¼256  256 mm; voxel size¼1 mm3; 176 slices;
GRAPPA acceleration factor¼2).
Data analysis was performed using the AFNI software package
(Cox, 1996). Functional scans were first de-obliqued (transformed
to the same cardinal orientation as the anatomical scan, AFNI
.058 3dWarp), then motion corrected by registering each volume to the
.004
.310
last functional volume of the session, which was acquired im-
.157 mediately preceding the anatomical scan, using a 3d rigid body
transformation (AFNI 3dvolreg). Outlier volumes caused by hard-
ware artifacts were identified as ones significantly deviating from
average image intensity using AFNI 3dToutcount using a threshold
of 10,000 outlier voxels, and subsequently removed from statis-
tical analyses using the CENSORTR option in AFNI 3dDeconvolve. A
single motion parameter was created by averaging the absolute
values of six motion parameters (roll, pitch, yaw, superior, left,
posterior directions), and this was included in the subject-level
statistical model as a variable of no interest.
Single-subject statistical maps were formed using a general
linear model (GLM) with the following six predictor functions: one
each for matching and mismatch trials of English runs, one each
for matching and mismatch trials of Mandarin runs, one re-
presenting reaction times and one representing motion. Only
correct trials were included in the analysis, with accuracy ranging
from 71.25% to 100% correct on the Mandarin task, and ranging
from 82.5% to 100% correct on the English task. Subjectwise sta-
tistical maps were then resampled to 1 mm3, and transformed to
stereotaxic space of Talairach and Tournoux (1988) via an auto-
matic registration procedure (least-squares cost function) and a
standard atlas brain. A 5 mm FWHM Gaussian spatial filter was
then applied (AFNI 3dmerge). Finally, a brain mask was calculated
and applied to each subject's GLM using an automatic procedure
(AFNI 3dSkullStrip). Paired samples t-statistical maps were ob-
tained for differences in activation of L1 and L2 trials, in order to
identify regions that were differentially involved in the processing
of each language. Subsequent analyses examined the association
between subjectwise regional brain activity and behavioural
measures obtained prior to scanning. First, difference in L1/L2
brain activity, using an L2 versus L1 contrast, was calculated.
Multiple linear regression was then performed on this subtraction
(AFNI 3dRegAna), using AoA and proficiency as independent pre-
dictor variables. This identified clusters of voxels in which the
difference between L1 and L2 activation was predicted by AoA
independent of L2 proficiency scores, and by L2 proficiency scores
independent of AoA. Specifically, this analysis revealed areas pre-
dicted by one behavioural variable while holding constant the
effects of the other variable. Region of Interest (ROI) masks were
then drawn using these areas, and by-subject beta values for the
L2 versus L1 contrast were extracted and plotted.
For contrasts, correction for multiple comparisons at po.001 was
achieved by setting a minimum cluster size of 195 voxels, obtained
using a 10,000 iteration Monte Carlo simulation (AFNI 3dClustSim) at
a voxelwise alpha level of po.001. Minimum cluster size after cor-
rection for multiple comparisons at po.05 for the multiple regres-
sion analyses was achieved by setting a minimum cluster size of
1318 voxels with a voxelwise alpha level of po.01.
2.5. DTI acquisition and processing
Diffusion-weighted images were acquired following the ana-
tomical scan. Images were acquired in the axial plane using an EPI
sequence (68 slices with 2 mm slice thickness, voxel size ¼
2.083  2.083 mm in-plane, matrix¼ 96  96  68, field of view¼
200 mm2, 64 diffusion directions with b ¼1000 s/mm2, TR ¼7.6 s,
TE ¼79 ms; GRAPPA acceleration factor¼3).
DTI scans were processed using the AFNI and FSL (FMRIB
18 E.S. Nichols, M.F. Joanisse / NeuroImage 143 (2016) 15–25

Software Library, FMRIB, Oxford, UK; Smith et al., 2004) software
packages. First, eddy current correction (eddy_correct in FSL) and
motion correction (AFNI 3dvolreg) were performed. Next, voxel-
wise statistical analysis of the fractional anisotropy (FA) data was
carried out using FSL TBSS (Tract-Based Spatial Statistics; Smith
et al., 2006). FA images were created by fitting a tensor model to
the raw diffusion (FDT in FSL), and then brain-extracted using (BET
in FSL; Smith, 2002). All subjects' FA data were then aligned into a
common space using the FSL nonlinear registration tool FNIRT
(Andersson et al., 2010), which uses a b-spline representation of
the registration warp field (Rueckert et al., 1999). Next, the mean
FA image was skeletonised with a threshold of 0.2, which re-
presents the centres of all tracts common to the group, and each
subject's aligned FA data was then projected onto this skeleton.
This skeletonised data was submitted to multiple linear regression
(AFNI 3dRegAna), again using AoA and proficiency as independent
predictor variables. This identified clusters of voxels in which the
FA was predicted by AoA independent of L2 proficiency scores, and
by L2 proficiency scores independent of AoA. Minimum cluster
size after correction for multiple comparisons at p o .001 for the
multiple linear regression analyses was achieved by setting a
minimum cluster size of 8 voxels with a voxelwise alpha level of
p o.01, again using a 10,000 iteration Monte Carlo simulation
(AFNI 3dClustSim). (Note the minimum cluster size is much smaller
for DTI than f MRI analyses. This is due to the use of a much
smaller search space in the DTI analyses, which represented a
white matter skeleton versus a spatially-smoothed whole-brain in
the f MRI analyses). For each significant cluster, probabilistic trac-
tography was performed subject-wise in native space using BED-
POSTX and FDT diffusion, with ROI masks used as waypoints. The
resulting tractography was then transformed to Talairach space
and averaged to form a thresholded map. To aid visualisation,
2 mm Gaussian blurring was applied to ROI masks before being
Table 3
Clusters of significant activation in L2–L1 contrast at a voxel-wise alpha of p o .001.
Contrast Region Talairach coordinates
x y
L1 Mandarin4 L2 R Middle fron- 34 35
English tal gyrus
R Paracentral 2  41
lobule
L2 English 4L1 L Superior  46  29
Mandarin temporal
gyrus
R Superior 51  15
temporal
gyrus
Note. Coordinates denote the location of peak activation. L/R¼ left/right.
overlaying on a stereotaxic brain atlas. Note that tractography is
used here for illustrative purposes, showing how significant FA
clusters relate to long-distance tracts; thus only sub-regions of
these tracts actually showed FA values that were significantly re-
lated by our behavioural measures.
3. Results
3.1. Behavioural results
Analyses of the behavioural data acquired prior to scanning
confirmed L2 proficiency was significantly weaker than L1 profi-
ciency (t(21)¼3.24, p¼ .003). We also observed a relative decoupling
of AoA and proficiency measures, marked by a null correlation in the
two measures (r¼.02, p¼.940). Corresponding analyses using age of
arrival in North America or number of years since arrival yielded
similarly null effects, suggesting that proficiency and AoA were
generally unrelated in our sample.
For the picture-word matching task performed during f MRI
scanning, there was no significant difference between Mandarin
and English trials in terms of RT (M¼ 1.27 s, SD ¼.32, M¼1.28 s,
SD ¼.62, respectively; t(21) ¼0.36, p o.941) or accuracy (M ¼94%,
SD ¼8.79, M¼95%, SD ¼4.03; t(21) ¼ .61, p ¼.517).
3.2. f MRI results
We first identified regions showing differences in activation for
L1 (Mandarin) versus L2 (English) trials. As listed in Table 3, results
revealed greater activity in the right middle frontal gyrus and right
paracentral lobule in L1 compared to L2, and greater activity in
bilateral superior temporal gyri (STG) in L2 compared to L1.
Multiple regression analysis was next performed to identify
how AoA and L2 proficiency predicted differences in activation for
L2 versus L1 trials. Note these regions might not overlap with
those identified in the direct L2 versus L1 contrast above because
Size
the degree of difference could differ across subjects as a function
(mm3) of either the AoA or proficiency variables. Indeed, as shown in
z Table 4, AoA uniquely predicted activity differences in seven areas,
notably the right parahippocampal gyrus, bilateral STG, and the
26 464
bilateral inferior frontal gyrus (Fig. 1A). Conversely, proficiency
50 331 uniquely predicted activity in the left parahippocampal gyrus and
the right cingulate gyrus (Table 4 and Fig. 1B). The relationship
9 5499 between AoA and key regions identified by the above analyses is
plotted in Fig. 2A; for all regions except the bilateral STG, early L2
7 5005 learners showed greater activation when performing the task in L1
compared to when performing the task in L2. As AoA increased,
the direction of the difference in activation reversed, and late L2
learners showed greater activation when performing the task in L2
compared to when performing the task in L1. Interestingly, in the

Table 4
Clusters of significant activation in multiple regression of AoA and proficiency in L2 English when compared to L1 Mandarin.

Predictor Region Talairach coordinates Size (mm3)

x y z

Age of acquisition L Superior temporal gyrus  37  60 18 4875

R Parahippocampal gyrus  64  14 8 3711
R Superior temporal gyrus 57  13 5 3097
L Uncus 45  54 15 2735
L Inferior frontal gyrus  30 37 6 2420
R Inferior frontal gyrus 28 3  10 1808
R Middle occipital gyrus  24 4 9 1325

Proficiency L Parahippocampal gyrus 41  13 9 1605

R Cingulate gyrus 52  38 19 1498
 E.S. Nichols, M.F. Joanisse / NeuroImage 143 (2016) 15–25 19

bilateral STG the difference in activation between L2 and L1 was

Fig. 1. Multiple regression of the difference in activation between the L2 English
versus L1 Mandarin tasks as a function of either (A) AoA or (B) L2 proficiency.
Statistical maps are thresholded at F ¼4.38, p¼ .05, overlaid on a stereotaxic brain
atlas, L ¼ left. Cluster sizes are reported in Table 4.
largest at early AoAs, with L1 activating more than L2, and ap-
proached zero with increasing AoA.
The relationship between proficiency and key regions identified
by the above analyses is plotted in Fig. 2B. For both areas, low pro-
ficiency L2 speakers showed greater activation when performing the
task in L1 compared to when performing the task in L2. However, as
proficiency increased, the reverse pattern appeared, with high pro-
ficiency L2 speakers showing greater activation when performing
the task in L2 than when performing the task in L1.
It is important to note that when examining proficiency, raw L2
proficiency score was used as a regressor, as opposed to the dif-
ference between L1 and L2 proficiency. However, L2 proficiency
and the L1–L2 proficiency difference are tightly correlated
(r ¼  .81, p o.001). Due to the nature of the values of the differ-
ence between L1 and L2 proficiency, a regression model using the
difference as a regressor was unable to converge, however plotting
the L1–L2 proficiency difference against betas extracted from the
left parahippocampal gyrus and right cingulate confirmed that
smaller differences between L1 and L2 proficiency, i.e., higher L2
proficiency, corresponded to greater activation levels for L2 versus
L1 in both areas.

Fig. 2. Scatterplots indicating A) The relationship between AoA, controlling for proficiency, and beta values for the difference in activation between the L2 English and L1
Mandarin tasks (L2 versus L1 contrast), in each of the regions predicted by AoA. B) The relationship between proficiency, controlling for AoA, and beta values for the L2
versus L1 contrast, in both regions predicted by proficiency. Positive beta values reflects L2 4L1, and negative beta values reflects L14 L2.
20 E.S. Nichols, M.F. Joanisse / NeuroImage 143 (2016) 15–25
3.3. DTI results
To examine how white matter integrity is related to L2 beha-
vioural measures, we performed multiple regression on the skele-
tonised FA data, allowing us to identify regions of white matter
whose FA was uniquely predicted by AoA or proficiency. Results of
the multiple regression are shown in Table 5 and Fig. 3. AoA was
positively associated with FA in four sections of tracts: two in the left
inferior longitudinal fasciculus (ILF), one in the fibres emerging from
the anterior midbody of the corpus callosum, and one in the arcuate
fasciculus. Proficiency was also positively associated with FA in three
sections of tracts: one in the right ILF, one in the right arcuate fas-
ciculus, and one in the forceps minor of the corpus callosum.
We next used probabilistic tractography to compute volumes of
white matter fibres extending from each ROI identified in the above
regression analysis. The group averages of the computed volumes for
each ROI are shown in Fig. 4. Two separate networks of white matter
were identified, with AoA being associated with the bilateral ILF, left
corpus callosum, and left arcuate fasciculus. This network was
composed of complimentary right and left hemisphere tracts, with
connections between hemispheres. In contrast, proficiency showed
an extensive right-lateralised network composed of the inferior
longitudinal fasciculus, arcuate fasciculus, and forceps minor, with
limited projections to the left hemisphere.
4. Discussion
Previous studies disagree as to whether the neural substrates of
first- versus second-language processing differ due to age of ac-
quisition (AoA) or proficiency. In the past, both factors have typi-
cally been studied independently of one another. Likewise, both
have typically been treated as discrete rather than continuous
variables. As such, there has been much inconsistency in the lit-
erature with respect to the cutoff points for early versus late age of
acquisition, and to low and high proficiency. Thus, results tend to
be difficult to interpret across studies.
The present study sought to isolate the contribution of both
variables to neural function and white matter microstructure
supporting second-language English in native speakers of Man-
darin. Of note, individuals in our sample showed a relative de-
coupling of the two variables of interest such that their AoA did
not significantly correlate with English proficiency level. This is
likely due to how English as a second language is taught in
mainland China, which provides minimal opportunity to use
English outside of the educational curriculum. Indeed, the popu-
lation studied here was not fully immersed in their L2 from the
time of first exposure (and perhaps not ever) and consequently
their proficiency outcomes varied significantly compared to po-
pulations where this is the case. Thus while many of these
Table 5
Clusters of white matter in which FA was predicted by AoA and proficiency in Mandarin–English L2 speakers.
Predictor Region
Age of acquisition L Corpus callosum
L Arcuate fasciculus
L Inferior longitudinal fasciculus
R Inferior longitudinal fasciculus
Proficiency R Inferior longitudinal fasciculus
R Arcuate fasciculus
R Forceps minor
individuals began learning English at an early age, the level of
proficiency achieved by those individuals was not uniformly high.
If the relationship were stronger, the independent contribution of
either variable would have been statistically harder to identify. The
weak relationship between the two variables provided us with the
opportunity to examine proficiency and AoA variables in a way
that was not confounded.
During f MRI scanning, individuals performed a lexical proces-
sing task in either Mandarin (L1) or English (L2). Groupwise ac-
tivity was significantly greater in L1 than in L2 in the right middle
frontal gyrus (MFG) and the right paracentral lobule. The right
MFG has been implicated in a number of processes, including re-
cognition memory (Ranganath et al., 2003; Rugg et al., 1996),
which may be higher in L1 than L2. The paracentral lobule is a key
part of the default mode network (Margulies et al., 2009), which
may be more engaged during a language task in which the subject
is more comfortable. Conversely, groupwise activity was sig-
nificantly greater in L2 than in L1 in both the left and right su-
perior temporal gyri, in line with previous studies of spoken word
processing in a second language (Dehaene et al., 1997; Kim et al.,
1997; Perani et al., 1998). However, due to the large amount of
variability in our sample, L2–L1 contrasts should be interpreted
with caution.
We next examined individual differences in second language
learning by performing a linear regression analysis on subjectwise
differences in L2 versus L1 activation in which we tested the in-
dependent effects of AoA and proficiency on neural activation. This
approach identified multiple brain regions that were independently
modulated by either AoA or L2 proficiency. We found that AoA un-
iquely predicted the degree of L2 involvement in the classic lan-
guage-processing network, namely the bilateral STG and IFG, such
that L2 activity relative to L1 activity in these regions increased as a
function of later AoA. Both the STG and IFG are known to be involved
in speech comprehension (Friederici et al., 2003; Grindrod et al.,
2008; Myers et al., 2009; Obleser et al., 2006; Scott and Johnsrude,
2003; Warren et al., 2006), with degree of activity in both areas
found to increase as a function of processing demands (Fridriksson
and Morrow, 2005; Noppeney et al., 2004; Obleser et al., 2007;
Sharp et al., 2010). Additionally, our results support the finding of
increased bilateral activation in L2 versus L1 (Centeno et al., 2014;
Park et al., 2012). This was demonstrated by the positive relationship
between AoA and activation in all areas identified in the regression,
suggesting that as AoA increases, there is additional recruitment of
right-hemisphere areas in addition to the typical left-hemisphere
areas. Thus, the increased activity observed in the present study may
reflect more effortful processing within basic speech perception
mechanisms as AoA increases.
AoA also modulated activity in the right parahippocampal gyrus.
The left parahippocampal gyrus has been implicated in semantic
memory and retrieval (Binder et al., 1997), and it is possible that
Talairach coordinates Size (mm3)
x y z
 12 20 49 21
 38 8 36 18
 34  63 25 9
28  77 30 9
44  24  15 10
8  13 13 10
36 48 1 8
 E.S. Nichols, M.F. Joanisse / NeuroImage 143 (2016) 15–25 21

Fig. 3. Multiple regression of the skeletonised FA values as a function of either (A) AoA or (B) L2 proficiency. Regions explaining significant variance are shown in red. To aid
visualisation, ROI masks were smoothed to 2 mm before being overlaid on the averaged FA skeleton. Statistical maps are thresholded at F¼ 14.92, p¼ .001 (corrected),
overlaid on a stereotaxic brain atlas, L¼ left. Cluster sizes are reported in Table 5.

AoA-related maturational constraints lead to increased recruitment
of right parahippocampal areas for semantic retrieval of L2. Indeed,
the right parahippocampal gyrus has been implicated in L2 proces-
sing, and found to be present in high proficiency late L2 learners
while processing their L2 but not their L1 (Perani et al., 1998).
The positive relationship between AoA and level of activation in
the language-processing network also supports previous research
showing increased activation in late learners (Isel et al., 2010; Kim
et al., 1997; Hernandez and Li, 2007; Wartenburger et al., 2003).
FMRI effects that are attributed to AoA could reflect proficiency
which is often confounded with AoA. Because proficiency was
explicitly controlled for in our analyses of AoA, we argue that ac-
tivation differences predicted by AoA reflect age-dependent effects
of plasticity for spoken word processing, rather than simply an
index of subsequent proficiency differences between early and late
L2 learners.
Our design also allowed us to similarly examine the influence
of proficiency independent of AoA, yielding a complementary
pattern of results in the left parahippocampal gyrus and right
cingulate gyrus. At lower proficiencies, activity in L2 was less than
activity in L1; however, as proficiency increased, activity in L2
became greater than activity in L1. One interpretation is that
lower-proficiency L2 speakers will tend to rely less on brain re-
gions involved L1 language processing. Indeed, Chee et al. (2004)
found increased left insula activation in high proficiency L2
speakers compared to low proficiency speakers, even though both
groups performed equally on a phonological working memory
task. Thus, greater engagement of language-related areas may
22 E.S. Nichols, M.F. Joanisse / NeuroImage 143 (2016) 15–25

Fig. 4. Probabilistic tractography based on white matter seeds identified by multiple regression (see Fig. 3 and Table 5), representing the independent contribution of AoA
(A) and L2 proficiency (B). Tractography is overlaid on a stereotaxic brain atlas, L ¼ left.

support enhanced, rather than more effortful, L2 processing, al-
lowing individuals to become more proficient (Callan et al., 2003,
2004). Although it is difficult to determine whether greater ac-
tivity reflects effort or effectiveness, the positive correlation be-
tween proficiency and L2–L1 difference shown in Fig. 2B supports
this conclusion, showing that increasing proficiency predicts
greater L2 activation than L1 activation in both the right cingulate
gyrus and the left parahippocampal gyrus.
Much less is known about how L2 ability is reflected in the
microstructure of white matter tracts supporting language. While
some studies have observed differences in white matter for
monolinguals versus bilinguals (Coggins et al., 2004; Cummine
and Boliek, 2013; Luk et al., 2011; Mohades et al., 2012), it remains
an open question whether these differences are influenced by the
age at which a L2 was learned, or the relative proficiency that an
individual has achieved in their dominant versus second language.
To address this, we located regions of white matter where the FA
values were independently predicted by AoA or proficiency. We
also computed white matter tracts extending from these ROIs as
an indication of the structural pathways involved in second lan-
guage processing. The results revealed that FA was positively
correlated with AoA in one section of the left corpus callosum,
which has been previously found to be larger in bilinguals than
monolinguals (Coggins et al., 2004), although Mohades et al.
(2012) failed to find differences in FA in this tract. FA was also
positively correlated with AoA in the left arcuate fasciculus, which
contains bidirectional connections between IFG and STG (i.e.,
Broca's area and Wernicke's area), and in the bilateral ILF, which
connects occipital regions of cortex to inferior temporal regions of
cortex through ventral pathways. This supports the present f MRI
results, which found increased activation with increasing AoA that
included the left IFG and STG, demonstrating that both local
 E.S. Nichols, M.F. Joanisse / NeuroImage 143 (2016) 15–25
activity and between-region connectivity are involved in L2 ac-
quisition and processing. The present results also support the
presence of a dual route in language connectivity containing both
dorsal and ventral regions (Duffau et al., 2005; Frey et al., 2008),
demonstrating that these pathways are affected by the age at
which a second language is learned.
FA positively correlated with proficiency in sections of the right
ILF, arcuate fasciculus, and forceps minor of the corpus callosum,
consistent with f MRI findings of greater right-hemisphere in-
volvement linked to L2 proficiency (Chee et al., 2001; Dehaene
et al., 1997; Wang et al., 2011). All three tracts have previously
been tied to bilingualism (Mohades et al., 2012; Luk et al., 2011),
although results have been variable. While a cross-sectional study
of bilingual and monolingual children found higher FA in the
anterior portion of the corpus callosum in monolinguals (Mohades
et al., 2012), these group differences were no longer present two
years later (Mohades et al., 2015). In contrast, Luk et al. (2011)
found higher FA in the corpus callosum in lifelong bilinguals
compared to monolinguals. Similarly, Mohades et al. (2012), (2015)
found no differences between monolinguals and bilinguals in the
arcuate fasciculus, while Luk and colleagues found increased FA in
both the SLF and ILF of lifelong bilinguals relative to monolinguals.
Although these studies did not examine proficiency and AoA, the
present study suggests that differences in white matter micro-
structure between bilinguals and monolinguals may be in part due
to both age effects and proficiency levels in L2.
We propose that these positive correlations reflect the differing
and independent contributions of AoA and proficiency to L2 pro-
cessing. The most intuitive finding is of a positive relationship
between proficiency and tract FA, such that higher proficiency in a
second language promotes higher white matter integrity among
brain regions supporting second-language learning. In contrast,
the positive relationship between AoA and FA is perhaps less in-
tuitive, however it suggests that older learners rely more heavily
on the bilateral ILF, left corpus callosum, and left arcuate fasciculus
to compensate for a late AoA. Indeed, higher FA values have been
found in the corpus callosum of bilingual individuals who acquired
their two languages sequentially relative to simultaneously (Mo-
hades et al., 2012). This positive relationship between AoA and FA
is in agreement with the present f MRI findings in which later AoA
led to an increase in activity, again suggesting increased effort or
usage of areas related to L2 processing.
Probabilistic tractography allowed us to determine the network
of tracts supporting the bilingual brain. Two separate sets white
matter tracts emerged, with those underlying AoA showing com-
plimentary tracts extending anterior-posterior in each hemisphere
and connecting inter-hemispherically. In contrast, areas of FA that
correlated with proficiency were right-lateralised, thus the net-
work revealed by probabilistic tractography was also generally
right-lateralised, although the tracks also projected contralaterally
through anterior portions of the corpus callosum. We propose that
the maturational constraints reflected here by AoA modulate
whole-brain white matter connections involved in L2 acquisition
and processing, while proficiency instead primarily modulates
connections in the right hemisphere. This is supported by previous
research showing that AoA predicts more widespread modulations
in the brain than proficiency, while proficiency tends to result in
the recruitment of right-hemisphere areas (Chee et al., 2001; De-
haene et al., 1997; Indefrey, 2006; Kim et al., 1997; Wang et al.,
2011).
The results help to inform present theories of why neural sig-
natures of L2 differ from those of L1. We identified a network of
regions in which differences in L1–L2 processing are predicted by
AoA independently of proficiency. However, we also identified two
regions, the right cingulate gyrus and the left parahippocampal
gyrus, whose L1–L2 differences were predicted by proficiency
23
independent of AoA. These results indicate that neuroanatomical
models of L2 processing must account for both AoA and profi-
ciency, with age-dependent and experience-dependent plasticity
affecting the L2 processing network and driving differences be-
tween L1 and L2.
The present study currently leaves open the nature of the re-
lationship between proficiency and white matter integrity. While
we only identified tract regions showing a positive association
between FA and proficiency, the direction of causation remains
unclear. The association may relate to the reinforcement of white
matter tracts as a result of more frequent practice and/or greater
L2 exposure – that is, all individuals might possess a similar initial
state of white matter coherence, but those who are generally
predisposed to being better second language learners will subse-
quently develop more coherent white matter tracts over time.
Alternatively, it might be that higher proficiency is itself a direct
consequence of superior pre-existing white matter integrity in
areas required for learning an L2, such that higher white matter
integrity individuals are predisposed to become more successful
L2 learners. We expect that longitudinal designs that better ac-
count for the independent contributions of AoA and proficiency
will shed more light on this distinction.
One potential concern about our design is also worth noting
here: because the two languages were presented in interleaved
alternating runs during scanning, it is possible that the differences
in proficiency are inflated due to language switching. For instance
individuals with greater language proficiency might also have had
greater facility in alternating between English and Mandarin from
one run to the next, thereby influencing the results. However, this
was not a rapid language-switching paradigm, rather each run
consisted of 3.5 min of trials in a single language. Likewise, sub-
jects were given breaks between languages and informed of which
language was upcoming. Based on this, we argue that the alter-
nating runs paradigm was likely not problematic, although we
cannot rule out its influence on the results. In addition, it should
be noted that due to the limited nature of the f MRI task, the
functional conclusions cannot be extended beyond the current
paradigm.
4.1. Conclusions
The goal of the present study was to examine both functional
and structural correlates of second language processing as a
function of age of acquisition and proficiency. Two separate net-
works of brain regions and white matter tracts predicted by these
two variables emerged, confirming that both play a role in how the
brain supports two languages.
These findings help to integrate previous results concerning the
functional and anatomical correlates of L2 learning. This is the first
study to map how AoA and proficiency independently explain
neural differences in second language. Of note, treating both AoA
and proficiency as continuous variables also overcomes the con-
cern with experimental designs that attempt to divide individuals
into binary groups such as ‘early’ versus ‘late’ L2 learners, and
‘high’ versus ‘low’ proficiencies. Inconsistencies in the literature
regarding where to draw the divide between early and late AoA
and high and low proficiency have made it difficult to compare
between studies. The current approach provides a clearer de-
scription as to how the brain is affected by these two variables.
This is also the first study we know of to merge measures of
grey and white matter organisation to address this important
question. Additionally, there are few studies using DTI to compare
monolinguals and bilinguals, and the results appear to be con-
flicting, in that they have yielded both positive and negative as-
sociations between white matter FA and L2 ability. The results of
the current study indicate that both functional activity and white
24 E.S. Nichols, M.F. Joanisse / NeuroImage 143 (2016) 15–25
matter integrity play a role in L2 learning, that the association
might be either positive or negative depending on the brain region
in question, and that ultimate success in L2 learning hinges both
on age-dependent and age-independent factors.
Funding
This work was supported by a Discovery Grant and Accelerator
Award from the Natural Sciences and Engineering Research Council.
Acknowledgements
We are grateful to Yi Xie for providing help with Mandarin
translations.
Appendix A. Supporting information
Supplementary data associated with this article can be found in the
online version at http://dx.doi.org/10.1016/j.neuroimage.2016.08.053.
References
Abutalebi, J., 2008. Neural aspects of second language representation and language
control. Acta Psychol. 128, 466–478. http://dx.doi.org/10.1016/j.
actpsy.2008.03.014.
Andersson, J.L.R., Jenkinson, M., Smith, S., 2010. Non-linear registration, aka spatial
normalisation. FMRIB Technical Report TR07JA2.
Baayen, R., Piepenbrock, R., Gulikers, L., 1995. CELEX2 LDC96L14. Web Download.
Linguistic Data Consortium, Philadelphia.
Bernaus, M., Gardner, R.C., 2008. Teacher motivation strategies, student percep-
tions, student motivation, and english achievement. Mod. Lang. J. 92, 388–401.
http://dx.doi.org/10.1111/j.1540-4781.2008.00753.x.
Binder, J.R., Frost, J.A., Hammeke, T.A., Cox, R.W., Rao, S.M., Prieto, T., 1997. Human
brain language areas identified by functional magnetic resonance imaging. J.
Neurosci. 17, 353–362.
Breining, B.L., Faria, A.V., Hillis, A.E., 2014. Neural regions supporting lexical pro-
cessing of objects and actions: a case series analysis. In: Conference Abstract:
Academy of Aphasia – 52nd Annual Meeting. Front. Psychol. http://dx.doi.org/
10.3389/conf.fpsyg.2014.64.00030.
Cai, Q., Brysbaert, M., 2010. SUBTLEX-CH: Chinese word and character frequencies
based on film subtitles. PLoS One 5, e10729. http://dx.doi.org/10.1371/journal.
pone.0010729.
Callan, D.E., Jones, J.A., Callan, A.M., Akahane-Yamada, R., 2004. Phonetic perceptual
identification by native- and second-language speakers differentially activates
brain regions involved with acoustic phonetic processing and those involved
with articulatory-auditory/orosensory internal models. Neuroimage 22,
1182–1194. http://dx.doi.org/10.1016/j.neuroimage.2004.03.006.
Callan, D.E., Tajima, K., Callan, A.M., Kubo, R., Masaki, S., Akahane-Yamada, R., 2003.
Learning-induced neural plasticity associated with improved identification
performance after training of a difficult second-language phonetic contrast.
Neuroimage 19, 113–124. http://dx.doi.org/10.1016/S1053-8119(03)00020-X.
Centeno, M., Koepp, M.J., Vollmar, C., Stretton, J., Sidhu, M., Michallef, C., Symms, M.
R., Thompson, P.J., Duncan, J.S., 2014. Language dominance assessment in a
bilingual population: validity of f MRI in the second language. Epilepsia 55,
1504–1511. http://dx.doi.org/10.1111/epi.12757.
Chee, M.W., Hon, N., Lee, H.L., Soon, C.S., 2001. Relative language proficiency
modulates BOLD signal change when bilinguals perform semantic judgments.
Neuroimage 13, 1155–1163. http://dx.doi.org/10.1006/nimg.2001.0781.
Chee, M.W., Tan, E.W., Thiel, T., 1999. Mandarin and English single word processing
studied with functional magnetic resonance imaging. J. Neurosci. 19,
3050–3056.
Chee, M.W.L., Soon, C.S., Lee, H.L., Pallier, C., 2004. Left insula activation: a marker
for language attainment in bilinguals. Proc. Natl. Acad. Sci. USA 101,
15265–15270. http://dx.doi.org/10.1073/pnas.0403703101.
Coccia, M., Bartolini, M., Luzzi, S., Provinciali, L., Ralph, M. a L., 2004. Semantic
memory is an amodal, dynamic system: Evidence from the interaction of
naming and object use in semantic dementia. Cognit. Neuropsychol. 21,
513–527. http://dx.doi.org/10.1080/02643290342000113.
Coggins, P.E., Kennedy, T.J., Armstrong, T.A., 2004. Bilingual corpus callosum
variability. Brain Lang. 89, 69–75. http://dx.doi.org/10.1016/S0093-934X(03)
00299-2.
Cox, R.W., 1996. AFNI: software for analysis and visualization of functional mag-
netic resonance neuroimages. Comput. Biomed. Res. 29, 162–173. http://dx.doi.
org/10.1006/cbmr.1996.0014.
Cummine, J., Boliek, C.A., 2013. Understanding white matter integrity stability for
bilinguals on language status and reading performance. Brain Struct. Funct. 218,
595–601. http://dx.doi.org/10.1007/s00429-012-0466-6.
Dehaene, S., Dupoux, E., Mehler, J., Cohen, L., Paulesu, E., Perani, D., van de Moor-
tele, P.F., Lehéricy, S., Le Bihan, D., 1997. Anatomical variability in the cortical
representation of first and second language. Neuroreport 8, 3809–3815. http:
//dx.doi.org/10.1097/00001756-199712010-00030.
Dräger, B., Breitenstein, C., Helmke, U., Kamping, S., Knecht, S., 2004. Specific and
nonspecific effects of transcranial magnetic stimulation on picture-word ver-
ification. E. J. Neurosci. 20 (6), 1681–1687. http://dx.doi.org/10.1111/
j.1460-9568.2004.03623.x.
Duffau, H., Gatignol, P., Mandonnet, E., Peruzzi, P., Tzourio-Mazoyer, N., Capelle, L.,
2005. New insights into the anatomo-functional connectivity of the semantic
system: a study using cortico-subcortical electrostimulations. Brain 128,
797–810. http://dx.doi.org/10.1093/brain/awh423.
Frey, S., Campbell, J.S., Pike, G.B., Petrides, M., 2008. Dissociating the human lan-
guage pathways with high angular resolution diffusion fiber tractography. J.
Neurosci. 28, 11435–11444, doi:28/45/11435 [pii]/r10.1523/JNEUROSCI.2388-
08.2008.
Fridriksson, J., Morrow, L., 2005. Cortical activation and language task difficulty in
aphasia. Aphasiology 19, 239–250. http://dx.doi.org/10.1080/
02687030444000714.
Friederici, A.D., Rüschemeyer, S.-A., Hahne, A., Fiebach, C.J., 2003. The role of left
inferior frontal and superior temporal cortex in sentence comprehension: lo-
calizing syntactic and semantic processes. Cereb. Cortex 13, 170–177. http://dx.
doi.org/10.1093/cercor/13.2.170.
Fromkin, V., Krashen, S., Curtiss, S., Rigler, D., Rigler, M., 1974. The development of
language in genie: a case of language acquisition beyond the “critical period.”.
Brain Lang. 1, 81–107.
Frost, R., Siegelman, N., Narkiss, A., Afek, L., 2013. What predicts successful literacy
acquisition in a second language? Psychol. Sci. 24, 1243–1252. http://dx.doi.org/
10.1177/0956797612472207.
Gardner, R.C., Masgoret, A.M., Tennant, J., Mihic, L., 2004. Integrative motivation:
changes during a year-long intermediate-level language course. Lang. Learn. 54,
1–34. http://dx.doi.org/10.1111/j.1467-9922.2004.00247.x.
Grindrod, C.M., Bilenko, N.Y., Myers, E.B., Blumstein, S.E., 2008. The role of the left
inferior frontal gyrus in implicit semantic competition and selection: an event-
related f MRI study. Brain Res. 1229, 167–178. http://dx.doi.org/10.1016/j.
brainres.2008.07.017.
Hernandez, A.E., Li, P., 2007. Age of acquisition: Its neural and computational me-
chanisms. Psychol. Bull. 133, 638–650. http://dx.doi.org/10.1037/
0033-2909.133.4.638.
Indefrey, P., 2006. A meta-analysis of hemodynamic studies on first and second
language processing: which suggested differences can we trust and what do
they mean? Lang. Learn. 56, 279–304. http://dx.doi.org/10.1111/
j.1467-9922.2006.00365.x.
Isel, F., Baumgaertner, A., Thran, J., Meisel, J.M., Buchel, C., 2010. Neural circuitry of
the bilingual mental lexicon: effect of age of second language acquisition. Brain
Cognit. 72, 169–180. http://dx.doi.org/10.1016/j.bandc.2009.07.008.
Kim, K.H., Relkin, N.R., Lee, K.M., Hirsch, J., 1997. Distinct cortical areas associated
with native and second languages. Nature 388, 171–174. http://dx.doi.org/
10.1038/40623.
Klein, D., Mok, K., Chen, J.K., Watkins, K.E., 2013. Age of language learning shapes
brain structure: a cortical thickness study of bilingual and monolingual in-
dividuals. Brain Lang. 131, 20–24. http://dx.doi.org/10.1016/j.bandl.2013.05.014.
Lambon Ralph, M.A., Patterson, K., 2008. Generalization and differentiation in se-
mantic memory. Ann. N. Y. Acad. Sci. 1124, 61–76. http://dx.doi.org/10.1196/
annals.1440.006.
Li, P., Legault, J., Litcofsky, K.A., 2014. Neuroplasticity as a function of second lan-
guage learning: anatomical changes in the human brain. Cortex 58, 301–324.
http://dx.doi.org/10.1016/j.cortex.2014.05.001.
Luk, G., Bialystok, E., Craik, F.I.M., Grady, C.L., 2011. Lifelong bilingualism maintains
white matter integrity in older adults. J. Neurosci. 31, 16808–16813. http://dx.
doi.org/10.1523/JNEUROSCI.4563-11.2011.
Margulies, D.S., Vincent, J.L., Kelly, C., Lohmann, G., Uddin, L.Q., Biswal, B.B., Vill-
ringer, A., Castellanos, F.X., Milham, M.P., Petrides, M., 2009. Precuneus shares
intrinsic functional architecture in humans and monkeys. Proc. Natl. Acad. Sci.
USA 106, 20069–20074. http://dx.doi.org/10.1073/pnas.0905314106.
Mechelli, A., Crinion, J.T., Noppeney, U., O’Doherty, J., Ashburner, J., Frackowiak, R.S.,
Price, C.J., 2004. Structural plasticity in the bilingual brain. Nature 431, 757.
http://dx.doi.org/10.1038/nature03017.
Mohades, S.G., Struys, E., Van Schuerbeek, P., Mondt, K., Van De Craen, P., Luypaert,
R., 2012. DTI reveals structural differences in white matter tracts between bi-
lingual and monolingual children. Brain Res. 1435, 72–80. http://dx.doi.org/
10.1016/j.brainres.2011.12.005.
Mohades, S.G., Van Schuerbeek, P., Rosseel, Y., Van De Craen, P., Luypaert, R., Bae-
ken, C., 2015. White-matter development is different in bilingual and mono-
lingual children: a longitudinal DTI study. PLoS One 10, 1–17. http://dx.doi.org/
10.1371/journal.pone.0117968.
Myers, E.B., Blumstein, S.E., Walsh, E., Eliassen, J., 2009. Inferior frontal regions
underlie the perception of phonetic category invariance. Psychol. Sci. 20,
895–903. http://dx.doi.org/10.1111/j.1467-9280.2009.02380.x.
Newman, A.J., Tremblay, A., Nichols, E.S., Neville, H.J., Ullman, M.T., 2012. The in-
fluence of language proficiency on lexical semantic processing in native and
late learners of English. J. Cognit. Neurosci. 24, 1205–1223.
 E.S. Nichols, M.F. Joanisse / NeuroImage 143 (2016) 15–25
Noppeney, U., Phillips, J., Price, C., 2004. The neural areas that control the retrieval
and selection of semantics. Neuropsychologia 42, 1269–1280. http://dx.doi.org/
10.1016/j.neuropsychologia.2003.12.014.
Obleser, J., Boecker, H., Drzezga, A., Haslinger, B., Hennenlotter, A., Roettinger, M.,
Eulitz, C., Rauschecker, J.P., 2006. Vowel sound extraction in anterior superior
temporal cortex. Hum. Brain Mapp. 27, 562–571. http://dx.doi.org/10.1002/
hbm.20201.
Obleser, J., Wise, R.J.S., Alex Dresner, M., Scott, S.K., 2007. Functional integration
across brain regions improves speech perception under adverse listening
conditions. J. Neurosci. 27, 2283–2289. http://dx.doi.org/10.1523/
JNEUROSCI.4663-06.2007.
Oldfield, R.C., 1971. The assessment and analysis of handedness: the Edinburgh
inventory. Neuropsychologia 9, 97–113. http://dx.doi.org/10.1016/0028-3932
(71)90067-4.
Pakulak, E., Neville, H.J., 2011. Maturational constraints on the recruitment of early
processes for syntactic processing. J. Cognit. Neurosci. 23, 2752–2765. http://dx.
doi.org/10.1162/jocn.2010.21586.
Park, H.R.P., Badzakova-Trajkov, G., Waldie, K.E., 2012. Language lateralisation in
late proficient bilinguals: a lexical decision f MRI study. Neuropsychologia 50,
688–695. http://dx.doi.org/10.1016/j.neuropsychologia.2012.01.005.
Patterson, K., Nestor, P.J., Rogers, T.T., 2007. Where do you know what you know?
The representation of semantic knowledge in the human brain. Nat. Rev.
Neurosci. 8, 976–987. http://dx.doi.org/10.1038/nrn2277.
Perani, D., Abutalebi, J., 2005. The neural basis of first and second language pro-
cessing. Curr. Opin. Neurobiol. 15, 202–206. http://dx.doi.org/10.1016/j.
conb.2005.03.007.
Perani, D., Dehaene, S., Grassi, F., Cohen, L., Cappa, S.F., Dupoux, E., Fazio, F., Mehler,
J., 1996. Brain processing of native and foreign languages. Neuroreport 7,
2439–2444.
Perani, D., Paulesu, E., Galles, N.S., Dupoux, E., Dehaene, S., Bettinardi, V., Cappa, S.F.,
Fazio, F., Mehler, J., 1998. The bilingual brain proficiency and age of acquisition
of the second language. Brain 121, 1841–1852.
Ranganath, C., Johnson, M.K., D’Esposito, M., 2003. Prefrontal activity associated
with working memory and episodic long-term memory. Neuropsychologia 41,
378–389. http://dx.doi.org/10.1016/S0028-3932(02)00169-0.
Rueckert, D., Sonoda, L.I., Hayes, C., Hill, D.L., Leach, M.O., Hawkes, D.J., 1999.
Nonrigid registration using free-form deformations: application to breast MR
images. IEEE Trans. Med. Imaging 18, 712–721. http://dx.doi.org/10.1109/
42.796284.
Rugg, M.D., Fletcher, P.C., Frith, C.D., Frackowiak, R.S., Dolan, R.J., 1996. Differential
activation of the prefrontal cortex in successful and unsuccessful memory re-
trieval. Brain 119, 2073–2083. http://dx.doi.org/10.1093/brain/119.6.2073.
25
Sahinkarakas, S., 2011. Young students' success and failure attributions in language
learning. Soc. Behav. Personal.: Int. J. 39, 879–885. http://dx.doi.org/10.2224/
sbp.2011.39.7.879.
Scott, S.K., Johnsrude, I.S., 2003. The neuroanatomical and functional organization
of speech perception. Trends Neurosci. 26, 100–107. http://dx.doi.org/10.1016/
S0166-2236(02)00037-1.
Sharp, D.J., Awad, M., Warren, J.E., Wise, R.J.S., Vigliocco, G., Scott, S.K., 2010. The
neural response to changing semantic and perceptual complexity during lan-
guage processing. Hum. Brain Mapp. 31, 365–377. http://dx.doi.org/10.1002/
hbm.20871.
Smith, S.M., Jenkinson, M., Woolrich, M.W., Beckmann, C.F., Behrens, T.E.J., Jo-
hansen-Berg, H., Bannister, P.R., De Luca, M., Drobnjak, I., Flitney, D.E., Niazy, R.
K., Saunders, J., Vickers, J., Zhang, Y., De Stefano, N., Brady, J.M., Matthews, P.M.,
2004. Advances in functional and structural MR image analysis and im-
plementation as FSL. Neuroimage 23 (Suppl 1), S208–19. http://dx.doi.org/
10.1016/j.neuroimage.2004.07.051.
Smith, S.M., 2002. Fast robust automated brain extraction. Hum. Brain Mapp. 17,
143–155. http://dx.doi.org/10.1002/hbm.10062.
Smith, S.M., Jenkinson, M., Johansen-Berg, H., Rueckert, D., Nichols, T.E., Mackay, C.
E., Watkins, K.E., Ciccarelli, O., Cader, M.Z., Matthews, P.M., Behrens, T.E.J., 2006.
Tract-based spatial statistics: voxelwise analysis of multi-subject diffusion data.
Neuroimage 31, 1487–1505. http://dx.doi.org/10.1016/j.
neuroimage.2006.02.024.
Talairach, J., Tournoux, P., 1988. Co-planar stereotaxic atlas of the human brain. In:
3-Dimensional Proportional System: An Approach to Cerebral Imaging.
Wang, Y., Xiang, J., Vannest, J., Holroyd, T., Narmoneva, D., Horn, P., Liu, Y., Rose, D.,
deGrauw, T., Holland, S., 2011. Neuromagnetic measures of word processing in
bilinguals and monolinguals. Clin. Neurophysiol. 122, 1706–1717. http://dx.doi.
org/10.1016/j.clinph.2011.02.008.
Warren, J.D., Scott, S.K., Price, C.J., Griffiths, T.D., 2006. Human brain mechanisms for
the early analysis of voices. Neuroimage 31, 1389–1397. http://dx.doi.org/
10.1016/j.neuroimage.2006.01.034.
Wartenburger, I., Heekeren, H.R., Abutalebi, J., Cappa, S.F., Villringer, A., Perani, D.,
Olgettina, V., 2003. Early setting of grammatical processing in the bilingual
brain. Neuron 37, 159–170.
Weber-Fox, C.M., Neville, H.J., 1996. Maturational constraints on functional spe-
cializations for language processing: ERP and behavioral evidence in bilingual
speakers. J. Cognit. Neurosci. http://dxdoi.org/10.1162/jocn.1996.8.3.231.
Weniger, D., Crelier, G.R., Alkadhi, H., Kollias, S.S., 2000. Picture-word matching as a
paradigm in determining regions of language processing: an f MRI study.
Neuroimage 11, S314. http://dx.doi.org/10.1016/S1053-8119(00)91246-1.