Constraints On Parallel Activation in Bilingual Spoken Language Processing: Examining Proficiency and Lexical Status Using Eye-Tracking
Constraints On Parallel Activation in Bilingual Spoken Language Processing: Examining Proficiency and Lexical Status Using Eye-Tracking
Constraints On Parallel Activation in Bilingual Spoken Language Processing: Examining Proficiency and Lexical Status Using Eye-Tracking
o
w
n
l
o
a
d
e
d
B
y
:
[
E
B
S
C
O
H
o
s
t
E
J
S
C
o
n
t
e
n
t
D
i
s
t
r
i
b
u
t
i
o
n
]
A
t
:
0
3
:
5
8
1
5
A
p
r
i
l
2
0
0
8
Constraints on parallel activation in bilingual spoken
language processing: Examining prociency and lexical
status using eye-tracking
Henrike K. Blumenfeld and Viorica Marian
Northwestern University, Evanston, IL, USA
During spoken word-recognition, bilinguals have been shown to access their
two languages simultaneously. The present study examined effects of language
proficiency and lexical status on parallel language activation. Language
proficiency was manipulated by testing German-native and English-native
bilingual speakers of German and English. Lexical status was manipulated by
presenting target words that either overlapped in form across translation
equivalents (cognate words) or did not overlap in form across translation
equivalents (English-specific words). Participants identified targets (such as
hen) from picture-displays that also included similar-sounding German
competitor words (such as Hemd, shirt). Eye-movements to German
competitors were used to index co-activation of German. Results showed
that both bilingual groups co-activated German while processing cognate
targets; however, only German-native bilinguals co-activated German while
processing English-specific targets. These findings indicate that high language
proficiency and cognate status boost parallel language activation in bilinguals.
Correspondence should be addressed to Henrike K. Blumenfeld, Department of
Communication Sciences and Disorders, 2240 Campus Drive, Northwestern University,
Evanston, IL 60208, USA. E-mail: k-blumenfeld@northwestern.edu
This work was supported by grants NICHD 1R03HD046952-01A1 and NSF BCS-0418495
to the second author, and by a Northwestern University Graduate Research Grant to the first
author.
Thanks go to Margarita Kaushanskaya, James Booth, Cynthia Thompson, and Judith Kroll
for insightful comments and discussion, and to Valerie Burt, Gayatri Menon, Naveen Malik,
Vridhi Chhabria, Nicole Kaligeropolous, Nadia Cone, Olga Boukrina, and Avital Rabin for their
assistance at various stages of the project. We are grateful to Manuel Carreiras and to two
anonymous reviewers for helpful comments and feedback on a previous version of this paper.
Parts of this work were presented at the Fifth International Symposium on Bilingualism, and at
the 27th Annual Meeting of the Cognitive Science Society.
LANGUAGE AND COGNITIVE PROCESSES
2007, 22 (5), 633660
#
2007 Psychology Press, an imprint of the Taylor & Francis Group, an Informa business
http://www.psypress.com/lcp DOI: 10.1080/01690960601000746
D
o
w
n
l
o
a
d
e
d
B
y
:
[
E
B
S
C
O
H
o
s
t
E
J
S
C
o
n
t
e
n
t
D
i
s
t
r
i
b
u
t
i
o
n
]
A
t
:
0
3
:
5
8
1
5
A
p
r
i
l
2
0
0
8
Recent evidence suggests that bilinguals activate their two languages in
parallel during language comprehension both in the auditory modality (e.g.,
Ju & Luce, 2004; Marian & Spivey, 2003a, 2003b; Schulpen, Dijkstra,
Schriefers, & Hasper, 2003; Spivey & Marian, 1999; Weber & Cutler, 2004)
and in the visual modality (e.g., Dijkstra & Van Heuven, 1998; Dijkstra, Van
Heuven, & Grainger, 1998; Grainger, 1993). Yet, parallel language activation
may be constrained by factors such as language proficiency (e.g., Weber &
Cutler, 2004), language exposure (e.g., Spivey & Marian, 1999), and acoustic
characteristics of the input signal (e.g., voice-onset time, Ju & Luce, 2004).
The goal of the present research was to investigate how language proficiency
and lexical status constrain parallel language activation during bilingual
spoken word recognition.
During spoken word recognition, multiple word candidates (i.e., cohort
members) that match the acoustic input become active, and as the input
unfolds over time, the best match (i.e., the target) is selected (e.g., Marslen-
Wilson, 1987). Tanenhaus, Spivey-Knowlton, Eberhard, and Sedivy (1995)
covertly measured competition between target-words and cohort-words by
using eye-tracking. Participants heard object names, and were asked to
identify these target objects from a set of items in a visual display. During
target identification, participants eye-movements to cohort members (hen-
ceforth competitors) reflected parallel activation of both items.
1
For example,
if participants heard the word marker, they were likely to also look at a marble
(see Figure 1A). This monolingual eye-tracking paradigm (e.g., Tanenhaus et
al., 1995) was extended to investigate whether cohort words in bilinguals two
languages become active in parallel during spoken word recognition (Marian
& Spivey, 2003a, 2003b; Spivey & Marian 1999; see Figure 1B). Bilinguals
heard a word in one language, and identified it from a set of objects that also
included a competitor from their other language. Findings showed that when
Russian-English bilinguals heard the word marker in English, they were likely
to also look at the Russian between-language competitor marka (stamp).
This finding of parallel language activation has since been replicated with
DutchEnglish bilinguals (Weber & Cutler, 2004), SpanishEnglish bilin-
guals (CansecoGonzales, Brick, Fischer, & Wagner, 2005; Ju & Luce, 2004),
FrenchEnglish bilinguals (Weber & Paris, 2004), and JapaneseEnglish
bilinguals (Cutler, Weber, & Otake, 2006). In the current study, we extended
the bilingual eye-tracking paradigm to investigate constraints on parallel
language activation. The influence of proficiency on parallel activation was
examined by testing two bilingual groups. One bilingual group consisted of
German-native bilinguals who were highly proficient in German; the other
1
For a linking hypothesis between linguistic processing and eye-movements, see Tanenhaus,
Magnuson, Dahan, and Chambers (2000).
634 BLUMENFELD AND MARIAN
D
o
w
n
l
o
a
d
e
d
B
y
:
[
E
B
S
C
O
H
o
s
t
E
J
S
C
o
n
t
e
n
t
D
i
s
t
r
i
b
u
t
i
o
n
]
A
t
:
0
3
:
5
8
1
5
A
p
r
i
l
2
0
0
8
Figure 1. Unfolding of the acoustic signal, and cohort activation within and across languages, as indexed by eye-movements. The top panel illustrates
Marslen-Wilsons Cohort Model (1987) in a monolingual scenario (Tanenhaus et al. , 1995). The bottom panel extends the model to a bilingual scenario,
as proposed by Marian (2000) and replicated by others (Canseco-Gonzalez et al., 2005; Cutler, et al., 2006; Ju & Luce, 2004; Marian & Spivey, 2003a,
2003b; Spivey & Marian, 1999; Weber & Cutler, 2004; Weber & Paris, 2004).
B
I
L
I
N
G
U
A
L
P
A
R
A
L
L
E
L
L
A
N
G
U
A
G
E
A
C
T
I
V
A
T
I
O
N
6
3
5
D
o
w
n
l
o
a
d
e
d
B
y
:
[
E
B
S
C
O
H
o
s
t
E
J
S
C
o
n
t
e
n
t
D
i
s
t
r
i
b
u
t
i
o
n
]
A
t
:
0
3
:
5
8
1
5
A
p
r
i
l
2
0
0
8
bilingual group consisted of English-native bilinguals who were less proficient
in German. The influence of lexical status on parallel language activation was
examined by presenting target words that either overlapped in form with their
translation equivalents (cognate words, such as hen, Henne in German) or did
not overlap in form with their translation equivalents (English-specific words,
such as dress, Kleid in German). The entire experiment was conducted in
English only, with no use of German before or during the study.
LANGUAGE PROFICIENCY AND
PARALLEL LANGUAGE ACTIVATION
During bilingual word-recognition, proficiency in the language irrelevant to
the task has been found to constrain parallel language activation. Parallel
language activation is more reliable when proficiency in the unused language
is high than when it is low (e.g., Jared & Kroll, 2001; Silverberg & Samuel,
2004; Van Hell & Dijkstra, 2002; Weber & Cutler, 2004). In the visual
modality, for instance, Silverberg and Samuel (2004) found that late
bilinguals with high L2 proficiency showed between-language form priming,
while late bilinguals with low L2 proficiency did not. Jared and Kroll (2001)
recruited FrenchEnglish and EnglishFrench bilinguals and compared the
degree to which they co-activated French phonology while reading aloud
English words. French phonology was activated by French-native bilinguals
(with higher proficiency in French), but not by English-native bilinguals
(with lower proficiency in French). Further, Van Hell and Dijkstra (2002)
tested trilinguals with a highly proficient second language and a less
proficient third language. On an L1 lexical decision task, participants co-
activated a highly proficient L2 while processing L1L2 cognates, but did
not co-activate a less proficient L3 while processing L1L3 cognates.
In the auditory modality, the role of proficiency in parallel language
activation has also been examined. Bilinguals listening to words in their
lower-proficiency language have consistently shown co-activation of their
higher-proficiency language (Marian & Spivey, 2003a, 2003b; Weber &
Cutler, 2004; Weber & Paris, 2004). However, bilinguals listening to words in
their higher-proficiency language have not always shown co-activation of
their lower-proficiency language. Some studies have found parallel activation
of a lower-proficiency language (Marian & Spivey, 2003a; Spivey & Marian,
1999), while others have not (Ju & Luce, 2004; Weber & Cutler, 2004). Extent
of immersion in the second language may be one explanation for these
differences between studies. Moreover, testing bilinguals in both languages
may confound research on proficiency effects due to differences in linguistic
structure and stimulus sets across languages. Another way to examine the
influence of proficiency on parallel language activation is to hold the testing
636 BLUMENFELD AND MARIAN
D
o
w
n
l
o
a
d
e
d
B
y
:
[
E
B
S
C
O
H
o
s
t
E
J
S
C
o
n
t
e
n
t
D
i
s
t
r
i
b
u
t
i
o
n
]
A
t
:
0
3
:
5
8
1
5
A
p
r
i
l
2
0
0
8
language constant and to vary proficiency in the unused language across
groups. Such a design would allow for testing language and stimuli to remain
constant across proficiency comparisons, so that confounds from these
sources could be ruled out. In the current study, two bilingual groups with
different proficiency levels in German were tested.
LEXICAL STATUS AND PARALLEL LANGUAGE ACTIVATION
In the bilingual lexicon, translation equivalents are connected via associative
links (Chen & Leung, 1989; De Groot, 1992; Kroll & Curley, 1988; Kroll &
Stewart, 1994). When bilinguals process words in one language, they co-
activate translation equivalents in the other language (e.g., Hermans,
Bongaerts, De Bot, & Schreuder, 1998). Previous studies have shown that
form overlap between cognate translation equivalents results in high
activation of both wordforms (e.g., Costa, Caramazza & Sebastian-Galles,
2000, Van Hell & Dijkstra, 2002). Therefore, cognate words with high form-
overlap between translation equivalents (e.g., English cactus, German
Kaktus) were used to examine whether processing of words with high cross-
linguistic overlap facilitated parallel language activation. Evidence support-
ing parallel activation of cognate translation equivalents includes findings of
stronger cognate translation priming (Gollan, Forster & Frost, 1997), faster
cognate translation times (De Groot, 1992; De Groot, Dannenburg, & Van
Hell, 1994), and more accurate cognate processing (Friel & Kennison, 2001;
Tokowicz, Kroll, DeGroot, & Van Hell, 2002) relative to words with unrelated
translation equivalents (noncognates). In general, a processing advantage for
cognates is well-established during word production (e.g., Costa et al., 2000;
De Groot, Borgwaldt, Bos, & Van den Eijnden, 2002; De Groot & Keijzer,
2000; Gollan & Acenas, 2004; Kohnert, 2004; Roberts & Deslauriers, 1999)
and comprehension (e.g., De Groot et al., 2002; De Groot & Keijzer, 2000;
Dijkstra et al., 1998; Dijkstra, Grainger, & Van Heuven, 1999; Lalor &
Kirsner, 2001; Lemho fer, Dijkstra, & Michel, 2004; Nakayama, 2002; Van
Hell & Dijkstra, 2002). Nevertheless, it remains unclear how cognate
processing influences overall activation of an unused language. In the current
study, we examined whether processing cognate targets would boost co-
activation of an unused language during bilingual word recognition. We used
English-specific target-words that did not overlap phonologically across
translation equivalents (e.g., shark, Hai in German), and cognate target-
words that did overlap phonologically across translation equivalents (e.g.,
guitar, Gitarre in German). Previous research has shown that translation
equivalents are co-activated more for words that share form than for words
that do not share form (e.g., De Groot, 1992; De Groot et al., 1994). Thus, in
the present study, German translation equivalents should be more active for
BILINGUAL PARALLEL LANGUAGE ACTIVATION 637
D
o
w
n
l
o
a
d
e
d
B
y
:
[
E
B
S
C
O
H
o
s
t
E
J
S
C
o
n
t
e
n
t
D
i
s
t
r
i
b
u
t
i
o
n
]
A
t
:
0
3
:
5
8
1
5
A
p
r
i
l
2
0
0
8
English cognate targets than for English-specific targets. In addition, German
translation equivalents of cognate targets would overlap with German
competitors andwould directly co-activate them. For example, the translation
equivalent Gitarre of the cognate target guitar would directly co-activate its
German competitor Gitter (bars); yet the translation equivalent Hai of the
English-specific target shark would not co-activate its German competitor
Schal (scarf). It follows, then, that in the cognate-target condition, German
translation equivalents would be more active and would also co-activate
German competitors more. In contrast, in the English-specific-target condi-
tion, German translation equivalents would be less active and would co-
activate German competitors less. As a result, when the target is a cognate,
German competitors would be activated via links between translation
equivalents, and via bottom-up acoustic input. However, when the target is
English-specific, German competitors would be activated via bottom-up
acoustic input only. Due to these differences in co-activation mechanisms, we
predicted that co-activation of German competitors would be stronger with
cognate targets than with English-specific targets.
CURRENT STUDY
In the current study, we investigated parallel activation of German during
English word recognition. Participants heard object names in English, and
identified them among four pictures that included a similar-sounding
German competitor. German was never used overtly. Co-activation of
German was probed covertly by tracking participants eye-movements toward
pictures of German competitors (relative to control items). To examine the
role of proficiency in parallel language activation, we recruited two groups
of late bilinguals. One group consisted of German-native bilinguals who
were highly proficient in German; the other group consisted of English-
native bilinguals who were less proficient in German. An English monolingual
control group was also tested. To examine the role of Lexical Status in
parallel language activation, we manipulated overlap between targets
translation equivalents. One condition consisted of English targets (with
phonologically unrelated German translations); the other condition consisted
of cognate targets (with phonologically similar German translations).
In addition, the Phonological Overlap between target and competitor
word-onsets was manipulated for a preliminary look at the role of cross-
linguistic overlap in parallel activation. Onset similarity between target
words and German competitors was either low (e.g., English b
all and
German B