Content uploaded by Leeann Stover
Author content
All content in this area was uploaded by Leeann Stover on Jun 15, 2021
Content may be subject to copyright.
EFFECTS OF LANGUAGE DOMINANCE
ON L1 RELATIVE CLAUSE PROCESSING
LEEANN STOVER,
1
MICHAEL C. STERN,
CASS LOWRY & GITA MARTOHARDJONO
Abstract
The present study investigated the effects of language dominance during
bilingual comprehension of relative clauses. We asked whether language
dominance, operationalized as a continuous variable, modulates whether/how
Spanish-English bilinguals exhibit a relative clause subject-object
processing asymmetry in their first-learned language, Spanish. Highly
proficient bilinguals with varying ages of arrival to the US completed a
language dominance questionnaire and a visual world eye-tracking
experiment with auditorily presented Spanish relative clauses. Results
revealed that higher Spanish dominance led to a larger processing
asymmetry while listening to both the relative clause and the matrix
predicate. However, rather than a facilitatory effect in subject relatives, this
asymmetry was primarily driven by a late negative effect in object relative
constructions. To account for these results, we propose that increased
dominance in the first-learned language leads to more active online
syntactic structure building, leading to a higher integration cost when an
expected parse fails.
1. Introduction
In recent literature on bilingualism, attempts have been made to formalize a
theoretical notion of language dominance as a super-construct which
subsumes a bilingual’s proficiency, use, and exposure to both languages
1
Corresponding author: LeeAnn Stover, Department of Linguistics, CUNY
Graduate Center, 365 Fifth Ave., New York, NY 10016; email:
lstover@gradcenter.cuny.edu
LeeAnn Stover, Michael C. Stern, Cass Lowry & Gita Martohardjono
207
(e.g., Montrul, 2015; Treffers-Daller, 2015). Although there is still no
consensus on a precise definition in the literature (see Meisel, 2007; Silva-
Corvalán & Treffers-Daller, 2015), most recent research agrees that language
dominance is (1) relative to both of a bilingual’s languages, (2)
multidimensional, and (3) continuous in nature. However, despite the general
acknowledgement that language dominance is a continuous construct,
studies of bilingual processing tend to treat it as a categorical variable (e.g.,
Fernández, 2003; Puig-Mayenco et al., 2018). While some researchers have
studied the effects of language dominance as a continuous measure on
linguistic knowledge (e.g., Bedore et al., 2012; Dunn & Tree, 2009; Gollan
et al., 2012), the link between continuous dominance and language
processing is much less understood (Robinson & Blumenfeld, 2018).
The present study explores the relationship between L1 processing and
language dominance. Focusing on the subject-object asymmetry in relative
clause processing, we conducted an eye-tracking experiment using the
Visual World Paradigm on Spanish-English bilinguals who share Spanish
as the L1 but live and work in an L2-dominant society. To look at a
continuum of language dominance and a variety of L1 experience, we
included participants with varying ages of arrival to the anglophone US. At
one end of the spectrum, Spanish heritage speakers learned Spanish as their
home language but grew up in an English-speaking society and are
generally dominant in their L2. On the other end of the spectrum lie first-
generation late arrivals who immigrated to the anglophone US after
adolescence. These speakers grew up and were educated in a Spanish-
speaking society and are generally L1-dominant. However, as language
dominance is gradient and dynamic (e.g., Birdsong, 2015), we do not
presume categorical dominance of any individuals in this study. Rather, we
use a relative dominance index to determine each participant’s language
dominance on the English-Spanish spectrum at the time of testing. This
allows us to operationalize dominance as a continuous variable and measure
its potential effects on L1 Spanish language processing with higher resolution.
2. Theoretical Background
2.1 Language dominance
Despite having long been present in the literature on bilingualism (Lambert,
1955), the multifaceted and complex construct of language dominance is
difficult to define and measure (for a review, see Treffers-Daller & Silva-
Corvalán, 2016). At the core of the concept of language dominance is the
idea that isolating one of a speaker’s languages is insufficient in depicting
Effects of Language Dominance on L1 Relative Clause Processing
208
the bilingual/multilingual experience. Language dominance is inherently
relative between all of an individual’s languages. One’s ‘dominant’
language is said to be the language with higher proficiency, use, exposure,
or a combination of any or all of these dimensions (e.g., Silva-Corvalán &
Treffers-Daller, 2015). However, language dominance is not categorical or
static, but rather a continuous measure that can shift over a lifetime. Domain
of use also plays a crucial role, as an individual can have different
dominance in oral vs. written language and in different situations, e.g. when
talking to their parents vs. talking to their friends, when talking about art vs.
talking about soccer, etc. (e.g., Grosjean, 2015). Dominance is clearly
multidimensional and not absolute, and as of yet there is no consensus as to
how different dimensions and domains interact with linguistic outcomes.
Similar to the difficulty in defining language dominance, its operationalization
and measurement are equally complex (see Montrul, 2015; Birdsong, 2015).
Most studies that measure language dominance rely on biographical
measures or self-reports, which are easily collected with a language
background questionnaire. This method can capture some of the
multidimensionality of language dominance in a way that objective
measures cannot by probing topics such as lifetime exposure and
proficiency across multiple domains (e.g., reading, writing, speaking,
listening). However, self-reported measurements vary in their reliability,
with some studies supporting their effectiveness (e.g., Luk & Bialystok,
2013) and others finding them misleading or unreliable (e.g., Dunn & Tree,
2009) as they may reflect individual language attitudes more than language
facility. More recently, there has been a call to use objective measures in
operationalizing language dominance (Montrul, 2015). These measures
vary from body-part naming tasks (e.g., O’Grady et al., 2009), other picture-
naming tasks (e.g., Gollan et al., 2015), sentence repetition (Flege et al.,
2002), self-paced reading (Fernández, 2003), and speech rate (e.g., Stevens,
2019), among others. However, these tasks are more time-consuming, may
not straightforwardly operationalize the intended construct, and are
inherently limited to one or two domains of language dominance.
Despite these difficulties in definition and measurement, language dominance
plays an important role in the bilingual experience. Research on bilingualism
has shown that lifetime and current language exposure, proficiency, and
other factors subsumed by the super-construct of language dominance
correlate with processes such as lexical access (e.g., O’Grady et al., 2009)
and spoken fluency (e.g., Dunn & Tree, 2009). Obtaining a composite
measure of dominance through a continuous, relative index that weighs both
languages allows us to operationalize language dominance as a continuous
LeeAnn Stover, Michael C. Stern, Cass Lowry & Gita Martohardjono
209
variable, which we can then correlate with online sentence processing
measures.
2.2 Relative clause processing
To examine the effects of language dominance on bilingual sentence
processing, we chose relative clauses as the syntactic structure of interest.
The relative clause has played a large role in both theoretical syntax and
psycholinguistic studies of language comprehension due to its suitability in
testing knowledge of recursion, competence vs performance distinctions,
and working memory limitations. Relative clauses are argued to be
syntactically unambiguous (Babyonyshev & Gibson, 1999), manipulatable
with simple word order changes (e.g., Grodner & Gibson, 2005), early-
acquired (e.g., McKee et al., 1998), and cross-linguistically universal
(Comrie & Fernández, 2012), reducing confounds of ambiguity resolution,
lexical factors, and language attrition. While some argue that there are other
complex influences on relative clause processing (e.g., MacDonald, 2013),
a wealth of experimental research attests that the relative clause is an
informative and useful structure for studying syntactic processing.
Many studies across a variety of methodologies and languages have shown
that object relative clauses, such as (1b), are more costly or difficult to
process than matched subject relative clauses, such as (1a) (see O’Grady,
2011 for review).
(1) a. Subject relative clause (SRC):
the student [that __ met the teacher]
b. Object relative clause (ORC):
the student [that the teacher met __ ]
Various syntactic, semantic, and psycholinguistic accounts attempt to
explain this asymmetry. Some syntactic/semantic theories posit that a
mismatch in thematic roles between the extracted noun in the matrix clause
and its trace in the relative clause incurs a processing cost, making ORCs
more difficult than SRCs (Sheldon, 1974). Others argue that intervening
feature-matched elements (in this case, the subject of the embedded clause)
disrupt the local relation between an extracted element and its trace
(Relativized Minimality: Rizzi, 1990). Influential psycholinguistic theories
attribute the asymmetry to either prominence factors, working memory
constraints, or filler-gap dependencies. For example, the active filler
hypothesis (Clifton & Frazier, 1989) posits that the parser will try to fill
potential gaps at the earliest point allowed by the grammar, which is
Effects of Language Dominance on L1 Relative Clause Processing
210
unproblematic for SRCs but causes a failed parse in ORCs because the first
possible gap is filled by the embedded noun phrase, necessitating a
reanalysis.
Regardless of the explanation, the presence of this SRC-ORC asymmetry is
well-documented in the literature.
2
Monolinguals of English (e.g., Traxler
et al., 2002), Spanish (Betancort et al., 2009), and many other languages
from various typological families have shown this SRC preference. This
asymmetry has also been found in bilinguals in their L1 (e.g., Madsen,
2018) as well as highly proficient second-language learners in their L2
(Juffs & Rodríguez, 2014).
In a recent study, Stern and colleagues (2019) investigated L1 relative
clause processing in Spanish heritage speakers as compared to first-
generation late bilinguals. The study found that the late bilingual group
demonstrated the expected SRC processing advantage (i.e. significantly
higher fixation proportions on the target image while hearing a SRC
compared to an ORC), while the heritage speaker group showed little
evidence of a processing asymmetry despite self-rating as highly proficient
in Spanish and performing similarly to the late bilinguals in behavioral
measures of accuracy and response times. Since the groups did not differ in
proficiency, the question arises as to what other factors could have driven
the observed difference. The current study looks at individual differences
among bilinguals with varying language experience in their L1 to probe the
effect of language dominance on subject and object relative clause
processing.
2.3 Bilingual processing in the Visual World
Based on Cooper's (1974) mind-eye theory that posits a closely time-locked
relationship between spoken language processing and eye fixations on a
visual scene, the Visual World Paradigm (VWP; Allopenna et al., 1998) has
become an increasingly-utilized methodology for examining online
language processing. VWP studies track the real-time location of participant
eye gaze on a visual display while they listen to spoken language (for a
review, see Huettig et al., 2011), allowing researchers to examine language
processing without confounds caused by varying literacy levels or
metalinguistic judgments (Tanenhaus et al., 1995). Research in the VWP
has shown that listeners use semantic and morphosyntactic cues to predict
upcoming linguistic information, showing anticipatory eye movements
2
For an alternative account, see MacDonald, 2013.
LeeAnn Stover, Michael C. Stern, Cass Lowry & Gita Martohardjono
211
towards images before the spoken word (e.g., Altmann & Kamide, 1999).
This methodology has also provided insight into how spoken input is
integrated with information retrieved from the visual environment (e.g.,
Grodner et al., 2010).
Studies in the VWP have also enhanced our understanding of bilingual
processing. Bilingual children have been shown to exhibit semantic
prediction in their L2 at a comparable level to their monolingual peers,
sometimes with more rapid prediction which suggests a possible bilingual
advantage over monolingual participants (Brouwer et al., 2017). Highly-
proficient bilinguals take advantage of morphosyntactic cues in the L2 as
well, utilizing gender marking (Dussias et al., 2013) to predict upcoming
linguistic information. However, the VWP has also revealed limitations in
L2 processing and prediction based on proficiency, L1 similarity, and
productive accuracy (e.g., Dussias et al., 2013; Lew-Williams & Fernald,
2010).
The modulating effects of L2 proficiency on L1 processing have also been
noted using the VWP, though to a limited degree. This has been evidenced
with Chinese-English bilinguals with high English proficiency, who showed
increased eye fixations when presented with smaller Chinese phonological
units that are not major processing units for spoken word recognition in
Chinese (Brouwer et al., 2017). An effect of L2 exposure on L1 processing
was also found by Stern et al. (2019), where Spanish heritage speakers
exhibited different processing patterns than late-US arrival Spanish-English
bilinguals while parsing subject and object relative clauses despite showing
comparable levels of comprehension. The methodology of Stern et al.
(2019) was replicated by the current study.
2.4 Present study
In this study, we sought to explore the effect of individual language
dominance on subject and object relative clause processing among Spanish
speakers with different L1 experiences. This study was an extension of Stern
et al. (2019), but rather than manipulating a group comparison we focused
on the effect of individual language dominance. To operationalize language
dominance, we chose the index provided by the Bilingual Language Profile
(Birdsong et al., 2012). Gaze fixation served as a proxy measurement of
language processing during a picture selection task in the VWP. In addition
to the effect of language dominance on target fixation, we also looked for
an effect of relative clause type to further probe the subject-object
processing asymmetry. Thus, our main research question was whether
Effects of Language Dominance on L1 Relative Clause Processing
212
language dominance, operationalized as a continuous variable, modulates
gaze movements in L1-Spanish relative clause processing.
We additionally analyzed participant accuracy and response time during the
picture selection task to further examine effects of language dominance and
sentence type. We made two predictions: that increased Spanish dominance
would 1) lead to increased evidence of a subject-object asymmetry in gaze
data measures, with significantly higher fixation proportions on the target
image during SRCs compared to ORCs, and 2) have minimal effect on
behavioral measures of comprehension accuracy and response time. These
predictions are based on findings from Stern et al. (2019), which found
greater evidence of the subject-object asymmetry in the gaze data of late
bilinguals compared to heritage speakers, while showing no group
differences in behavioral measures. We will first present the results of the
language dominance questionnaire, followed by the results of the eye-
tracking experiment including both gaze data and behavioral measures.
3. Language dominance
3.1 Methods
3.1.1 Participants
Fifty-nine Spanish-English bilingual adults with normal or corrected-to-
normal hearing and vision (aged 19-55: M = 27.25, SD = 8.37) participated
in this study. Data from forty-one of these participants was reported in the
previous group-level analysis in Stern et al. (2019). All participants self-
rated as fluent in both Spanish and English, spoke primarily Spanish with
their caregivers until at least age 10, and resided in New York City at the
time of testing. In other words, all participants were highly proficient
bilingual adults whose childhood home language was Spanish but who live
and work in a society where English is the majority language.
To represent a continuum of language dominance, participants were chosen
to range widely in their age of arrival to the anglophone US. In our sample,
21 participants were born in the continental US (n = 11) or arrived before
age 9 (n = 10). These participants were classified as heritage speakers in
Stern et al. (2019), as they spent their entire childhood in an environment
where their home language was the societal minority language. They had
LeeAnn Stover, Michael C. Stern, Cass Lowry & Gita Martohardjono
213
minimal formal education in their L1 (Spanish),
3
and as a group tended to
be English dominant. Another 20 participants were raised in regions where
Spanish is the societal majority language, were educated in Spanish, and did
not arrive to the anglophone US until age 17 or older. These participants
were classified as late bilinguals by Stern et al. (2019), as they did not live
in an English-majority society until adulthood and are mostly dominant in
Spanish. Finally, to bridge the continuum of language dominance, a third
L2-English learner group was tested (‘Middle’ group: n = 18) consisting of
participants who moved to an English-dominant society between ages 10
and 16. These participants vary more in their language exposure and
dominance, as they have received education in both Spanish and English
and became immersed in an English-speaking society during adolescence.
Together, these participants represent a wide spectrum of relative language
dominance between Spanish and English.
3.1.2 Procedure and analysis
For the dominance measurement task, participants completed the Bilingual
Language Profile (BLP; Birdsong et al., 2012).
4
This is a four-module, 19-
item questionnaire which probes language history, use, proficiency, and
attitudes. Weighted answers for the four modules generate a subtractive
score of relative language dominance, with positive scores indicating
increased Spanish dominance and negative scores indicating increased
English dominance. This dominance score is not intended to be absolute or
categorical but rather relative to an individual’s experience in both
languages. Scores close to zero are more reflective of a balanced bilingual
than of a bilingual who is clearly dominant in one language over the other.
It is crucial to interpret this score as a continuous and relative index of
language dominance.
3
Spanish is referred to throughout this paper as the L1 of all participants. However,
four participants did report having exposure to English from birth, and could be
classified as simultaneous bilinguals with two L1s. The labeling of Spanish as the
L1 is not intended to diminish the distinction between simultaneous and sequential
learners, but rather for simplicity’s sake as this is not relevant to the research
question of this study.
4
The BLP was administered at different times. Most participants completed the BLP
the same day as the eye-tracking experiment. However, some participants had
already filled out the BLP during a previous experiment (within 12 months of the
current study) and were not asked to re-complete it.
Effects of Language Dominance on L1 Relative Clause Processing
214
3.2. Results
To illustrate the continuum of dominance represented in this study, each
participant’s relative language dominance score by age of arrival is plotted
in Figure 1. We include groups in the visualization not because it is an
experimental variable, but rather to show the range of dominance scores
across the traditional group categories. Heritage speakers are represented
with green dots, late bilinguals with purple dots, and the late childhood
arrivals with orange dots. Means and 95% confidence intervals are also
shown for each group.
Figure 1: Age of arrival and relative language dominance
Overall, Spanish dominance increases as age of arrival to the anglophone
US increases. That is, participants who were born in the US or arrived in
early childhood tend to be more English dominant, while participants who
immigrated during adolescence or adulthood tend to be more Spanish
dominant. Additionally, despite traditional definitions of heritage speakers
in the literature which assume that bilinguals raised with early exposure to
the societal majority language will have stronger dominance in that
language, there are three early arrivals who have positive scores, indicating
Spanish dominance. There are also many participants with scores close to
zero in this dataset, indicating a balance between Spanish and English rather
than a clear, categorically-dominant language. These results demonstrate
the utility of the continuous, relative measure of language dominance over
LeeAnn Stover, Michael C. Stern, Cass Lowry & Gita Martohardjono
215
a categorical or absolute one. This measure is sensitive to small variation,
thus making it a suitable index for language dominance to assess its effects
on relative clause processing.
4. Eye-tracking experiment
4.1 Method
4.1.1 Participants
The fifty-nine participants in this experiment are the same ones previously
described in Section 3.1.1, as all participants who were administered the
BLP also completed the eye-tracking experiment.
4.1.2 Stimuli
Participants were presented with 40 complex Spanish sentences: 10 items
per experimental condition (subject relative, SRC; and object relative
clauses, ORC) and 20 fillers. All items contained varying combinations of
the same five noun phrases, which were anthropomorphic animals with
masculine gender in Spanish. An example of an SRC stimulus is shown in
(2), and an ORC in (3).
5
Relative and matrix verbs were all transitive, and
all relative clauses were subject embedded across both stimuli conditions.
Subject relative clause (SRC)
(2) El conejo, que ____ abraza al perro, cepilla al oso.
the.M rabbit that hug.3SG to-the.M dog brush.3SG to-the.M bear
‘The rabbit, that ___ hugs the dog, brushes the bear.’
Object relative clause (ORC)
(3) El perro, que el conejo abraza _____, cepilla al oso.
the.M dog that the.M rabbit hug.3SG brush.3SG to-the.M bear
‘The dog, that the rabbit hugs _____, brushes the bear.’
5
As a reviewer aptly pointed out, there is another construction for expressing ORCs
in Spanish where the word order does not differ from SRCs by manipulating the
DP/PP introducing the embedded noun phrase. For example, ‘El perro, que abraza
el conejo, cepilla al oso’ shares the same meaning as the ORC in (4) In fact, this
manipulation has been adopted in eye-tracking studies (e.g., Betancort et al., 2009).
This alternation primarily manipulates differential object marking, which has been
shown to be a difficult grammatical feature for Spanish heritage speakers (e.g.,
Montrul & Bowles, 2009). For this reason, we chose the ORC manipulation that
alters word order to reduce possible misinterpretation.
Effects of Language Dominance on L1 Relative Clause Processing
216
Sentences were presented auditorily with simultaneous presentation of a
three-image picture array. The use of auditorily-presented stimuli reduces
confounds that may arise from varying L1 literacy levels among these
participants with different language experiences in Spanish. The positions
of the image types were counterbalanced across trials. One image
corresponded to the linguistic stimulus, one distractor image always
corresponded to the stimulus until the matrix verb (‘Consistent’ distractor),
and the ‘Other RC’ distractor always corresponded to the reverse
interpretation of the relative clause (e.g., if the stimulus was an SRC, then
the distractor would depict the ORC interpretation of the matrix subject).
Figure 2 shows an example of a three-image visual display that is consistent
with example (2) above.
Figure 2. Sample visual display during experimental trial
4.1.3 Procedure
Each trial began with a black cross fixation marker appearing at the top-
center of the screen. When participants clicked on the cross, they were
shown three images and asked to familiarize themselves with the images.
After a two-second delay the cross reappeared, and when ready participants
clicked again to hear the auditory stimulus. Participants then selected the
image that best represented the aurally-presented stimulus with a mouse
click. After a short practice session, the experiment began with stimuli
presented in a pseudorandomized order.
6
Gaze fixations were recorded
throughout each trial at a sampling rate of 60 Hz using a Tobii TX300 eye-
6
Participants first completed a portion of the experiment where relative clauses were
embedded in intransitive matrix sentences, but these results are not reported in the
present paper. After a short break, participants’ eye movements were recalibrated
and the second half of the eye-tracking experiment was completed with transitive
matrix clauses, which is the focus of this study.
LeeAnn Stover, Michael C. Stern, Cass Lowry & Gita Martohardjono
217
tracker, and the experiment was presented with E-Prime 2.0 (Schneider et
al., 2002).
4.1.4 Analysis
Following the analysis of Stern et al. (2019), gaze data was divided into four
temporal regions of the stimulus. This is illustrated in Figure 3.
Figure 3. Division of auditory stimuli into temporal regions
Region 1 was coded from the onset of the sentence until the onset of the first
word after the relativizer que. Linguistically, this information is equivalent
across both conditions. No information is provided to the participant that
would allow them to eliminate any of the three images in this region. During
Region 2 (the Relative Clause Region), participants hear the onset of the
first word after the relativizer que through the onset of the matrix verb. For
SRCs the first word of this region is the subordinate verb, while for ORCs
the first word is the determiner el (which begins the subordinate noun
phrase). Information provided during the Relative Clause Region would
allow participants to eliminate the ‘Other RC’ distractor. Region 3 (the
Matrix Predicate Region) provides the remaining information required to
eliminate the ‘Consistent’ Distractor and converge upon the target image,
and for both conditions this region extends from the onset of the matrix verb
to sentence offset. Finally, Region 4 is the time window from spoken
sentence offset until participants click on an image. No linguistic
information is provided in the final region. The two regions of interest for
this study are the Relative Clause Region and the Matrix Predicate Region.
To analyze gaze data, only observations for which the eye tracker was at its
highest validity level were retained (0 on the 0-4 scale output by the eye
tracker: 8.9% of the data was removed). Furthermore, two participants with
average total gaze fixation proportions of less than 30% on any image across
all regions and stimuli were removed entirely from gaze data analysis.
Proportion of fixation on the target image within each region from accurate
trials only was then used as the dependent variable for gaze models. Two
behavioral measures were also analyzed in addition to gaze data:
comprehension accuracy and response time. Participant-average accuracy
Effects of Language Dominance on L1 Relative Clause Processing
218
was calculated by relative clause type, and comprehension accuracy was
operationalized as dichotomous accuracy (0,1) for modeling. One highly
Spanish-dominant participant was removed as an outlier from analyses
because their average accuracy across conditions was more than three
standard deviations lower than the mean. Response time, measured from
sentence offset until participants clicked on an image, was log-transformed
to address the non-Gaussian distribution of the data, and outlying response
times of less than 50ms or greater than 20,000ms were removed. All
analyses, visualizations, and models were run using R (R Core Team, 2019).
4.2 Results
Our results are presented in two sections: gaze data results and behavioral
results. We begin each section by providing relevant descriptive statistics.
Then, we present inferential statistics for the effects of relative clause type
and language dominance on the dependent measures described above.
Finally, we provide a brief summary of all findings.
4.2.1 Gaze data
A course-grained analysis of target gaze fixation proportion by region is
plotted in Figure 4. This shows the proportion of fixation on the target image
by condition for each of the four temporal regions, including means and
within-subject adjusted 95% confidence intervals (see Morey, 2008). The
dashed line represents chance, assuming that participants are looking at one
of the three images on the screen. As expected, target fixation increases over
time as participants gain more information allowing them to converge on
the target image. Figure 5 shows target fixation proportions for SRCs and
ORCs in the two regions of interest; the Relative Clause Region and the
Matrix Predicate Region. Mean target fixation proportion is higher for SRCs
than ORCs in the Relative Clause Region (SRC: M = 0.421, SD = 0.521;
ORC: M = 0.374, SD = 0.527) but not in the Matrix Predicate Region (SRC:
M = 0.549, SD = 0.525; ORC: M = 0.544, SD = 0.557).
LeeAnn Stover, Michael C. Stern, Cass Lowry & Gita Martohardjono
219
Figure 4. Target proportion fixation by condition for each temporal region
Effects of Language Dominance on L1 Relative Clause Processing
220
Figure 5. Target fixation proportion in the Relative Clause Region (left) and Matrix
Predicate Region (right)
Proportion of gaze fixation on the target image during the Relative Clause
and Matrix Predicate Regions is modeled using mixed effects beta
regression models in R using the gamlss function (gamlss R package; Rigby
& Stasinopoulos, 2005). Beta regression models are argued to be flexible
for modeling limited range data such as proportion data (Johnson et al.,
1995). Fixation proportions are adjusted for beta regression by adding 10-17
to all values of 0 and subtracting 10-17 from all values of 1. Model predictors
include sentence type (SRC, ORC) which was sum contrast coded with SRC
as “-1” and ORC as “1”, language dominance which was scaled and
centered, and the interaction between the two factors. The gamlss function
affords the estimations of both means (μ; mu) and standard deviations (σ;
sigma), as well as the inclusion of random intercepts and slopes for random
effects (i.e., participants and items). We compared four models for the two
regions of interest using a generalized Akaike information criterion (GAIC).
LeeAnn Stover, Michael C. Stern, Cass Lowry & Gita Martohardjono
221
These models contrasted the benefit of the inclusion of a sentence type
random slope for items,
7
and the inclusion of fixed effects to model sigma.
The model with the lowest AIC for the Relative Clause Region included by-
participant and by-item random intercepts, as well as a sentence type
random slope for items, with no fixed effect for sigma. For the Matrix
Predicate Region, the model with the lowest AIC included fixed effects, by-
participant and by-item random intercepts, and sentence type random slopes
for items for both mu “μ” and sigma “σ”. While the sum-contrast coding
described above allowed us to model the effect of each predictor across
levels of the other, we also conducted follow-up models using treatment
contrasts to further probe the effect of language dominance at each level of
the sentence type predictor.
Relative Clause Region
Table 1. Effects of RC type and language dominance on target fixation
proportions- Relative Clause Region
Parameters
Fixed effect
Estimate
SE
t
P
μ
(Intercept)
-0.008
0.043
-0.182
0.856
link = logit
Sentence type (ST)
-0.084
0.042
-1.975
0.049
Language dominance (LD)
0.005
0.042
0.117
0.907
ST:LD
-0.099
0.042
-2.368
0.018
LD effect on SRCs
0.104
0.058
1.792
0.074
LD effect on ORCs
-0.094
0.060
-1.564
0.118
σ (link =
logit)
(Intercept)
2.641
0.032
83.680
0.000
The output of the Relative Clause Region model is shown in Table 1, and a
visualization of the relationship between language dominance and target
fixation is shown in Figure 6 (panel on the left). For SRC items, there is a
pattern of increased target fixation proportions as Spanish dominance
increases. The opposite pattern is found for ORCs: Spanish dominance
correlates negatively with target fixation proportion. Modeling shows no
7
Models with more complex random structure, particularly with dominance random
slopes or any random slope for participants did not converge.
Effects of Language Dominance on L1 Relative Clause Processing
222
main effect of language dominance on target fixation proportions in the
Relative Clause Region, while sentence type does show a main effect such
that target fixation proportions are higher for SRCs than ORCs. The
interaction between these two factors is also a significant predictor. Follow
up models show that the effect of language dominance has opposite
directionality for the SRC and ORC conditions, with neither significantly
driving the interaction more than the other, although the effect is marginal
for SRC sentences. Thus, the interaction appears to be driven by both the
increase in Spanish dominance as target fixation increases during SRCs as
well as the decrease in Spanish dominance as target fixation increases
during ORCs.
Matrix Predicate Region
Table 2. Effects of RC type and language dominance on target fixation
proportions- Matrix Predicate Region
Parameters
Fixed effect
Estimate
SE
t
p
μ
(Intercept)
0.061
0.043
1.405
0.160
link = logit
Sentence type (ST)
-0.017
0.043
-0.391
0.696
Language dominance (LD)
-0.118
0.043
-2.758
0.006
ST:LD
-0.117
0.043
-2.738
0.006
LD effect on SRCs
-0.001
0.058
-0.015
0.988
LD effect on ORCs
-0.236
0.063
-3.755
0.000
σ
(Intercept)
2.858
0.031
91.344
0.000
link = logit
Sentence type (ST)
0.112
0.031
3.580
0.000
Language dominance (LD)
0.033
0.030
1.101
0.271
ST:LD
0.056
0.030
1.841
0.066
LD effect on SRCs
-0.022
0.041
-0552
0.581
LD effect on ORCs
0.089
0.045
1.983
0.048
Effects of relative clause type and language dominance on target fixation
proportions for the Matrix Predicate Region are modeled in Table 2 and
LeeAnn Stover, Michael C. Stern, Cass Lowry & Gita Martohardjono
223
visualized in Figure 6 (panel on the right). Similar to the Relative Clause
Region, the Matrix Predicate Region shows a positive correlation trend
between target fixation proportions and Spanish dominance for SRCs and a
negative correlation for ORCs. While relative clause type shows no main
effect on target fixation, language dominance is a significant predictor in
this region such that as Spanish dominance increases, proportion fixation to
the target image decreases. Conversely, increased English dominance
predicts increased target fixation proportions across conditions. The model
also revealed a significant interaction between language dominance and
sentence type, which follow up models show to be largely driven by the
negative effect of increased Spanish dominance on gaze fixation
proportions to the target image while parsing a sentence with an ORC. The
model of sigma revealed a significant effect of sentence type and a
marginally significant interaction between relative clause type and language
dominance. Target fixations were significantly more variable for ORCs than
SRCs, and increased Spanish dominance increased variance during ORC
processing but not SRC processing. In the Matrix Predicate Region,
increased Spanish dominance is detrimental to target fixation during ORCs.
Figure 6. Target fixation proportions by Spanish dominance for Regions 2 and 3,
where positive scores indicate greater Spanish dominance.
Effects of Language Dominance on L1 Relative Clause Processing
224
4.2.2 Behavioral data
Accuracy and log-transformed response time results by condition are
visualized in Figure 7, with violin plots including means and 95%
confidence intervals (adjusted for within-subject designs). Overall,
participants were highly accurate in both conditions, though participants
were more accurate on items with SRCs (M = 0.936, SD = 0.32) than those
with ORCs (M = 0.850, SD = 0.47). In turn, mean response time was lower
for the SRCs (M = 7.568, SD = 1.27) than for ORCs (M = 7.787, SD = 1.17).
Figure 7. Comprehension accuracy and log-transformed response time by condition
Response accuracy is modeled through a generalized linear mixed effects
regression using the glmer function (lme4 package; Bates et al., 2014) with
a binary dependent variable coded for correct and incorrect responses.
Predictors in the model include sentence type (SRC vs. ORC), participants’
relative language dominance, and the interaction between the two. Sentence
type is sum contrast coded, while the continuous predictor of language
dominance is scaled and centered. We opted for models with maximal
random structure justified by the design and followed recommendations
given (Barr et al., 2013) whenever confronted with non-convergence. The
LeeAnn Stover, Michael C. Stern, Cass Lowry & Gita Martohardjono
225
model is reported in Table 3, which includes by-item and by-participant
random intercepts as well as their corresponding random slope for sentence
type.
Table 3. Effects of RC type and language dominance on comprehension
accuracy
Estimate
SE
x2
p(x2)
(Intercept)
2.973
0.261
Sentence type (ST)
-0.653
0.204
9.009
0.003
Language dominance (LD)
0.106
0.203
0.275
0.600
ST:LD
-0.239
0.182
1.714
0.191
Based on significance reporting from a chi-square statistic using a model
comparison approach, sentence type is the only predictor with a main effect
on accuracy. Participants are significantly more accurate on SRCs than
ORCs. Language dominance and the interaction between the two predictors
do not show a main effect on accuracy.
Log-transformed response times are modeled using a linear mixed effects
regression also with the lme4 R package (Bates et al., 2014). The results
from the model are presented in Table 4. We follow the same approach as
for the accuracy data, and the final model includes predictors of sentence
type and language dominance (both as described above) as well as the
interaction between the two, with by-item and by-participant random
intercepts as well as their corresponding random slope for sentence type.
Table 4: Effects of RC type and language dominance on log-
transformed response time
Estimate
SE
x2
p(x2)
(Intercept)
7.711
0.115
Sentence type (ST)
0.110
0.044
4.845
0.028
Language dominance (LD)
0.225
0.097
5.089
0.024
ST:LD
0.038
0.028
1.791
0.181
Significance scores from chi-square model comparisons reveal a main effect
of sentence type such that ORCs have longer response times than SRCs, and
Effects of Language Dominance on L1 Relative Clause Processing
226
a main effect of language dominance such that increased Spanish
dominance leads to increased response times.
Figure 8. Log-transformed response time by language dominance for SRCs and
ORCs
4.3 Summary of findings
In the gaze data, increased Spanish dominance led to greater evidence of a
subject-object asymmetry in relative clause processing, driven primarily by
a late negative effect of increased Spanish dominance on ORC processing
(in the Matrix Predicate Region). Language dominance did not have a main
effect on target gaze fixation proportions in the Relative Clause Region, but
a significant interaction in this region showed that increased Spanish
dominance led to higher target fixation proportions for SRCs and to lower
target fixation proportions for ORCs. Relative clause type also had a main
effect in the Relative Clause Region, such that target fixation proportions
were greater for SRCs than ORCs. In the Matrix Clause Region, language
dominance predicted target fixation proportions while sentence type in itself
did not. A significant interaction in the Matrix Predicate Region showed that
as Spanish dominance increased, target fixation proportions increased for
SRCs but decreased for ORCS, and the interaction was driven mainly by
the negative effect on ORCs. Sentence type was the only significant
LeeAnn Stover, Michael C. Stern, Cass Lowry & Gita Martohardjono
227
predictor of comprehension accuracy, such that SRCs had significantly
higher accuracy than ORCs. Finally, both sentence type and language
dominance had significant effects on log-transformed response times. ORCs
led to greater response times than SRCs, and increased Spanish dominance
also led to slower response times.
5. Discussion
Overall, the prediction that increased Spanish dominance leads to greater
evidence of a subject-object processing asymmetry is supported. The
processing asymmetry is present to a higher degree among participants with
increased Spanish dominance, as evidenced by the interaction between
language dominance and relative clause type in both regions. Interestingly,
the increased SRC/ORC asymmetry among participants with greater
Spanish dominance is driven more by a late ORC processing cost than an
early SRC benefit. In the Relative Clause Region (Region 2), English-
dominant participants are looking at the target at almost the same proportion
for SRCs and ORCs (see Figure 6). Conversely, Spanish-dominant
participants diverge in their target fixations for SRCs and ORCs by the
Relative Clause Region. There is minimal evidence that this reflects an SRC
advantage in the Relative Clause Region though, as higher Spanish
dominance only leads to slightly higher target fixation proportions. This
pattern holds for SRCs in the Matrix Predicate Region (Region 3), where
increased Spanish dominance offers only the slightest increase in target gaze
fixation. On the other hand, there is a steep decline in target fixation
proportions in this region for ORCs as Spanish dominance increases.
We posit that the detrimental effect of Spanish dominance on ORC
processing in the matrix predicate region is due to increased integration
costs for participants with higher Spanish dominance when an SRC is
predicted (along the lines of the active filler hypothesis) and the missed
parse forces a reanalysis. While this study did not explicitly measure this
prediction, results are compatible with an interpretation that participants
who are not dominant in the testing language are not utilizing active
prediction and online resources to the extent that test-language-dominant
participants are. This was suggested in Stern et al. (2019) as a group-level
effect of differential online resources, such that the heritage speakers (who
were generally less Spanish-dominant) predicted less actively than the late
bilinguals (who were generally more Spanish-dominant). The current study
indicates that this group-level difference was in fact driven by an effect of
L1 language dominance on L1 processing. This interpretation would explain
Effects of Language Dominance on L1 Relative Clause Processing
228
the slight advantage on SRCs with increased Spanish dominance, and the
more substantial detrimental effect on ORCs.
The behavioral results were slightly unexpected. Comprehension accuracy
was not affected by language dominance, which is consistent with previous
findings that accuracy is a less sensitive measure for highly proficient
bilingual populations with subtle differences in relative language strength
(e.g., O’Grady et al., 2009). However, participants with higher Spanish
dominance responded slower to Spanish stimuli than participants with
higher English dominance. While no significant interaction was found,
numerical trends suggest that the processing demand of ORCs in Spanish
may affect response times among Spanish-dominant bilinguals more than
English-dominant participants.
This study demonstrates that operationalizing language dominance as a
continuous, relative measure has the potential to reveal insights into
bilingual language processing beyond those afforded by categorical or
absolute measures. While many studies have found that language
dominance correlates with different measures of language proficiency,
performance, and comprehension (see Section 2), most previous studies
have still treated language dominance categorically. That is, groups have
been defined as dominant in Language A or dominant in Language B.
However, as the theoretical literature on language dominance emphasizes,
the construct is inherently gradient with different individual degrees of
language dominance. Operationalizing it as such allows us to more carefully
explore the effects of language dominance on other measures, which could
benefit the field of bilingual processing immensely.
6. Conclusion
This study probed the effect of language dominance on the online
processing and offline comprehension of subject and object relative clause
constructions. Spanish-English bilinguals along a continuum of language
dominance who share Spanish as the L1 but live in an English-dominant
society completed an eye-tracking study with aurally-presented Spanish
stimuli during a picture selection task. Our results demonstrate that
language dominance impacts bilinguals’ eye movements as they are
presented aural stimuli in their first-learned language while looking at
corresponding images. This influence extends beyond group differences
between heritage speakers and late bilinguals as found in Stern et al. (2019)
and shows that individual language dominance plays an important role in
bilingual relative clause processing. Participants with higher Spanish
LeeAnn Stover, Michael C. Stern, Cass Lowry & Gita Martohardjono
229
dominance seem to integrate aural information quickly during stimuli with
subject relatives and have a delayed convergence on the correct image
during object relative constructions. On the other hand, participants with
lower Spanish dominance (i.e., more English dominant) do not seem to
demonstrate the expected processing asymmetry between subject and object
relative clauses despite comparable levels of comprehension to Spanish-
dominant counterparts.
This study revealed a detrimental effect of ORCs that disproportionately
affects the eye movements and response times of highly Spanish-dominant
individuals as compared to less Spanish-dominant participants. This effect
of language dominance could possibly be due to differential processing
strategies in the dominant and non-dominant languages, as more active
online prediction among individuals with greater Spanish dominance could
possibly be leading to higher integration costs. However, further probing of
this topic is crucial to understanding the interaction between language
dominance and bilingual processing. Importantly, this study demonstrates
the utility of continuous, relative measures of language dominance in the
investigation of bilingual language processing.
Acknowledgements
We are grateful to the vibrant Spanish-speaking community in NYC which
has inspired and allowed us countless ways to explore the bilingual
experience, and to those members who participated in this study. Many
thanks to Ernesto Guerra for his significant contribution to data analysis.
We would also like to thank Christen N. Madsen II, Richard G. Schwartz,
and Second Language Acquisition Lab research assistants Daniela Castillo,
Daniel Choconta, Christina Dadurian, Andrea Monge, Omar Ortiz, Matthew
Stuck, and Armando Tapia.
References
Allopenna, P. D., Magnuson, J. S., & Tanenhaus, M. K. (1998). Tracking
the time course of spoken word recognition using eye movements:
Evidence for continuous mapping models. Journal of Memory and
Language, 38(4), 419–439.
Altmann, G. T., & Kamide, Y. (1999). Incremental interpretation at verbs:
Restricting the domain of subsequent reference. Cognition, 73(3), 247–
264.
Effects of Language Dominance on L1 Relative Clause Processing
230
Babyonyshev, M., & Gibson, E. (1999). The complexity of nested structures
in Japanese. Language, 423–450.
Barr, D. J., Levy, R., Scheepers, C., & Tily, H. J. (2013). Random effects
structure in mixed-effects models: Keep it maximal. Journal of Memory
and Language, 68(3), 255–278.
Bates, D., Mächler, M., Bolker, B., & Walker, S. (2014). Fitting linear
mixed-effects models using lme4. ArXiv Preprint ArXiv: 1406.5823.
Bedore, L. M., Peña, E. D., Summers, C. L., Boerger, K. M., Resendiz, M.
D., Greene, K., Bohman, T. M., & Gillam, R. B. (2012). The measure
matters: Language dominance profiles across measures in Spanish–
English bilingual children. Bilingualism: Language and Cognition,
15(3), 616–629.
Betancort, M., Carreiras, M., & Sturt, P. (2009). The processing of subject
and object relative clauses in Spanish: An eye-tracking study. Quarterly
Journal of Experimental Psychology, 62(10), 1915–1929.
Birdsong, D. (2015). Dominance in bilingualism: Foundations of
measurement, with insights from the study of handedness. In C. Silva-
Corvalán & J. Treffers-Daller (eds.) Language Dominance in
Bilinguals: Issues of Measurement and Operationalization (pp. 85–
105). Cambridge: Cambridge University Press.
Birdsong, D., Gertken, L. M., & Amengual, M. (2012). Bilingual language
profile: An easy-to-use instrument to assess bilingualism. COERLL,
University of Texas at Austin.
Brouwer, S., Özkan, D., & Küntay, A. C. (2017). Semantic prediction in
monolingual and bilingual children. In E. Blom, L. Cornips & J.
Schaeffer (eds.) Cross-linguistic Influence in Bilingualism (pp. 49–74).
Amsterdam: John Benjamins.
Clifton, C., & Frazier, L. (1989). Comprehending Sentences with Long-
Distance Dependencies. In G.N. Carlson, M.K. Tanenhaus (eds.)
Linguistic Structure in Language Processing. Studies in Theoretical
Psycholinguistics, vol 7. Dordrecht: Springer.
Comrie, B., & Fernández, Z. E. (2012). Relative Clauses in Languages of
the Americas: A Typological Overview. Amsterdam: John Benjamins.
Cooper, R. M. (1974). The control of eye fixation by the meaning of spoken
language: A new methodology for the real-time investigation of speech
perception, memory, and language processing. Cognitive Psychology,
6(1), pp. 84-107.
Dunn, A. L., & Tree, J. E. F. (2009). A quick, gradient Bilingual Dominance
Scale. Bilingualism: Language and Cognition, 12(03), 273-289.
Dussias, P. E., Valdés Kroff, J. R., Guzzardo Tamargo, R. E., & Gerfen, C.
(2013). When Gender and Looking Go Hand in Hand: Grammatical
LeeAnn Stover, Michael C. Stern, Cass Lowry & Gita Martohardjono
231
Gender Processing In L2 Spanish. Studies in Second Language
Acquisition, 35(2), 353–387.
Fernández, E. M. (2003). Bilingual Sentence Processing: Relative Clause
Attachment in English and Spanish. Amsterdam: John Benjamins.
Flege, J. E., Mackay, I. R. A., & Piske, T. (2002). Assessing bilingual
dominance. Applied Psycholinguistics, 23(04), 567–598.
Gollan, T. H., Starr, J., & Ferreira, V. S. (2015). More than use it or lose it:
The number-of-speakers effect on heritage language proficiency.
Psychonomic Bulletin & Review, 22(1), 147–155.
Gollan, T. H., Weissberger, G. H., Runnqvist, E., Montoya, R. I., & Cera,
C. M. (2012). Self-ratings of spoken language dominance: A multi-
lingual naming test (MINT) and preliminary norms for young and aging
Spanish-English bilinguals. Bilingualism: Language and Cognition,
15(3), 594.
Grodner, D., & Gibson, E. (2005). Consequences of the serial nature of
linguistic input for sentenial complexity. Cognitive Science, 29(2), 261–
290.
Grodner, D. J., Klein, N. M., Carbary, K. M., & Tanenhaus, M. K. (2010).
“Some,” and possibly all, scalar inferences are not delayed: Evidence
for immediate pragmatic enrichment. Cognition, 116(1), 42–55.
Grosjean, F. (2015). The Complementarity Principle and its impact on
processing, acquisition, and dominance. In C. Silva-Corvalán & J.
Treffers-Daller (eds.) Language Dominance in Bilinguals: Issues of
Measurement and Operationalization (pp. 66–84). Cambridge:
Cambridge University Press.
Huettig, F., Rommers, J., & Meyer, A. S. (2011). Using the visual world
paradigm to study language processing: A review and critical
evaluation. Acta Psychologica, 137(2), 151–171.
Johnson, N. L., Kotz, S., & Balakrishnan, N. (1995). Continuous univariate
distributions. New York: John Wiley.
Juffs, A., & Rodríguez, G. A. (2014). Second language sentence processing.
London: Routledge.
Lambert, W. E. (1955). Measurement of the linguistic dominance of
bilinguals. Journal of Abnormal Psychology, 50(2), 197–200.
Lew-Williams, C., & Fernald, A. (2010). Real-time processing of gender-
marked articles by native and non-native Spanish speakers. Journal of
Memory and Language, 63(4), 447–464.
Luk, G., & Bialystok, E. (2013). Bilingualism is not a categorical variable:
Interaction between language proficiency and usage. Journal of
Cognitive Psychology, 25(5), 605–621.
Effects of Language Dominance on L1 Relative Clause Processing
232
MacDonald, M. C. (2013). How language production shapes language form
and comprehension. Frontiers in Psychology, 4, 226.
Madsen II, C. N. (2018). De-centering the monolingual: A
psychophysiological study of heritage speaker language processing.
Ph.D Dissertaion. The Graduate Center, CUNY.
McKee, C., McDaniel, D., & Snedeker, J. (1998). Relatives children say.
Journal of Psycholinguistic Research, 27(5), 573–596.
Meisel, J. M. (2007). The weaker language in early child bilingualism:
Acquiring a first language as a second language? Applied
Psycholinguistics, 28(3), 495.
Montrul, S. (2015). Dominance and proficiency in early and late
bilingualism. In C. Silva-Corvalán & J. Treffers-Daller (eds.) Language
Dominance in Bilinguals: Issues of Measurement and Operationalization
(pp. 15–35). Cambridge: Cambridge University Press.
Montrul, S., & Bowles, M. (2009). Back to basics: Incomplete knowledge
of Differential Object Marking in Spanish heritage speakers.
Bilingualism: Language and Cognition, 12(3), 363–383.
Morey, R. D. (2008). Confidence intervals from normalized data: A
correction to Cousineau (2005). Reason, 4(2), 61–64.
O’Grady, W. (2011). Relative clauses: Processing and acquisition. In E. Kid
(ed.) The Acquisition of Relative Clauses: Processing, Typology and
Function. Amsterdam: John Benjamins 13–38.
O’Grady, W., Schafer, A. J., Perla, J., Lee, O.-S., & Wieting, J. (2009). A
psycholinguistic tool for the assessment of language loss: The HALA
project. Language Documentation & Conservation, 3(1), pp. 100-112.
Puig-Mayenco, E., Cunnings, I., Bayram, F., Miller, D., Tubau, S., &
Rothman, J. (2018). Language dominance affects bilingual performance
and processing outcomes in adulthood. Frontiers in Psychology, 9,
1199.
R Core Team. (2019). R: A language and environment for statistical
computing. R Foundation for Statistical Computing, Vienna, Austria.
https://www.R-project.org/
Rigby, R. A., & Stasinopoulos, D. M. (2005). Generalized additive models
for location, scale and shape, Applied Statistics, 3(54), 507–554.
Rizzi, L. (1990). Relativized minimality. Cambridge, MA: The MIT Press.
Robinson, J. J., & Blumenfeld, H. K. (2018). Language dominance predicts
cognate effects and inhibitory control in young adult bilinguals.
Bilingualism: Language and Cognition, 22(5), 1–17.
Schneider, W., Eschman, A., & Zuccolotto, A. (2002). E-Prime: User’s
guide. Pittsburg, PA: Psychology Software Incorporated.
LeeAnn Stover, Michael C. Stern, Cass Lowry & Gita Martohardjono
233
Sheldon, A. (1974). The role of parallel function in the acquisition of
relative clauses in English. Journal of Verbal Learning and Verbal
Behavior, 13(3), 272–281.
Silva-Corvalán, C., & Treffers-Daller, J. (2015). Digging into dominance:
A closer look at language dominance in bilinguals. In C. Silva-Corvalán
& J. Treffers-Daller (eds.) Language Dominance in Bilinguals: Issues
of Measurement and Operationalization (pp. 1–14). Cambridge:
Cambridge University Press.
Stern, M. C., Madsen II, C. N., Stover, L. M., Lowry, C., & Martohardjono,
G. (2019). Language history attenuates syntactic prediction in L1
processing. Journal of Cultural Cognitive Science, 3(2), 235–255.
Stevens, L. S. (2019). Heritage Speaker and Late Bilingual L2 Relative
Clause Processing and Language Dominance Effects. Master’s Thesis.
The Graduate Center, CUNY.
Tanenhaus, M. K., Spivey-Knowlton, M. J., Eberhard, K. M., & Sedivy, J.
C. (1995). Integration of visual and linguistic information in spoken
language comprehension. Science, 268(5217), 1632–1634.
Traxler, M. J., Morris, R. K., & Seely, R. E. (2002). Processing subject and
object relative clauses: Evidence from eye movements. Journal of
Memory and Language, 47(1), 69–90.
Treffers-Daller, J. (2011). Operationalizing and measuring language
dominance. International Journal of Bilingualism, 15(2), 147–163.
Treffers-Daller, J. (2015). Language dominance: The construct, its
measurement, and operationalization. In C. Silva-Corvalán & J.
Treffers-Daller (eds.) Language Dominance in Bilinguals: Issues of
Measurement and Operationalization (pp. 235–265). Cambridge:
Cambridge University Press.
Treffers-Daller, J., & Silva-Corvalán, C. (2016). Language dominance in
bilinguals: Issues of measurement and operationalization. Cambridge:
Cambridge University Press.
