Content uploaded by John W. Schwieter
Author content
All content in this area was uploaded by John W. Schwieter on Apr 11, 2021
Content may be subject to copyright.
Language Learning ISSN 0023-8333
Concept Selection and Developmental
Effects in Bilingual Speech Production
John Schwieter
Wilfrid Laurier University
Gretchen Sunderman
Florida State University
The present study investigates the locus of language selection in less and more proficient
language learners, specifically testing differential predictions of La Heij’s (2005) concept
selection model (CSM) and Kroll and Stewart’s (1994) revised hierarchical model
(RHM). Less and more proficient English dominant learners of Spanish participated
in a Stroop translation task that included semantically related and unrelated word or
picture distracters. The results for the more proficient learners provide support for the
CSM as well as the RHM. The results for the less proficient learners provide support for
the RHM and demonstrate the continued reliance on lexical level links and the difficulty
in accessing the conceptual store during second language production. The selection by
proficiency model of bilingual speech production is discussed.
Keywords language selectivity; bilingual speech production; lexical processing;
concept selection
When monolinguals are asked to name an object that has two interchangeable
names, such as couch or sofa, the unnamed alternative has been found to be
active and competing for selection, almost as if the other word is on the tip
of their tongues (Jescheniak & Schriefers, 1998; Peterson & Savoy, 1998).
Monolinguals rarely suffer this type of competition because few word pairs,
such as couch and sofa, exist. However, for bilinguals, almost each and every
word in one language has a translation equivalent in the other language. These
translation equivalents can be thought of as cross-language synonyms. Indeed,
like the monolingual research, previous research on language production with
proficient bilinguals suggests that lexical candidates in the nontarget language
Correspondence concerning this article should be addressed to John Schwieter, Wilfrid Laurier
University, 75 University Ave. W., Waterloo, ON, Canada, N2L 3C5. Internet: jschwieter@wlu.ca
Language Learning 59:4, December 2009, pp. 897–927 897
C
2009 Language Learning Research Club, University of Michigan
Schwieter and Sunderman Concept Selection
Figure 1 Model of bilingual language production. Adapted from “First Language Use
in Second Language Production,” by N. Poulisse and T. Bongaerts, 1994, Applied Lin-
guistics, 15, pp. 36–57. Copyright Oxford University Press, 1994; and Word Production
in a Foreign Language, by D. Hermans, 2000, unpublished master’s thesis, University
of Nijmegen.
are available well into the process of speech planning (e.g., Costa, Miozzo, &
Caramazza, 1999; Hermans, Bongaerts, de Bot, & Schreuder, 1998).
One model of bilingual word production based on work by Poulisse and
Bongaerts (1994) and Hermans (2000) is shown in Figure 1 and illustrates the
process a bilingual goes through when naming a pictured object. The model
in Figure 1 adapts monolingual production models (e.g., Levelt, 1989; Levelt,
Roelofs, & Meyer, 1999) for the bilingual case. Three levels1of representation
are depicted: the conceptual level, the lemma level, and the phonological level.
The model represents an English-Spanish bilingual naming the object chair in
English. First, at the conceptual level, semantic features related to the meaning
Language Learning 59:4, December 2009, pp. 897–927 898
Schwieter and Sunderman Concept Selection
of the pictured object are activated. Information about the intended language of
production—in this case English—is also hypothesized to be active at this level.
At the next level, lemmas, or abstract lexical representations, are specified for
each of the lexical alternatives in the two languages. Both language alternatives,
chair in English and silla in Spanish, as well as lemmas that are semantically
related to the meaning of the object and share the same grammatical class, such
as table in English and mesa in Spanish, will be activated to differing degrees
at the lemma level. Finally, at the phonological level, the sound representation
of chair is specified.
In the bilingualism literature, the cross-language competition in bilingual
word production that occurs at the lemma level has recently been referred to
as the “hard problem” (Finkbeiner, Gollan, & Caramazza, 2006). Specifically,
the “hard problem” occurs when similar levels of activation are sent from the
conceptual level to the lexical level (i.e., words in both languages), leading to a
decision to select the correct word in the appropriate language that maps onto
the target concept. How does a bilingual solve this “hard problem”?
One model of bilingual speech production—La Heij’s (2005) concept se-
lection model (CSM)—proposes a solution to the hard problem. This model
argues that choosing between two competing words at the lexical level during
speech production is not a difficult task for bilinguals because the problem has
already been solved at the conceptual level. This model assumes that due to
the fact that the language of production is specified during the preverbalization
stage of speech production, more activation is sent to the lexicon of the target
language, making the selection of the target word a much easier process. La
Heij posited the notion of a “language cue” (i.e., speaking in either the first
language [L1] or the second language [L2]) that is similar to Levelt’s (1989)
hypothesis that information regarding aspects such as register is part of the
preverbal message. Thus, the basic assumption of La Heij’s model is that target
language selection occurs before lexical retrieval (i.e., at the conceptual level
and during preverbalization). In other words, when a picture is to be named in
English, although some Spanish words are activated, the English ones are only
considered in the lexical selection process.
Figure 2 illustrates an example of how lexical selection proceeds in bilin-
guals according to the CSM. When an English language learner of Spanish is
asked to name a picture of a chair in English, the preverbal message specifies
the concept and the target language. Consequently, the target language’s lexical
nodes receive more activation from these cues than the nontarget lexical nodes
(shown by dark and dotted lines, respectively). Once activation has been sent
to the lexical level, the correct word candidate is chosen based on activation
899 Language Learning 59:4, December 2009, pp. 897–927
Schwieter and Sunderman Concept Selection
Figure 2 Lexical selection according to La Heij’s (2005) concept selection model.
Adapted from “Lexical Access in Bilingual Speakers: What’s the (Hard) Problem?”
by M. Finkbeiner, T. Gollan, and A. Caramazza, 2006, Bilingualism: Language and
Cognition, 9(2), pp. 153–166. Copyright Cambridge University Press, 2006.
levels and, in turn, activates the corresponding phonemes at the phonologi-
cal level. Again, although words in the nontarget language are activated, the
English ones are only considered in the lexical selection process. Language se-
lection has taken place at the conceptual level. Empirical support from Stroop
translation tasks using this theoretical framework has been found for proficient
bilinguals (Bloem & La Heij, 2003; Bloem, van den Boogaard, & La Heij,
2004). However, it is not clear whether the hard problem can be solved ac-
cording to the CSM for less proficient language learners. An important factor,
namely the strength of and reliance on the L1, may make the CSM’s solution
to the hard problem potentially untenable for less proficient learners.
In order to understand lexical retrieval and production in less proficient
learners, we turn to the revised hierarchical model, (RHM) (Kroll & Stewart,
Language Learning 59:4, December 2009, pp. 897–927 900
Schwieter and Sunderman Concept Selection
L1 L2
concepts
lexical
links
conceptua
l
links
conceptual
links
Figure 3 The revised hierarchical model. Adapted from “Category Interference in
Translation and Picture Naming: Evidence for Asymmetric Connections Between
Bilingual Memory Representations,” by J. Kroll and E. Stewart, 1994, Journal of Mem-
ory and Language, 33, pp. 149–174. Copyright Elsevier, 1994.
1994). The RHM (see Figure 3) is a developmental model of the lexicon that
captures the interlanguage connections between lexical and conceptual repre-
sentations as learners become more proficient in their L2. The model proposes
independent lexical representations for words in each language, but an inte-
grated conceptual system via a lexical link, especially for less proficient L2
learners. Words in the L1 are assumed to have direct access to their respective
meanings, whereas words in the L2 are hypothesized to be associated to their
translation equivalents. The model also assumes that word-to-concept connec-
tions are stronger for the L1 than for the L2 for all but the most proficient and
balanced bilinguals. With increasing proficiency in the L2, the RHM further
assumes that the strength of word-to-concept connections for the L2 increases
and the presence of lexically mediated processing decreases.
To illustrate, during early stages of language learning for an English learner
of Spanish, the Spanish word perro is hypothesized to be associated with the
translation equivalent dog in English. The English word dog will have privileged
access to the meaning; thus, the word-to-concept connection is stronger in the
L1 than in the L2. As proficiency increases in the L2, the model hypothesizes
that the connection between perro and the concept will strengthen and the
dependency on the L1 translation equivalent will diminish. In other words,
only as proficiency increases will learners be able to access concepts directly.
901 Language Learning 59:4, December 2009, pp. 897–927
Schwieter and Sunderman Concept Selection
Although a great deal of evidence supports the claim that proficient bilin-
guals are able to process the L2 conceptually (de Groot, Dannenberg, & Van
Hell, 1994; La Heij, Hooglander, Kerling, & Van der Velden, 1996; Zeelenberg
& Pecher, 2003), the evidence for conceptual mediation in less proficient L2
learners is mixed. Some studies suggest that the ability to conceptually mediate
is available quite early in learning (Altarriba & Mathis, 1997; de Groot & Poot,
1997; Frenck-Mestre & Prince, 1997; Potter, So, Von Eckardt, & Feldman,
1984; Schwieter, 2007, 2008), whereas other studies suggest that it de-
velops in stages (Dufour & Kroll, 1995; Schwieter & Sunderman, 2008;
Talamas, Kroll, & Dufour, 1999) and still others argue that beginning L2
learners are unable to conceptually mediate (e.g., Kroll & Stewart, 1994; Sholl,
Sankaranarayanan, & Kroll, 1995). On the one hand, results from prior studies
have suggested that models of L2 vocabulary acquisition assume some kind of
lexical mediation at initial stages of word learning (Hall & Ecke, 2003; Jiang,
2000; Swain, 1997), whereas additional studies have found that language dom-
inance (Heredia, 1997) and maturational constraints (Weber-Fox & Neville,
1996) may modulate whether an L2 learner can directly access the conceptual
store by relying on L2 links (see Heredia & Altarriba, 2001, for further details.)
In the end, conceptual mediation at beginning levels of proficiency remains
a point of contention in the literature. However, the reliance on lexical links
within the RHM is less controversial. Previous research (Ferr´
e, Sanchez-Casas,
& Guasch, 2006; Schwieter, 2007, 2008; Schwieter & Sunderman; Talamas
et al., 1999) has consistently found that beginning bilinguals rely on the trans-
lation equivalent during L2 processing, whereas more advanced bilinguals are
less reliant on the lexical level links.
It is precisely because of less proficient learners’ reliance on the L1, as
predicted by the RHM, that language selection at the conceptual level may
be impossible for them. For less proficient learners, activation may indeed
spread to the lexical level, thus pushing the hard problem from the conceptual
to the lexical level.2In other words, whereas the CSM argues that proficient
bilinguals are able to rely on higher linguistic cues at the conceptual level to
establish the language of production, the RHM predicts that L1 lexical links
may be needed to mediate conceptual processing. To illustrate, imagine that a
beginning English language learner of Spanish is asked to translate the Spanish
word perro “dog” to English. According to the RHM, in order for the language
learner to be able to access the meaning of the word perro, the learner must first
translate it to English. At low levels of L2 proficiency, a language learner’s L2
lexical links may not be developed enough to be able to directly access the L2
word’s meaning. Moreover, because a language learner uses those L1 lexical
Language Learning 59:4, December 2009, pp. 897–927 902
Schwieter and Sunderman Concept Selection
links to access concepts, the relative activation of the L1 will be higher for less
proficient language learners compared to more proficient learners.
It is apparent that the CSM and the RHM posit competing predictions with
respect to whether language selection could even occur at the conceptual level.
The CSM argues that bilinguals are able to select language at the conceptual
level; the RHM predicts that the necessary reliance on L1 lexical links at low
levels of proficiency will essentially impede language selection from taking
place at the conceptual level and will force language selection down to the
lexical level. Given that all of the research on the CSM has been conducted
with highly proficient bilinguals and given that the role of proficiency in the
process of language selectivity has been of recent interest (Costa, 2005; Costa,
Santesteban, & Ivanova, 2006; Kroll, Bobb, & Wodniecka, 2006; Schwieter,
2007, 2008; Schwieter & Sunderman, 2008), this study tests the differential
predictions of these two models in a word translation task with both less
and more proficient L2 learners. We use predictions from two rather distinct
models—a bilingual word production model (CSM) and a developmental model
of the bilingual lexicon (RHM)—but argue, as others have (e.g., Finkbeiner,
Almeida, Janssen, & Caramazza, 2006), that lexical access and word production
go hand in hand. Although many speech production models take the highly
proficient bilingual case as the starting point, we argue that a developmental
account is necessary and that only by jointly examining predictions from both
perspectives are we able to accomplish this goal.
Accordingly, the present experiment is a Stroop translation task in which 54
English learners of Spanish with differing levels of L2 proficiency translated
words from their L2 to their L1. Critical variables included context pictures and
words that were either semantically related or unrelated to the target word. By
examining the effects of context modality and semantic relatedness on L2-to-
L1 translation, inferences were made regarding the locus of language selection
during speech production among less and more proficient language learners. In
the sections that follow, we discuss the logic of the Stroop translation task and
the semantic-relatedness effect that has been used with proficient bilinguals to
test the locus of language selection. We then present our research questions
and hypotheses regarding language selection for bilinguals of various levels of
proficiency.
Testing the Concept Selection Model
Research testing the CSM has used a word translation variant of the Stroop
task. In this task, a target word is presented with a word or picture distracter.
903 Language Learning 59:4, December 2009, pp. 897–927
Schwieter and Sunderman Concept Selection
The task is to translate the word as quickly as possible while ignoring the
distracters. By manipulating the relationship of the distracters and the words
to be translated, it is possible to infer the nature of the processes that are
active at a given point in planning the spoken utterance. In fact, two studies in
particular (La Heij et al., 1990, 1996) demonstrated that simply changing the
distracters from words to pictures will produce a complete reversal of semantic
interference to semantic facilitation. Essentially, this paradox has been referred
to as the semantic-relatedness effect, in which the presence of a semantic
relation between a target stimulus and a context has a facilitative effect in some
cases and causes interference in others (Neumann, 1986).
La Heij et al. (1990) used semantically related and unrelated distracter words
presented either at a stimulus onset asynchrony3(SOA) of +140 ms or −400
ms in a translation Stroop task and asked highly proficient Dutch-English bilin-
guals to articulate the translation of the stimulus words from the L2 (English)
to the L1 (Dutch). This direction of translation is sometimes referred to in the
literature as “backward translation.” For example, the to-be-translated English
word spoon was either accompanied by the Dutch word vork “fork” or by an un-
related Dutch context word. Although related distracter words caused semantic
interference at SOAs of +140 ms, such interference disappeared at SOAs of
−400 ms. The researchers explained this lack of interference by suggesting that
the semantic interference effect is modulated by a rapidly decaying phonologi-
cal representation of the distracter words. La Heij et al. (1996) also investigated
backward translation in a Stroop translation experiment but used picture dis-
tracters. Using the above example, the distracter for the to-be-translated word
spoon would be a picture of a fork or an unrelated picture. In their two exper-
iments testing the semantic-relatedness effect, there was semantic facilitation
when participants translated words from the L2 to the L1 in the presence of
semantically related pictures (in Experiment 3, related context =817 ms and
unrelated context =846 ms; in Experiment 4, related context =865 ms and
unrelated context =910 ms). This suggests support for conceptual mediation.
Given that La Heij et al.’s (1996) participants were not balanced bilinguals, the
semantic facilitation found in backward translation and the lack of support for
lexical mediation in bilingual speech production posed a challenge to Kroll and
Stewart’s (1994) claim that L2-to-L1 translation was lexically mediated rather
than conceptually mediated, although Kroll and Stewart’s results are based on
categorized list experiments, whereas La Heij’s results are based on the Stroop
word translation experiments. See Kroll and Tokowicz (2001) for further dis-
cussion of this issue and other alternative explanations for the contrary findings
in these two studies. Indeed, La Heij et al. (1996) reported data that did not
Language Learning 59:4, December 2009, pp. 897–927 904
Schwieter and Sunderman Concept Selection
include evidence for lexical mediation, and due to this lack of support, they
postulated conceptual mediation in all bilinguals.
Because La Heij et al.’s (1990) study revealed that backward translation
was impeded by the presence of semantically related distracter words com-
pared to unrelated distracter words and La Heij et al.’s (1996) study reported
that backward translation was facilitated by the presence of semantically related
distracter pictures compared to unrelated pictures, Bloem and La Heij (2003)
and Bloem et al. (2004) sought to explore the possibility that context modal-
ity was responsible for the difference in polarity of the semantic-relatedness
effect. In fact, these later two studies provide the primary empirical evidence
supporting the CSM. Bloem and La Heij’s study was a Stroop translation task
in which English words were shown individually to highly proficient Dutch-
English bilinguals. Each English stimulus (not target) word was accompanied
by either a semantically related Dutch word or picture. The participants were
told to ignore the context (distracter) word or picture and to only translate the
English word to Dutch. The effect, known as the semantic-relatedness effect, is
calculated by subtracting the mean reaction times (in milliseconds) for related
contexts from the mean reaction times for unrelated contexts. Bloem and La
Heij’s study showed that translation was quicker (765 ms) for unrelated word
contexts than for related word contexts (793 ms). The opposite was true for pic-
tures: Unrelated picture contexts led to slower translation (797 ms) than related
picture contexts (769 ms). Their results supported the notion that L2-to-L1
translation was impeded when accompanied by a semantically related context
word (−28 ms) but facilitated when accompanied by a semantically related
context picture (+28 ms).
Bloem and La Heij (2003) interpreted their results as evidence that context
pictures activated their conceptual representations but did not activate their
names. Under this assumption, the language of production is established at the
conceptual level, making lexical selection a fairly simple process at the lexical
level. In addition to localizing language selection at the conceptual level, the
findings from Bloem and La Heij were threefold: (a) Semantic facilitation is
found at the conceptual level; (b) semantic interference is localized at the lexical
level; and (c) only one concept is selected for lexicalization (the one specified
in the preverbal message).
Further support for the CSM has come from Bloem et al. (2004). These
researchers followed the same experimental procedure as Bloem and La Heij
(2003) but manipulated the SOA in order to observe whether previous effects
were due to the time that distracters were presented to the bilinguals. Indeed,
their study reported that context words that induce semantic interference at a
905 Language Learning 59:4, December 2009, pp. 897–927
Schwieter and Sunderman Concept Selection
SOA of +200 ms create semantic facilitation at a SOA of −400 ms. Bloem
et al. suggested that the decay of activation of lexical representations must be
stronger than the decay of conceptual representations.
The support for the claims put forth by the CSM suggesting that the locus
of language selection occurs at the conceptual level has been based on highly
proficient bilinguals. However, we reiterate that this assumption implies that
bilinguals are able to directly access the conceptual system without relying
on L1 lexical links. Kroll and Stewart’s (1994) RHM suggests that at lower
proficiency levels, L2 words are associated with their L1 translation equivalents
and conceptual mediation occurs via those lexical links. Thus, the RHM raises
questions of whether the CSM is able to account for lexical selection when
bilinguals are more dependent on lexical level links and may not have direct
access to the conceptual system (i.e., for language learners). Given that two
distinct experimental tasks were used by Kroll and Stewart and Bloem and La
Heij (2003) and Bloem et al. (2004) and given that the Stroop word translation
task has not been conducted with less proficient language learners, we argue
that this experimental task will provide a clear test of the predictions regarding
lexical and conceptual mediation. In a Stroop word translation experiment, it
seems plausible that less proficient language learners should be able to ignore
distracter pictures if, for them, lexical access is not a conceptually mediated
procedure but one of reliance on lexical links. During L2-to-L1 translation,
less proficient language learners, according to the RHM, access words during
bilingual speech production via the “route” that is lexically mediated, not
conceptually. Furthermore, if the claims of the RHM hold, we should expect
that more proficient language learners may have less reliance on lexical links
and thus may be more able to conceptually mediate.
Although Bloem and La Heij (2003) and Bloem et al.’s (2004) studies
provide strong support for a language selective view of L2 production, for
which the locus of language selection is at the conceptual level (i.e., the CSM),
it is not clear whether the predictions of the CSM can accurately account
for lexical selection in less proficient L2 learners. The present study tests the
differing predictions of this model and the RHM and explores the effects of
proficiency on the locus of language selection in bilingual speech production.
We expect that, in accordance with the RHM, less proficient learners will be
more reliant on lexical level links to access conceptual representations and the
magnitude of such reliance will be greater for the less proficient learners as
compared to the more proficient learners. More proficient learners may be able
to conceptually mediate and therefore will be able to select language at the
conceptual level, thus supporting the CSM and the RHM.
Language Learning 59:4, December 2009, pp. 897–927 906
Schwieter and Sunderman Concept Selection
Present Study
The experiment conducted in the current study was based on the work of Bloem
and La Heij (2003) and extensions were made to investigate whether proficiency
modulates the locus of language selection. In accordance with Bloem and La
Heij, a speeded word translation task was performed to investigate language se-
lectivity in language learners. In this Stroop translation task, learners translated
words presented in the L2 to their L1. Critical distracter items were displayed
in the form of L1 words or pictures and were either semantically related or
unrelated to the target word. For example, in the related context, the target
word casa “house” was presented with either the word barn or a picture of a
barn. In the unrelated context, the target word casa was presented with either
the word axe or a picture of an axe. Bloem and La Heij’s study showed that
semantic facilitation or interference depended on the type of distracter (picture
or word). When bilinguals translated in the presence of distracter words, this
induced interference (i.e., participants took longer to translate words in the
related condition as opposed to the unrelated condition.) However, facilitation
occurred in the presence of semantically related pictures (i.e., participants were
faster in the related condition as opposed to the unrelated condition). In other
words, context pictures were activating their conceptual representations but not
activating their names. This finding was explained by the notion that language
selectivity is a language-specific procedure in which a language cue plays an
intricate role in modulating activation levels at the lexical level.
Given the predictions put forth by the RHM that specify proficiency as a
critical factor when accessing the conceptual system, the present experiment
explores whether the CSM applies at various proficiency levels. Specifically,
does the claim of the CSM that language selection takes place at a conceptual
level hold for learners who are less proficient? If it does, then we would expect
to find the same semantic-relatedness effect for context pictures as in Bloem
and La Heij’s (2003) study among both the more and less proficient learners.
However, if the RHM’s prediction that less proficient language learners rely
on lexical links and are therefore unable to conceptually mediate holds, then
we would expect little or no effect of the picture condition and, instead, expect
to see greater interference in the word condition. In other words, because
the less proficient learner is lexically mediating and therefore automatically
sending increased activation to the lexical level, language selection will not be
able to take place at the conceptual level. An additional possibility, given the
nature of the bilinguals used in the current study (English dominant learners of
Spanish) compared to the highly proficient Dutch-English bilinguals in Bloem
907 Language Learning 59:4, December 2009, pp. 897–927
Schwieter and Sunderman Concept Selection
and La Heij’s study, is that the threshold of proficiency needed to demonstrate
conceptual mediation in this task may be quite high and out of the reach of our
language learners.
Method
Participants and Grouping
A total of 54 participants were recruited from lower and upper division Spanish
courses and from the faculty body at a large state university in the United States.
All participants were English dominant learners of Spanish. An independent
measure of proficiency was used to establish two experimental groups (less
and more proficient). This measure was a production task that estimated the
L2 proficiency level based on spontaneously and rapidly generating as many
words as possible related to a given semantic category in the less dominant
language.4A total of 10 categories (countries, clothing, animals, academic
majors, colors, fruits, vegetables, things with wheels, musical instruments, and
sports) were taken from the work of Gollan, Montoya, and Werner (2002) and
were individually verbalized to each participant. Upon hearing the category,
each participant was given 60 s to name as many items within that category as
he could in the L2. A total semantic fluency (proficiency) score was calculated
by adding all responses from each of the 10 semantic categories. Each word was
only counted once and, therefore, words that may have been repeated during
the proficiency measure were not included in the participant’s semantic fluency
(proficiency) score. Essentially, this score represented the total number of words
articulated in the L2 for all 10 semantic categories during a total of 10 min.
The median score (100) from the proficiency measure was used to split
the participants into two experimental groups: less proficient (N=29) and
more proficient (N=25). In addition, a language history questionnaire was
administered to provide the researchers with important information regarding
the participants’ language use and experiences in addition to their self-ratings in
English and Spanish over a variety of factors (reading, writing, speaking, etc.).
These self-ratings were based on a 10-point scale, with 1 representing “least
proficient” and 10 representing “most proficient.” Once the two experimental
groups were established based on the semantic fluency (proficiency) scores, an
ANOVA was conducted on the participants’ self-ratings. The results mirrored
what the proficiency measure had already established: The more proficient
group reported significantly higher ratings ( p<.05) than the less proficient
group for its L2 reading, writing, speaking, comprehension, and overall usage.
There were no significant differences between the less proficient and more
Language Learning 59:4, December 2009, pp. 897–927 908
Schwieter and Sunderman Concept Selection
proficient groups for their English language abilities. This was expected given
that all participants were dominant in English. Information from the language
history questionnaire and the proficiency measure can be seen in Table 1.
Materials
Thirty-two English words of high frequency were taken from the study by
Bloem and La Heij (2003). The translation equivalents of these words were
also of high frequency in Spanish. Overall, the English and Spanish words
were of similar mean language frequencies (1.92 and 1.80, respectively; Davies,
2008a, 2008b) and none of these words were cognates or had any orthographic
similarities to the target. For each of the words chosen, a related picture and
its lexical representation were chosen as distracter items (related contexts).5
Additionally, an unrelated picture and its lexical representation were paired
with each of the target words (unrelated contexts). See the Appendix for a list
of all materials.
The pictures used as distracter items were line drawings from Snodgrass
and Vanderwart’s (1980) standardized picture list and were identically matched
in size. The to-be-translated words in the L2 were always presented in the
exact middle of a computer screen in black, lowercase letters against a white
background. With respect to their screen position, the distracter items were
always placed directly to the right and above the stimulus word. In Bloem and
La Heij’s (2003) study, context words were presented below the exact center
of stimulus words and the context pictures were presented on the point of
fixation. The researchers identified this as a limitation in their study design
because it may have led to a possible larger visual masking effect induced by
pictures. This limitation was corrected in the current experiment, and for the
sake of consistency, both pictures and words were placed in the same place
with relation to the stimulus word.
Procedure
The present experimental procedure was run on a Macintosh computer using
PsyScope (Cohen, MacWhinney, Flatt, & Provost, 1993) to present stimuli and
to record reaction times (RTs). Before the experiment began, all participants
went through a series of practice trials in which each participant was tested
individually and given verbal instructions in addition to instructions on the
computer screen. All 32 target words were individually presented on a computer
screen in related or unrelated conditions in the L2. Participants were told to
quickly verbalize into a microphone the translation of each stimulus word into
their L1 while ignoring any other items that were presented on the screen. Two
909 Language Learning 59:4, December 2009, pp. 897–927
Schwieter and Sunderman Concept Selection
Tab l e 1 Participant characteristics
Averag e SD
Less More Less More
proficient proficient Difference proficient proficient Fp
Age mean 23.28 26.84 3.56 6.36 8.74 2.99 .09
Semantic fluency score 82.03 134.72 52.69 11.74 35.57 56.63 .00
English readinga9.41 9.52 0.11 0.83 0.82 0.22 .64
English writing 9.20 9.30 0.10 0.79 1.17 0.02 .89
English speaking 9.59 9.40 0.19 0.73 1.08 0.56 .46
English listening 9.69 9.32 0.37 0.54 1.18 2.29 .14
English overall 9.54 9.36 0.18 0.50 1.00 0.73 .40
Spanish reading 6.00 8.04 2.04 1.60 1.51 22.88 .00
Spanish writing 5.38 7.24 1.86 1.70 1.59 17.10 .00
Spanish speaking 5.79 8.16 2.37 2.00 1.55 23.00 .00
Spanish listening 6.69 8.48 1.79 2.09 1.19 14.30 .00
Spanish overall 5.77 8.04 2.27 1.70 1.26 30.18 .00
aThe 10 rows of English and Spanish scores represent the participants’ self-assessment of their two language abilities based on a 10-point
scale (1 =least proficient, 10 =most proficient).
Language Learning 59:4, December 2009, pp. 897–927 910
Schwieter and Sunderman Concept Selection
randomly ordered lists of 64 target words were presented to each participant. In
each list, every target word was presented two times: once in a related context
and once in an unrelated context.6At least 10 trials were placed between
the first and second presentation of any given stimulus word. Half of the
participants started with a trial accompanied with a picture distracter and half
started with a word distracter. The ordering of the related and unrelated contexts
was randomized. The experiment began with a fixation point represented by an
asterisk in the middle of the computer screen for 500 ms. Next, the stimulus
word and the context word or picture appeared simultaneously and remained
on the screen until the participant responded. If no response occurred after
2,000 ms, the next trial started. The latter was repeated until all target words
were presented twice (after the 64th trial). A fixation point was displayed again
in the center of the computer screen to let the participant know that he had
finished the first list. A 2-min break was given to the participant and the second
list began in the same sequence as the first list.
Results
Only correct responses were included in the RT analyses; correct and incorrect
responses were both included in the error analyses. Accuracy was coded as
either correct (if the participant cor rectly translated the target) or incorrect (if the
participant incorrectly translated the target). With regard to accuracy, the reader
will recall that participants translated stimulus words into their more dominant
language (English). Although there was little concern for mispronunciation of
the target, if a participant made any type of mistake (stutter, mispronunciation,
incorrect language, etc.), the item was recorded as incorrect. RTs faster than
300 ms and slower than 2,000 ms were removed from the analyses and treated
as outliers. Most of these were due to a technical malfunction in which the
microphone did not register the participant’s response. This amount represented
1.5% of the total data collected. In addition, reaction times that were 2.5 SDs
above or below the participant’s mean RT for each of the overall conditions
were excluded from the analyses and treated as outliers. These data equaled
3.8% of the total data collected. Mean scores across the four contexts were
computed for each participant. The descriptive statistics for reaction time,
accuracy, and the semantic-relatedness effect for the two proficiency groups
tested appear in Table 2. The reader will recall that the semantic relatedness
effect is calculated by subtracting the mean RTs for related contexts from those
of the unrelated contexts. This was done for both word and picture contexts.
911 Language Learning 59:4, December 2009, pp. 897–927
Schwieter and Sunderman Concept Selection
Tab l e 2 Mean reaction time (RTs), accuracy (Acc), and semantic relatedness effect (SRE)
Bloem & La Heij’s (2003)
Less proficient More proficient highly proficient bilinguals
Word Picture Word Picture Word Picture
RT Acc RT Acc RT Acc RT Acc RT Acc RT Acc
Unrelated 1,063 96.3 1,140 96.0 889 97.9 928 96.6 765 98.4 797 97.8
Related 1,090 97.3 1,139 96.2 906 97.3 905 97.3 793 98.0 769 98.2
SRE −27 +1−17 +23∗−28∗+28∗
∗Significant (p<.05).
Language Learning 59:4, December 2009, pp. 897–927 912
Schwieter and Sunderman Concept Selection
For ease of comparison, we also present the results from the study by Bloem
and La Heij (2003) as a point of reference.
In the following subsection we first present separate repeated measures
2×2 ANOVAs using participant means as random factors, with Context (word
or picture) and Relatedness (related or unrelated) as within-group factors in
order to draw direct comparisons with the study by Bloem and La Heij (2003).
We then present the omnibus repeated-measures 2 ×2×2 ANOVA using
participant means as random factors with Proficiency (less or more) as the
between-group factor and Context (word or picture) and Relatedness (related
or unrelated) as within-group factors. Measures of effect sizes (η2) are also
reported in the results.
Response Latencies
As a point of departure, we first describe the overall pattern of results. Table 2
shows the mean RTs and accuracy for the less and more proficient language
learners in the related and unrelated conditions for both pictures and words. As
Table 2 indicates, the pattern of lexical interference and semantic facilitation
appears to be different for the less and more proficient L2 learners. The less
proficient learners suffered more lexical interference (−27 ms) compared to
the more proficient learners (−17 ms). In terms of semantic facilitation, the
less proficient learners experienced very little semantic facilitation (+1ms)
in comparison to the more proficient learners (+23 ms). Moreover, the data
for more proficient learners appear to show a pattern of results similar to the
study by Bloem and La Heij (2003), with a reversal from semantic facilitation
to interference depending on the context: Picture distracters caused facilitation
and word distracters induced interference. The data for the less proficient
learners does not show the same reversal effect. In the following we report the
statistical test of the patterns reported earlier.
In order to directly compare the results of the study by Bloem and La
Heij (2003) to our results, we will analyze the data from the more and less
proficient learners separately. As discussed in the introduction, Bloem and La
Heij found an interaction between context and relatedness. Additional tests
revealed that the semantic interference effect for words was significant as well
as the semantic facilitation effect for pictures. In other words, in Table 2,
the semantic relatedness effects (SREs) for both words and pictures for the
study by Bloem and La Heij represent significant differences and illustrate the
reversal from interference to facilitation depending on modality. The pattern
of results for our more proficient learners looks very similar to Bloem and
La Heij’s results. Indeed, we find a significant interaction between context and
913 Language Learning 59:4, December 2009, pp. 897–927
Schwieter and Sunderman Concept Selection
800
900
1,000
1,100
1,200
Lower Higher
L2 proficiency
RT in ms
Words
Pictures
Figure 4 The effects of context and proficiency.
relatedness, F(1, 24) =9.37, p<.01, η2=.281. However, paired-samples
ttests were performed on the subject means and revealed that the SRE was
only significant for the picture condition, t(24) =2.36, p<.05. There was
no significant difference between the related and unrelated trials in the word
condition despite the 17-ms difference between the two. The patterns of results
for our less proficient learners did not parallel Bloem and La Heij’s results.
There was no interaction between context and relatedness ( p>.05). There
was a main effect of context for the less proficient learners, F(1, 28) =14.03,
p<.001, η2=.334, indicating that translation was faster in the presence of
words as opposed to pictures.
As we turn to the omnibus ANOVA that now includes the variable of
proficiency, we find that there was a significant interaction between context
and proficiency, F(1, 53) =4.47, p<.05, η2=.079 (see Figure 4). To
further investigate this interaction, paired-samples ttests were performed on the
subject means and revealed that the less proficient language learners performed
differently in the word and picture conditions. The less proficient learners were
significantly faster translating in the word condition (1,076 ms) compared to
the picture condition (1,139 ms), t(28) =−3.75, p<.001. Again, this parallels
the above finding that the less proficient learners were highly sensitive to lexical
information that resulted in facilitated translation.
There was also a significant interaction between context and relatedness,
F(1, 53) =7.84, p<.01, η2=.131 (see Figure 5). Paired-samples ttests were
performed to further explore this interaction and revealed that related words
were processed significantly slower (1,004 ms) compared to unrelated words
(982 ms), t(53) =−0.208, p<.05. The three-way interaction of relatedness,
Language Learning 59:4, December 2009, pp. 897–927 914
Schwieter and Sunderman Concept Selection
900
950
1,000
1,050
Words Pictures
Context Modality
RT in ms
Unrelated
Related
Figure 5 The effects of context and relatedness.
context, and proficiency failed to reach significance, F(1, 53) =0.22, p=.64,
η2=.004. Finally, the more proficient learners were able to translate from
the L2 to the L1 much faster than the less proficient learners overall. Indeed,
regardless of context or relatedness, the mean RTs for the less proficient lear ners
was 1,108 ms and it was 907 ms for more proficient learners, leading to a main
effect of proficiency, F(1, 53) =11.79, p<001, η2=.184.
Error Analyses
No significant main effects or interactions were reported in the error analyses.
As in the study by Bloem and La Heij (2003), the error rates were considered
too low to allow for a meaningful analysis.
Discussion
In this study, a Stroop translation task was conducted in which less and more
proficient language learners translated words from their L2 to their L1 while
faced with semantically related and unrelated distracter words and pictures.
Bloem and La Heij (2003) argued that for highly proficient bilinguals, con-
text pictures activate their conceptual representations but do not activate their
names. Therefore, these researchers suggest that when bilinguals are faced
with distracters in the form of pictures, competition can be avoided because
the target concept has already been established and, thus, is the only one that
is considered during lexicalization. In fact, language selection will be facili-
tated by the presence of a picture. The present study investigated whether more
and less proficient language learners would be able to select the language of
production at the conceptual level.
915 Language Learning 59:4, December 2009, pp. 897–927
Schwieter and Sunderman Concept Selection
In the first set of analyses given earlier, we compared the less proficient and
more proficient learners to the bilinguals in the study by Bloem and La Heij
(2003). We found that the more proficient learners in our study are both distinct
and similar to Bloem and La Heij’s Dutch-English bilinguals. Bloem and La
Heij’s study showed that L2-to-L1 translation was impeded when accompanied
by a semantically related context word (−28 ms) but facilitated when accom-
panied by a semantically related context picture (+28 ms). Our study reveals
that L2-to-L1 translation was not significantly impeded when accompanied by
a semantically related context word (despite the −17 ms) but it was facilitated
when accompanied by a semantically related context picture (+23 ms). Thus,
in the spirit of Bloem and La Heij’s claim that context pictures activate their
conceptual representations but do not activate their names, our results, if reli-
able, suggest that in our more proficient L2 learners, who are nonetheless less
proficient than those of Bloem and La Heij, conceptual representations, but
perhaps also names, were activated. Thus, the results of our analyses suggest
that more proficient learners increasingly use concept mediation but perhaps
not exclusively (i.e., the SRE of −17 ms for more proficient learners suggests
that they are perhaps relying on word association despite their higher profi-
ciency). These results suggest that the gradual shift from word association to
concept mediation is affected by access routes for each lexical item and may
differ depending on the configuration and automatization of the learner’s net-
work.7In the end, the finding of facilitation in the picture condition for the
more proficient learners (and the lack thereof for the less proficient learners)
provides support for the CSM in the sense that more proficient language learn-
ers were able to select language at the conceptual level. The translation of the
less proficient learners was not facilitated by the presence of pictures and, thus,
did not show evidence of selecting language at the conceptual level. In other
words, the data for the less proficient learners do not appear to support the
CSM.
When we turn to the second set of analyses that examine whether the less
and more proficient learners are different from each other with respect to the
word and picture condition, we find that, overall, the less proficient learners are
significantly slower to translate compared to the more proficient, regardless of
context or relatedness. Recall that the direction of translation within the task
was L2 to L1. Given that these learners are native speakers of English, it is
improbable that difficulties in articulation are at the root of the production delay.
The more likely interpretation is that the L2-to-L1 access routes are slower and
not yet automatic, as the links between L2 and L1 words are weaker compared
to the connections in the more proficient learners.
Language Learning 59:4, December 2009, pp. 897–927 916
Schwieter and Sunderman Concept Selection
We also find that less proficient language learners translated faster in the
context of distracter words than in the context of distracter pictures, regardless
of relatedness. One interpretation of this result is that the L1 context words help
prime the target language regardless of the relatedness of the context word. It
is possible that such a language prime helps to limit the activation flow to
the L1 sublexicon (see also Potter et al., 1984, for a similar argument). These
results support the prediction of the RHM that less proficient learners lexically
mediate. Thus, the L2-to-L1 links, as predicted by the RHM, are highly salient
for the less proficient learners.
When we investigate whether the less and more proficient language learn-
ers are different from each other with respect to the SRE, we ultimately fail to
find a three-way interaction among relatedness, context, and proficiency. There
are two potential reasons for this lack of interaction. First, there was an inter-
action between context and relatedness that pertained to the word condition,
not the picture condition. Both groups of learners suffered more interference
with related words, thus demonstrating the strong lexical links for both of these
groups. There was also an interaction between context and proficiency indicat-
ing that the less proficient learners were indeed more affected overall by words
as compared to pictures. Thus, the word condition, particularly for the less
proficient, is driving the interaction effects. The SRE for pictures for the more
proficient learners is simply not strong enough within the context of the omnibus
ANOVA. Second, given the nature of a 2 ×2×2 ANOVA with 54 partici-
pants, it is possible that there was not sufficient statistical power to reveal the
differential SRE across the two proficiency levels.
However, despite the lack of three-way interaction, our study has provided
evidence that the role of proficiency is important in language selection during
bilingual lexical access. Theoretically, we can see this effect if we examine these
three groups along a continuum: less proficient learners from our study, more
proficient learners from our study, and Bloem and La Heij’s (2003) bilinguals.
Although one might have anticipated the proficiency threshold to occur between
the less and more proficient learners in our study, we instead found that the
results from our more proficient learners in our study were similar to the results
from the Dutch-English bilinguals, but only with respect to language selection
at the conceptual level. The more proficient learners still exhibited a reliance on
lexical level information that was not present in the Dutch-English bilinguals.
Thus, the results suggest that the claims of the CSM do not hold for less
proficient learners, but they do hold for the more advanced learners of the
language. Our findings then also provide support for the RHM for both the less
and more proficient learners. The data suggest that L1 links (word association)
917 Language Learning 59:4, December 2009, pp. 897–927
Schwieter and Sunderman Concept Selection
are strongest when an individual is less proficient in the language and that only
with increased proficiency do learners begin to conceptually mediate.
How is it that less proficient learners resolve the “hard problem” of lan-
guage selection if they are not able to reconcile it through conceptual selection
mechanisms such as a language cue? The reader will recall that the “hard
problem” occurs when similar levels of activation are sent from the conceptual
level to the lexical level, prompting a decision to select the correct word in
the appropriate language that maps onto the target concept. What other cogni-
tive mechanisms can be called upon to specify the language of production? It
would be premature to claim that picture contexts simply are easier to ignore
than word contexts and that these effects are stronger among more proficient
language learners. There is more than likely a cognitive procedure that is ac-
tive during lexicalization. Previous research has investigated the critical role of
proficiency in lexical selection within Green’s (1986, 1998) inhibitory control
model (ICM) and within the context of shifting from inhibitory control mech-
anisms to language selective mechanisms (Costa, 2005; Costa & Santesteban,
2004; Costa, Santesteban, et al., 2006; Schwieter, 2007; Schwieter &
Sunderman, 2008). Essentially, the ICM is a model that explains how bilingual
speech production is lexically mediated. Its claims suggest that the nontarget
words are suppressed to a degree that is proportionate to the level of activation
based on language tags (i.e., the more activation that is sent to the nontarget lan-
guage, the more inhibition will be needed). Because of the asymmetrical system
sizes of the L1 and L2, naturally more inhibition is applied to the L1 when the
L2 is the language of production. The opposite is true for the smaller L2 system:
Less inhibition is called upon when the L1 is the language of production. Based
on these theories and their empirical support, we therefore have an alternative
explanation to concept selection for less proficient language learners, namely
inhibitory control. Taking into account previous studies showing strong support
for conceptual mediation among highly proficient bilinguals, the present study
has highlighted the importance of reliance on L1 lexical links and, in particular,
to what extent this reliance is modulated by L2 proficiency. We present next a
model that captures how target language selection is achieved in less and more
proficient L2 learners and bilinguals while entertaining the idea that language
selection and lexical processing may be lexically and conceptually mediated as
specified by L2 proficiency.
In Figure 6, we present the selection by proficiency (SbP) model (Schwieter,
2007; Schwieter & Sunderman, 2008). This is a bilingual speech production
model that incorporates both selective and nonselective elements as well as
conceptual elements, all of which are modulated by proficiency. This model is
Language Learning 59:4, December 2009, pp. 897–927 918
Schwieter and Sunderman Concept Selection
Figure 6 The selection by proficiency model. Adapted from “Language Switching in
Bilingual Speech Production: In Search of the Language-Specific Selection Mechanism”
by J. W. Schwieter and G. Sunderman, 2008, The Mental Lexicon, 3(2), pp. 214–238.
Copyright John Benjamins, 2008.
essentially a visual representation of a bilingual speaker attempting to name
a picture in his L2. Although the present study reported here was a L2-to-
L1 translation task, the production model presented here captures the role of
proficiency as well as inhibitory control in the process of lexical and conceptual
mediation. In this model, L2 learners move along a proficiency continuum in
a bidirectional manner to account for learning and language attrition. Thus, at
the top of the model, two representations of the bilingual mental lexicon and
their relationship to the conceptual system are illustrated: The one on the left
919 Language Learning 59:4, December 2009, pp. 897–927
Schwieter and Sunderman Concept Selection
represents a less proficient language learner and the one on the right shows a
more proficient L2 learner.
Before we explain both sides of the model separately, it is important to
note a few general aspects of the SbP model. First, the differential strengths
of association between the conceptual links and lexical links are illustrated
by the thickness of the arrows in the model. For instance, the darker arrows
represent stronger links and the dashed arrows represent either weaker links
or, as in the right portion of the model, those that may be present but not nec-
essarily called upon. Second, this model demonstrates that for less proficient
L2 learners, the mechanisms that help select the target language occur at the
lexical level and consist of inhibitory control. This is different in the case of
the highly proficient bilinguals (such as the bilinguals in Bloem & La Heij’s,
2003, study). In the latter, reliance will be on language selective mechanisms in
the form of the language cue and other higher linguistic information apparent
during preverbalization. However, in certain situations or instances, the highly
proficient bilingual may need to call upon mechanisms of inhibitory control
again (as shown by the longest diagonal arrow). The shift in reliance to lan-
guage selective mechanisms, however, cannot be achieved for less proficient
L2 learners until they become much more proficient.
We now turn to how the SbP model moves through the lexicalization and
language selection procedure during bilingual speech production as modulated
by L2 proficiency. As can be seen on the left side of the model, for less proficient
L2 learners there is a stronger association between concepts and the L1 words
mapped onto them. The links between L2 words and their concepts are much
weaker. As in the RHM, less proficient learners are more reliant on lexical
links than conceptual links when producing in their L2. Therefore, when a less
proficient English language learner of Spanish attempts to name a picture of
a cat in Spanish, the learner must have to do so by associating the concept to
its L1 lexical item first. Under this assumption, the less proficient language
learner also must rely on mechanisms of inhibitory control at the lexical level
to select the language of production. These language task schemata then inhibit
the irrelevant language’s competing words so that the accurate production of
gato can be spoken.
The right part of the model represents speech production for highly profi-
cient bilinguals. Unlike less proficient learners, there is a stronger association
between concepts and words inboth the L1 and L2. In essence, highly proficient
bilinguals are able to conceptually mediate in both languages, calling upon dif-
ferent types of control mechanisms than were required at lower proficiency
Language Learning 59:4, December 2009, pp. 897–927 920
Schwieter and Sunderman Concept Selection
levels. For example, when a highly proficient bilingual attempts to name a
picture of a cat in L2 Spanish, the learner may now directly access the word
mapped onto the concept instead of having to rely upon the L1 translation. Be-
cause competition between languages is no longer confined to the lexical level,
highly proficient bilinguals are able to rely on a language cue at the conceptual
level to select the language of production.
By putting into perspective the less and more proficient L2 learners from
our study and highly proficient bilinguals from previous studies, the SbP model
allows for us to explain their processing differences graphically and, more
importantly, as a developmental procedure. For instance, we did not find a
reversal of the semantic relatedness effect for the less proficient learners and we
used this as evidence suggesting that they were highly reliant upon lexical links
and more than likely utilizing mechanisms of inhibitory control. As shown in the
SbP model, less proficient language learners are not able to rely on the language
cue and, in turn, rely on inhibitory control at the lexical level. Contrarily, because
there was facilitation in the picture condition for the more proficient learners
in the present study and highly proficient bilinguals from previous studies,
the right-hand side of the SbP model is more language selective in nature.
However, the more proficient learners in our study—despite the fact that at first
glance they appeared to be conceptually mediating—provided evidence that,
like less proficient learners, they were also relying on lexical links. Thus, the
possibility of lexical mediation is maintained in the SbP model by specifying
that when a more proficient learner is not able to conceptually mediate, the
learner must revert to reliance on a lexical-level procedure such as inhibitory
control.
In addition to making theoretical extensions to the RHM and the CSM
with a proficiency-dependent component, the SbP model combines issues of
conceptual selection, reliance on L1 lexical links, and inhibitory control in a
comprehensive developmental model of lexical access and bilingual speech
production. Moreover, the importance of this model lies in its separation of
the two issues of conceptual retrieval and inhibitory control that, based on the
data presented here, are two salient, yet unrelated factors in L2 production.
Although previous studies have also suggested that the underlying mechanisms
involved in language selection shift from inhibitory control mechanisms to lan-
guage selective mechanisms (Costa, 2005; Costa & Santesteban, 2004; Costa,
Santesteban, et al., 2006; Schwieter, 2007), they have failed to take into account
the importance of reliance on lexical links. The SbP model, thus, incorporates
the predictions of the RHM as a means to explain such a reported shift.
921 Language Learning 59:4, December 2009, pp. 897–927
Schwieter and Sunderman Concept Selection
Conclusion
In this study, we tested the predictions of two important models of bilingual
speech production: the CSM and the RHM in a Stroop translation task. The re-
sults suggest that although the more proficient L2 learners performed similarly
to the highly proficient bilinguals from previous studies, the CSM’s solution to
the “hard problem” does not hold for less proficient learners. It appears that less
proficient learners were unable to rely upon the language cue to establish the
language of production at the conceptual level. Although they eventually are
able to activate the appropriate lexicon, as demonstrated by their accuracy in
translation (i.e., target productions were just as accurate as the more proficient
learners’), they do so via a different route. The pattern of data reported here
for the less proficient learners supports the claim of the RHM that learners rely
on lexical links to access concepts. Additionally, although all learners showed
reliance on lexical links to some degree, the more proficient learners appeared
to solve the “hard problem” in accordance with the predictions of the CSM. We
also discussed our results vis-`
a-vis the bilingual speech production model, the
SbP model, which captures the selective nature of lexical processing in highly
proficient bilinguals and the nonselective nature in less proficient L2 learners.
Future research must continue to address the variability of the mechanisms
involved in concept selection and lexical processing as modulated by internal
factors such as proficiency and external factors such as linguistic environment,
interlocutor, and so forth. It will be critical for future research to continue
examining L2 learners and how the ability to produce language changes along
a developmental continuum. Furthermore, additional research is merited that
seeks to explore the issue of whether more proficient language learners revert
back to or rely upon mechanisms of inhibitory control. If this is the case, what
factors (i.e., social contexts, experimental task demands, etc.) contribute to
such regression or reliance?
Revised version accepted 16 October 2008
Notes
1 There is some debate in the field with respect to conceptual and semantic level
representations. See Francis (1999) and Pavlenko (2000) for a discussion of this
issue.
2 See Green’s (1986, 1998) inhibitory control model for an account of how
between-language competition for proficient bilinguals is resolved if the “hard
problem” continues to the lexical level.
Language Learning 59:4, December 2009, pp. 897–927 922
Schwieter and Sunderman Concept Selection
3 Stimulus onset asynchrony (SOA) refers to the time interval between the beginning
of one stimulus and the beginning of another. For example, an SOA of −400 ms
means that a distracter item/prime appears 400 ms before the presentation of the
target stimulus and +140 ms means that it appears 140 ms after the presentation of
the target stimulus.
4 An anonymous reviewer pointed out that the measure of L2 proficiency used in the
present study is perhaps confounded with a speaker’s fluency of articulation
regardless of language. We agree that there are individual differences in a speaker’s
fluency of articulation. It may be beneficial for future research using this task as a
measure of proficiency to also collect L1 measures and partial out any effects of
individual differences.
5 We would like to thank an anonymous reviewer for pointing out that Kessler,
Treiman, and Mullennix (2002) have shown that voiceless, posterior, and obstruent
consonants were detected later than others in word recognition tasks. Four of the
target words in the present study began with these sounds in question. Furthermore,
another anonymous reviewer pointed out that 2 of the 64 distracter-target pairs
(pencil-pen, strawberry-cherry) were phonologically related. Although not
controlled for in this study, interested scholars can read the work by Tyler, Tyler, and
Burnham (2005) for ways to avoid this limitation.
6 Participants translated each target word four times during the entire experimental
procedure. Although we attempted to control for a possible priming effect by
placing 10 trials before repeating any given trial, it is possible that there was a
priming effect. We argue that such facilitation would be independent of the relation
between the distracter and the to-be-named target, which is the variable of critical
interest. However, this remains an empirical question for future research.
7 We would like to thank an anonymous reviewer for this observation. This idea is
also represented in our SbP model in Figure 6.
References
Altarriba, J., & Mathis, K. M. (1997). Conceptual and lexical development in second
language acquisition. Journal of Memory and Language, 36, 550–568.
Bloem, I., & La Heij, W. (2003). Semantic facilitation and semantic interference in
word translation: Implications for models of lexical access in language production.
Journal of Memory and Language, 48, 468–488.
Bloem, I., Van Den Boogaard, S., & La Heij, W. (2004). Semantic facilitation and
semantic interference in language production: Further evidence for the conceptual
selection model of lexical access. Journal of Memory and Language, 51, 307–323.
Cohen, J., MacWhinney, B., Flatt, M., & Provost, J. (1993). PsyScope: An interactive
graphic system for designing and controlling experiments in the psychology
laboratory using Macintosh computers. Behavior Research Methods, Instruments
and Computers, 25, 257–271.
923 Language Learning 59:4, December 2009, pp. 897–927
Schwieter and Sunderman Concept Selection
Costa, A. (2005). Lexical access in bilingual production. In J. F. Kroll & A. M. B. de
Groot (Eds.), Handbook of bilingualism: Psycholinguistic approaches
(pp. 308–325). New York: Oxford University Press.
Costa, A., Miozzo, M., & Caramazza, A. (1999). Lexical selection in bilinguals: Do
words in bilingual’s lexicon compete for selection? Journal of Memory and
Language, 41, 365–397.
Costa, A., & Santesteban, M. (2004). Lexical access in bilingual speech production:
Evidence from language switching in highly proficient bilinguals and L2 learners.
Journal of Memory and Language, 50, 491–511.
Costa, A., Santesteban, M., & Ivanova, I. (2006). How do highly proficient bilinguals
control their lexicalization process? Inhibitory and language-specific selection
mechanisms are both functional. Journal of Experimental Psychology: Learning,
Memory, and Cognition, 32(5), 1057–1074.
Davies, M. (2008a). Corpus of American English: 360 million words. Available online
at http://www.americancorpus.org
Davies, M. (2008b). Corpus del Espa˜
nol: 100 million words. Available online at
http://www.corpusdelespanol.org
de Groot, A. M. B., Dannenburg, L., & Van Hell, J. G. (1994). Forward and backward
word translation by bilinguals. Journal of Memory and Language, 33, 600–629.
de Groot, A. M. B., & Poot, R. (1997). Word translation at three levels of proficiency in
a second language: The ubiquitous involvement of conceptual memory. Language
Learning, 47, 215–264.
Dufour, R., & Kroll, J. F. (1995). Matching words to concepts in two languages: A test
of the concept mediation model of bilingual representation. Memory and Cognition,
23, 166–180.
Fer r ´
e, P., S´
anchez-Casas, R., & Guasch, M. (2006). Can a horse be a donkey? Semantic
and form interference effects in translation recognition in early and late proficient
and nonproficient Spanish-Catalan bilinguals. Language Learning, 56, 571–608.
Finkbeiner, M., Almeida, J., Janssen, N., & Caramazza, A. (2006). Lexical selection in
bilingual speech production does not involve language suppression. Journal of
Experimental Psychology: Learning, Memory and Cognition, 32(5), 1075–1089.
Finkbeiner, M., Gollan, T., & Caramazza, A. (2006). Lexical access in bilingual
speakers: What’s the (hard) problem? Bilingualism: Language and Cognition, 9(2),
153–166.
Francis, W. (1999). Cognitive integration of language and memory in bilinguals:
Semantic representation. Psychological Bulletin, 125(2), 193–222.
Frenck-Mestre, C., & Prince, P. (1997). Second language autonomy. Journal of
Memory and Language, 37, 481–501.
Gollan, T. H., Montoya, R. I., & Werner, G. A. (2002). Semantic and letter fluency in
Spanish–English bilinguals. Neuropsychology, 16(4), 562–576.
Green, D. (1986). Control, activation, and resource: A framework and a model for the
control of speech in bilinguals. Brain and Language, 27, 210–223.
Language Learning 59:4, December 2009, pp. 897–927 924
Schwieter and Sunderman Concept Selection
Green, D. (1998). Mental control of the bilingual lexico-semantic system.
Bilingualism: Language and Cognition, 1, 67–81.
Hall, C. J., & Ecke, P. (2003). Parasitism as a default mechanism in L3 vocabulary
acquisition. In J. Cenoz, B. Hufeisein, & U. Jessner (Eds.), The multilingual lexicon
(pp. 71–85). Dordrecht: Kluwer Academic.
Heredia, R. (1997). Bilingual memory and hierarchical models: A case for language
dominance. Current Directions in Psychological Science, 6, 34–39.
Heredia, R., & Altarriba, J. (2001). Bilingual language mixing: Why do bilinguals
code-switch? Current Directions in Psychological Science, 10, 164–168.
Hermans, D. (2000). Word production in a foreign language. Unpublished master’s
thesis, University of Nijmegen, Nijmegen.
Hermans, D., Bongaerts, T., de Bot, K., & Schreuder, R. (1998). Producing words in a
foreign language: Can speakers prevent interference from their first language?
Bilingualism: Language and Cognition, 1, 213–229.
Jescheniak, J. D., & Schriefers, K. I. (1998). Discrete serial versus cascading
processing in lexical access in speech production: Further evidence from the
coactivation of near-synonyms. Journal of Experimental Psychology: Learning,
Memory, and Cognition, 24, 1256–1274.
Jiang, N. (2000). Lexical representation and development in a second language.
Applied Linguistics, 21, 47–77.
Kessler, B., Treiman, R., & Mullennix, J. (2002). Phonetic biases in voice key response
time measurements. Journal of Memory and Language, 47, 145–171.
Kroll, J., Bobb, S., & Wodniecka, Z. (2006). Language selectivity is the exception, not
the rule: Arguments against a fixed locus of language selection in bilingual speech.
Bilingualism: Language and Cognition, 9(2), 119–135.
Kroll, J., & Stewart, E. (1994). Category interference in translation and picture
naming: Evidence for asymmetric connections between bilingual memory
representations. Journal of Memory and Language, 33, 149–174.
Kroll, J., & Tokowicz, N. (2001). The development of conceptual representation for
words in a second language. In J. L. Nicol (Ed.), One Mind, two languages:
Bilingual language processing (pp. 49–71). Cambridge, MA: Blackwell.
La Heij, W. (2005). Selection processes in monolingual and bilingual lexical access. In
J. F. Kroll & A. M. B. de Groot (Eds.), Handbook of bilingualism: Psycholinguistic
approaches (pp. 289–307). New York: Oxford University Press.
La Heij, W., de Bruyn, E., Elens, E., Hartsuiker, R., Helaha, D., & van Schelven, L.
(1990). Orthographic facilitation and categorical interference in a word-translation
variant of the Stroop task. Canadian Journal of Psychology, 44(1), 76–83.
La Heij, W., Hooglander, A., Kerling, R., & Van der Velden, E. (1996). Nonverbal
context effects in forward and backward translation: Evidence for concept
mediation. Journal of Memory and Language, 35, 648–665.
Levelt, W. J. M. (1989). Speaking: From intention to articulation. Cambridge, MA:
MIT Press.
925 Language Learning 59:4, December 2009, pp. 897–927
Schwieter and Sunderman Concept Selection
Levelt, W. J. M., Roelofs, A., & Meyer, A. S. (1999). A theory of lexical access in
speech production. Behavioral and Brain Sciences, 22, 1–75.
Neumann, O. (1986). Facilitative and inhibitory effects of “semantic relatedness”
(Report No. 111/1986). University of Bielefeld, Bielefeld, Germany.
Pavlenko, A. (2000). New approaches to concepts in bilingual memory. Bilingualism:
Language and Cognition, 3(1), 1–4.
Peterson, R. R., & Savoy, P. (1998). Lexical selection and phonological encoding
during language production: Evidence for cascaded processing. Journal of
Experimental Psychology: Learning, Memory, and Cognition, 24, 539–557.
Potter, M. C., So, K. F., Von Eckardt, B., & Feldman, L. B. (1984). Lexical and
conceptual representation in beginning and more proficient bilinguals. Journal of
Verbal Learning and Verbal Behavior, 23, 23–38.
Poulisse, N., & Bongaerts, T. (1994). First language use in second language
production. Applied Linguistics, 15, 36–57.
Schwieter, J. (2007). A psycholinguistic investigation of language selectivity in
bilingual speech production. Dissertation Abstracts International, 68(9). (UMI No.
3282661)
Schwieter, J. W. (2008). The effects of bilingual type on language selectivity. In M.
Montero, P. Chamness Miller, & J. Watze (Eds.), Readings in language studies:
Vol. 1. Language across disciplinary boundaries. New York: The International
Society for Language Studies Inc.
Schwieter, J. W., & Sunderman, G. (2008). Language switching in bilingual speech
production: In search of the language-specific selection mechanism. The Mental
Lexicon, 3(2), 214–238.
Sholl, A., Sankaranarayanan, A., & Kroll, J. F. (1995). Transfer between picture
naming and translation: A test of asymmetries in bilingual memory. Psychological
Science, 6, 45–49.
Snodgrass, J. G., & Vanderwart, M. (1980). A standardized set of 260 pictures: Norms
for name agreement, image agreement, familiarity, and visual complexity. Journal
of Experimental Psychology: Human Learning and Memory, 6, 174–215.
Swain, M. (1997). The influence of the mother tongue on second language vocabulary
acquisition and use. In N. Schmitt & M. McCarthy (Eds.), Vocabulary: Description,
acquisition and pedagogy (pp. 156–180). Cambridge: Cambridge University Press.
Talamas, A., Kroll, J., & Dufour, R. (1999). Form related errors in second language
learning: A preliminary stage in the acquisition of L2 vocabulary. Bilingualism:
Language and Cognition, 2, 45–58.
Tyler, M. D., Tyler, L., & Burnham, D. K. (2005). The delayed trigger voice key: An
improved analogue voice key for psycholinguistic research. Behavior Research
Methods, 37(1), 139–147.
Weber-Fox, C. M., & Neville, H. J. (1996). Maturational constraints on functional
specializations for language processing: ERP and Behavioral evidence in bilingual
speakers. Journal of Cognitive Neuroscience, 8(3), 231–256.
Language Learning 59:4, December 2009, pp. 897–927 926
Schwieter and Sunderman Concept Selection
Zeelenberg, R., & Pecher, D. (2003). Evidence for long-term language repetition
priming in conceptual implicit memory tasks. Journal of Memory and Language,
49, 80–94.
Appendix
Target Words and Distracter Items by Condition Type
Related context Unrelated context
Translation (picture or word (picture or word
Stimulus (produced) distracter) distracter)
´
arbol tree leaf chair
avi ´
on airplane train deer
brazo arm nose eagle
burro donkey camel eye
caballo horse deer foot
cama bed dresser fork
casa house barn axe
cebolla onion asparagus glove
cintur´
on belt paints hand
cuchara spoon fork leaf
cuchillo knife axe lemon
dedo finger glove camel
estrella star moon lettuce
fresa strawberry cherry moon
gato cat mouse orange
labios lips hand asparagus
l´
apiz pencil pen nose
manzana apple orange cloud
mesa table chair paints
naranja orange peach barn
oreja ear eye peach
paraguas umbrella cloud mouse
pato duck eagle cherry
perro dog bear pear
pierna leg foot pen
puerta door window ruler
sand´
ıa watermelon pear train
tijeras scissors ruler shirt
uvas grapes lemon dresser
vaca cow zebra window
vestido dress shirt bear
zanahoria carrot lettuce zebra
927 Language Learning 59:4, December 2009, pp. 897–927
