Content uploaded by Aina Casaponsa
Author content
All content in this area was uploaded by Aina Casaponsa on Feb 02, 2018
Content may be subject to copyright.
How do bilinguals identify the language of the words they read?
Aina Casaponsa1, Manuel Carreiras1,2 and Jon Andoni Duñabeitia1
1 BCBL. Basque Center on Cognition, Brain and Language; Donostia, Spain
2 Ikerbasque, Basque Foundation for Science; Bilbao, Spain
Contact information:
Aina Casaponsa Galí
Basque Center on Cognition, Brain and Language (BCBL)
Paseo Mikeletegi 69, 2nd floor
20009 Donostia, SPAIN
a.casaponsa@bcbl.eu
Telephone: +34943309300
1
Abstract
How do bilinguals detect the language of the words they read? Recent electrophysiological
research using the masked priming paradigm combining primes and targets from different
languages has shown that bilingual readers identify the language of the words within
approximately 200 milliseconds. Recent evidence shows that language-detection mechanisms
vary as a function of the orthographic markedness of the words (i.e., whether or not a given word
contains graphemic combinations that are not legal in the other language). The present study
examined how the sub-lexical orthographic regularities of words are used as predictive cues.
Spanish-Basque bilinguals and Spanish monolinguals (control group) were tested in an Event-
Related Potential (ERP) experiment, using the masked priming paradigm. During the
experiment, Spanish targets were briefly preceded by unrelated Spanish or Basque words.
Unrelated Basque words could contain bigram combinations that are either plausible or
implausible in the target language (Spanish). Results show a language switch effect in the N250
and N400 components for marked Basque primes in both groups, whereas, in the case of
unmarked Basque primes, language switch effects were found in bilinguals but not
monolinguals. These data demonstrate that statistical orthographic regularities of words play an
important role in bilingual language detection, and provide new evidence supporting the
assumptions of the BIA+ extended model.
Keywords: Bigrams, Bilingualism, Multilingual reading, Orthographic cues, Masked language
switch cost priming, Event-related potentials.
2
1. Introduction
It is well-established that when bilinguals process words, they unconsciously access both of their
languages, even when set in a single language context (Dijkstra et al., 2000; Dimitropoulou et al.,
2011a, 2011b; Duyck et al., 2007; Duyck and Warlop, 2009; Grossi et al., 2012; Hoshino and
Thierry, 2012; Ng and Wicha, 2013; Midgley et al., 2008; Perea et al., 2008; Spalek et al., 2014;
Schwartz et al., 2007; Thierry an Wu, 2004; Thierry and Wu, 2007; Van Heuven et al., 1998,
2008; Zhang et al., 2011). However, bilinguals need to know in which language they are reading
to correctly retrieve the meaning of words. Consider the case of false friends (words that strongly
overlap between languages but have different meanings). For instance, the word pie refers to a
culinary preparation encased in pastry in English and ´foot’ in Spanish, which cannot be
disambiguated without knowing in which language the word is presented (i.e., language
attribution).
Despite the importance of language attribution in bilinguals, the mechanisms by which a given
word form is associated with a given language remain unclear. The main aim of this study is to
provide new evidence about the mechanisms underlying language identification by exploring the
time-course of unconscious and automatic language switch effects in reading. More specifically,
we investigated whether language-specific sub-lexical cues can drive language attribution.
Several studies have suggested that in an ambiguous language context (i.e., when languages have
similar scripts; e.g., Spanish and Catalan) language information is accessed through the lexical
representations of words (Chauncey et al, 2008, 2011; Dijkstra et al., 1999; Dijkstra and Van
Heuven, 2002; Midgley et al., 2009a, 3009b; Von Studnitz and Green, 2002). In this case, effects
of language identification result from top-down modulations from the language nodes feeding
information back to the lexical units. However none of these studies have explored whether these
effects can be modulated by sub-lexical information.
3
In many language combinations sharing the same script (e.g., Dutch and English or Basque and
Spanish), one can easily find words that are orthographically marked (i.e., words that contain
sub-lexical features that are only plausible in one language). For example, the Basque word
neska -´girl´, contains the bigram “sk” which is illegal in Spanish. At the same time, it is
relatively easy to find words that are orthographically unmarked, such as the Basque word mutil -
´boy´, which complies with Spanish orthotactics1). Along these lines, Van Kesteren et al. (2012)
showed that language-specific orthography can guide language decisions in bilinguals, and
proposed a direct link between sub-lexical information and language membership. In their study,
Norwegian-English bilinguals completed a series of language decision tasks featuring marked
and unmarked Norwegian and English words. The authors demonstrated that language
membership can be accessed via lexical representations, but critically also via sub-lexical levels
(see also Casaponsa et al., 2014; Casaponsa and Duñabeitia, in press; Lemhöfer et al., 2011;
Orfanidou and Sumner, 2005; Vaid and Frenck-Mestre, 2002). Therefore, Van Kesteren et al.
(2012) proposed an extension to the Bilingual Interactive Activation Plus model (BIA+ ; Dijkstra
and Van Heuven, 2002). The addition to the model is a sub-lexical language node, which can be
accessed directly through excitatory connections from sub-lexical levels and then be read out by
the task/decision system. Thus, an orthographically marked word containing language-specific
orthographic cues can lead to effective language attribution solely on the basis of sub-lexical
information. Although the number and nature of the connections of the sub-lexical language
node with other levels of word processing remain underspecified, the BIA+ extended model
highlights the importance of language-specific orthography in reading (see Casaponsa and
Duñabeitia, in press, for review).
1 Orthotactics is a term used to define the orthographic characteristics of the existent vocabulary in a
specific language that focuses on the sequences of letter combinations that are allowed in that language.
4
The present study aims to shed light on the influence of orthographic markedness by tracking the
time-course of language identification in bilingual word recognition, by investigating the types
of cues that help bilingual readers detect the language code. The masked priming language-
switching paradigm combined with Event-Related Potential (ERP) recording is an extremely
valuable approach for researchers exploring the time-course of language identification. Using
such approach, it has been shown that bilinguals can automatically and unconsciously detect the
language of a word within 200 milliseconds of stimulus onset (Chauncey et al., 2008, 2011;
Dunabeitia et al., 2010b). In masked priming ERP studies exploring language detection,
participants are typically presented with primes for 50 ms (i.e., below the threshold of conscious
recognition) and unrelated targets that are in the same or different languages (e.g., casa-DOG vs.
house-DOG, where casa means ‘house’ in Spanish). This way, the only relationship between
targets and primes is the language code, which allows studying the cost associated with an
unconscious language switch.
One of the first studies exploring this language switch effect using an electrophysiological
masked priming paradigm was that of Chauncey et al. (2008) testing French-English bilinguals.
They found that switch trials elicited more negative-going waveforms than non-switch trials in
the windows of two components, the N250 and N400. The authors suggested that these increased
negativities for switch compared to non-switch trials resulted from automatic top-down
modulation from the language node feeding back to lexical representations. Thus, the language
nodes of the prime words are presumed to inhibit the target language lexical representations in
switch trials and/or to enhance the activation of the same-language lexical representations in
non-switch trials (see also Duñabeitia et al., 2010b). Other ERP studies have shown similar
patterns (Martin et al., 2012; Midgley et al., 2009a; Proverbio et al., 2004; Van Der Meij et al.,
2011), but to date no ERP study has investigated the contribution of orthographic cues.
5
However, the time-course of visual word recognition based on results from experiments using
masked priming relies on the interpretation of the N250 as typically associated with the mapping
of orthographic units onto orthographic word forms (see Grainger & Holcomb, 2009, for a
review), and the N400 with a later stage of processing sensitive to interactions between lexical
forms and semantic representations. Therefore, the former component reflects sub-lexical and
lexical information integration while the later reflects lexical-semantic integration. Thus it is
possible that bilinguals use both lexical and sub-lexical routes to access language membership,
giving rise to the early effects found in masked language-switching priming paradigm (in line
with the parallel language identification routes sketched by Van Kesteren et al., 2012). Along
these lines, one could predict that under masked priming conditions in a seemingly monolingual
context, a fast-operating sub-lexical route highly sensitive to cross-language orthographic
regularities would initially detect conflicting graphemic chunks in the masked primes from a
different language. This is consistent with the study by Casaponsa and Duñabeitia (in press) in
which marked masked primes did not yield significant lexico-semantic effects on target word
processing. Despite the theoretical and experimental plausibility of this hypothesis, to our
knowledge no masked language-switching priming experiment has manipulated orthographic
markedness. In fact, masked language-switching priming experiments published so far have used
cross-script priming, which warrants the presence of sub-lexical cues (e.g., Hoshino et al., 2010;
Dimitropoulou et al., 2011b), or same-script priming without controlling for orthographic
markedness (e.g., Chauncey et al., 2008; Dimitropoulou et al., 2011a, 2011b; Midgley et al.,
2009b, Perea et al., 2008). Here we explore the effects associated with sub-lexically marked and
unmarked switch trials and compared them to non-switch trials.
6
In a nutshell, the current study explores how highly proficient Spanish-Basque bilinguals access
language information about the words they read. Our guiding hypothesis is that bilinguals do not
rely solely on lexical-level information but also on sub-lexical statistical regularities. We used
the ERP masked language-switching priming paradigm while participants performed a go/no-go
semantic categorization task. Target words were always in the dominant language (L1-Spanish),
and the masked primes were either in the dominant or in the non-dominant language (L1 or L2).
Critically, language-specific orthographic markedness of the bigrams of L2 primes was
manipulated as a function of the L1 (target) orthography to explore how and when sub-lexical
information modulates language identification.
According to the dual-route model of language identification (BIA+ extended, Van Kesteren et
al., 2012), if within-word statistical regularities guide bilingual language detection, then Basque
(L2) primes that are formed by Basque-specific marked bigrams should modulate the masked
language switch cost effect in initial stages of target word processing (i.e., N250) and produce
greater switch costs as compared to masked primes that follow the orthotactic rules of the target
language (i.e., unmarked Basque primes). Furthermore, under the assumption that sub-lexical
information is not informative enough to detect unmarked primes as belonging to Basque,
language switch effects are only expected (if any) at later stages of processing (e.g., in the N400
window; see Chauncey et al., 2008).
In addition to the bilingual group tested in Experiment 1, a second experiment was conducted
with a control group of Spanish monolinguals without any prior knowledge of Basque using the
same materials and procedure (Experiment 2). This monolingual control group was tested in
order to ensure that the differential masked language switch effects found for orthographically
marked and unmarked words in the bilingual group could not be accounted for by the saliency of
the sub-lexical orthotactics of the marked items. We expected that marked Basque prime words
7
would be processed as non-words (i.e., strings containing illegal combinations that cannot be
recognized as potential words), while unmarked Basque prime words were expected to act as
pseudo-words (i.e., strings that lack a referent in the mental lexicon even though they follow
their language orthotactics). Therefore, marked Basque primes were expected to elicit greater
negativities than unmarked Basque primes, based on sub-lexical feature analysis (see Grossi &
Coch, 2005). At the same time, we expected Basque marked and unmarked primes to elicit
similar effects in the N400 window due to the absence of a lexical referent in Spanish (e.g.,
Grainger & Holcomb, 2009; Holcomb & Grainger, 2006). For the same reason, Basque words
should elicit greater negativities than Spanish control words at this later stage of processing.
2. Results
2.1 Experiment 1: Spanish (L1) - Basque (L2) bilinguals.
2.1.1. Behavioral results
Bilingual participants correctly categorized 92.18% (SD=4.01) of the Spanish animal names
when these words were presented as targets. Furthermore, when the animal names were
presented as masked primes in the prime visibility test, none of the participants reported
consciously perceiving them (or any other word), confirming that participants were unaware of
the existence and nature of the masked primes (percentage of false alarms in the task: .59%,
SD=.45; percentage detected primes in the prime visibility test: .47%, SD=2.18).
2.1.2. ERP results
The different language priming conditions split by bigram markedness are plotted in Figure 1
together with the voltage maps for the marked and unmarked switch effects in the different time
windows of interest (see also Table 1 for means and standard deviations).
-Insert Table 1 around here-
2.1.2.1. 180-260 ms (N250)
8
The main effect of Language (switch vs. non-switch trials) was significant [F(1,20)=7.48,
p=.013, η2p=.27], showing more negative-going waveforms for switch (i.e., L2!L1) than for
non-switch trials (i.e., L1!L1). Critically, the interaction between Language and Bigram was
significant [F(1,20)=4.50, p=.047, η2p=.18]. Planned comparison showed a modulation of the
language switch cost effect as a function of the orthographic markedness of the Basque masked
primes. A significant switch cost effect was found for marked primes compared to their L1
controls [F(1,20)=11.67, p=.003, η2p=.37], and not for the unmarked Basque primes [F<1, p>.5,
η2p=.02] (see Figure 1). All other main effects and interactions did not approach significance
[ps>.73].
2.1.2.2. 350-500 ms (N400)
The factors Language and Bigram significantly interacted with each other [F(1,20)=5.80; p=.026,
η2p=.23]. Follow-up comparisons revealed a similar pattern to that found in the previous time
window: a significant switch cost effect was found for the marked Basque primes [F(1,20)=7.36,
p=.013, η2p=.27], but not for the unmarked Basque primes [F<1, p>.8, η2p<.01]. All other main
effects and interactions did not reach significance [ps>.13].
-Insert Figure 1 around here-
2.1.3. Discussion
Bilingual participants showed a masked language switch cost effect in two main negative-going
components (the N250 and N400) replicating and extending the findings reported in previous
studies (Chauncey et al., 2008, 2011; Duñabeitia, Dimitropoulou, Uribe-Etxebarria, Laka, &
Carreiras, 2010). We found a large and long-lasting switch cost effect only for those Basque
words (L2) that were made of bigrams that were implausible in the L1 (Spanish). This masked
language switch cost effect started as early as 200 milliseconds after the target word presentation
(see Figure 1), and lasted for around 400 milliseconds. Critically, we did not find any significant
9
switch cost effect for unmarked L2 words in any epoch (see Figure 1; see also Appendix for the
visualization of the effects across the entire epoch).
2.2. Experiment 2: Spanish monolinguals
2.2.1. Behavioral results
Monolinguals correctly detected 91.84% (SD=4.25) of the animals (percentage of false alarms in
the task: .38%, SD=.60). None of the participants reported consciously perceiving the animal
names (or any other word) when these were presented as masked primes, confirming that
participants were unaware of the existence and nature of the masked primes (percentage of
primes detected in the prime visibility test= .32%, SD=1.00).
2.2.2. ERP results
The different language priming conditions split by bigram markedness are plotted in Figure 2
together with the voltage maps for the marked and unmarked effects in the different time
windows of interest (see also Table 2 for means and standard deviations).
-Insert Table 2 around here-
2.2.2.1. 180-260 ms (N250)
The main effect of Language (switch vs. non-switch trials) was significant [F(1,20)=66.81,
p<.001, η2p=.77], showing more negative-going waveforms for target words preceded by primes
in the other language than for non-switch trials. This factor was partially modulated by the factor
Region [F(2,40)=3.38, p=.07, η2p=.14, ε=.60]. Even though all three regions showed more
negative-going waveforms for switch than non-switch trials, the effect sizes were greater over
central and posterior regions [anterior: F(1,20)=25.69, p<.001, η2p=.56; central: F(1,20)=74.00,
p<.001, η2p=.79; posterior: F(1,20)=80.57, p<.001, η2p=.80]. A marginal effect of Bigram was
also found [F(1,20)=3.68, p=.07, η2p=.15], and the interaction between this factor and Language
was marginally significant [F(1,20)=3.97, p=.06, η2p=.17], suggesting that the magnitude of the
10
language effect was slightly modulated by the violation of L1 orthotactic rules (see Figure 2).
Marked Basque masked primes elicited partially larger effects than unmarked Basque masked
primes, but importantly these two conditions showed significant switch cost effects
[F(1,20)=38.51, p<.001, η2p=.66, and F(1,20)=11.34, p=.003, η2p=.36, respectively]. All other
interactions did not reach significance [ps>.35].
2.2.2.2. 350-500 ms (N400)
The main effect of Language was significant [F(1,20)=13.01, p=.002, η2p=.39], showing a
general language switch cost effect. Switch trials were associated with more negative-going
waveforms than non-switch trials. Critically, this effect was not modulated by the Bigram factor
[interaction: F<.1, p>.85, η2p<.001], showing that in the N400 time window monolingual
participants showed similar switch cost effects for marked and unmarked Basque primes.
Besides, a marginal three way interaction was found [F(2,40)=3.28, p=.07, η2p=.14, ε=.68].
However, follow-up comparisons did not show differential Language effects as a function of
Bigrams in any of the Regions [anterior: F(1,20)=.84, p=.37, η2p=.04; central: F(1,20)=.24,
p=.63, η2p=.01; posterior: F(1,20)=1.12, p=.30, η2p=.05]. All other main effects and interactions
did not result significant [ps>.31].
-Insert Figure 2 around here-
2.2.3. Discussion
Targets preceded by prime words in an unknown language showed more negative-going
waveforms than targets preceded by words in the same language in both time windows (N250
and N400). Interestingly, this effect was slightly greater for non-word primes (marked Basque
words) than for pseudo-word primes (unmarked Basque words) in the N250 time window. While
the interaction in the N250 window only approached significance, it should be kept in mind that
in sharp contrast with the results from the bilingual sample tested in Experiment 1, both
11
conditions (orthographically marked and unmarked Basque primes) elicited larger negativities as
compared to same-language conditions in Experiment 2. Besides, and again in contrast to the
results from Experiment 1, robust N400 effects were found both for orthographically marked and
unmarked Basque primes as compared to the Spanish control prime words. These results suggest
that monolingual participants are also sensitive to L1 orthotactic violations at initial stages of
visual word recognition. Moreover, these results show that monolingual readers are able to detect
that a given word does not belong to their native language as soon as 200 ms, even when these
strings follow the orthotactic rules of their language, demonstrating that some form of lexical
access occurs within this epoch (see also Carreiras et al., 2009a, 2009b; Coch and Mitra, 2010;
Duñabeitia et al., 2009; Grossi and Coch, 2005; Hoshino et al., 2010; Massol et al., 2011;
Midgley et al., 2009b; Morris et al., 2007; Proverbio et al., 2009).
3. General discussion
The present study was designed to explore how bilinguals identify the language of the words
they read. We investigated whether the magnitude of masked language switch cost effects is
determined by language-dependent orthographic regularities. We aimed at exploring the
influence of sub-lexical orthographic cues on the ERP patterns related to automatic and
unconscious masked language switch effects that have been replicated recently with different
bilingual samples in different languages (e.g., Chauncey et al., 2008; Dunabeitia et al., 2010a).
To this end, highly proficient Spanish (L1) – Basque (L2) bilinguals were presented with L1
target words preceded by either unrelated L1 or L2 masked words (Experiment 1). Critically, the
orthotactic plausibility in the L1 (Spanish; target language) of L2 masked prime words was
manipulated. In addition to the bilingual group, a group of Spanish monolinguals was also tested
for control purposes (Experiment 2).
12
The results showed a masked language switch cost effect in initial stages of reading (around 200
ms), which was also present at the later stages of visual word processing for Spanish-Basque
highly proficient bilinguals. Importantly, these effects were modulated as a function of the
orthographic markedness or specificity of Basque (L2) words for bilingual participants. They
showed masked language switch effects only for those Basque words that were orthographically
implausible in Spanish (i.e., orthographically marked items), and no effects for those Basque
primes whose bigrams were plausible in Spanish orthography (i.e., orthographically unmarked
items; see Appendix to better visualize these effects). Therefore, we demonstrated that bilingual
readers unconsciously rely on basic orthographic cues in order to discriminate the words that do
not correspond to the target language. Furthermore, results from Experiment 2 confirmed that
this differential switch cost effect based on orthographic markedness is specific to bilinguals,
since monolinguals showed strong effects for both unmarked and marked masked strings. These
effects were slightly greater for strings that violated Spanish orthotactic regularities only at early
stages of visual word processing (see Appendix).
Previous behavioral studies have shown that bilingual readers are highly sensitive to the bigram
frequencies of words (Casaponsa, et al 2014; Casaponsa and Duñabeitia, in press; Grainger and
Beauvillain, 1987; Thomas and Allport, 2000; Vaid and Frenck-Mestre, 2002; Lemhöfer et al.,
2004, 2008, 2011; Van Kesteren et al., 2012; see also Grainger and Van Heuven, 2003;
Dandurand et al., 2011; Hauk et al., 2006, 2008; Whitney and Cornelissen, 2008; Whitney et al.,
2011, for evidence in monolinguals), and that orthographic information is explicitly used to
identify the language of the words. Furthermore, recent data indicates that bilinguals use this
information to speed up word recognition in ambiguous language contexts (e.g., Casaponsa et al.,
2014) and to access language-selective or language-unspecific lexical representations (e.g.,
Casaponsa and Duñabeitia, in press), corroborating the hypothesis of two parallel routes to
13
language identification (i.e., the lexical and sub-lexical routes; see Van Kesteren et al., 2012).
The current study demonstrates that the manner in which the bilingual visual word recognition
system detects the language of words is modulated by orthographic statistical information (i.e.,
orthotactic regularities) associated with the probabilistic combination of letters that are allowed
in the context language(s). Furthermore, the current data suggest that language detection occurs
at early automatic stages of visual word processing, even under circumstances in which the
language code switch is not consciously perceived (i.e., under masked priming conditions),
extending preceding evidence from conscious reading paradigms (Casaponsa et al., 2014;
Grainger and Beauvillain, 1987; Lemhöfer et al., 2011; Vaid and Frenck-Mestre, 2002; Van
Kesteren et al., 2012) and reinforcing the view of the automaticity of these processes (see also
Casaponsa and Duñabeitia, in press).
Following this view, these results suggest that language switch cost effects in the N250 and
N400 components for L2-marked primes are mainly guided by an automatic sub-lexical analysis
in our sample of bilinguals. As suggested in the recent extension of the BIA+ model (Van
Kesteren et al., 2012), the involvement of the sub-lexical route to determine language
membership information is critical. When bilingual participants perform a task in their L1, the
language-dependent sub-lexical specificities of unconsciously perceived L2 primes that violate
the orthographic regularities of the target language give rise to early incongruity effects due to
the saliency of these mismatching representations. The processing of marked L2 prime words is
then impoverished, due to the effort needed to map these incongruent L2 sub-lexical
representations onto L1 lexical forms. The high similarity between the results associated with
L1-incongruent Basque marked masked primes from the bilingual and monolingual samples are
in line with this hypothesis. The orthotactic specificity of these strings would make monolinguals
process them as non-words lacking a referent in their lexicon. In a parallel manner, bilinguals
14
would initially detect the differential orthographic regularities of these items, and even if the
strings are represented in their lexicon(s), their processing would get bogged down at sub-lexical
levels (i.e., language-selective lexical access for orthographically marked items; see Casaponsa
and Duñabeitia, in press, for evidence supporting this hypothesis). These findings also fit with
the predictions of the BIA+ extended model, which suggests that the presence of
orthographically salient and distinctive units would immediately activate the language nodes via
sub-lexical language information, and these would be read out by the task decision system. In the
context of the current experiment in which participants were only aware of the presence of L1
words (namely, the targets), L2 marked masked primes represent a conflict for the task decision
system given that their orthotactic regularities do not match those of the target language. This
conflict gives rise to the early language-mismatch EEG effects found in the N250 epoch.
In contrast, when sub-lexical cues are not available (i.e., unmarked Basque primes), lexical
access is the only key to detect whether the target was preceded by a prime belonging to the
same or different language. Considering the existence of a single integrated lexicon and the task
demands at hand, bilinguals were expected to show similar co-activation and/or competition
effects for orthographically unmarked words from a different language and for words from the
same language. This way, the possibility to observe reliable switch cost effects for
orthographically unmarked items is highly reduced, given that the mismatch at the lexico-
semantic level between these items and their controls in the non-switch conditions is highly
similar (i.e., they are equally unrelated to the targets; see Van Wijnendaele and Brysbaert, 2002,
for similar results during L1 and L2 masked primes with highly proficient bilinguals). Our
results for unmarked L2 items are also in line with the BIA+ extended model. In the absence of
sub-lexical information pointing to a specific language, L2 and L1 are granted language-
independent lexical access to an integrated lexicon, and language identification occurs at (post-)
15
lexical stages of processing (see also Casaponsa and Duñabeitia, in press). Thereby, similar
effects are expected to emerge for L1 and orthographically unmarked L2 masked primes, as
demonstrated in the current study. However, in the case of the monolinguals, unmarked L2
words were pseudo-words in essence (i.e., orthographically unmarked strings lacking lexical
representation), and they elicited greater negativities due to the effort needed to map the existent
orthographic representations onto the inexistent lexical entry (Grainger and Holcomb, 2009).
Thus, targets preceded by primes that do not exist in the lexicon (orthographically unmarked
items) give rise to greater conflict than those preceded by primes with existing lexical
representations (L1 items).
In sum, this study is illustrative of how bilinguals process different languages during reading and
of how multilingual reading is grounded in language selection criteria that are based on both
basic-level orthographic regularities and on higher-level lexical dimensions. As evidenced by the
early effects found only for the marked L2 words, it is suggested that masked language switch
priming effects are highly sensitive to the language-dependent sub-lexical features of words.
Furthermore, unconsciously perceived L2 primes that matched the orthotactic regularities of the
L1 yielded similar conflict to that produced by L1 primes (always unrelated to the target),
evidencing the nature of language-independent lexical access in highly proficient bilinguals.
These data support the recent extension of the BIA+ model proposed by Van Kesteren et al.
(2012) and recent results showing the relevance of sub-lexical cues in automatic and unconscious
language-selective and language-nonselective activation by Casaponsa and Duñabeitia (in press).
We conclude that the bilingual visual word recognition system is strongly influenced by sub-
lexical stages of processing, and that language-dependent orthotactic combinatorial rules play an
important role in bilingual lexical access and language identification.
4. Materials and Methods
16
4.1 Experiment 1
4.1.1. Participants
Twenty-one right-handed native Spanish speakers (10 women; mean age=21, SD= 2.07)
completed this experiment as part of the bilingual sample. All of them were recruited from the
bilingual pool of participants from the Basque Country. Spanish was their dominant language,
but they were also highly proficient in Basque as calculated by self-ratings (see Table 3), and
acquired Basque very early in life (mean=2.48, SD=1.66). Their overall self-rated proficiency in
Spanish (L1) ranged from 9 to 10 (mean= 9.73, SD=0.46) and their overall Basque (L2)
proficiency ranged from 7 to 8 (mean=7.56, SD=0.60). Further self-ratings regarding speaking,
understanding, writing and reading skills showed a significant difference in the relative
proficiency between languages (i.e., Spanish-dominant; ps<.001). All participants had normal or
corrected-to-normal vision and participated voluntarily in this experiment in exchange for
monetary compensation. None of the participants reported neurological or psychiatric disorders.
Written informed consent was obtained from all participants prior to the experimental session in
accordance with guidelines approved by the Ethics and Research Committees of the Basque
Center on Cognition, Brain and Language. The study was also performed in accordance with the
ethical standards set in the Declaration of Helsinki.
-Insert Table 3 around here-
4.1.2. Stimuli
Three hundred and forty Spanish (L1) words were selected as targets (e.g., cuento [story]), taken
from the Spanish B-Pal database (Davis and Perea, 2005) and were split into two groups
balanced for word frequency, length, number of neighbors, AoA and concreteness scores (see
Table 4). Primes were either unrelated Basque (L2) words (340 items) or unrelated Spanish (L1)
words (340 items) selected from the E-Hitz database (Perea et al., 2006) and B-Pal, respectively.
17
Half of the Spanish targets were preceded by 1) an orthographically marked unrelated Basque
word containing implausible bigrams when measured in the L1 (e.g., neska [girl]; “sk” is an
implausible bigram in Spanish), or by 2) a Spanish unrelated word (e.g., bolsa [bag]). The other
half of Spanish target words could be preceded by 1) an orthographically unmarked unrelated
Basque word containing only plausible Spanish bigrams (e.g., mutil [boy]), or by 2) a Spanish
unrelated word (e.g., cerebro [brain]). See Table 4 for means and SD of the factors balanced
within and across languages.
To control for the L1 plausibility of the L2 primes (i.e., their orthographic markedness), we first
obtained all the possible Spanish bigrams from the Spanish LEXESP (Sebastián-Gallés et al.,
2000) and Basque SYLLABARIUM (Dunabeitia et al., 2010a) databases respectively, and
calculated their log10 frequencies (see the B-Pal database for a similar procedure). These
Spanish bigram frequencies were used to split Basque words into two sets, depending on whether
their mean bigram frequencies when measured according to the Spanish bigram distribution
resulted above or below the mean score. The unmarked Basque primes exclusively included
words whose mean bigram frequency fell above the mean of this distribution of Spanish bigrams.
Similarly, the marked Basque set only included words whose mean bigram frequency fell below
the mean. Furthermore, in order to better capture the differences between sets, another restriction
was imposed: All the Basque words from the unmarked set included bigrams that appeared at
least 10 times in the Spanish lexicon and all the Basque words from the marked set included at
least one bigram that appeared less than 10 times in the Spanish lexicon. This double-check at
the individual bigram frequency level and at the mean bigram frequency level provides a reliable
measure to assess the bigram probability of appearance or plausibility of Basque words in
Spanish and also allowed us to match Basque and Spanish word sets in terms of their bigram
frequencies within and across languages (see Table 4). Critically, the unmarked primes were as
18
frequent when measured according to the Spanish bigram count as the words from the Spanish
sets. Therefore, the unmarked Basque words were orthographically very similar to the Spanish
vocabulary.
Finally, prime-target orthographic overlap was also balanced across conditions by matching the
Levenshtein distance scores of all the possible prime-target combinations (Mean=8.38, SD=1.86;
all ps>.71; Minimum number of Levenshtein edits = 5). In addition, two lists were created and
counterbalanced across participants, so that each target word appeared only once in each list but
in a different prime condition in each of them. Participants were randomly assigned to each list
and priming conditions were evenly distributed across and within lists (340 critical prime-target
pair words in each list, with 85 words pairs per condition in each list). All these pairs were used
as no-go trials in a semantic categorization task. An additional set of 60 Spanish animal names
were also included (15% of the final trials) for the go-trials and for the prime visibility test in
each of the lists. The prime visibility test consisted of the inclusion of the animal names as
masked primes, followed by the presentation of non-animal targets, and it is typically used in
EEG masked priming studies in order to make sure that participants are not aware of the
presence and nature of the masked primes (see, among many others, Molinaro, Duñabeitia,
Marín-Gutiérrez, and Carreiras, 2010, for a similar procedure). (Note that if participants respond
to these masked animal names, it would suggest that they consciously perceive the briefly
presented primes). The complete list of materials used in this experiment can be accessed
through http://www.bcbl.eu/materials/acasaponsa/Appendix_Materials_C&C&D.pdf
-Insert Table 4 around here-
4.1.3. Procedure
All participants were tested individually in a quiet room. Visual stimuli were presented using
Presentation software (Version 4.6, Neurobehavioral systems, Inc.) on a 15” CRT monitor set to
19
a refresh rate of 90Hz. Stimuli were displayed at high contrast in white letters on a black
background. On each trial, a forward mask consisting of a row of hash marks (#) was presented
for 500ms. Next, the prime was presented in 25pt lowercase Courier New and stayed on the
screen for 50 ms. The prime was immediately followed by the presentation of the target stimulus
in 25pt uppercase Courier New. The target remained on the screen for 500 ms. The inter-trial
interval varied randomly between 900 and 1100 ms. After this interval, an asterisk was presented
for 1000 ms in order to allow for participant blinks. Participants were instructed to press the
space bar in the keyboard whenever they saw an animal name on the screen and to read all other
words passively without responding to them. The fact that critical stimuli did not require an over
response (i.e., semantic categorization go/no go task) granted a ERPs signal free from muscular
artifacts. Participants were not informed of the presence of the primes. Trial presentation order
was randomized across participants. Each participant received a total of 20 practice trials
(representative of the conditions in the critical trials) prior the experimental trials. Task
instructions (and interactions with the participants) were given in their L1 (Spanish). The
experimental session lasted for approximately 20 minutes (excluding participant preparation).
4.1.4. EEG recording procedure
The electroencephalogram was recorded from 27 active electrodes (plus Ground) held in place
on the scalp by an elastic cap (ElectroCap International, Eaton, USA, 10-10 system). Eye
movements and blinks were monitored with four further electrodes providing bipolar recordings
of the horizontal (Heog-, Heog+) and vertical (Veog-, Veog+) electrooculogram (EOG). Another
two electrodes were attached over the right mastoid bone (reference) and over the left mastoid
bone (recorded actively to monitor for differential mastoid activity). All EEG electrode
impedances were maintained below 5 kΩ (impedance for eye electrodes was less than 10 kΩ).
20
The EEG signal was sampled continuously throughout the experiment with a sampling rate of
500 Hz and digitally off-line re-referenced to linked mastoids.
4.1.5. Data analysis
Ocular artifacts were corrected using Independent Component Analysis (ICA). Based on
previous literature, the ICA algorithm used was Infomax (Gradient) Restricted Biased. High-pass
filter of 0.01 Hz was applied before the ICA procedure, and a low-pass filter of 30 Hz was
applied after ICA. Averaged ERPs time-locked to target onset were formed off-line from trials
free of ocular and muscular artifacts (85.27% of the data; rejected trials were equally distributed
across conditions). Baseline correction was performed using the averaged EEG activity in the
100 ms preceding the onset of the target stimuli. ERPs were quantified by taking the mean
amplitude of each participant and electrode in two temporal epochs corresponding to the two
primary components of interest found to be sensitive to masked language switch cost ERP
effects: the N250 (covering a time window between 180 and 260 ms post-target onset) and the
N400 (represented by a time window between 350 and 500 ms). These two epochs of interest are
in accordance with those reported in earlier studies on a similar topic using a parallel procedure
(e.g., Chauncey et al., 2008, 2011; Dunabeitia et al., 2010b; Holcomb and Grainger, 2006;
Hoshino et al., 2010; Midgley et al., 2009b). These two epochs were analyzed separately in two
sets of repeated measures analyses of variance (ANOVAs). 21 out of the 27 active electrodes
were used for the analysis, creating the factor Region with three different levels formed by the
averaged amplitude of 7 adjacent electrodes: Anterior (Fp1|Fp2|F7|F3|Fz|F4|F8), Central
(FC5|FC1|FC2|FC6|C3|Cz|C4|) and Posterior (CP5|CP1|CP2|CP6|P3|Pz|P4). Together with the
Region factor, the two other factors associated with the design were included in the analyses:
Language (2 levels: switch|non-switch) and Bigram (2 levels: marked|unmarked). Greenhouse-
Geisser correction was applied for departure from sphericity (Greenhouse and Geisser, 1959).
21
Greenhouse-Geisser epsilon value (ε) is provided only when different from 1, informing that
there was a violation of the assumption of sphericity, and the corrected p-value is therefore
reported. Effect-size was estimated using the partial eta-squared coefficient η2p (Cohen, 1973;
Haase, 1983).
4.2. Experiment 2
4.2.1. Participants
Twenty-one right-handed native Spanish speakers (15 women; mean age=20.86, SD= 3.23) with
normal or corrected-to-normal vision participated voluntarily in the experiment as the
monolingual sample. All of them were recruited from a pool of monolingual participants from
the University of Murcia. In contrast to the Basque Country, where both Spanish and Basque are
the official languages, Murcia is an autonomous community of Spain where Spanish is the only
official language. None of the participants had any knowledge of Basque at all. Their overall
self-rated proficiency in Spanish ranged from 8 to 10 (mean= 9.48, SD=0.75). All participants
had normal or corrected-to-normal vision and participated voluntarily in this experiment in
exchange for course credits. None of the participants reported neurological or psychiatric
disorders. Written informed consent was obtained from all participants prior to the experimental
session in accordance with guidelines approved by the Ethics and Research Committees of the
Basque Center on Cognition, Brain and Language. The study was also performed in accordance
with the ethical standards set in the Declaration of Helsinki.
4.2.3. Materials, Procedure and Data Analysis. These were the same as in Experiment 1.
22
Acknowledgements
This research was partially supported by the Spanish Government (CSD2008-00048, PSI2012-
32123 and PSI2012-31448), the European Research Council (ERC-AdG-295362) and the
Basque Government (PI2012-74). The authors are grateful to Guillaume Thierry, Margaret
Gillon Dowens, Ileana Quiñones and Ainhoa Bastarrika for their helpful comments on earlier
drafts and for stimulating discussion on the interpretation of the results.
23
References
Carreiras, M., Duñabeitia, J. A. & Molinaro, N. (2009a). Consonants and vowels contribute
differently to visual word recognition: ERPs of relative position priming. Cerebral
Cortex, 19, 2659 – 2670. doi: 10.1093/cercor/bhp019
Carreiras, M., Gillon-Dowens, M., Vergara, M. & Perea, M. (2009b). Are vowels and consonants
processed differently? ERP evidence with a delayed letter paradigm. Journal of Cognitive
Neuroscience. 21, 275-288. doi: 10.1162/jocn.2008.21023
Casaponsa, A., Carreiras, M., & Dunabeitia, J. A. (2014). Discriminating languages in bilingual
contexts: the impact of orthographic markedness. Front Psychol, 5, 424. doi:
10.3389/fpsyg.2014.00424
Casaponsa, A., & Duñabeitia, J. A (in press). Lexical organization of language-ambiguous and
language-specific words in bilinguals. Quarterly Journal of Experimental Psychology.
Chauncey, K., Grainger, J., & Holcomb, P. J. (2008). Code-switching effects in bilingual word
recognition: a masked priming study with event-related potentials. Brain Lang, 105(3),
161-174. doi: 10.1016/j.bandl.2007.11.006
Chauncey, K., Grainger, J., & Holcomb, P. J. (2011). The role of subjective frequency in
language switching: an ERP investigation using masked priming. Mem Cognit, 39(2),
291-303. doi: 10.3758/s13421-010-0006-7
Coch, D., & Mitra, P. (2010). Word and pseudoword superiority effects reflected in the ERP
waveform. Brain Res, 1329, 159-174. doi: 10.1016/j.brainres.2010.02.084
Cohen, J. (1973). Eta-squared and partial eta-squared in fixed factor ANOVA designs.
Educational and psychological measurement.
Dandurand, F., Grainger, J., Dunabeitia, J. A., & Granier, J. P. (2011). On coding non-
contiguous letter combinations. Front Psychol, 2, 136. doi: 10.3389/fpsyg.2011.00136
24
Davis, C. J., & Perea, M. (2005). BuscaPalabras: A program for deriving orthographic and
phonological neighborhood statistics and other psycholinguistic indices in Spanish.
Behavior Research Methods, 37(4), 665-671.
Dijkstra, T., Grainger, J., & Van Heuven, W. J. (1999). Recognition of cognates and interlingual
homographs: The neglected role of phonology. Journal of Memory and Language, 41(4),
496-518.
Dijkstra, T., Timmermans, M., & Schriefers, H. (2000). On being blinded by your other
language: Effects of task demands on interlingual homograph recognition. Journal of
Memory and Language, 42(4), 445-464.
Dijkstra, T., & Van Heuven, W. J. B. (2002). The architecture of the bilingual word recognition
system: From identification to decision. Bilingualism: Language and Cognition, 5(03).
doi: 10.1017/s1366728902003012
Dimitropoulou, M., Dunabeitia, J. A., & Carreiras, M. (2011a). Masked translation priming
effects with low proficient bilinguals. Mem Cognit, 39(2), 260-275. doi: 10.3758/s13421-
010-0004-9
Dimitropoulou, M., Dunabeitia, J. A., & Carreiras, M. (2011b). Two words, one meaning:
evidence of automatic co-activation of translation equivalents. Front Psychol, 2, 188. doi:
10.3389/fpsyg.2011.00188
Dunabeitia, J. A., Cholin, J., Corral, J., Perea, M., & Carreiras, M. (2010b). SYLLABARIUM:
an online application for deriving complete statistics for Basque and Spanish
orthographic syllables. Behav Res Methods, 42(1), 118-125. doi: 10.3758/BRM.42.1.118
Duñabeitia, J. A., Dimitropoulou, M., Uribe-Etxebarria, O., Laka, I., & Carreiras, M. (2010a).
Electrophysiological correlates of the masked translation priming effect with highly
25
proficient simultaneous bilinguals. Brain Res, 1359, 142-154. doi:
10.1016/j.brainres.2010.08.066
Duñabeitia, J. A., Molinaro, N., Laka, I., Estévez, A., & Carreiras, M. (2009). N250 effects for
letter transpositions depend on lexicality: ‘casual’ or ‘causal’? NeuroReport, 20(4), 381-
387. doi: 10.1097/WNR.0b013e3283249b1c
Duyck, W., Assche, E. V., Drieghe, D., & Hartsuiker, R. J. (2007). Visual word recognition by
bilinguals in a sentence context: evidence for nonselective lexical access. J Exp Psychol
Learn Mem Cogn, 33(4), 663-679. doi: 10.1037/0278-7393.33.4.663
Duyck, W., & Warlop, N. (2009). Translation priming between the native language and a second
language: new evidence from Dutch-French bilinguals. Exp Psychol, 56(3), 173-179. doi:
10.1027/1618-3169.56.3.173
Grainger, J., & Beauvillain, C. (1987). Language blocking and lexical access in bilinguals. The
Quarterly Journal of Experimental Psychology, 39(2), 295-319.
Grainger, J., & Holcomb, P. J. (2009). Watching the Word Go by: On the Time-course of
Component Processes in Visual Word Recognition. Lang Linguist Compass, 3(1), 128-
156. doi: 10.1111/j.1749-818X.2008.00121.x
Grainger, J., & Van Heuven, W. J. (2003). Modeling letter position coding in printed word
perception. Ment Lex, 1-24.
Greenhouse, S. W., & Geisser, S. (1959). On methods in the analysis of profile data.
Psychometrika, 24(2), 95-112.
Grossi, G., & Coch, D. (2005). Automatic word form processing in masked priming: an ERP
study. Psychophysiology, 42(3), 343-355. doi: 10.1111/j.1469-8986.2005.00286.x
26
Grossi, G., Savill, N., Thomas, E., & Thierry, G. (2012). Electrophysiological cross-language
neighborhood density effects in late and early english-welsh bilinguals. Front Psychol, 3,
408. doi: 10.3389/fpsyg.2012.00408
Haase, R. F. (1983). Classical and partial eta square in multifactor ANOVA designs. Educational
and psychological measurement, 43(1), 35-39.
Hauk, O., Davis, M. H., Ford, M., Pulvermuller, F., & Marslen-Wilson, W. D. (2006). The time
course of visual word recognition as revealed by linear regression analysis of ERP data.
NeuroImage, 30(4), 1383-1400. doi: 10.1016/j.neuroimage.2005.11.048
Hauk, O., Davis, M. H., & Pulvermuller, F. (2008). Modulation of brain activity by multiple
lexical and word form variables in visual word recognition: A parametric fMRI study.
NeuroImage, 42(3), 1185-1195. doi: 10.1016/j.neuroimage.2008.05.054
Holcomb, P. J., & Grainger, J. (2006). On the time course of visual word recognition: an event-
related potential investigation using masked repetition priming. J Cogn Neurosci, 18(10),
1631-1643. doi: 10.1162/jocn.2006.18.10.1631
Hoshino, N., Midgley, K. J., Holcomb, P. J., & Grainger, J. (2010). An ERP investigation of
masked cross-script translation priming. Brain Res, 1344, 159-172. doi:
10.1016/j.brainres.2010.05.005
Hoshino, N., & Thierry, G. (2012). Do Spanish-English Bilinguals have Their Fingers in Two
Pies - or is It Their Toes? An Electrophysiological Investigation of Semantic Access in
Bilinguals. Front Psychol, 3, 9. doi: 10.3389/fpsyg.2012.00009
Lemhöfer, K., & Dijkstra, T. (2004). Recognizing cognates and interlingual homographs: Effects
of code similarity in language-specific and generalized lexical decision. Memory &
Cognition, 32(4), 533-550.
27
Lemhöfer, K., Dijkstra, T., Schriefers, H., Baayen, R. H., Grainger, J., & Zwitserlood, P. (2008).
Native language influences on word recognition in a second language: a megastudy. J
Exp Psychol Learn Mem Cogn, 34(1), 12-31. doi: 10.1037/0278-7393.34.1.12
Lemhöfer, K., Koester, D., & Schreuder, R. (2011). When bicycle pump is harder to read than
bicycle bell: effects of parsing cues in first and second language compound reading.
Psychon Bull Rev, 18(2), 364-370. doi: 10.3758/s13423-010-0044-y
Martin, C. D., Costa, A., Dering, B., Hoshino, N., Wu, Y. J., & Thierry, G. (2012). Effects of
speed of word processing on semantic access: the case of bilingualism. Brain Lang,
120(1), 61-65. doi: 10.1016/j.bandl.2011.10.003
Massol, S., Midgley, K. J., Holcomb, P. J., & Grainger, J. (2011). When less is more: feedback,
priming, and the pseudoword superiority effect. Brain Res, 1386, 153-164. doi:
10.1016/j.brainres.2011.02.050
Midgley, K. J., Holcomb, P. J., & Grainger, J. (2009a). Language effects in second language
learners and proficient bilinguals investigated with event-related potentials. J
Neurolinguistics, 22(3), 281-300. doi: 10.1016/j.jneuroling.2008.08.001
Midgley, K. J., Holcomb, P. J., & Grainger, J. (2009b). Masked repetition and translation
priming in second language learners: a window on the time-course of form and meaning
activation using erps. Psychophysiology, 46(3), 551-565. doi: 10.1111/j.1469-
8986.2009.00784.x
Midgley, K. J., Holcomb, P. J., Van Heuven, W. J., & Grainger, J. (2008). An
electrophysiological investigation of cross-language effects of orthographic
neighborhood. Brain Res, 1246, 123-135. doi: 10.1016/j.brainres.2008.09.078
28
Molinaro, N., Duñabeitia, J.A., Marín-Gutiérrez, A. & Carreiras, M. (2010). From numbers to
letters: Feedback regularization in visual word recognition. Neuropsychologia, 48 (5),
1343-1355. doi: 10.1016/j.neuropsychologia.2009.12.037
Morris, J., Frank, T., Grainger, J., & Holcomb, P. J. (2007). Semantic transparency and masked
morphological priming: an ERP investigation. Psychophysiology, 44(4), 506-521. doi:
10.1111/j.1469-8986.2007.00538.x
Ng, S., & Wicha, N. Y. (2013). Meaning first: a case for language-independent access to word
meaning in the bilingual brain. Neuropsychologia, 51(5), 850-863. doi:
10.1016/j.neuropsychologia.2013.01.017
Orfanidou, E., & Sumner, P. (2005). Language switching and the effects of orthographic
specificity and response repetition. Memory & Cognition, 33(2), 355-369.
Perea, M., Duñabeitia, J. A., & Carreiras, M. (2008). Masked associative/semantic priming
effects across languages with highly proficient bilinguals. Journal of Memory and
Language, 58(4), 916-930. doi: 10.1016/j.jml.2008.01.003
Perea, M., Urkia, M., Davis, C. J., Agirre, A., Laseka, E., & Carreiras, M. (2006). E-Hitz: a word
frequency list and a program for deriving psycholinguistic statistics in an agglutinative
language (Basque). Behav Res Methods, 38(4), 610-615.
Proverbio, A. M., Adorni, R., & Zani, A. (2009). Inferring native language from early bio-
electrical activity. Biol Psychol, 80(1), 52-63.
Proverbio, A. M., Leoni, G., & Zani, A. (2004). Language switching mechanisms in
simultaneous interpreters: an ERP study. Neuropsychologia, 42(12), 1636-1656.
Schwartz, A. I., Kroll, J. F., & Diaz, M. (2007). Reading words in Spanish and English: Mapping
orthography to phonology in two languages. Language and Cognitive Processes, 22(1),
106-129. doi: 10.1080/01690960500463920
29
Sebastián-Gallés, N., Martí, M., Carreiras, M., & Cuetos, F. (2000). LEXESP: Una base de datos
informatizada del español. Universitat de Barcelona, Barcelona.
Spalek, K., Hoshino, N., Wu, Y. J., Damian, M., & Thierry, G. (2014). Speaking two languages
at once: Unconscious native word form access in second language production. Cognition,
133(1), 226-231.
Thierry, G., & Wu, Y. J. (2004). Electrophysiological evidence for language interference in late
bilinguals. NeuroReport, 15(10), 1555-1558. doi: 10.1097/01.wnr.0000134214.57469.c2
Thierry, G., & Wu, Y. J. (2007). Brain potentials reveal unconscious translation during foreign-
language comprehension. Proc Natl Acad Sci U S A, 104(30), 12530-12535. doi:
10.1073/pnas.0609927104
Thomas, M. S. C., & Allport, A. (2000). Language Switching Costs in Bilingual Visual Word
Recognition. Journal of Memory and Language, 43(1), 44-66. doi:
10.1006/jmla.1999.2700
Vaid, J., & Frenck-Mestre, C. (2002). Do orthographic cues aid language recognition? A
laterality study with French–English bilinguals. Brain and Language, 82(1), 47-53.
Van Der Meij, M., Cuetos, F., Carreiras, M., & Barber, H. A. (2011). Electrophysiological
correlates of language switching in second language learners. Psychophysiology, 48(1),
44-54. doi: 10.1111/j.1469-8986.2010.01039.x
Van Heuven, W. J., Dijkstra, T., & Grainger, J. (1998). Orthographic neighborhood effects in
bilingual word recognition. Journal of Memory and Language, 39(3), 458-483.
Van Heuven, W. J., Schriefers, H., Dijkstra, T., & Hagoort, P. (2008). Language conflict in the
bilingual brain. Cerebral Cortex, 18(11), 2706-2716.
30
Van Kesteren, R., Dijkstra, T., & de Smedt, K. (2012). Markedness effects in Norwegian-English
bilinguals: task-dependent use of language-specific letters and bigrams. Q J Exp Psychol
(Hove), 65(11), 2129-2154. doi: 10.1080/17470218.2012.679946
Van Wijnendaele, I., & Brysbaert, M. (2002). Visual word recognition in bilinguals:
phonological priming from the second to the first language. Journal of Experimental
Psychology: Human Perception and Performance, 28(3), 616.
Von Studnitz, R. E., & Green, D. W. (2002). Interlingual homograph interference in German–
English bilinguals: Its modulation and locus of control. Bilingualism: Language and
Cognition, 5(01), 1-23.
Whitney, C., Bertrand, D., & Grainger, J. (2011). On coding the position of letters in words: a
test of two models. Exp Psychol, 59(2), 109-114. doi: 10.1027/1618-3169/a000132
Whitney, C., & Cornelissen, P. (2008). SERIOL Reading. Language and Cognitive Processes,
23(1), 143-164. doi: 10.1080/01690960701579771
Zhang, T., Van Heuven, W. J., & Conklin, K. (2011). Fast automatic translation and
morphological decomposition in Chinese-English bilinguals. Psychol Sci, 22(10), 1237-
1242. doi: 10.1177/0956797611421492
31
Appendix
Distribution of the masked language effects across time.
A) Bilingual group
Marked effect Unmarked effect
B) Monolingual group
Note: p-values distribution across the trial for each time-point collapsed across all electrodes
and transformed to a logarithmic scale for visualization purposes. Blue color reflects p-values
above uncorrected significance level (α=0.05). Red color reflects p-values above the level of
significance corrected for multiple comparisons using Bonferroni adjustment (see Dunn, 1961).
The value of α-adjusted was 0.000125, corresponding to the 400 comparisons, one for each time-
point (i.e., 2 ms).
p-values distribution(log)
p-values distribution (log)
p-values distribution (log)
p-values distribution (log)
32
Figure captions
Figure 1. Experiment 1: A) ERPs associated with the language switch (thick lines) and non-
switch conditions (thin lines) in the marked (upper) and unmarked (lower) bigram sets. B)
Topographical distribution of the masked language switch cost effects in terms of amplitude
differences between the unrelated Basque primes and unrelated Spanish primes. Differences are
plotted separately for the averaged activity between 180 and 260 ms (N250) and 350 to 500 ms
(N400).
Figure 2. Experiment 2: A) ERPs associated with the language switch (thick lines) and non-
switch conditions (thin lines) in the marked (upper) and unmarked (lower) bigram sets. B)
Topographical distribution of the masked language switch cost effects in terms of amplitude
differences between the unrelated Basque primes and unrelated Spanish primes. Differences are
plotted separately for the averaged activity between 180 and 260 ms (N250) and 350 to 500 ms
(N400).
33
!
!
!
Table 1. Voltatge means and standard deviations for each condition and time window used in Experiment 1.
Marked
Unmarked
Spanish
Basque
Spanish
Basque
N250
Anterior
Central
Posterior
Anterior
Central
Posterior
Anterior
Central
Posterior
Anterior
Central
Posterior
Mean
3.98
2.44
1.26
2.95
1.21
.02
3.53
1.96
.96
3.43
1.82
.57
SD
2.56
2.54
3.04
2.74
2.41
2.82
2.62
2.62
3.11
3.01
2.94
3.24
N400
Mean
.42
.15
.95
-.13
-.40
.48
-.12
-.50
.39
.04
-.45
.34
SD
2.37
2.31
2.22
2.35
2.30
2.56
2.75
2.78
2.55
2.73
2.81
2.71
!
34
Table 2. Voltatge means and standard deviations for each condition and time window used in Experiment 2.
Marked
Unmarked
Spanish
Basque
Spanish
Basque
N250
Anterior
Central
Posterior
Anterior
Central
Posterior
Anterior
Central
Posterior
Anterior
Central
Posterior
Mean
3.89
2.93
1.49
2.61
1.25
-.02
3.94
2.78
1.39
3.35
1.97
.54
SD
3.36
3.28
2.36
2.67
2.74
2.47
2.81
2.66
2.13
2.87
2.72
2.18
N400
Mean
-.22
.19
.46
-.91
-.59
.07
-.45
-.10
.38
-.87
-.72
-.31
SD
2.31
2.25
2.07
2.18
2.12
1.98
2.33
2.31
1.91
1.94
1.58
1.75
!
35
Table 3. Mean levels of L1 (Spanish) and L2 (Basque) language proficiency calculated according to participants' self-ratings (in
a 1-to-10 scale) for Experiment 1 and 2. Standard deviations are provided within parentheses. All between-language
comparisons resulted in ps<0.001.
Language proficiency
Experiment 1 (Bilinguals)
Experiment 2 (Monolinguals)
Spanish
Basque
Spanish
Basque
Speaking
9.91 (0.29)
7.22 (0.97)
9.67 (0.58)
-
Understanding
9.86 (0.35)
8.45 (1.26)
9.54 (0.74)
-
Writing
9.81 (0.39)
7.55 (1.14)
9.57 (0.68)
-
Reading
9.86 (0.35)
8.55 (1.14)
9.55 (0.84)
-
General self-perception
9.73 (0.46)
7.55 (0.60)
9.48 (0.75)
-
36
Table 4. Mean values for each sub-lexical, lexical and semantic factor of the L1 (Spanish) and L2 (Basque) word used in Experiment 1 and 2 split by condition. Standard deviations are provided within parentheses.
Within-language and across language p-values are also provided.
PRIME WORDS
TARGET WORDS
BASQUE
SPANISH
Between-language Comparisons
(p-values)
SPANISH
Marked
Unmarked
p-values
Marked
control
Unmarked
control
p-values
Marked
Unmarked
Marked
Unmarked
p-values
Word Frequency
52.00
(114.53)
47.36
(109.53)
0.70
44.65
(81.17)
42.56
(74.86)
0.81
0.50
0.64
43.07
(62.05)
38.47
(48.98)
0.45
Word Length
6.62
(1.83)
6.81
(2.22)
0.35
6.81
(1.81)
6.82
(1.77)
0.98
0.33
0.98
7.65
(2.27)
7.63
(2.11)
0.94
Number of Orthographic Neighbors
1.42
(1.62)
1.55
(0.35)
0.54
1.53
(2.74)
1.69
(3.01)
0.61
0.67
0.64
1.96
(3.60)
1.69
(3.00)
0.45
Age of Acquisition
3.22
(0.49)
3.23
(0.50)
0.82
3.19
(0.56)
3.19
(0.61)
0.98
0.55
0.45
3.17
(0.57)
3.18
(0.60)
0.86
Word Concreteness
4.09
(0.89)
4.12
(0.86)
0.81
4.05
(0.81)
4.07
(0.85)
0.80
0.65
0.65
3.85
(0.82)
3.79
(0.84)
0.53
Spanish Bigram Frequency
1.72
(0.3)
2.97
(0.24)
0.00
2.49
(0.30)
2.46
(0.33)
0.42
0.00
0.34
Basque Bigram Frequency
2.88
(0.18)
2.89
(0.20)
0.68
Number of Spanish-Implausible Bigrams
2.35
(0.93)
0
(0)
0.00
Note: Frequency ratings were obtained from the B-Pal and E-Hitz databases. Bigram frequencies were obtained from the raw number of appearances of the bigrams in the Spanish LEXESP and Basque SYLLABARIUM corpus in a within-word
position-independent manner. The critical factor Number of Spanish-Implausible Bigrams refers to the number of bigrams from the Basque words that appear less than 10 times in the whole Spanish corpus.
37
Averaged anterior electrodes Averaged central electrodes Averaged posterior electrodes
Fp1|Fp2|F7|F3|Fz|F4|F8 FC5|FC1|FC2|FC6|C3|Cz|C4 CP5|CP1|CP2|CP6|P3|Pz|P4
Fp1|Fp2|F7|F3|Fz|F4|F8 FC5|FC1|FC2|FC6|C3|Cz|C4 CP5|CP1|CP2|CP6|P3|Pz|P4
Averaged anterior electrodes Averaged central electrodes Averaged posterior electrodes
A) Language switch and non-switch conditions for the Bilingual group
L2L1 L1L1
Unmarked set Marked set
B) Topography of the masked language effects
Unmarked
Marked
N250 N400
38
Averaged anterior electrodes Averaged central electrodes Averaged posterior electrodes
Fp1|Fp2|F7|F3|Fz|F4|F8 FC5|FC1|FC2|FC6|C3|Cz|C4 CP5|CP1|CP2|CP6|P3|Pz|P4
Fp1|Fp2|F7|F3|Fz|F4|F8 FC5|FC1|FC2|FC6|C3|Cz|C4 CP5|CP1|CP2|CP6|P3|Pz|P4
Averaged anterior electrodes Averaged central electrodes Averaged posterior electrodes
L2L1 L1L1
A) Language switch and non-switch conditions for the Monolingual group
Unmarked set Marked set
B) Topography of the masked language effects
Unmarked
Marked
N250 N400
