Bilingualism: Language and Cognitive control in bilinguals: Effects of

Cognition
language experience and individual variability
cambridge.org/bil
Michela Bonfieni1, Holly P. Branigan2, Martin J. Pickering3 and Antonella Sorace4
1
University of Edinburgh; 2University of Edinburgh; 3University of Edinburgh and 4University of Edinburgh

Research Article Abstract

Cite this article: Bonfieni M, Branigan HP, We report a study that investigated executive functions in four groups of participants that var-
Pickering MJ, Sorace A (2020). Cognitive ied in bilingual language experience, using a task that measured two theoretically motivated
control in bilinguals: Effects of language
mechanisms of cognitive control (proactive and reactive control). Analyses of accuracy
experience and individual variability.
Bilingualism: Language and Cognition 23, based on aggregated measures suggested an advantage in early highly proficient bilinguals
219–230. https://doi.org/10.1017/ over late passive bilinguals. However, when we factored in individual variability using
S1366728918001086 mixed-model regression with a full random effect structure, we only found a marginal effect
of language experience. Our results emphasise the importance of including individual vari-
Received: 1 September 2017
Revised: 7 October 2018 ability when studying bilingualism, and highlight a fundamental consideration in research
Accepted: 5 November 2018 on the relation between language and attention – namely, the need for a theory-driven
First published online: 18 February 2019 approach to measuring cognitive control through laboratory tasks.
Key words:
bilingual experience; individual variability;
cognitive control; AX-CPT; mixed-model
regression Introduction

Address for correspondence: Michela The relation between the bilingual linguistic experience and cognitive control has been the
Bonfieni, Email: s1471546@sms.ed.ac.uk object of extensive research over the last 15 years. The acquisition and use of more than
one language provide an ideal context for the study of cognitive plasticity, because the two lan-
guages of a bilingual are always active to some degree and interact with one another (Marian &
Spivey, 2003; Blumenfeld & Marian, 2013; Thierry & Wu, 2007; Wu & Thierry, 2010; Thierry
& Sanoudaki, 2012). The mechanisms underlying the ability to select the relevant language
and to inhibit the irrelevant one may lead to a transfer of abilities to other cognitive domains,
such as the ones responsible for selective attention and goal orientation, i.e., executive func-
tions. Therefore, some aspects that characterise the linguistic experience may result in cogni-
tive enhancement on non-verbal tasks engaging cognitive control. The hypothesis of a
relationship between bilingual experience and cognitive control has been the subject of
extended research and controversy, as we discuss below; for this reason, in this study we con-
sider theoretical and methodological aspects of that research that may limit its empirical gen-
eralizability. Specifically, we compare different groups of bilinguals that represent a range of
bilingual experiences, in order to identify what critical variables may affect cognitive abilities;
in addition, we adopt a theoretically motivated experimental task that targets specific aspects
of cognitive control, and we employ analytical techniques that account for the effects of indi-
vidual variability.
The neurosciences and cognitive psychology provide evidence for a relationship between
language processing and executive functions and for brain differences between bilinguals
and monolinguals. There are overlaps and patterns of dynamic connectivity between brain
areas dedicated to language processing and to cognitive control (Fedorenko & Thompson-
Schill, 2014; Fedorenko, 2014). Patterns of cortical activation, thickness and connectivity spe-
cific to bilinguals correlate with properties such as age of language acquisition and language
proficiency (Buchweitz & Prat, 2013; Abutalebi, Della Rosa, Ding, Weekes, Costa & Green,
2013; Ye and Zhou, 2009; García-Pentón, Pérez-Fernández, Iturria-Medina, Gillon-Dowens
& Carreiras, 2014; Klein, Mok, Chen & Watkins, 2014). In addition, monolingual and bilingual
participants show different patterns of activation during cognitive control tasks (Stocco & Prat,
2014; Rodríguez-Pujadas, Sanjuán, Ventura-Campos, Román, Martin, Barceló, Costa & Ávila,
2013). These findings attest that specific aspects of the bilingual experience have a widespread
impact on the brain’s functionality.
In contrast, behavioural evidence for advantages in cognitive abilities related to the bilin-
gual experience is less conclusive and highly controversial. Many studies have compared
monolinguals and bilinguals using tests such as the Simon task, the flanker task, and the
Stroop task, which engage attentional processes as they require the selection of an appropriate
© Cambridge University Press 2019 response in cases of conflicting information. Some of these found that bilinguals performed
better than monolinguals and therefore support a ‘bilingual advantage’ (Bialystok, Craik,
Klein & Viswanathan, 2004; Bialystok, Craik & Luk, 2008; Costa, Hernandez &
Sebastián-Gallés, 2008; Costa & Sebastián-Gallés, 2014; Bialystok, Craik & Luk, 2012).
However, others did not find any such effect (Paap & Greenberg, 2013; Paap & Sawi, 2014;
220 Michela Bonfieni, Holly P. Branigan, Martin J. Pickering and Antonella Sorace

Gathercole, Thomas, Kennedy, Prys, Young, Vinas Guasch, “segregational approach” to executive functions (or “divide and
Roberts, Hughes & Jones, 2014; Paap, 2014). These divergent conquer approach”; Stocco & Prat, 2014), which tries to separate
results may be the consequence of variables such as socio- and address single mechanisms of cognitive control, has been cri-
economic status or immigrant status, or effects of small sample ticised (Hartsuiker, 2015; Gade, 2015). For instance, some studies
sizes (Paap, Johnson & Sawi, 2015). have shown differences between monolinguals and bilinguals in
But these potential confounds only represent the tip of the ice- measures of disengagement of attention, rather than in inhibition
berg of two theoretical challenges in the study of bilingualism: the (Grundy & Keyvani-Chahi, 2017). Recent findings highlight the
large variability within and between bilingual groups, and the lack “unity and diversity” of executive functions mechanisms (Miyake
of a theory-driven approach to measuring cognitive control & Friedman, 2012): that is to say, the correlations between distinct
through laboratory tasks. In addition, this research also faces components of cognitive control such as updating, shifting and
the main problem for the study of executive functions: individual inhibition. These components dynamically adapt to the specific
variability, i.e., the fact that the ability to control attention varies demands of different interactional contexts, and differ greatly across
significantly across individuals (Braver, Gray & Burgess, 2007; situations as well as individuals (Green & Abutalebi, 2013).
Braver, 2012). We now elaborate on these three points in turn. Accordingly, some studies have used approaches such as latent-
First, rather than a dichotomous distinction between bilinguals variable analysis to find the common properties measured by
and monolinguals, the bilingual experience can be better under- executive functions tasks (Friedman, 2016). But these approaches
stood as a continuum, multi-variate dimension (Luk & are data-driven, i.e., do not make explicit reference to the individual
Bialystok, 2013; Bak, 2016). Bilingualism is in fact associated components that are recognised by theories of executive functions.
with a diversity of experiences in which multiple variables play Therefore it seems that the choice of the dependent variable in
a role (e.g., early or late age of acquisition, high or low profi- laboratory studies is not always based on a principled approach
ciency). The particular type(s) of experience that may affect cog- to executive functions and the specific components, beyond inhib-
nitive abilities such as executive functions need to be identified ition, that could be implicated in bilingual language processing
along these dimensions. At the same time, though, they are likely (Jared, 2015).
to interact with one another to create unique and diversified Consistent with the “unity and diversity” approach, Braver and
experiential profiles, and obscuring their impact on non-linguistic colleagues have proposed an explicit dual-component model of
cognitive aspects. It is important, therefore, to examine the role of cognitive control: the dual mechanisms framework (Braver et al.,
each dimension of the bilingual experience (e.g., age of acquisi- 2007; Braver, 2012). This model was originally elaborated to answer
tion, proficiency, exposure); however, a significant body of to the question of individual variability in executive functions.
research on bilingualism presents mixed bilingual samples (i.e., According to this framework, cognitive control operates through
groups of individuals with different language combinations and two separate components: ‘proactive control’ and ‘reactive control’.
backgrounds, broadly matched for age of acquisition and profi- ‘Proactive control’ is specialised to the active maintenance of goal-
ciency, e.g., Bialystok, Craik & Ruocco, 2006; Bialystok, Craik & relevant information, which directs attention, perception and
Ryan, 2006, Bialystok et al., 2008; Morales, Gómez-Ariza & action. ‘Reactive control’ is engaged as a ‘late correction’ mechanism
Bajo, 2013; Moradzadeh, Blumenthal & Wiseheart, 2014; Paap after a sudden event that re-directs attention, similar to the inhibi-
& Sawi, 2014) or ‘monolingual’ participants who know an add- tory mechanism put forth by Green (1998). Importantly, Braver and
itional language, albeit with low to medium proficiency (e.g., colleagues argue that the existence of distinct, but interconnected,
Bialystok et al., 2004; Bialystok, Craik & Ryan, 2006; Marzecová, components of cognitive control allow information processing to
Bukowski, Correa, Boros, Lupiáñez & Wodniecka, 2013; be optimized in a flexible way, because each control mechanism is
Morales et al., 2013; Paap & Greenberg, 2013; Paap & Sawi, 2014). associated with a cognitive cost. Proactive control is highly reliable
Secondly, research on bilingual cognitive control has been but cognitively expensive, because it requires sustained activation of
hampered by the lack of a theory-driven approach to measures of contextual information. In contrast, reactive control activates rele-
cognitive control. Tasks used in such research have little convergent vant information only transiently; so it is less expensive, but poten-
validity, in that the measures they provide are poorly correlated, as tially unreliable. The dynamics of these two components are also
highlighted by studies on bilinguals (Paap & Sawi, 2014) and responsible for the variability in control strategies within and across
monolinguals: for instance, the Stroop and the Simon effects may individuals, and as such provide an explanation for the individual
not correlate because they engage cognitive control processes in dif- variability that is central to the “unity and diversity” account.
ferent ways, as reflected by the fact that they have different time- The dual mechanisms framework is potentially relevant for the
courses (Pratte, Rouder, Morey & Feng, 2010; Speckman, Rouder, study of bilingualism not only because it overcomes the limita-
Morey & Pratte, 2008). In the flanker task, differences between tions of the Inhibitory Control Model, as mentioned above, but
bilingual and monolingual participants depend on the manipula- also because it reflects models of language control in language
tion of the amount of conflict that the task presents (Costa, switching. Studies on language mixing such as Ma, Li and
Hernández, Costa-Faidella & Sebastián-Gallés, 2009). In addition, Guo (2016) and Wu and Thierry (2017) associate a proactive
most research has used tasks that are ‘impure’, in the sense that mechanism of language control to mix costs (i.e., the difference
they involve cognitive components other than executive functions, between naming latencies in a single-language context and in a
such as spatial attention and a variety of perceptual and motor mixed-language context in language switching paradigms) and
mechanisms (Valian, 2014). an inhibitory mechanism to switch cost (i.e., the difference
Researchers originally adopted these tasks because they between naming latencies when switching languages in successive
assumed that the relationship between executive functions and trials in language switching paradigms).
bilingualism is based on one mechanism, namely inhibition, as The dual mechanisms framework has been evaluated in differ-
proposed by the Inhibitory Control Model (Green, 1998). ent populations in both neuroimaging and behavioural studies.
According to this model, bilinguals inhibit the language they are Proactive and reactive control correlate with flexible patterns of
not using at every level of linguistic representation. However, this activation of the prefrontal cortex in neurologically normal
Bilingualism: Language and Cognition 221

adults (Braver, Paxton, Locke & Barch, 2009). Moreover, the aggregated accuracy scores showed that bilinguals made fewer
AX-Continuous Performance Task (AX-CPT), a task of continu- errors than monolinguals on the “AY” trials, and that the groups
ous performance designed to measure the interplay of these two did not differ on the other types of trial (“AX”, “BX”, “BY”)
control mechanisms, revealed differences between younger and (Morales et al., 2013). To examine whether the bilingual advan-
older adults (Braver et al., 2009; Paxton, Barch, Storandt & tage was the result of better reactive control alone, Morales and
Braver, 2006). These findings suggest that people differ in the colleagues also administered a stop-signal task. This task specific-
extent to which they modulate proactive and reactive control to ally addresses reactive control by requiring participants to respond
optimize performance (Braver, Barch, Keys, Carter, Cohen, to stimuli but to suppress their response when a stop signal is pre-
Kaye, Janowsky, Taylor, Yesavage, Mumenthaler, Jagust & Reed, sented. In this task, they found no differences between the two
2001; Braver et al., 2007, 2009). groups, suggesting that better performance on the “AY” trials
Specifically, the AX-CPT presents participants with sequences indeed reflects a superior modulation of two cognitive control
of letters, which include pairs of cues and probes. Participants processes. In a second study, they found the same pattern of
have to press “yes” if they see an X (probe) following an A results with respect to accuracy (but not with respect to reaction
(cue). For any other cue-probe combination, they have to press times) and extended them through the analysis of ERP compo-
“no”. Moreover, between the cue and the probe a sequence of let- nents related to reactive control, which showed differential activa-
ters appear as distractors, and participants have to press “no” to tion between bilingual and monolingual participants (Morales
each of them (see Fig. 1). There are four combinations of cues et al., 2015).
and probes: “AX” trials (correct cue and correct probe); “AY” Taken together, the existing evidence suggests that to
trials (correct cue but incorrect probe, where Y stands for any adequately address the relationship between bilingualism and
probe other than X); “BX” trials, in which the cue is incorrect executive control, it is necessary both to adopt an explicit
but the probe is correct (B stands for any cue other than A), model of the relationship between language control and executive
and “BY” trials, in which neither the cue nor the probe is correct. functions, and to use a task (such as the AX-CPT) that can dis-
In the AX-CPT task, “AX” trials occur 70% of the time in criminate the relevant components. Nonetheless, the selection
order to bias participants to respond “yes”; “AY”, “BX”, “BY” of an appropriate task alone may not be sufficient: evidence
trials each occur 10% of the time (and therefore their frequency about a modulation of cognitive abilities dependent on language
is matched). In “AY” trials, participants first invoke proactive con- experience may also be susceptible to substantial individual vari-
trol to keep in memory the A cue and be prepared to respond ability in executive functions.
“yes”, but then they need to suppress this tendency when they Individual variability is a main challenge in the study of execu-
see the Y probe – that is to say, they need to engage reactive con- tive functions. One way to take individual variability into account
trol. In “BX” trials, in contrast, participants tend to answer “yes” is to use appropriate sample sizes. In these respects, Morales
when they see the X probe, but they can suppress this tendency by et al.’s (2013, 2015) conclusions may be affected by the small sam-
relying on the information provided by the B cue, i.e., through ple sizes (in the first study they examined 21 bilinguals and 23
proactive control alone. Both “AY” and “BX” trials therefore monolinguals, in their second study they tested 25 bilinguals
engage proactive control, but “AY” trials also engage reactive con- and 27 monolinguals). A stronger approach to addressing individ-
trol (Paxton et al., 2006). Finally, “BY” trials can be considered as ual variability is to factor it into data analysis. Mixed-model
baseline trials, as neither the cue nor the probe prompt a “yes” ANOVA, as used by Morales et al., is a widespread analytical tech-
response. Like the majority of executive functions task, the nique, but it allows only the specification of by-subject random
AX-CPT also involves perceptual and motor mechanisms, and effects (or by-item random effects). Mixed-effects models, in con-
the mapping between reactive and proactive control components trast, allow for the specification of complete, theoretically moti-
and type of trial has received criticism (Grundy & Timmer, 2016); vated random effects structures (Barr, Levy, Scheepers & Tily,
however, this task seems to allow the assessment of how indivi- 2013). Studies that are based on ANOVA, as in much research
duals combine the two (proactive and reactive) control mechan- on bilingualism and executive functions (e.g., Bialystok &
isms in order to respond appropriately to the different trials. Martin, 2004; Prior & MacWhinney, 2009; Mishra, Hilchey,
Morales and colleagues (Morales et al., 2013; Morales, Yudes, Singh & Klein, 2012; Blumenfeld & Marian, 2014), may therefore
Gómez-Ariza & Bajo, 2015) used evidence from this task to argue be limited in their ability to determine the effects of individual
that bilinguals showed an advantage over monolinguals in their variability in the critical components of executive functions.
ability to modulate proactive and reactive control. Their hypoth- Critically, their conclusions may result from the unwarranted
esis is in line with studies on language switching (Ma et al., 2016; attribution of the variability present in their data to the group
Wu & Thierry, 2017) that highlight the importance of both pro- level, rather than to the individual level.
active and reactive control mechanisms in language selection. Moreover, ANOVA is based on the aggregation of data-points,
Specifically, Morales and colleagues hypothesized that the lan- and it misrepresents accuracy data as normally distributed;
guage selection mechanism responsible for suppressing irrelevant mixed-effects models, instead, are adequate to the analysis of
linguistic representations is related to reactive control, whereas the binomial data such as accuracy (Barr et al., 2013, Dixon, 2008).
ability to monitor the context and to maintain activation of the The analysis of aggregated accuracy data using ANOVA, com-
relevant language is related to proactive control, and moreover bined with reduced sample size, as in Morales et al. (2013,
that the two mechanisms need to be combined to manage two 2015), contributes to increases in Type I error rates (i.e., false
languages efficiently. Consequently, they predicted that bilinguals positives).
would show different patterns of performance on the AX-CPT Our study targets these problematic aspects in research on
task from monolinguals. bilingualism and executive functions by adopting a theoretically
In one study, they administered the AX-CPT to a group of motivated experimental test of executive functions (i.e., the
monolinguals and to a group of highly proficient early bilinguals AX-CPT) and analytical techniques that are robust to inter-
with different language combinations. Their analysis of individual variability. By doing so, we ask whether any group
222 Michela Bonfieni, Holly P. Branigan, Martin J. Pickering and Antonella Sorace

Fig. 1. Design of the AX-CPT: procedure (top) and types of trials (bottom). Adapted from Morales et al. (2013).

differences stand up to an appropriate factorization of individual Sardinian Passive bilinguals reported on average limited product-
variability through the use of mixed-model regression and a com- ive proficiency in Sardinian, but high comprehension abilities,
plete random effect structure. Moreover, we compare patterns of and consistent passive exposure (in particular oral) throughout
performance across bilingual populations that differ between their lifetime.
each other with respect to important aspects of their linguistic In contrast to Italian–Sardinian bilinguals, for our Italian–
experience, such as age of acquisition and proficiency. We also English bilinguals, high L2 proficiency was the result of formal
adopt larger sample sizes than many previous studies, such as education and of extensive, albeit recent, immersion (average
Morales et al. (i.e., n > 30 in each group; see Paap, Johnson & length of residence in an English speaking country was 3.5
Sawi, 2015, for review and discussion). years, see section below). Finally, our Italian late passive bilingual
In order to understand the role of specific dimensions of the participants also learnt English in school, but did not have
bilingual experience, we compare four groups of Italian bilinguals advanced proficiency in English nor in any language other than
whose experience ranges from early (i.e., they acquired their two Italian, and no experience of prolonged immersion in an
languages before the age of 6) highly proficient bilingualism, to English-speaking environment. However, they all had a basic or
late (i.e., they acquired their second language after childhood) medium proficiency in English, as required in school and univer-
low proficient bilingualism. Specifically, we compared early highly sity, and a consistent experience of passive use of the language (in
proficient bilinguals (Italian–Sardinian), late highly proficient particular written) throughout their studies. This last group pre-
bilinguals (Italian–English), early passive bilinguals (Italian– sents a linguistic experience that locates it on a low end of a con-
Sardinian Passive), and late passive bilinguals (Italian late passive tinuum of bilingual experiences (passive, late bilingualism). The
bilinguals). With respect to Sardinian full and passive bilinguals, inclusion of this group of participants reflects the fact that com-
so far only two studies have addressed the cognitive effects of parisons should be based on specific dimensions of the linguistic
bilingualism in the Sardinian context. Focusing on children, experience of participants, in order to determine how these
Lauchlan and colleagues found an advantage among Italian– dimensions may affect cognitive abilities. Moreover, the inclusion
Sardinian children, with respect to Italian monolinguals, in a cog- of this group reflects the pervasive nature of multilingualism, and
nitive control test and in a vocabulary test (but not in a digit span the empirical limitations of a dichotomous approach to bilingual-
test nor in an arithmetic test, Lauchlan, Parisi & Fadda, 2012). ism (i.e., bilingual vs monolingual).
Another study similarly showed only limited differences in lin- We hypothesise that the AX-CPT task is sensitive to differ-
guistic and cognitive tests between bilingual and monolingual ences in cognitive control, and may reveal differences between
children (Garraffa, Beveridge & Sorace, 2015). As a minority lan- our bilingual groups, in relation to their different experiences
guage, Sardinian is learnt and used informally, mainly at home (age of acquisition, active and passive proficiency). Specifically,
and with friends, whereas Italian is the main language used at we examine if there is an advantage in accuracy among one or
work and to access the media, and the medium of education. more groups in the “AY” condition, which measures the ability
Our Italian–Sardinian highly proficient bilinguals reported learn- to combine the two mechanisms of cognitive control, while we
ing both Italian and Sardinian during childhood, being fluent in expect all groups to perform equally well on “AX”, “BX” and
both languages and using them daily. In contrast, our Italian– “BY” trials (which do not implicate both control mechanisms).
Bilingualism: Language and Cognition 223

If group differences based on linguistic experience are more participants reported never having become fluent in Sardinian1.
prominent than individual variability in executive functions mea- 5 other participants were tested but excluded from the analysis
sures, these differences should emerge also after we have excluded (2 over 40 years of age, 2 for history of linguistic impairment,
explanations in terms of individual variability, i.e., the overall 1 for performance lower than 20% on all types of trial).
variability across individuals (e.g., overall faster or slower RT), 4) Italian late passive bilinguals (N = 58, 36 females), mean age
but also – and crucially – the variability across individuals in 24.5 (SD = 2.5, range 20–35). These participants were recruited
the relative performance across conditions (e.g., variability across and tested at the University of Milan Bicocca, Italy. They
individuals in relative differences in accuracy in each condition reported a basic or medium proficiency in English, but no
compared to baseline). experience of prolonged immersion in the language; however,
Therefore, we use Morales et al.’s (2013) procedure and ini- they reported using English for their studies and to access the
tially adopt their analysis, i.e., an ANOVA on participants’ overall media. 1 participant reported never having learnt English, and
proportion of accurate responses. We then examine how the 6 participants reported never having become fluent in English2.
inclusion of individual variability affects the pattern of results,
by adopting a mixed-model regression analysis to examine accur- First, from the point of view of linguistic experience, the groups dif-
acy on individual trials, and comparing different random effect fered in terms of exposure to Italian and Sardinian or English, pro-
structures. ficiency in their L2, and frequency of switching between their
languages (see table 1). These differences revealed that Italian–
Sardinian full bilinguals and Italian–English bilinguals were highly
2. Method proficient bilinguals, that Italian–Sardinian passive bilinguals were
less proficient bilinguals, that Italian–Sardinian full and passive
2.1 Participants
bilinguals were early bilinguals, and that Italian–English bilinguals
A total of 200 participants were included in this study, divided were late bilinguals. Finally, Italian participants tested in Milan
in four groups. The common selection criteria were being a native were late, passive bilinguals, rather than monolinguals.
Italian speaker, age (between 18 and 40 years old) and having no Second, mean age and years of education (used as a proxy for
history of language or cognitive impairment. All participants socio-economic status) differed across groups. In addition, self-
completed a Language History Questionnaire that provided rated Italian proficiency was comparable among all Sardinian par-
measures of their proficiency and exposure to their different ticipants and Italian participants tested in Milan, whereas Italian–
languages (Marian, Blumenfeld & Kaushanskaya, 2007; Luk & English participants gave higher ratings of their Italian proficiency.
Bialystok, 2013), rated on Likert scales from 1 to 7 (where 1 Questionnaire responses showed a relation between age, years of
is the minimum). Table (1) shows the differences across the education, and self-rated Italian proficiency. Specifically, the num-
groups. ber of years of education was correlated with ratings of Italian pro-
ficiency (speaking, writing, listening, and reading, all r > 0.261, all
1) Italian–English bilinguals (N = 53, 34 females), mean age 26 p < .001). Age was also correlated to years of education (r = 0.298,
years (SD = 5.6, range 18–40). These participants were p < .001), and to Italian writing (r =.179, p = .010) and reading
Italian native speakers who have been living in Scotland on proficiency (r = .139, p = .048), as well as to L2 listening proficiency
average for 3.7 years (SD = 3.5, range: 6 months–18 years) (r = .169, p = .010). For this reason, and in order to exclude the
and were fluent in both Italian and English. They reported confounding effects of age and years of education on the perform-
to be dominant in Italian and had acquired English in primary ance on the AX-CPT task, these two measures were regressed out
school. These participants were recruited through the from the analysis (see next section, and the limitations section for
University of Edinburgh and through the Italian community further discussion of these potential confounds).
in Edinburgh. One more participant was tested but later
excluded from the analysis because of performance lower
than 20% on all types of trial; another participant was tested 2.2 Procedure and design
but then excluded from the analysis as they reported being All participants were tested individually in a quiet room. The
an early, balanced bilingual. experimental session involved the AX-CPT, the Language
2) Italian–Sardinian bilinguals (N = 46, 23 females), mean age History Questionnaire, two linguistic tasks for the highly profi-
30.5 years (SD = 6.6, range 18–39). These participants were cient bilinguals (total duration 90 minutes), and one linguistic
tested in different locations in Sardinia. They were recruited task for the passive bilinguals (total duration 60 minutes), for
through word of mouth and social networks; in addition to the purpose of a separate study. The order of the tasks was
common recruitment criteria, these participants were required
to be fluent speakers of Sardinian. A further 9 participants 1
Nevertheless, all participants reported some passive proficiency in Sardinian and no
were tested and excluded from the analysis (7 over 40 years consistent active usage. With regards to participants who reported never learning the lan-
of age, one for interruption of the task, and one for an error guage, it appears that they may have perceived a discrepancy between passive proficiency
in the administration of the tasks). and knowledge of the language, or that they may have interpreted ‘learning Sardinian’ as
implying formal instruction. With regards to the fact that some, but not all, participants
3) Italian–Sardinian passive bilinguals (N = 43, 34 females),
reported fluency, it can be the case that some of them answered in relation to their active
mean age 27.8 years (SD = 6, range 19–40). These participants proficiency – the fact that they cannot speak Sardinian fluently, while others may have
were tested and recruited in Sardinia, also through word of answered in relation to their ability to follow a conversation entirely in Sardinian, for
mouth and social networks; in addition to common recruit- example, albeit responding in Italian.
2
ment criteria, these participants were required to know All these participants had some proficiency in English, and reported no active use.
However, they all read texts in English on a daily basis in their studies, and occasionally
Sardinian but not being active or fluent speakers of it. All attend talks in English. As in the case of Italian–Sardinian Passive bilinguals, it appears
participants reported some proficiency in Sardinian, although that these participants may also have interpreted fluency in different ways, and that at
7 participants reported never having ‘learnt Sardinian’; 25 least one participant interpreted learning as implying active fluency.
224 Michela Bonfieni, Holly P. Branigan, Martin J. Pickering and Antonella Sorace

Table 1. Mean and SD (in parentheses) for age (years) and years of education, self-rated language proficiency, exposure, and age of acquisition (AoA, years) (Likert
scales 1–7). Values marked with (†) represent means ignoring missing values.

Italian-English Italian-Sardinian Italian-Sardinian Passive Italian late passive

Age (years) 26.15 (5.64) 30.48 (6.53) 27.88 (5.95) 24.52 (2.58)
Years of Education 17.68 (2.74) 15.48 (3.56) 15.42 (2.81) 16.22 (1.67)
L1 AoA 0.15 (0.98) 0.43 (1.05) 0.00 (0.00) 0.07 (0.41)
L1 AoA Fluent 2.98 (0.84) 3.52 (1.52) 3.07 (0.26) 3.71 (0.50)
L1 Speaking 6.54 (0.63) 6.11 (0.80) 5.98 (0.60) 5.91 (0.78)
L1 Writing 6.24 (0.97) 6.07 (0.90) 5.81 (0.76) 5.98 (0.87)
L1 Listening 6.79 (0.41) 6.54 (0.62) 6.16 (0.78) 6.38 (0.88)
L1 Reading 6.71 (0.49) 6.48 (0.66) 6.21 (0.77) 6.14 (0.78)
L1 Exposure 4.25 (0.81) 4.90 (1.04) 6.29 (0.33) 6.41 (0.44)
†
L2 AoA 7.79 (2.94) 0.93 (1.76) 4.09 (4.43) 7.73 (2.90)†
L2 AoA Fluent 18.18 (5.84) 8.15 (7.23) 10.63 (6.40)† 15.47 (4.08)†
L2 Speaking 5.56 (0.82) 5.83 (0.93) 3.26 (1.56) 3.71 (1.44)
L2 Writing 5.49 (0.95) 5.02 (1.61) 2.51 (1.47) 3.93 (1.41)
L2 Listening 6.00 (0.89) 6.46 (0.66) 4.79 (1.74) 3.97 (1.64)
L2 Reading 6.28 (0.71) 6.04 (1.21) 4.26 (1.72) 4.50 (1.52)
L2 Exposure 3.89 (0.76) 3.51 (1.01) 2.01 (0.82) 1.93 (0.81)
Switch frequency 4.92 (1.79) 5.20 (1.69) 3.16 (1.72) 2.28 (1.36)

systematically counterbalanced across participants: among highly 3. Results

proficient bilingual participants (total n = 99), 28 took the
AX-CPT as their first task, 30 took it as their second, and 41 as As “AX” trials were more frequent than the other types of trials,
their third; among passive bilinguals (total n = 101), 48 took the separate analyses were carried out on accuracy and reaction times
AX-CPT as their first task, and 53 took it as their second. The (RT) in “AX” trials, and on accuracy and RT in “AY”, “BX”, and
other two tasks, for the highly proficient bilingual participants, “BY” trials (Morales et al., 2013, 2015; Braver et al., 2001); RT for
were also counterbalanced in order. To control for any possible incorrect trials were excluded from the analysis. For each analysis,
effect of order of administration, we coded the order of the we regressed out age, years of education, and order of tasks by
AX-CPT task for each participant as a categorical variable with fitting a regression model on accuracy and RT with these three
three levels, and regressed it out from all our analyses, in the variables as predictors. The residuals of these models were then
same way as we dealt with age and years of education (see next used as the dependent variable for further analyses (Coco &
section). All tasks were presented on a 13’’ laptop, 60 cm away Keller 2015).
from the participants’ eyes, in comparable light conditions; the We analysed the data in two ways. First, we analysed overall
instructions and the Language History Questionnaire were in proportions of accurate responses in each condition following
Italian. All participants signed a consent form and were reim- the analysis reported by Morales et al. (2013), i.e., ANOVA, in
bursed £7/h in Scotland and €7/h in Italy for their participation. order to investigate whether there was a difference in accuracy
We adopted the version of the AX-CPT previously described. As between groups when variability between individuals and
mentioned, the AX-CPT presents fast sequences of letters in four variability within individuals across conditions was not taken
types of trials (“AX”, “AY”, “BX”, “BY”, where Y stands for any into account. Second, we examined how the factorization of
probe other than X, and B stands for any cue other than A). individual variability affected the results, by running a mixed-
Letters were presented one by one on a black screen for 300ms, model regression on the residuals of accuracy as a binomial
with an interval between them of 1000ms, so that 4900ms elapsed variable, with a maximal random structure. The motivation to
between the cue and the probe. The task involved 100 trials (70 do so was to implement a better model of accuracy data and
“AX”, 10 “AY”, 10 “BX”, 10 “BY”). The sequence of trials and to use a larger number of data-points to include a more
the sequences of distractors (i.e., any 3 letters except A and X, complete and theoretically motivated random effects structure:
and K and Y for visual similarity) between the cues and the specifically, one that specifies a random intercept for
probes were randomized for each participant. Half the partici- subject and a random slope for condition by subject (Barr
pants pressed the z key for “yes” and the m key for “no”; the et al., 2013; Dixon, 2008). This random effects structure follows
other half pressed m for “yes” and z for “no”. The experiment the hypothesis that not only does performance vary between
lasted approximately 13 minutes and was preceded by on-screen individuals, but also that the difference in performance in
instructions, examples, and a practice session which included 10 each condition varies across individuals. Raw measures of accur-
practice trials. Half the way through the experiment, participants acy and RT are presented in Table 2 and 3 and visualised in
were invited to take a break. Figure 2.
Bilingualism: Language and Cognition 225

Table 2. Mean accuracy (proportions) and SD (in parentheses) across conditions and groups.

Condition Italian-English Italian-Sardinian Italian-Sardinian Passive Italian late passive

AX 0.93 (0.23) 0.89 (0.29) 0.89 (0.29) 0.9 (0.26)

AY 0.74 (0.40) 0.78 (0.34) 0.69 (0.40) 0.64 (0.43)
BX 0.87 (0.23) 0.84 (0.26) 0.86 (0.24) 0.85 (0.35)
BY 0.93 (0.14) 0.97 (0.09) 0.93 (0.15) 0.94 (0.14)

Table 3. Mean Reaction Times (ms) and SD (in parentheses) across conditions and groups.

Condition Italian-English Italian-Sardinian Italian-Sardinian Passive Italian late passive

AX 318 (49) 342 (78) 320 (72) 320 (51)

AY 465 (81) 464 (116) 436 (115) 454 (108)
BX 246 (135) 284 (117) 260 (87) 257 (114)
BY 270 (101) 281 (115) 278 (113) 262 (108)

Fig. 2. Accuracy and Reaction Times (ms) on the probe across conditions and groups. Bars = SD.
226 Michela Bonfieni, Holly P. Branigan, Martin J. Pickering and Antonella Sorace

Fig. 4. Model fit of residuals of RT in “AY”, “BX”, “BY”. IS: Italian-Sardinian, IE:
Fig. 3. Model fit of residuals of accuracy in “AY”, “BX”, “BY”. IS: Italian-Sardinian, IE: Italian-English, ISP: Italian-Sardinian Passive, ILP: Italian Late Passive. Bars = 95% C.I.
Italian-English, ISP: Italian-Sardinian Passive, ILP: Italian Late Passive. Bars = 95% C.I.

3.1 Analysis of accuracy proportions linear mixed-model regression including group and condition as
fixed effects (Figure 4). There was a main effect of condition (p
We first analysed accuracy as overall proportions of accurate < .001), with longer RT in “AY” (β = 183.768, SE = 14.644, t =
responses (i.e., aggregated over individual observations), adopting 12.546) with respect to “BY”. The effect of group was not signifi-
mixed regression models with a random intercept for subject. cant (p = .870), and there was no interaction between group and
These mixed-model regressions are equivalent to repeated- condition (p = .390). We also ran a mixed-model regression on
measure ANOVAs, following Morales et al. (2013). We analysed un-aggregated RT with a full random effect structure (specified
“AX” trials separately from “AY”, “BY”, “BX” trials. For “AX”, as in the models presented in the next section). The results of
we fitted a mixed-model regression with a random intercept for this analysis were comparable to the results of the repeated meas-
subject and group as fixed effect. This analysis showed no differ- ure ANOVA.
ence between the groups (p = .148).
For the analysis of “AY”, “BY”, “BX” conditions, we fitted a
mixed-model regression with a random intercept for subject, and 3.3 Binomial mixed-model regression of accuracy
group and condition as fixed effects (Figure 3). We found a main Our second analysis of accuracy aimed to evaluate whether the
effect of condition (p < .001): accuracy was significantly lower in results obtained through the analysis of aggregated scores would
the “AY” condition (β = −0.180, SE = 0.030, t = −5.859) and in hold after the inclusion of individual variability, i.e., random
the “BX” condition (β = −0.123, SE = 0.030, t = −4.024), compared effects structure modelling variability between individuals, as
to the “BY” condition (which constitutes the baseline). In these well as variability between individuals across conditions.
trials, the effect of group was not significant (p = .140), but Therefore, we ran a further analysis on accuracy as a binomial
the interaction between condition and group was significant dependent variable. We first regressed out age, years of education
(p = .003). Pairwise comparison (Tukey’s test) showed that Italian and order of trials, as in our first analysis.
late passive bilinguals were significantly worse on the “AY” condi- For the “AX” condition, we fitted a mixed-model regression
tion than Italian-Sardinian bilinguals (Estimate = −0.128, SE = specifying a by-subject intercept and group as the fixed effect.
0.033, z-value = −3.830, adjusted p < .01); Italian late passive As in our first analysis, we found no effect of group (p = .129).
bilinguals were marginally worse than Italian–English participants For the “AY”, “BY”, and “BX” conditions, we fitted a mixed-
(Estimate = −0.099, SE = 0.032, z-value = −3.081, adjusted p = .084). model regression specifying a by-subject intercept and a condition
Groups did not differ either in the “BX” condition (all adjusted by subject slope. Group and condition were the fixed effects.
p > .977) or in the “BY” condition (all adjusted p > .999). The effect of condition was significant (p < .001): performance
in “AY” and in “BX” was significantly worse than in “BY”
(respectively: β = −1.384, SE = 0.225, t = −6.134; β = −0.937, SE
3.2 Analysis of reaction times
= 0.194, t = −4.815). The effect of group was not significant
With regards to RT, we fitted two comparable linear mixed-model (p = .438), but the interaction between condition and group was
regressions, equivalent to repeated-measure ANOVA (i.e., with significant (p = .019). However, pairwise comparison with
only by-subject random intercept) on aggregated RT. In RT in Tukey’s test showed that, in the “AY” condition, there was no dif-
“AX” trials, we found no difference between groups (p = .502). ference between groups. In particular, the difference between
For RT in “AY”, “BY”, “BX” conditions, we fitted a comparable Italian–Sardinian bilinguals and Italian late passive bilinguals
Bilingualism: Language and Cognition 227

was only marginally significant (Estimate = −0.898, SE = 0.290, the Italian–Sardinian bilingual group performing better than the
z-value = −3.091, adjusted p = .076). No difference was found Italian late passive group on the “AY” condition, but showing
across groups on “BX” and “BY” conditions (all adjusted comparable performance on the “AX”, “BX” and “BY” conditions.
p > .971), suggesting that the interaction between groups and The Italian–English bilingual group performed marginally better
conditions was led by differences, across groups, on different on this condition with respect to the late passive group. Better
conditions, but not in each condition individually examined. performance on the “AY” condition – all other conditions being
To discriminate the specific contribution of the random effects equal – can be argued to reflect the ability to adjust proactive
structure we tested two further models. First, to demonstrate that and reactive control mechanisms to adapt to the context, follow-
the inclusion of both a random intercept by subject and a random ing the assumption of a trade-off between the different mechan-
slope for condition by subject was the critical factor affecting isms of cognitive control. These results are compatible with
the generalizability of the interaction between groups and previous claims for the effect of the bilingual experience on the
conditions on “AY” trials, we compared this model to a model of flexible engagement and modulation of mechanisms of cognitive
the residuals of accuracy (after the regression of age, years of control (Morales et al., 2013, 2015; Green & Abutalebi, 2013).
education and order of tasks) that included only a random intercept Importantly, among our four bilingual groups, we found a dif-
by subject (i.e., did not include a random slope for condition by ference between early, highly proficient bilinguals on the one
subject). While no differences were found across groups on “BX” hand, and late, passive bilinguals on the other. We therefore
and “BY” conditions (all adjusted p > .96), the performance of extended the results of Morales et al. (2013, 2015), by identifying
the Italian late passive group on “AY” trials was significantly the contribution of specific aspects of the bilingual experience on
worse than the performance of the Italian–Sardinian group the modulation of control processes. Specifically, high proficiency
(Estimate = −0.898, SE = 0.219, z-value = −4.098, adjusted in both active and passive modalities was related to better per-
p < .01), and so was the performance of the Italian–English group formance, but only early highly proficient bilinguals seemed to
with respect to the Italian late passive group (adjusted p = .017). perform significantly better than late, low proficient passive bilin-
In a further model that eliminated the random structure altogether guals, whereas highly proficient late bilinguals did not. This sug-
(i.e., included neither a random intercept by subject, nor a random gests that early age of acquisition and high proficiency (in both
slope for condition by subject), not only did both highly proficient active and passive modalities) may result in cognitive effects,
bilingual groups show an advantage over the late passive group but that each of these variables, individually examined, may not
(Italian late passive bilinguals – Italian–Sardinian bilinguals: relate to better performance on cognitive control. This result
Estimate = −0.898, SE = 0.164, z-value = −5.467, adjusted p < .01; highlights the interaction of different dimensions of the bilingual
Italian late passive bilinguals – Italian–English bilinguals: experience, and the importance of focusing on these dimensions
Estimate = −0.749, SE = 0.158, z-value = −4.742, adjusted p < .01), in the study of the relation between bilingualism and executive
but Italian–Sardinian bilinguals also performed significantly functions. The same analytical approach, however, did not show
better on “AY” trials than the Italian–Sardinian passive bilinguals a difference between groups with respect to RT, contra Morales
(Italian–Sardinian passive bilinguals – Italian–Sardinian bilinguals: et al.’s (2013) results, but in keeping with Morales et al. (2015).
Estimate = −0.626, SE = 0.176, z-value = −3.549, adjusted p = .019). Second, we evaluated the generalizability of these findings, not
Again, no difference was found across groups on “BX” and “BY” only by using different populations and larger sample sizes than
conditions (all adjusted p > .8). in Morales et al. (2013), but also by investigating whether group
differences remained when we included an accurate measure of
individual variability in the analysis, based on the hypothesis
4. Discussion
that individual variance in executive functions may represent an
The first aim of this study was to examine the effect of the bilin- important confound in group comparisons, and affect the gener-
gual experience on cognitive control abilities, using a task whose alizability of the findings. We therefore analysed raw accuracy, i.e.,
structure was theoretically motivated by an established model of accuracy in binomial format rather than as proportion scores,
executive functions and its proposed relation to language control using a mixed-model regression, that allowed us to model both
in bilinguals. Specifically, we compared the performance of four random variability between subjects (by-subject intercepts) as
different bilingual groups, which differed with respect to age of well as individual variability in performance across conditions
acquisition and proficiency, on the AX-CPT, a task of continuous (random slopes for condition by subject). This analysis supported
performance previously used to evaluate the dual-mechanism the pattern and direction of data that we found in the analysis
framework of cognitive control (Braver et al., 2007; Braver over proportions of accurate responses; but, critically, it did not
2012). The second aim was to evaluate whether group differences show a significant difference between groups on the “AY” condi-
previously found using the same task stand up to the factorization tion (i.e., while the interaction between group and condition was
of individual variability, and how they relate to specific differences still significant, the pairwise comparison between groups in each
in type of bilingual experience (along the dimensions of age of condition was not).
acquisition and proficiency). We now discuss our results relating To discriminate the contribution of the random effects struc-
to these aims in turn, and then discuss the limitations of our ture to the analysis of this type of data, we compared the full ran-
study. dom effects model to a by-subject-intercept-only model, as well as
First, in a series of analyses that aggregated accuracy over indi- to a model with no random structure at all. When the random
vidual observations only using by-subject intercepts as a measure effects structure was simplified in this way, the results suggested
of individual variability, we found a group difference in perform- group differences. The by-subject-intercept-only model suggested
ance between Italian–Sardinian bilinguals and Italian late passive an advantage in favour of both highly proficient bilingual groups
bilinguals, consistent with previous studies (Morales et al., 2013, with respect to the late passive group. The model with no random
2015). Specifically, we found a significant interaction between effects structure further suggested an advantage for the Italian–
group and condition in the accuracy of our participants, with Sardinian active bilinguals over the Italian–Sardinian Passive
228 Michela Bonfieni, Holly P. Branigan, Martin J. Pickering and Antonella Sorace

bilinguals (in addition to an advantage for both groups over the which can be considered a proxy for socio-economic status in
late passive group). Taken together, these analyses show that the the Italian context – the participants in the Italian–English
exclusion of individual variability is directly related to the gener- group and in the Italian late passive group were university stu-
alisability of group differences. dents, primarily at postgraduate level in the former group, and
Hence, this comparison highlights the importance of consider- at the graduate level in the latter group. Student status is linked
ing individual variability in the study of the relationship between to the context of recruitment, which happened through word of
language and cognitive control, both methodologically and theor- mouth in Sardinia, and primarily through university recruitment
etically. Analyses that did not consider such variability (i.e., in channels in Scotland and in Italy. Age and student status may
which the random effects structure was reduced) produced results obviously have important relationships with measures of execu-
that were consistent with a group difference in proficiency, inde- tive functions, language processing and general intelligence,
pendent of age of acquisition, and – when the random effects while context of recruitment may relate to attitudes and motiva-
structure was completely eliminated – an advantage of highly pro- tions towards participation in the experiments (e.g., participants
ficient bilinguals over low proficient ones. But as our analyses in Sardinia may have been more intrinsically motivated while par-
show, the exclusion of individual variability misleadingly flattens ticipants recruited through university channels may have been
the differences between our bilingual groups, and inflates the more extrinsically motivated). While controlling more strictly
effect of group averaging, a statistical artefact not uncommon in for these differences at the recruitment stage would have been
psychological research (Speelman & McGann, 2013; Speelman ideal, we controlled for the possible effects of differences in age
& Muller Townsend, 2015). By doing so, it also inflates Type I and in level of education by analysing the correlation within
error. Thus, the exclusion of individual variability can result in responses to the language history questionnaire across groups,
a spurious link between individual aspects of the bilingual experi- and by regressing out these predictors from the analysis – i.e., per-
ence (e.g., age of acquisition, language proficiency) and perform- forming our analyses on the variance not explained by these fac-
ance in cognitive control. Consequently, our findings demonstrate tors. However, it is important to notice how further research in
that the inappropriate factorization on individual variability can this field needs to address these aspects in a more controlled way.
ultimately obscure the contribution of these specific dimensions To conclude, our study identifies an explicit theoretical model
to a model of bilingual language control, as well as of a model and a reliable task that suggest a possible relationship between
of the bilingual mind in terms of cognitive plasticity. specific aspects of the bilingual experience (early age of acquisi-
While our study suggests important implications for future tion and high proficiency in both active and passive modalities)
research on bilingualism and non-linguistic abilities, it also pre- and cognitive control abilities. However, our study does not sup-
sents various limitations. Specifically, the four groups of partici- port the unequivocal existence of cognitive effects related to the
pants we tested did not only differed in terms of age of bilingual experience, as we found no more than a marginal
acquisition and proficiency, but also in terms of language dis- trend in favour of early, highly proficient bilinguals over late pas-
tance, contexts of use, as well as in other ways unrelated to sive ones: the effects of the bilingual experience may not be strong
their bilingual experience. We first address the linguistic differ- enough to over-ride the effects of individual variability on execu-
ences and then the non-linguistic ones. tive functions. Therefore, our study highlights the empirical
With regards to language distance, Italian and Sardinian are of aspects that limit our ability to measure the effects of bilingualism
course more closely related than Italian and English (from the on general cognition: as we show, this type of investigation cannot
points of view of typology, syntax, morphology and phonology). be meaningfully pursued without taking into account individual
In addition, Italian, Sardinian and English do not have the variability, which represents a major challenge in the study of
same status, as Sardinian – albeit official – is a minority language. executive functions. These two results – the identification of a the-
Language distance for sure represents an important factor for oretical model and of a laboratory task, on the one hand, and the
bilingual language processing; however, its effects on cognitive demonstration of the role of individual variability in the study of
control are undocumented, and thereby represent an interesting bilingualism, on the other – can inform theoretical and methodo-
venue for future research. logical choices for future research on the cognitive effects of the
With regards to contexts of use, Italian and English were used bilingual experience.
in both formal and informal context in the Italian–English and
Acknowledgements. Testing and recruiting in Sardinia has been possible
the Italian late passive group, whereas in the Italian–Sardinian
thanks to the help of Giuseppe Corongiu, Francesco Cheratzu, su
groups, Italian was typically associated with formal contexts,
Coordinamentu pro su Sardu Ufitziale, Daniela Corongiu, Giuseppe Melis,
and Sardinian with informal ones (and with informal learning Salvatore Serra, Dolores Lai, Maria Antonietta Pinna, Maria Leonarda
too). The effects of contexts of use have been related to the Corredda, Francesca Sini, Giovanna Bosu, Immacolata Salis. We thank
engagement of cognitive control components (Green & Maria Teresa Guasti, Francesca Foppolo and Mirta Vernice for helping testing
Abutalebi, 2013). We operationalised contexts of use in terms of and recruiting in Milan. This project has received funding from the European
active versus passive proficiency and exposure – a distinction Union’s Seventh Framework Programme for research, technological develop-
that we considered in the comparison between our groups. ment and demonstration under grant agreement no. 613465.
However, developing a quantitative measure of this aspect of
the bilingual experience would undoubtedly be useful for future
research.
Finally, our groups also presented differences in age, level of References
education, and context of recruitment. With respect to age, Abutalebi J, Della Rosa PA, Ding G, Weekes B, Costa A and Green DW
while all participants were aged between 18 and 40, the partici- (2013) Language proficiency modulates the engagement of cognitive control
pants tested in Sardinia were on average older than the partici- areas in multilinguals. Cortex 49, 905–911. doi:10.1016/j.cortex.2012.08.018
pants tested in Scotland (Italian–English group) and in Italy Bak TH (2016) Cooking pasta in La Paz. Linguistic Approaches to Bilingualism
(Italian late passive group). With respect to the level of education – 6, 699–717. doi:10.1075/lab.16002.bak
Bilingualism: Language and Cognition 229

Barr DJ, Levy R, Scheepers C and Tily HJ (2013) Random effects structure Friedman NP (2016) Research on individual differences in executive func-
for confirmatory hypothesis testing : Keep it maximal. Journal of Memory tions: Implications for the bilingual advantage hypothesis. Linguistic
and Language 68, 255–278. doi:10.1016/j.jml.2012.11.001 Approaches to Bilingualism 6, 535–548. doi:10.1075/lab.15041.fri
Bialystok E and Martin MM (2004) Attention and inhibition in bilingual chil- Gade M (2015) On tasks and cognitive constructs for the bilingual (non-)
dren: evidence from the dimensional change card sort task. Developmental advantage. Cortex 73, 347–348. doi:10.1016/j.cortex.2015.07.017
Science 7, 325–339. doi:10.1111/j.1467-7687.2004.00351.x García-Pentón L, Pérez Fernández A, Iturria-Medina Y, Gillon-Dowens M
Bialystok E, Craik FIM and Luk G (2012) Bilingualism: consequences for and Carreiras M (2014) Anatomical connectivity changes in the bilingual
mind and brain. Trends in Cognitive Sciences 16, 240–50. doi:10.1016/ brain. NeuroImage 84, 495–504. doi:10.1016/j.neuroimage.2013.08.064
j.tics.2012.03.001 Garraffa M, Beveridge M and Sorace A (2015) Linguistic and cognitive skills
Bialystok E, Craik FIM and Ruocco AC (2006) Dual-modality monitoring in in Sardinian-Italian bilingual children. Frontiers in Psychology 6, 1–15.
a classification task: the effects of bilingualism and ageing. Quarterly doi:10.3389/fpsyg.2015.01898
Journal of Experimental Psychology (2006), 59(11), 1968–83. http://doi. Gathercole VC, Thomas EM, Kennedy I, Prys C, Young N, Vinas Guasch N,
org/10.1080/17470210500482955 Roberts EJ, Hughes EK and Jones L (2014) Does language dominance
Bialystok E, Craik FIM and Ryan J (2006) Executive control in a modified affect cognitive performance in bilinguals? Lifespan evidence from pre-
antisaccade task: Effects of aging and bilingualism. Journal of Experimental schoolers through older adults on card sorting, Simon, and metalinguistic
Psychology. Learning, Memory, and Cognition 32(6), 1341–54. http://doi. tasks. Frontiers in Psychology 5, 11. doi:10.3389/fpsyg.2014.00011
org/10.1037/0278-7393.32.6.1341 Green D (1998) Mental control of the bilingual lexico-semantic system. Bilingualism:
Bialystok E, Craik FIM, Klein R and Viswanathan M (2004) Bilingualism, Language and Cognition 1, 67–81. doi:10.1017/S1366728998000133
aging, and cognitive control: Evidence from the Simon task. Psychology Green DW and Abutalebi J (2013) Language control in bilinguals: The adap-
and Aging 19, 290–303. doi:10.1037/0882-7974.19.2.290 tive control hypothesis. Journal of Cognitive Psychology 25, 515–530.
Bialystok E, Craik F and Luk G (2008) Cognitive control and lexical access in doi:10.1080/20445911.2013.796377
younger and older bilinguals. Journal of Experimental Psychology-Learning Grundy JG and Keyvani-Chahi AK (2017) Post-conflict slowing effects in mono-
Memory and Cognition 34, 859–873. doi:10.1037/0278-7393.34.4.859 lingual and bilingual children. Developmental Science 20. doi:10.1111/
Blumenfeld HK and Marian V (2013) Parallel language activation and cognitive desc.12488
control during spoken word recognition in bilinguals. Journal of Cognitive Grundy JG and Timmer K (2016) Cognitive control mechanisms underlying per-
Psychology 25, 37–41. http://doi.org/10.1080/20445911.2013.812093 formance differences between monolinguals and bilinguals. In JW Schwieter
Blumenfeld HK and Marian V (2014) Cognitive control in bilinguals: (Ed.), The cognitive control of multiple languages: Empirical studies and new
Advantages in Stimulus-Stimulus inhibition. Bilingualism 17, 610–629. directions. Amsterdam, The Netherlands: John Benjamins Publishing Company
doi:10.1017/S1366728913000564 Hartsuiker RJ (2015) Why it is pointless to ask under which specific circum-
Braver TS (2012) The variable nature of cognitive control: a dual mechan- stances the bilingual advantage occurs. Cortex 73, 336–337. doi:10.1016/
isms framework. Trends in Cognitive Sciences 16, 106–13. doi:10.1016/ j.cortex.2015.07.018
j.tics.2011.12.010 Jared D (2015) What is the theory? Cortex 73, 361–3. doi: 10.1016/
Braver TS, Barch DM, Keys BA, Carter CS, Cohen JD, Kaye JA, j.cortex.2015.07.009.
Janowsky JS, Taylor SF, Yesavage JA, Mumenthaler MS, Jagust WJ and Klein D, Mok K, Chen JK and Watkins KE (2014) Age of language learning
Reed BR (2001) Context processing in older adults: evidence for a theory relat- shapes brain structure: A cortical thickness study of bilingual and monolingual
ing cognitive control to neurobiology in healthy aging. Journal of Experimental individuals. Brain and Language 131, 20–24. doi:10.1016/j.bandl.2013.05.014
Psychology: General 130, 746–763. doi:10.1037/0096-3445.130.4.746 Lauchlan F, Parisi M and Fadda R (2012) Bilingualism in Sardinia and
Braver TS, Gray JR and Burgess GC (2007) Explaining the Many Varieties of Scotland: Exploring the cognitive benefits of speaking a “minority” lan-
Working Memory Variation: Dual mechanisms of cognitive control. In guage. International Journal of Bilingualism 17, 43–56. doi:10.1177/
ARA Conway, C Jarrold, MJ Kane, A Miyake and JN Towse (eds), 1367006911429622
Variation in Working Memory. Oxford: Oxford University Press, (pp. Luk G and Bialystok E (2013) Bilingualism is not a categorical variable:
76–106). Interaction between language proficiency and usage. Journal of Cognitive
Braver TS, Paxton JL, Locke HS and Barch DM (2009) Flexible neural Psychology 25, 605–621. doi:10.1080/20445911.2013.795574
mechanisms of cognitive control within human prefrontal cortex. Ma F, Li S and Guo T (2016) Reactive and proactive control in bilingual word
Proceedings of the National Academy of Sciences of the United States of production: An investigation of influential factors. Journal of Memory and
America 106, 7351–7356. doi:10.1073/pnas.0808187106 Language 86, 35–59. http://doi.org/10.1016/j.jml.2015.08.004
Buchweitz A and Prat C (2013) The bilingual brain: flexibility and control in Marian V and Spivey M (2003) Competing activation in bilingual language
the human cortex. Physics of Life Reviews 10, 428–43. doi:10.1016/ processing: Within- and between-language competition. Bilingualism:
j.plrev.2013.07.020 Language and Cognition 6, 97–115. doi:10.1017/S1366728903001068
Coco MI and Keller F (2015) Integrating Mechanisms of Visual Guidance in Marian V, Blumenfeld HK and Kaushanskaya M (2007) The Language
Naturalistic Language Production. Cognitive Processing 16, 131–150. Experience and Proficiency Questionnaire (LEAP-Q): assessing language
doi:10.1007/s10339-014-0642-0 profiles in bilinguals and multilinguals. Journal of Speech, Language, and
Costa A and Sebastián-Gallés N (2014) How does the bilingual experience Hearing Research 50, 940–967. doi: 10.1044/1092-4388(2007/067)
sculpt the brain? Nature Reviews. Neuroscience 15, 336–45. doi:10.1038/ Marzecová A, Bukowski M, Correa Á., Boros M, Lupiáñez J and Wodniecka Z
nrn3709 (2013) Tracing the bilingual advantage in cognitive control: The role of flexibil-
Costa A, Hernández M and Sebastián-Gallés N (2008) Bilingualism aids con- ity in temporal preparation and category switching. Journal of Cognitive
flict resolution: evidence from the ANT task. Cognition 106, 59–86. Psychology 25(5), 586–604. http://doi.org/10.1080/20445911.2013.809348
doi:10.1016/j.cognition.2006.12.013 Mishra RK, Hilchey MD, Singh N and Klein RM (2012) On the time course of
Costa A, Hernández M, Costa-Faidella J and Sebastián-Gallés N (2009) On exogenous cueing effects in bilinguals: higher proficiency in a second language
the bilingual advantage in conflict processing: now you see it, now you is associated with more rapid endogenous disengagement. Quarterly Journal of
don’t. Cognition 113, 135–149. doi:10.1016/j.cognition.2009.08.001 Experimental Psychology 65, 1502–10. doi:10.1080/17470218.2012.657656
Dixon P (2008) Models of accuracy in repeated-measures designs. Journal of Miyake A and Friedman NP (2012) The Nature and Organization of Individual
Memory and Language 59, 447–456. doi:10.1016/j.jml.2007.11.004 Differences in Executive Functions: Four General Conclusions. Current
Fedorenko E (2014) The role of domain-general cognitive control in lan- Directions in Psychological Science 21, 8–14. doi:10.1177/0963721411429458
guage comprehension. Frontiers in Psychology 5, 1–17. doi:10.3389/ Moradzadeh L, Blumenthal G and Wiseheart M (2014) Musical Training,
fpsyg.2014.00335 Bilingualism, and Executive Function: A Closer Look at Task Switching
Fedorenko E and Thompson-Schill SL (2014) Reworking the language network. and Dual-Task Performance. Cognitive Science 1–29. http://doi.org/10.
Trends in Cognitive Sciences 18, 120–127. doi:10.1016/j.tics.2013.12.006 1111/cogs.12183
230 Michela Bonfieni, Holly P. Branigan, Martin J. Pickering and Antonella Sorace

Morales J, Gómez-Ariza CJ and Bajo MT (2013) Dual mechanisms of cogni- Brain Areas More Than Monolinguals to Perform Non-Linguistic
tive control in bilinguals and monolinguals. Journal of Cognitive Psychology Switching Tasks. PLoS One 8, e73028. doi:10.1371/journal.pone.0073028
25, 531–546. doi:10.1080/20445911.2013.807812 Speckman PL, Rouder JN, Morey RD and Pratte MS (2008) Delta Plots and
Morales J, Yudes C, Gómez-Ariza CJ and Bajo MT (2015) Bilingualism Coherent Distribution Ordering. The American Statistician 62, 262–266.
modulates dual mechanisms of cognitive control: Evidence from ERPs. doi:10.1198/000313008X333493
Neuropsychologia 66, 157–169. doi:10.1016/j.neuropsychologia.2014.11.014 Speelman CP and McGann M (2013) How mean is the mean? Frontiers in
Paap KR (2014) The role of componential analysis, categorical hypothesising, Psychology 4, 1–12. doi:10.3389/fpsyg.2013.00451
replicability and confirmation bias in testing for bilingual advantages in Speelman CP and Muller Townsend KL (2015) Attaining automaticity in the
executive functioning. Journal of Cognitive Psychology 26(3), 242–255. visual numerosity task is not automatic. Frontiers in Psychology 6, 1–6.
http://doi.org/10.1080/20445911.2014.891597 doi:10.3389/fpsyg.2015.01744
Paap KR and Greenberg ZI (2013) There is no coherent evidence for a bilin- Stocco A and Prat CS (2014) Bilingualism trains specific brain circuits
gual advantage in executive processing. Cognitive Psychology 66, 232–58. involved in flexible rule selection and application. Brain and Language
doi:10.1016/j.cogpsych.2012.12.002 137, 50–61. doi:10.1016/j.bandl.2014.07.005
Paap KR and Sawi O (2014) Bilingual advantages in executive functioning : Thierry G and Sanoudaki E (2012) Activation syntaxique non-sélective à la lan-
problems in convergent validity, discriminant validity, and the identifica- gue chez le bilingue précoce. Revue Française de Linguistique Appliquée 17,
tion of the theoretical constructs, Frontiers in psychology 5, 1–15. 33–48.
doi:10.3389/fpsyg.2014.00962 Thierry G and Wu YJ (2007) Brain potentials reveal unconscious translation
Paap KR, Johnson HA and Sawi O (2015) Bilingual advantages in executive during foreign-language comprehension. Proceedings of the National
functioning either do not exist or are restricted to very specific and undeter- Academy of Sciences of the United States of America 104, 12530–12535.
mined circumstances. Cortex 73. doi: 10.1016/j.cortex.2015.04.014. doi: 10.1073/pnas.0609927104
Paxton JL, Barch DM, Storandt M and Braver TS (2006) Effects of environ- Valian V (2014) Bilingualism and cognition. Bilingualism: Language and
mental support and strategy training on older adults’ use of context. Cognition 18, 3–24. doi:10.1017/S1366728914000522
Psychology and Aging 21, 499–509. doi:10.1037/0882-7974.21.3.499 Wu YJ and Thierry G (2010) Chinese – English Bilinguals Reading English
Pratte M, Rouder J, Morey R and Feng C (2010) Exploring the differences in dis- Hear Chinese, The Journal of Neuroscience 30, 7646–7651. doi: 10.1523/
tributional properties between Stroop and Simon effects using delta plots. JNEUROSCI.1602-10.2010.
Attention, Perception & Psychophysics 72, 2013–2025. doi:10.3758/APP Wu YJ and Thierry G (2017) Brain potentials predict language selection
Prior A and Macwhinney B (2009) A bilingual advantage in task switching. before speech onset in bilinguals. Brain and Language 171, 23–30. http://
Bilingualism: Language and Cognition 13, 253–262. doi:10.1017/ doi.org/10.1016/j.bandl.2017.04.002
S1366728909990526 Ye Z and Zhou X (2009) Executive control in language processing.
Rodríguez-Pujadas A, Sanjuán A, Ventura-Campos N, Román P, Martin C, Neuroscience and Biobehavioral Reviews 33, 1168–1177. doi:10.1016/
Barceló F, Costa A and Ávila C (2013) Bilinguals Use Language-Control j.neubiorev.2009.03.003
Copyright © Cambridge University Press 2019