Other Topics The Bilingual Brain

involving two groups of French-Engh'sh bitinguals who had

• Meaningful pictures: that a single person spoke both Languages ftuently, how would
• Speech production: name each picture—first the top
equally high L2 proficiency but differed in L2 age of acquisition, these hugely different consonant repertoires be implemented
with one group being called "simultaneous" because they learned
one, then the bottom left one, and finally the bottom in his or her brain? (For a pertinent study which suggests that
both languages from birth, and the other group being called
right one learning a second language increases the size of one's auditory
"sequential" because they learned L2 after the age of five. The
• Button press: make a semantic judgment by indicating cortex, see Ressel et at., 2012.)
experimental condition required the subjects to read aloud simple
which of the two bottom pictures is more closely
sentences in either French or English, whereas the baseline Lexical semantks
related to the top one
condition required them to view strings ofXs as if they were
• Meaningful words: In Chapters 8 and 9 we encountered considerable evidence
sentences. For each group the researchers investigated whether, for the Grounded Cognition Model, which maintains that
• Speech production: read each word aloud—first the
Figure 17.5 Results from Parker Jones et al.'s (2012) fMRI study of after subtracting out the baseline condition, there was greater
top one, then the bottom left one, and finally the the meanings of object nouns (e.g., dog, apple, spatula)
bilingual speech production in subjects with high L2 brain activity when reading aloud in L2 (i.e., English) than in
bottom right one and action verbs (e.g., kick, grab, bite) tend to be anchored
proficiency. Relative to the button press conditions, the
LI (i.e., French). The simultaneous group did not display any
• Button press: make a semantic judgment by indicating speech production conditions engaged many left temporal
in anatomicalty segregated modality-specific systems for
significant differences at ati; however, as shown in Figure 17.6,
which of the two bottom words is more closely related and frontaL regions that have previously been associated perception and action, such that a great deal of semantic
with talking. However, in the five regions shown here, the the sequential group evinced greater activity not only in the left
to the top one processing involves the partial reconstruction of sensory and
magnitude of activity was significantly greater in biLinguals posterior IFG and ventraL premotor cortex, both of which have
• Meaningless pictures: motor states in different parts of the brain, usually in a rapid,
than monolinguals during picture naming and reading
• Speech production: say "1,2,3" while viewing first the
been linked with phonetic encoding (see Chapter 6 for details), automatic, and unconscious manner, but with sensitivity to
aloud than during semantic judgments for pictures and
but also in the Left posterior fusiform gyms near the Visual
top picture, then the bottom left one, and finally the words. PTr, pars triangularis of inferior frontal gyrus; POp, task, context, and individual experience. Alt of the empirical
pars opercularis of inferior frontaL gyrus; PrC, precentral Word Form Area, which has been linked with orthographic-to- studies that we discussed in those chapters involved
bottom right one
gyrus; STG, superior temporal gyrus; PT, planum temporale. phonological transformations (see Chapter 15 for details). These
• Button press: make a visual judgment by indicating monolingual subjects, but several fMRI experiments with
(From Parker Jones et al., 2012, p. 898.) Adapted and results indicate that even when a person's general proficiency in
which of the two bottom pictures is identical to the bilingual subjects have shown that the Grounded Cognition
reproduced permission of Oxford University Press. OUP is
L2 is quite high, if that person acquired L2 several years after LI,
top one not responsible or in any way liable for the accuracy of Model also applies fairly well to words in L2 (for reviews see
the adaptation. The Licensee is solely responsible for the he or she tends to recruit the neural substrates of speech-motor Kuhne & Gianelti, 2017; Monaco et aL, 2019).
• Meaningless symbol sequences:
adaptation in this publication. planning more heavily for L2 than Ll. Such observations help
• Speech production: say "1,2,3" while viewing first the In one fMRI study, for instance. De Grauwe et al. (2014)
to explain why it is so hard for late learners of L2 to speak that administered a texicaL decision task to native German speakers
top sequence, then the bottom left one, and finally
some of the same regions that are sensitive to processing demands language without at least a mild foreign accent.
the bottom right one who achieved high proficiency in Dutch as adults. As indicated in
during monolingual speech production—an idea that fits well with Lastly, it's worth noting that further fMRI research is needed
• Button press: make a visual judgment by indicating previous chapters, a Lexical decision task simply requires subjects
the lesion study by Hope et aL (2015) that we discussed earlier. to explore the degree to which LI and L2 speech sounds
which of the two bottom sequences is identical to to distinguish between real words and pseudowords, so whatever
In further analyses, Parker Jones et al. (2012) compared the diverge in the brains of bi'tinguats whose two languages have
the top one semantic processing is evoked by real words is mostly reftexive
response properties of the five regions across conditions in order substantially different phoneme inventories. For instance, as
rather than deliberate. The researchers found that, compared to
to get a better sense of which stage(s) of speech production mentioned in Box 5.1 in Chapter 5, Rotokas, which is spoken the recognition of semantically similar abstract verbs in both
ALL of the subjects performed the verbal tasks in English, and a
they likely subserve. In two regions—namely, POp and PTr—the in Papua New Guinea, has only six consonants, whereas !X6o, languages (e.g., German zogern 'hesitate' and Dutch aarzelen
subgroup of the bilinguals also performed the verbal tasks in LI,
effect of bUingualism was greater for both naming pictures and which is spoken in Botswana, has 122. In the unlikely event
but always on a different day. hesitate'), the recognition of semanticatty similar action verbs
Relative to the button press conditions, the speech production reading aloud than for saying "1,2,3," which supports the view in both Languages (e.g., German werfen 'throw' and Dutch gooien
conditions elicited remarkably similar distributions of activity that these regions contribute more to accessing, selecting, and 'throw') engaged some of the same precentraL motor areas and
across many left temporal and frontaL regions that have preparing word forms than to overtly articulating them (see postcentral somatosensory areas. These results suggest that for
previously been associated with talking. However, in five different Chapter 6 for details). In the other three regions—namely, PrC, highly competent bilinguals, the meanings of closely related
regions the magnitude of activity was significantly greater in STG, and PT—group differences were present not only for naming concrete words in LI and L2 have shared neural substrates in
bilinguals than monolinguaLs during picture naming and reading pictures and reading aloud but also for saying "1,2,3," which modality-specific corticaL regions. As one would expect, however,
aloud than during semantic judgments for pictures and words. supports the view that these regions underlie articulation per se other work suggests that if L2 proficiency is relatively Low, the
These regions are shown in Figure 17.5 and Listed below: (PrC) and the subsequent processing of auditory feedback (STG sensory/motor grounding of L2 words is likewise relatively weak
and FT; again, see Chapter 6 for details). Parker Jones et al. (Birba et at., 2020; see also Zhang et aL, 2020b).
(2012, p. 901) state their main conclusion as follows: In sum, In Chapters 8 and 9 we also saw that the muttifarious
• Pars trianguLaris (PTr) of IFG
the effect of bHingualism on regional activations in naming and Figure 17.6 Results from Berken et aL's (2015) fMRI study of bUinguat semantic features of concrete words appear to be integrated by
• Pars opercularis (POp) of IFG
reading is best explained by greater demands on processes of^ speech production in subjects that had high L2 proficiency transmodaL hubs such as the anterior temporal lobes (ATLs). Once
• Precentral gyrus (PrC)
word retrieval, articulation, and postarticuLatory momtonng but differed in L2 age of acquisition. When reading aloud
• Superior temporal gyrus (STG) again, all of that literature was based on data from monolingual
simple sentences in LI and L2, simultaneous bih'nguals
are in common with word processing in monolinguals. subjects, but there is some evidence that the left ATL also
• Planum temporale (PT) (i.e., those who learned L2 together with LI) did not
What about age of acquisition? Although not many funct display any activation differences; however, as shown here. represents unified lexicat concepts in L2. For instance, Crinion
neuroimaging studies have attempted to isolate the potential sequential bilinguals (i.e., those who learned L2 after age et al. (2006a) found that in high-proficiency German-English
As Parker Jones et al. (2012) point out, other fMRI research
effects of this factor on speech production, there is some five) displayed greater activity for L2 than LI, specifically in
has shown that when monolinguals generate words, the signal and Japanese-Engh'sh bih'nguals, priming effects were elicited
the left premotor cortex, inferior frontal gyrus, and fusiform
evidence that it does modulate activation patterns in cortical in the left ATL by pairs of semanticaLly related words such as
strength m almost all of these areas increases as the frequency gyrus. These effects were found after subtracting out a
areas associated with phonetic encoding. This was shown, f°r control condition that involved viewing strings of XXXs as if trout-salmon (compared to pairs of semanticalty unrelated words
of the target words decreases (Chee et aL, 2002; Graves et al.,
instance, by Berken et al. (2015), who conducted an fMRI they were sentences. (From Berken et at., 2015, p. 214.) such as trout-horse), regardless of whether the words in each pair
2007). This suggests that bilingualism amplifies the workload for
Other Topics

The Bilingual Brain

its age of acquisition, the more it overlaps with LI in core
overlapping activation in this general territory for both LI and
language regions. For instance, in an fMRI study involving It's also worth noting, however, that other fMRI studies of
L2, there was also clearly some segregated activation. Moreover.
highly fluent and early-exposed Spanish-Engh'sh bih'nguals, bilingual grammatical processing have failed to find significant
even though both languages engaged the controt-retated left
WiLlms et aL (2011) demonstrated with multivariate pattern activation differences between LI and L2 in core language
DLPFC, L2 engaged it together with another control-related
analysis that in several cortical areas subserving verb inflection regions for subjects with both late L2 acquisition and high L2
region, namely the left inferior parietaL cortex, suggesting
the detailed activation patterns for the two languages were proficiency. For example, Weber et at. (2016) compared the
greater effort for grammatical processing in L2 than Ll (these
statistically indistinguishable (see also Consonni et aL, 2013). processing of complex Dutch sentences in monoh-nguat Dutch
regions are discussed ,-n more detail in the section called
Other fMRI studies have found that the neural topography "Language control in bitinguals"). speakers and in German-Dutch bitinguats who did not become
of bilingual grammatical processing is significantly influenced competent in Dutch until after the age of 18. They found that
Importantly, a few other fMRI studies of bilingual
by differences between LI and L2 in both proficiency and age the two groups processed the sentences in essentialty'thesame
grammatical processing have found substantial effects of
of acquisition. An instructive example is a study by Golestani left-hemisphere language network, though the bitingualsubjects
Figure 17.7 Results from Correia et al.'s (2014) fMRI study of bilingual L2 age of acquisition independently of L2 proficiency. For
semantic processing. Multivariate pattern analyses revealed et al. (2006). They administered two tasks to native French also recruited some control-related regions. The fact that these
instance, Wartenburger et al. (2003) administered a'syntactic
that the left anterior temporal Lobe (ATL) was among a speakers who achieved only moderate competence in English results conflict with those described in the previous paragraph"
J"d9menttask to both "earty-acquisition & high-proficiency"
restricted set of corticat regions in which the activity after the age often through explicit instruction. In the is just one manifestation of the many inconsistencies in the'
patterns that could discriminate between four different (EAHP) ,bit!nguals and "bte-acquisition & high-proficiency;
main experimental condition, the task was to covertly create relevant Literature, highlighting the need for further research.
English words for animals—specifically, bull, duck, (LAMP) bilinguats. When they contrasted the'artivation
horse, and shark — could also discriminate between the grammatically well-formed sentences from sequences of three As a final point, the question arises as to whether the"
patterns for L2 against those for LI in each group, they found
corresponding Dutch words for animals—specifically, stier, to five words in either French or English; and in the baseline activation patterns that bitinguats display during grammatical
that although L2 did not elicit any unique activity at all
eend, paard, and haai. (From Correia et al., 2014, p. 335.)
condition, the task was just to covertly read sequences of tasks vary as a function of differences in "the mc^hotogicat"
Copyright 2014 3oao Correia, et at. in the EAHP group, it significantly engaged several corticat
three to five words in either French or English (examples of the and/or syntactic complexity of the two languages'. So far-
and subcortica^regions in the LAHP group, most notably the
were from the same Language or different languages. In addition, stimuli are, -in French, amis marcher plage 'friends walk beach', very little functional neuroimaging research has addressed
left posterior IFG. These results are depicted in Figure 17.9.'
as shown in Figure 17.7, using muLtivariate pattern analysis and in English, family eat meal). The results of the subtraction thi-s issue, but there is some evidence that these sorts
Similarly, Kovelman et at. (2008) reported that dunng a
Correia et aL (2014) found that in a restricted set of cortical of the word reading condition from the sentence creation of differences do matter. For instance, inftertionatverb
syntartic judgment task in English, the left posterior'IFG was
regions that included the left ATL, the fine-grained activity condition are shown for LI (i.e., French) in Figure 17.8A and morphotogy is more complex in Persian than English, and
recruited more strongly by bilinguats who learned English tate
patterns that could discriminate between four English words for L2 (i.e., English) in Figure 17.8B. As in other studies that Momenian et al. (2018) recently found that wheri Pe'rsian-
but well as^ their L2, than by monolinguals who learned'English
for animals—specifically, bull, duck, horse, and shark— could conducted similar subtractions, the left posterior IFG turned early and fluentty as their Ll. English bilinguals performed the same type of sentence
also discriminate between the corresponding Dutch words for out to be recruited for grammatical processing (see also Figures 9enerat10n task !" both lan9"ages, there was greater activity
animals—specifically, stier, eend, paard, and haai — and vice 12.12-12.15 in Chapter 12). But while there was certainly some m BA44 for Persian than Engtish-a difference that most
versa. Given, for example, the ATL activity pattern triggered by tikely reflected the greater morphotogical demands of the
one of the four Dutch words, it was possible to determine that former language than the latter. Much more work is "needed,
A

m^
word's identity with above-chance accuracy based solely on the however, to understand how cross-b-nguistic disparities'm"
unique ATL "signatures" of the semantically matching English grammatical structure affect the bilingual brain.
words. These results provide further support for the view that •"><
the meanings of translation equivalents in different languages
have remarkably similar cortical representations, with obvious r/^ (-'.-" r^ Evidence from electrophysiology

relevance to bilinguaLism (see also Zinszer et al., 2016; Van de Another source of data about the degree to which Ll
Putte et aL, 2017, 2018; Evans et aL, 2019). and L2 have shared cortkat underpinnings is intracraniat
And yet, while there are certainly many translation etectrophysiotogy-in particular, direct electrical stimulation
equivalents between languages, there is also a tremendous during awake neurosurgery. This anatomically precise brain"
amount of cross-Linguistic semantic diversity (see the discussion mapping method has been applied to bilinguals with increasing
of lexical concepts in Chapter 6 and the introductory sections frequency over the past 30 years; however/the vast majority"

>C] .~;1

%?p<^
-;;
of Chapters 8 and 9; see also Malt & Majid, 2013; Kemmerer, of studies have several limitations. For one thing, they only
2019a, 2019d). Hence, an important direction for future fMRI investigated whether stimulation disrupts oral object naming,
work on bilingualism will be to explore the implications of such without distinguishing between different types of errors (e.g.
diversity for the neural substrates of LI and L2 word meanings anomia, semantic paraphasia, phonemic paraphasia, dysarthria,
^y^J^: ^^'.^^t.'
(for a psycholinguistic perspective see PavLenko, 2014; and for a or speech arrest; see Figure 2.23 in Chapter 2) and without
recent fMRI study see Nichols et al., 2021). exploring other aspects of language. In addition, very few
studies have carefully considered the potential influences of
Figure 17.8 Results from Golestani et al.'s (2006) fMRI study of
w'^ 17.9 Results from Wartenburger et aL's (2003) fMRI study of
Morphology and syntax bilingual grammatical processing in subjects who proficiency, age of acquisition, and manner of acquisition.
bilingual grammaticat processing in (A) subjects with early
Lastly, we turn to functional neuroimaging studies of bilingual achieved only moderate competence in L2 after the age It is perhaps for these reasons that the results have been
acquisition and high proficiency (EAHP) and (B) subjects
of 10 through explicit instruction. Activation patterns
grammatical processing. The findings in this literature are, with bte acquisition and high proficiency (LAMP). In each rather mixed, ranging from complete neural overlap of Ll and
in French-English bilinguals are shown for the contrast
unfortunately, quite mixed. It should come as no surprise, panel, syntactic judgments in L2 are contrasted against L2, through a cortical organization that has mostly common"
between sentence generation and word reading in Ll
syntactic judgments in LI. (From Wartenburqer et'al.
however, that the greater the proficiency in L2 and the earlier and L2 (B). (From Golestani et aL, 2006, p.1034.) 2003, p. 166.) ' - "' """'
but partly distinct areas for Ll and L2, to one that has even
more spatial separation of Ll and L2 (for a review of the early
 otherropics

uterat^see^,,

5;£5&^—^..
FernSndez.r'Zgual brain, we'n"^'115 Broach',
^n^Tr[oeuo et ^(2oT^focus °"^Zln revea^

; iS^SS^SK ^5?^
Th^Co^

:;;e:^'S:,:W^X°T^^^
^5S?S£:S5S £§'SS5^-
i£5Bi££5S:t
,::tes^^. '"""^
M:l:leo^MC::^
^^^as^^d^^pic^
55§iSS^£SSS
m'^ecoLcowca's'^ ^^;lst/'e P'c^reZ"tyoc(rs ^

S!^5^SS?WW s^Sl^S^
"'^^''^
^^^^^sla^lsl
.SiSSS?
^S^SIgi:
U9'ca's^9eo^
^S§S!^SS!?'
^^S^'S^^r l§sterSill-"
' apP^to'^etween ^ren^hers ^ n^n c^

:SS-sS^S^^'
/
/

^5§5§S5'
)n pf -,) ~" clectr;'c,i/ ...i... 'Jlt":;>-
;
[etaL'^89~b^ai Station,
^e Con,"! salr)e time.''^/'loa'1dth7^s (FernSnde^rner.res^

^SiSS5^d """"•" ^^^a^- one e;<^o'/'^.reffar(/^ZS%/?'70t easnl a,cknowledg^,

SSSSsiS^
so"'e ~w^lu: "'ost of thaZyeepkenters/' ^ authors

§sss§js§
B
, .'s^S=^-^ '"dc the onset"o7p^sdaily
i^-s=T<pr^ . "^ofp^r^
9^......
.'at.e-37^1^arffer'^ ''s^.'r'"1'nj>

^
wh^tZrea lan9ua9es7Zer ofsites (97'}a^a1srupted by ^^"^."^
*"<•"•>.',. """i"

IS
'et".ln',':,IM'"»
»T^,""-"'"-1

^5S:5§SS^' rZ^::esi ^ .y>

-/

^t^^,'tjcrspecti^s

reg3r^sth."lvesurvey^a,
Ll^ ?!/e^°fZ/asu^n^/ a
Other Topics
The Bilingual Brain

Convergence Hypothesis clearly predicts parallel impairment; direct support, however, derives from neurosdentific research involves the degree to which LI and L2 have overlapping neural
on the Learning ofso-catled "model languages," as described the speaker may need to decide, based the communicati've
however, during the initial stages of their disease, the patients substrates. Numerous findings from different brain mapping
in Box 17.1. Interestingly, another source of evidence, one situation and his or her goats, whether to refer to a house with
were significantly more impaired in L2 than LI. Why? The methods suggest that L2 often recruits many of the same
that we have not yet encountered, involves Parkinson's disease the English word house or the Italian word casa. What cognitive
answer is still unknown (but it may depend on hidden aspects perisylvian regions as LI. However, the degree of overlap is
(PD). This neurodegenerative disorder gradually degrades the processes are required to make such decisions, and how are
of the patients' Language histories). Another discovery that significantly influenced by proficiency, age of acquisition, and
they implemented in the bilingual brain? Furthermore, how are
seems to challenge the Convergence Hypothesis comes from procedural system (specifically the basal ganglia) while sparing manner of acquisition, and these factors frequently interact in
the declarative system. Hence, for those bilingual PD patients bitinguats able to efficiently switch back and forth between
Fernandez-Coelto et al.'s (2017) direct electrical stimulation complex ways are not yet understood. Moreover, for bitmguats
languages as their social circumstances change?
study. As Indicated above, across 13 patients stimulation at who learned L2 explicitly, the Dedarative/Procedural Model with low L2 proficiency and/or late L2 acquisition, the efficient
predicts that there should be an atypical pattern of differential We touched briefly on some issues involving bilingual
several sites disrupted object naming more frequently in early- use of L2 relies heavily on the additional recruitment of control-
impairment such that grammatical knowledge and processing cognitive control at several points in the previous section,
than tate-acquired languages (see again Figure 17.11), and in related regions. Although contemporary theoretical frameworks
is disrupted more for LI than L2. And precisely this type of but now we wilt put this topic in the spotlight. During the past
one patient who was fluent in five Languages, stimulation at can explain many of the currently available findings, some
pattern has in fact been documented by several studies, in few years, a substantial amount of research has examined the
some sites induced double dissodations between early- and results remain mysterious and a variety of issues have yet to be
psychological and neurobiotogicat aspects of language control
late-acquired languages (see again Figure 17.12). Together, keeping with the theory (Fabbro & Paradis, 1995; Zanini et aL, resolved. Future directions include exploring the specific neural
in bitinguats. But because it would be beyond the purview of
these findings suggest that for people who know more than 2004, 2010;Johari et aL, 2013). effects of more sensitive, gradient measures of bilingualism
this chapter to survey ati of this work, we will focus on the
one language, the underlying corticaL architecture may be more On the other hand, this framework has trouble (e.g., Perani et at., 2017; Sulpizio et aL, Z020a), making greater
approach taken by Green and Abutatebi, since it is especially
intricate that the Convergence Hypothesis assumes. accommodating several findings. For instance, Hope et at. use of muttivariate fMRI pattern analysis to identify distinct
welt-developed (Green & Abutatebi, 2013; Abutatebi & Green
(2015) argue that the Declarative/Procedural Model cannot response profiles for LI and L2 in the same cortical areas (e.g.
2016; see also Calabria et at., 2018; for reviews from other
The Declarative/ 'Procedural Model account for their results regarding relationships between Xu et aL, 2017a), and taking into account the social. context of
perspectives see Hervais-Adetman et al., 2011; Baus et at.
The second framework comes from a broader neurolinguistic behavioral data and tesion data. According to their line L2 acquisition (Li & Jeong, 2020).
2015; Costa et aL, 2016; Branzi et al, 2018; Calabria et at..
theory called the Dedarative/Procedural Model. It was originatLy of reasoning, the theory assumes at least some degree of
2019; Grant et aL, 2019; Green & Krolt, 2019; Li & Dong, 2020).
developed by Michael Ullman, a professor at Georgetown macroscopic neural segregation of LI and L2, especially for
University, in 2001, but it has undergone several refinements grammatical knowledge. Hence, it predicts that the bilingual
since then (UUman, 2001, 2004, 2016; for specific application patients in the study, for whom English was the non-native L2, Language control in bi'linguals Different cognitive demands of different
to L2 see Utlman, 2020; and for a similar approach see Paradis, should have displayed correspondences between deficits and interactional contexts
2009). The main assumption is that the two most well- lesion sites that were partially distinct from those displayed by In Chapter 6 we noted that monotinguals must sometimes
In a paper that elaborates the psychological aspects of
established learning and memory systems in the human brain the monolingual native English-speaking patients. Such group choose between alternative words, like couch vs. sofa. For
their approach. Green and Abutalebi (2013) discuss the
play crucial, yet largely different, roles in Language acquisition, differences did not emerge, however, and as a consequence bilinguats, however, the challenge of texicat selection is often
different cognitive demands that bitinguals face in different
with significant implications for both LI and L2. The dedarative the theory is threatened. Now, this argument ignores one of much greater, because even when they are only using one
interactional contexts, which are defined as recurrent patterns
system is anchored in circuits connecting the mediat temporal the theory's other assumptions—in particular, that the degree language, competing representations in the other language are
of conversational exchange. We wilt restrict our attention
lobes (including the hippocampus and related structures) with of neural segregation of LI and L2 is greatly reduced if an frequently activated too. For instance, as shown in Figure 17.13,
to two of them: the duat-language context and the dense
other cortical regions, and it is known to be vital for learning individual teamed L2 implicitly (i.e., in a similar manner as LI). when two Engtish-Itatian bilinguats are having a conversation,
code-switching context.
idiosyncratic facts and their associations. Hence, it is predicted And yet, as mentioned earlier, 24 of the 33 bilinguats had a
to be particularly important for learning the arbitrary sound- mean age of L2 acquisition of 10.7, which suggests that many
meaning pairs that constitute words, not only in LI but also in if not most of them Learned L2 explicitly in school rather than

L2. In contrast, the procedural system is anchored in circuits implicitly through natural communication. So the theory stilt
connecting the basal ganglia with the frontaL [obes, and it seems to predict at least some group differences in the relevant
is known to be vital for learning motor and cognitive skills tesion sites, when in fact none were found.
as well as rules and sequences. Hence, it is predicted to be This framework faces other difficulties as well. For instance,
particularly important for Learning the combinatorial principles it cannot easily explain the same data from bilingual PPA that
that constitute regular morphology and syntax,certainty in LI, challenge the Convergence Hypothesis (Malcomb et al., 2019).
but not necessarily in L2. According to the hypothesis, if L2 is And contrary to the theory's predictions, many functional
acquired in the same implicit manner as LI before the age of neuroimaging studies have found that the temporal-lobe
about seven, the relevant grammatical knowledge is most likely dedarative system does not often contribute to grammatical
to be implemented by the procedural system, just like for LI. But processing in bilinguals who learned L2 late and in an exptic't
if L2 is acquired Later in life through explicit instruction, that manner (for a review see Abutalebi & Della Rosa, 2012).
same knowledge is more likely to be stored in the dectarative
system instead. A qualification, however, is that if L2 proficiency
Summary
continues to grow, the grammatical knowledge may become
increasingly automatized, with greater mobilization of the We've covered quite a lot of empirical and theoretical work
procedural system. (For additional details see Ultman, 2020.) on the organization of the bilingual brain. Now let's wrap UP
this long and detailed section by briefly pulling together the 'l7'13 ^enn^ol9"sh^a"anylLTa.ts arehav"19.iconversation'the speaker mayneed to decide' based °" the communicative context and
Like the Convergence Hypothesis, this framework is
hllhw go,ak:whetherto refer to a .house withthe En9l"h word ^(Torth'e Man"wo7d'^.°^s^lusteraTeroTS
compatible with much of the data discussed above. The most main themes. One of the key questions in this field ofinqu"y control in bningual,sm/(From-Abutalebi7DeUaRosa^20ylZp>>u52u8;)Zpyu^ wjZC^TsZSZes one kind of C09nitive
Other Topics
The Bilingual Brain

Dual-language context freely draw upon both Languages, making use of whatever Dual-language context
words and grammatical constructions come readily to mind. 3ones et al. (2012) found that even bilinguals with high L2
In the dual-language context, a bilingual speaker uses both of According to Abutalebi and Green (2016), the different forms
Green and Abutalebi (2013) call this form of cognitive control proficiency activated the left posterior IFG more strongly than
his or her languages but with different conversation partners, of cognitive control that are necessary in the dual-language
opportunistic planning, and they point out that one of its major monotinguals when naming pictures and reading words aloud,
sometimes switching between Languages when switching context are typically orchestrated by the set of corta'cal
aspects is the ability to rapidly coordinate and integrate items regardless of whether they performed the tasks in Ll or L2
between addressees. For example, my wife happens to be a and subcortical structures shown in the Lower left panel of
from different Languages. As one would expect, for bilinguals (see again Figures 17.4-17.5 and the accompanying text). In
Russian-English bilingual, but I am a monotinguat English Figure 17.14. Here, however, the dual-language context is
with low/moderate L2 proficiency, dense code-switching often addition, Sebastian et aL's (2011) meta-analysis of numerous PET
speaker, and her mother, father, and sister are all monoLingual referred to as the "multiple language interactionat context." The
involves trying to produce fluent and well-formed sentences and fMRI studies revealed that bilinguals with only low/moderate
Russian speakers. Suppose that she and her mother are relevant functional-anatomical relationships are quite complex
in L2, but periodically getting stuck and falling back on more L2 proficiency tend to activate the left DLPFC—as well as the Left
conversing in Russian by themselves when I suddenly enter and not yet understood very deeply, but some insight can be
familiar and accessible items in LI. ACC, which is discussed below—much more during L2 than LI
the room and approach them, at which time she greets me in gained by returning to the four kinds of cognitive control listed
processing (see again Figure 17.2B and the accompanying text).
English and quickly tells me what they've been talking about. above and considering their putative neural substrates in light
And Golestani et at. (2006) discovered that when bitinguals with
Despite the fact that she has no trouble performing these of evidence from various sources.
Neural substrates only moderate L2 proficiency generated sentences, both the left
actions, they actually require a tremendous amount of cognitive First, maintaining the goat of using one language rather
Following up on their earlier paper, AbutaLebi and Green DLPFC and the left SMG were activated more strongly for L2 than
control. According to Green and Abutalebi (2013), the following than another is believed to depend primarily on certain left
(2016) discuss the neural substrates of Language control in LI (see again Figure 17.8B and the accompanying text). These
kinds of executive functions are necessary: lateral prefrontat and parietal regions—in particular, both
bilingualism, again with reference to the dual-language context findings reflect the mental effort that bitinguals must exert in
the posterior IFG and the DLPFC in the frontal tobe, and the
and the dense code-switching context. The images shown in order to keep their attention focused on using one language
• Goof maintenance: While conversing with her mother, SMG in the parietal tobe. Collectively, these regions, as well
Figure 17.14 not only indicate the rough locations of the most rather than another (see also Reverberi et aL, 2015, 2018).
she must continually select words and grammatical as the homotogous areas in the right hemisphere, mediate
critical cortical and subcortical structures, but also illustrate Evidence that the left lateral prefrontaL cortex not only
constructions in Russian and suppress or avoid intrusions intelligent goal-directed behavior in a wide range of linguistic
which networks of interconnected structures are thought to be contributes to this kind of cognitive control, but is actually
from English. and nonlinguistic domains, and for this reason they are core
essential for it, comes from clinical studies showing that bilingual
• Salient cue detection: When I approach, she must recognize recruited in the two contexts. components of what is sometimes catted the "multiple demand
patients who have suffered damage to this large territory are
this event as a social signal that bears on the immediate network" (Duncan, 2010, 2013; Cole et aL, 2014; Assem et al.
prone to switch involuntarily from one language to another, even
communicative situation. 2020). Hence, it is hardly surprising that they have been
when interacting with people who know only one of the two
• Conflict monitoring: After detecting the cue, she must implicated in the executive process of continually selecting
languages. For instance, in a frequently cited case study, Fabbro
register the conflict between her current use of Russian words and grammatical constructions in one language white
et at. (2000) reported a 56-year-old FriuLian-ItaLian speaker who
with her mother and her need to use English with me. suppressing or avoiding those in another.
had a large tumor that was centered in the white matter beneath
• Task switching: When she shifts attention away from her In the previous section we encountered several functional
the left inferior, middle, and superior frontat gyri (Figure 17.15).
mother and toward me, she must inhibit Russian and neuroimagmg studies that support this view. For instance, Parker
Remarkably enough, he did not display any aphasic symptoms,
promote English.

Because she is highly proficient in both Russian and English,

she can carry out all of these mental operations without much
difficulty. Not surprisingly, however, for bih'nguals with Low/
moderate L2 proficiency, many of these forms of cognitive
control are more taxing.

Dense code-swi'tching context MULTIPLE IAN6UASE DENSE CODE

INTERACTIONAL CONTEXT SWITCHING CONTEXT
In the dense code-switching context, bilingual conversation
ACC/pre-SMA

;:1 ^7'""
partners who share the same two languages frequently «i,hnrc|
•RiB^tlFC

b-^—
interleave words and grammatical constructions from both
msula —.^Bawleanel'a
languages in single utterances. In the following example from rlcaudatc/putamcn} •v caudate/putamen

an EngLish-Tagalog bilingual, English words are underlined to ^. I

help distinguish them from Tagalog words (Green & Abutalebi, Key;
Speech pipeline
2013, p. 518): Wala akong cash pang grocery ngayon, if you Control network

want, bukas na lang, ipagdadme pa kita! English translation:

"I do not have cash for groceries today, if you want, tomorrow,
Figure 17.14 Brain regions related to language control (top) and
I will even drive you there!" Whereas the relationship between their functional interactions in two different situationat

a bilingual's two languages is competitive in the dual-language contexts (bottom). Details are described in the main
text. Note that the dual-language context is referred
context, it is cooperative in the dense code-switching context. e 17'15 LZ" °hLa«S! wl^.^ulsLwlnvoluntarylan9uage switchin9due to a teft frontat tum°r- The damage affected the white
to here (in the bottom Left panel) as the multiple nmT^beRn^,tle.l!ftJnfer'OLmddle:-an^
A speaker does not have to concentrate on using just one Language interactional context. (From Abutalebi ^55glhIs'S£d with perm'ssion from Bm put)l''shin3^ro'UJP"Ltud- ^py°righ't'© 200UO<01:CB^LPUubL^ng^omuphaLt)tbdro et al" 2000'
Language while conversing with a particular person, but can & Green, 2016, p. 691.)
Other Topics

The Bilingual Brain

but he did exhibit a compulsive tendency to alternate between Language currently being used, he or she must recognize this
utterances in LI and utterances in L2. The researchers described conflict and get ready to switch to the language that the
this as follows (Fabbro et aL, 2000, p. 652): person does know. Abutalebi and Green (2016) suggest that
BOX 17.2 The neural substrates of simultaneous interpretation
conflict monitoring relies on the dorsaL ACC, and as mentioned
When requested to speak in LI, the patient produced 58.3% above, this region is part of the saLience network, often being One of the most difficult forms of language control is
model see Dong & Li, 2020).
of the utterances in that language, 40.6% in the unwanted coactivated with the ventral anterior insula. More generally, a simultaneous interpretation (SI). This is when an interpreter
translates spoken sentences from a source language to a" In a valiant effort to identify the neural substrates of SI.
L2, and only 1.1% of mixed utterances. When requested to wealth of data suggests that the dorsal ACC is very sensitive to
target language in real time, so that the natural flow of Hervais:Adelman et at- (2015) condurted an fMRI study with
speak in L2, the patient produced 56.5% of utterances in L2 conflict and surprise, is heavily recruited during effortful tasks,
50 participants, all of whom were highly proficient in three"
discourse is kept more or less intact. SI obviously places
and 43.5% of utterances in LI. No utterances with elements and directly influences adaptive executive processes in the
or more languages, and many of whom had recently enrolled
of the two languages were recorded in this session. lateraL prefrontaL cortex—processes that include choosing which tremendous demands not only on language comprehension
w the Master's program for conference interpretation
and production, but also on language switching verbal
language to use (Shenhav et al., 2017). Indeed, considerable
working memory, and divided attention (for a processing
offered by the Faculty of Translation and Interpretation at
Interestingly, he was always aware that he should speak in a support for the involvement of the dorsal ACC in handling
the University of Geneva. As shown in Figure 17B2^,'there
particular language, and he often commented on, or apologized language conflict comes from functional neuroimaging studies
for, his inappropriate switches. Thus, although his large Left (Sulpizio et aL, 2020b). [ Instruction: LISTEN
frontal lesion seems to have Left his basic linguistic abilities The fourth and final kind of cognitive control in the
intact, it disrupted the goal-oriented control mechanisms dual-language context involves actually switching from one PL Lis^n to sentence in source language
that are necessary to maintain attention on one Language Language to another. This is a multistep process that consists
and inhibit another (see also Kong et aL, 2014). In this of disengaging from the language currently being used, shifting
No Response
connection, it's also notable that direct electrical stimulation attention to the other language, and engaging it. Given the
of the Left DLPFC was recently found to trigger involuntary complexity of this process, it is not surprising that several
language switches in bilingual neurosurgical patients different cortical and subcortical regions have been associated
6.9s ~9s
(Sierpowska et aL, 2018). with it (for meta-analyses of functional neuroimaging studies
The second kind of cognitive control that is often required see Luk et aL, 2012, and SuLpizio et aL, 2020b). But because a TR:2.1s~

in the dual-language context is salient cue detection—that is, detailed review of the relevant literature would take us too far Instruction: SHADOW
registering a change in the interactive situation that signals afield, we will just take a quick Look at what appear to be the
[Listen to sentence in source language"
an upcoming need to switch languages. Such events frequently most critical regions and their likely roles.
involve the approach of a person who does not speak the
language currently in use. Detecting cues of this nature may • The left inferior parietal lobule: Nearly 100 years ago, Potzl
Repeat in source language
depend especially on the ventral anterior sector of the insula. (1925) observed that bitinguals who suffered damage to
(The fact that Abutalebi and Green [2016] did not mark this the left inferior parietal lobule manifested a pathological
structure as belonging to the control network in the lower left fixation on one language, such that disengaging from it
panel of Figure 17.14 was an oversight.) This insular region and was extremely difficult. This finding, together with data
the dorsal ACC, which is discussed next, are core components from recent fMRI studies. Led Abutalebi and Green (2007,
of what is sometimes called the "salience network" (Seeley 2016) to suggest that during language switching the Left Instruction: INTERPRET
et aL, 2007; Menon & Uddin, 2010; Seeley, 2019). This network inferior parietal lobule helps the speaker "unlock" attention
[Listen to sentence in source language
is critical for perceiving and reacting to socially significant from the language initially being used.
stimuli and for imbuing them with emotional weight, with the • The left caudate nucleus in the basal ganglia: We noted
insular region being the input node and the dorsal ACC being earlier, specifically in connection with Ullman's Dedarative/ Convert to target language
the output node. Seeley (2019, p. 9880) recently noted that Procedural Model, that the basat ganglia contribute to
one important function of the insular region is to "orchestrate motor and cognitive skills as well as rules and sequences.
switching to a new task set... in response to shifts in the What matters here, however, is that the basal gangh'a
salience landscape." And of course this fits with the proposal also facilitate some executive functions, with the caudate
TR: 2.1s
that in bilinguals the insular region plays a key role in preparing nucleus playing an important role in shifting between
other brain areas to switch from one language to another. "mental sets" (Grahn et aL, 2008). With regard to
Evidence for this view comes from Sutpizio et al. (2020b), who bih'ngualism, an fMRI study by Abutalebi et at. (2013)
conducted a meta-analysis of numerous functional neuroimaging showed that the left caudate was not only activated
studies of language switching and found the insuLa to be during Language switching, but was activated most
strongly when subjects switched from their most proficient
reliably engaged.
'''" "•2-1 ^^Sn;£."5»u:;S,';S^^^^^^
The third kind of cognitive control in the duaL-language
context is conflict monitoring, and it follows naturally from
Language to their least proficient one. In addition, as
described in Box 17.2, the Left caudate has been implKated ^srs^l= S^^S£2t?^^:^rssiS!t:^^ .
^contained^arrati^B^^^n^cu^dZS^^^^^^^S^^!"^
salient cue detection. In short, once a bilingual speaker has in what may be the most challenging form of language
t,meth^coindde^^^;^'^Z^^n:;^^;^^S^^^^c^n^toc,
registered the approach of a person who does not know the switching—namely, moment-to-moment translation from

(continued)
532