Abstract

We investigate how L1 phonology and semantics affect processing of interlingual homographs by manipulating language context before, and auditory input during, a visual experiment in the L2. Three experiments contained German–English homograph primes (gift = German “poison”) in English sentences and was performed by German (L1) learners of English (L2). Both reaction times and event-related brain potentials were measured on targets reflecting the German meaning of the interlingual homograph. In Experiment 1, participants viewed a pre-experiment English film, then half of the participants (n = 16) heard noise and the other half (n = 16) heard German pseudowords during the experiment; in Experiment 2, participants (n = 16) viewed a pre-experiment German film then heard noise; and in Experiment 3, participants (n = 16) viewed the pre-experiment English film then heard real German words. Those who had viewed the English film then heard noise during Experiment 1 showed no L1 influence. Those who saw the English film but heard German pseudowords during Experiment 1, or viewed the German film before and heard noise during Experiment 2, showed L1 influence as indicated by N400 priming of L1-related targets in the first half of the experiment. This suggests that a pre-experiment film in the L1 or the presence of L1 phonology during the experiment slowed down adjustment to the L2 task. In Experiment 3 with real L1 words in the background, N400 priming of L1 meanings was observed throughout the entire experiment for lower-proficiency participants. We discuss our findings in terms of context types that affect L1-to-L2 adjustment.

INTRODUCTION

At our institute, almost everyone speaks more than one language, and people switch from the L1 (usually German) to the L2 (English) all the time. Intuitively, most people feel that the surrounding context can make a switch easy or difficult. For example, if we view a presentation in German, then have a discussion in English afterwards, zooming in to or adjusting to the L2 is fairly hard, at least at first. But if a lecture was just viewed in the L2 English, the previous language context—the semantic, phonological setting—helps in adjusting to a discussion afterward in the L2. Evidence for this intuitive idea that preceding context affects adjustment from the L1 to the L2 was recently reported with reaction times (RTs) and event-related brain potentials (ERPs) in Elston-Güttler, Gunter, and Kotz (2005). But what about the actual background that we hear when adjusting from the L1 to the L2? Let's assume that after the English lecture, we are discussing something in English (which should be fairly easy due to the previous context), but across the room, we can hear a loud discussion in German—we can't make out what is being said, but we can hear German-specific phonology. Does this make our adjustment to English harder? Even more extreme, assume that we can make out what is being said in the loud German discussion: Would this semantic and phonological distraction prevent us from fully adjusting to English? Is there neuropsychological evidence that supports these intuitions that bilinguals have from their everyday experience?

In the present study, we investigate the role of previous language context, but along with the additional layer of background phonology, in zooming in from the L1 to the L2. Several behavioral and neurocognitive studies suggest that a person who can speak two languages utilizes mental lexicons containing L1 and L2 words that are highly interconnected (e.g., Kotz & Elston-Güttler, 2004; Williams, 1994) so we are concerned with so-called cognitive control (see Dijkstra & Van Heuven, 2002 for a review), or how well or poorly a bilingual can activate only the appropriate language at the appropriate time to the appropriate degree. In Elston-Güttler, Gunter, et al. (2005), we proposed the new term zooming in because it is unique in two respects. First, as we are concerned with changes from one entire language in a monolingual context to another and staying there, we are not referring to so-called language switching or translating, which is defined by constantly changing back and forth between less contextually rich events such as sentences or trials (for recent neuroimaging studies in this area, see Proverbio, Leoni, & Zani, 2004; Hernandez, Dapretto, Mazziotta, & Bookheimer, 2001; Price, Green, & Von Studnitz, 1999). Second, the zooming-in process, by definition, has to change over time, or adjust, as more and more stimuli from the new language create a new language mode setting. This idea is based on the theory of language mode as defined by Grosjean (for a recent discussion, see Grosjean, 2001), where a bilingual finds him or herself in a monolingual, bilingual, or in-between setting based on a particular processing situation. In their Bilingual Interactive Activation+ (from now on referred to as BIA+) model of bilingual word recognition, Dijkstra and Van Heuven (2002) state explicitly that task demands and processing situations can change over time, and that the system needs to recalibrate accordingly.

Nonselective Lexical Access Affected by Different Context Types

Cognitive control in bilingual language recognition has been investigated extensively using interlingual homographs, or words that are highly similar or identical in form but differ in meaning across languages (e.g., chef refers to “cook” in English and “boss” in German). Such studies investigate whether both meanings are activated by bilinguals, indicating nonselective access, or whether context can cause the L1 or L2 to be deactivated, suggesting selective access. Although some studies argue that only the language-appropriate meaning is activated (Rodriguez-Fornells, Schmitt, Kutas, & Münte, 2002; Price et al., 1999; Gerard & Scarborough, 1989; Scarborough, Gerard, & Cortese, 1984), there is a great deal of data that support a nonselective bilingual word recognition system (Chen & Ho, 1986 in Stroop interference; Smith & Kirsner, 1982 in bilingual picture–word distracter tasks; De Bruijn, Dijkstra, Chwilla, & Schreifers, 2001; Van Heste, 1999; Grainger & Dijsktra, 1992; Beauvillain & Grainger, 1987 in primed lexical decision tasks [LDTs]; De Groot, Delmaar, & Lupker, 2000; Dijkstra, Grainger, & Van Heuven, 1999; Van Heuven, Dijkstra, & Grainger, 1998 in nonprimed LDTs).

However, task manipulations, such as whether an experiment is mixed or all-L2, make a big difference in whether recognition is really nonselective (cf. Dijkstra, Timmermans, & Schreifers, 2000; De Moor, 1998; Dijkstra, Van Jaarsveld, & Brinke, 1998). For instance, in nonprimed LDTs in which semantic processing is not crucial, an all-L2 stimuli list composition results in what appears to be selective access, whereas mixed-stimuli lists show nonselective access (cf. De Moor, 1998; Dijkstra et al., 1998, but see Rodriguez-Fornells et al., 2002). Even if there is nonselective access in a mixed list, however, Von Studnitz and Green (2002) showed that other-language interference can decrease over time and that participants' knowledge that homographs were included in the experiment also improved control. This implies that the locus of bilingual control is indeed external to the actual lexico-semantic system, a principle adopted in the BIA+ model (Dijkstra & Van Heuven, 2002).

However, in semantic priming LDTs, in which semantic processing is more crucial, even all-L2 stimuli lists show nonselective access (Paulmann et al., 2006; Elston-Güttler, 2000; Dijkstra et al., 1999; Van Heste, 1999; for effects of the L2 on the L1, see Van Hell & Dijkstra, 2002). With Dutch–English bilinguals, Van Heste (1999) showed a priming effect for homograph-translation trials (e.g., brand–fire where brand means “fire” in Dutch) when compared to homograph-control trials. With German–English bilinguals, Elston-Güttler (2000) reported priming in pairs such as chef–boss in an all-L2 English stimuli list.

Recent ERP studies involving semantic priming of interlingual homograph meanings have also provided strong evidence of nonselective access, even with context manipulations or in all-L2 experiments. In ERP studies, semantic priming is measured using the N400 component, a negative-going waveform about 400 after stimulus onset that is more negative for words that are difficult to integrate into a preceding context. For example, the N400 for a target boy, preceded by a semantically related prime junior, is less negative than the N400 for a target preceded by an unrelated prime (for L2-priming studies, see Kotz & Elston-Güttler, 2004). Using prime–target pairs such as chef–boss or gift–poison, Paulmann et al. (2006) reported that German learners of English showed RT and ERP N400 priming of L1 meanings (boss), regardless of whether the all-English experiment was presented after a film narrated in English (L2) or in German (L1). This suggested that in single-word processing, more global language or task manipulations do not really affect nonselective access. In a similar vein, De Bruijn et al. (2001) reported that Dutch–English bilinguals showed significant RT and ERP N400 priming of targets reflecting the English reading of a preceding Dutch–English homograph prime. Crucially, the effect was obtained regardless of whether the pair was preceded by an English-only or Dutch-only word (i.e., there was priming of heaven in both the triplets house–angel–heaven and zaak–angel–heaven, where angel means “sting” in Dutch and zaak means “case” or “shop”). From the basis of these data, one could conclude that the effects of linguistic context directly preceding a trial are minimal or that the mere presence of Dutch words in the experiment could have prevented the task system from essentially ruling out the L1. Also, one could imagine that the contexts provided by the film in the Paulmann et al. (2006) study, and the word preceding the prime–target trials in the De Bruijn et al. study were simply not enough to prevent influence of the context-irrelevant language, especially in single-word processing.

Lexical Access of Words in Sentences and Language Context

However, how do manipulations such as preceding language context or background phonology affect bilingual lexical access when interlingual homographs appear in full syntactic and semantic contexts, namely, in L2 sentences? The BIA+ model (Dijkstra & Van Heuven, 2002) accounts for such linguistic context effects by assuming that sentence context affects the so-called word identification system. Here, we assume as in BIA+ that task manipulations affect threshold levels, but linguistic contexts can add an additional dimension in context changes, so task and language manipulations may affect the system differently when interlingual contexts are processed as single words versus in sentences.

This assumption is most likely accurate. In a sentence RT–ERP study (Elston-Güttler, Gunter, et al., 2005), the very same interlingual homograph primes as in the Paulmann et al. (2006) study completed English sentences as in “The woman gave her friend an expensive gift” (control prime: item). The sentence-final primes were followed by targets (i.e., poison) for lexical decision, and as in the Paulmann et al. study, half of the participants viewed a film narrated in German, whereas the other half viewed the film in English before performing the same all-English LDT. To address the question of language mode or task effects over time, the first and second halves of the experiment were analyzed. Results showed that only the group who viewed the German film before the experiment showed L1 influence, or significant RT and ERP priming of L1-related targets, and only in the first half of the experiment. The group who viewed the English film before the English experiment did not show any L1 influence on processing as RT and ERP priming of L1 meanings was not observed.

The film and block effects can be explained using aspects of the BIA+ model of bilingual word recognition, which draw upon ideas in previous models such as the BIA (Dijkstra et al., 1998; Van Heuven et al., 1998) and the Inhibitory Control (IC) model by Green (1998; see also Von Studnitz & Green, 2002). In Elston-Güttler, Gunter, et al. (2005), we suggested that in the first experimental half for those who viewed the German film, the task setting as defined in the BIA+ suddenly changes from “monolingual L1” to “monolingual L2,” and that there is a residual influence from the L1 as the system attempts to “zoom in to” the new L2 language setting. By the second half of the experiment, time and additional L2 input have helped the system to fully recalibrate, thereby raising decision thresholds high enough to eliminate L1 influence. In the single-word study (Paulmann et al., 2006), no such block effect, alone or in interaction with film and/or semantic priming, was observed. Because both experiments made use of the same prime–target pairs, we suggest that the additional sentence context provided in the Elston-Güttler, Gunter, et al. study helped to recalibrate the system more effectively. The BIA+ model of word recognition presumes that access to the bilingual lexicon is nonselective, but theoretically allows this access to be influenced by linguistic (e.g., sentence context) and nonlinguistic contexts (e.g., task-oriented factors such as the narrative of the film).

If sentence context, combined with a preceding language context, affects the bilingual word recognition system, then what about other types of contexts that are highly relevant to language use in the real world, namely, acoustic context? In the present study, we investigate the role that purely phonological information (L1 pseudowords) and semantic/phonological information (real L1 words) play in zooming in to the L2 when presented in the background during an all-L2 experiment. In the present study, we keep the all-English experiment constant and control for the language context created before the experiment by showing the same film in English or German that was used in Paulmann et al. (2006) and Elston-Güttler, Gunter, et al. (2005). What we manipulate, however, is the background stimuli (unattended channel) that participants hear during execution of the visual lexical decision experiment (attended channel): Participants either hear noise, German (L1) phonology only in the form of pseudowords, or L1 semantics/phonology in the form of real German words. The general idea behind the chosen paradigm is that the manipulation of the information content of the to-be-ignored message will show up as interference effects in the to-be-attended channel when processing resources/structures are shared (Broadbent, 1982). To put things in perspective of the present experiment, suppose the zooming-in to L2 is affected by L1 phonology; one would expect to find a zooming-out when the L1 phonology is presented cross-modally, which will lead to interference during the all-L2 lexical decision experiment when an interlingual homograph is presented as prime word.

The present paradigm resembles the so-called irrelevant speech effect (for a seminal study, see Colle & Welsh, 1976), where impairment in memory performance (i.e., immediate serial recall) is found as a result of irrelevant background speech1 presented during encoding (cf. LeCompte, Neely, & Wilson, 1997; Salamé & Baddeley, 1982, 1989). There have been several reports in the literature that meaningful speech disrupts recall more than meaningless speech (i.e., typically speech presented backward; Oswald, Tremblay, & Jones, 2000; Neely & LeCompte, 1999; LeCompte et al., 1997). There is even some evidence that the content of the irrelevant speech affects its potency. Buchner and Erdfelder (2005), for instance, report that high frequent words cause less memory impairment than low frequent words, whereas Neely and LeCompte (1999) present data that serial recall is affected to a larger extent when the irrelevant speech contains strong free associates than when the irrelevant words were dissimilar to items to be recalled. Particularly interesting for the present study is that there have also been some reports which show that phonological similarity within the irrelevant acoustic stream has an impact on recall. The more similar the speech or consonant–vowel–consonant syllables are, the less impact there is on recall (Hughes, Tremblay, & Jones, 2005; Jones & Macken, 1995). It is important to note that the irrelevant speech effect is typically explored in the memory domain and is specifically found for serial recall and not recall in general (cf. Jones, 1995). As the irrelevant speech effect even extends to sequence learning, serial order information appears to be an essential component of the effect (Farley, Neath, & Allbritton, 2007). To summarize, studies on serial recall have shown that semantic and phonological information presented in the irrelevant acoustic channel has clear memory effects on information presented in the visual channel. For the present experiment, it is relevant to know whether such effects can also be found in situations where irrelevant speech is presented during language processing.

Several studies have shown a detrimental impact of irrelevant speech on language comprehension. Oswald et al. (2000), for instance, showed that the reading comprehension of short passages is affected by irrelevant speech. Similarly, disruption was found during proofreading (Venetjoki, Kaarlela-Tuomaala, Keskinen, & Hongisto, 2006; Jones, Miles, & Page, 1990). In a series of five experiments, Martin, Wogalter, and Forlando (1988) showed that unattended speech interfered with reading comprehension of passages from a graduate record examination practice book. Interestingly enough, they also found that unattended phonological background information did not interfere. On the one hand, this seems to contrast the serial recall findings as discussed above, but on the other hand, Martin et al. (1988) suggest that this inconsistency probably had to do with the type of information subjects have to process in order to accomplish the primary task. In reading for comprehension, a participant has to remember the content of a passage and not its verbatim content, whereas during serial recall, this (phonological) verbatim information is highly task relevant. Because of this difference, serial recall will show interference of unattended phonological information and reading for comprehension will not. In the following three experiments, we explore the background stimuli manipulations just discussed as well as critical replications of the ERP block and film effects reported in Elston-Güttler, Gunter, et al. (2005).

EXPERIMENT 1: PHONOLOGICAL L1 INTERFERENCE

The aim of Experiment 1 is to test whether purely phonological information in the L1 (German) played in the background (unattended channel) during an all-L2 (English) visual experiment (attended channel) with interlingual homographs causes L2 learners to have difficulty zooming in to the L2. To test this, we used the same stimuli from Elston-Güttler, Gunter, et al. (2005), sentences ending with interlingual homograph primes (or a control prime) and followed by targets that reflect the L1 homograph meaning (see Table 1 for examples). During the same visual experiment, half of the participants (n = 16) heard German nonderivational pseudowords played continuously in the background, and the other half heard noise (i.e., the same pseudowords played in reverse and not resembling any known language). As a control condition to compare to the L1 phonological input, we chose to use the reversed pseudowords as opposed to silence, as this ensured that the two conditions differed only in terms of whether the background sounds contained L1 phonology.

Table 1. 

Examples of Stimuli Materials by Condition for Sentence-final Interlingual Homograph Primes, Control Primes, and Targets

Condition
Sentence
Target
Sentence 1 for Gift = “Poison” in German 
REL The woman gave her friend a pretty gift POISON 
UREL The woman gave her friend a pretty shell POISON 
 
Sentence 2 for Gift= “Poison” in German 
REL On holiday he found the perfect gift POISON 
UREL On holiday he found the perfect shell POISON 
 
Sentence 1 for Tag = “Day” in German 
REL Joan used scissors to remove the tag DAY 
UREL Joan used scissors to remove the label DAY 
 
Sentence 2 for Tag = “Day” in German 
REL The shoplifter ripped off the tag DAY 
UREL The shoplifter ripped off the label DAY 
Condition
Sentence
Target
Sentence 1 for Gift = “Poison” in German 
REL The woman gave her friend a pretty gift POISON 
UREL The woman gave her friend a pretty shell POISON 
 
Sentence 2 for Gift= “Poison” in German 
REL On holiday he found the perfect gift POISON 
UREL On holiday he found the perfect shell POISON 
 
Sentence 1 for Tag = “Day” in German 
REL Joan used scissors to remove the tag DAY 
UREL Joan used scissors to remove the label DAY 
 
Sentence 2 for Tag = “Day” in German 
REL The shoplifter ripped off the tag DAY 
UREL The shoplifter ripped off the label DAY 

We predict that if purely phonological information in the L1 affects zooming in to the L2, then RT and N400 priming of targets reflecting the L1 meanings of interlingual homographs should be obtained in the all-English task during which German phonology can be heard, at least in the first half of the experiment (presumably before threshold levels are fully calibrated). In Elston-Güttler, Gunter, et al. (2005), we obtained such RT and ERP priming of L1 meanings, but only when a German (L1) film was played before the experiment and only in the first half of the experiment (nothing was played in the background during the experiment). In the version in which an English (L2) film served to create an L2 context before the film, no L1 effects on L2 processing were obtained. Because we wanted to create a situation in which the L1 is not active unless due to our experimental manipulation (background phonology vs. noise), and because we wanted to control for global language context, we presented all participants with the English film before Experiment 1.

Methods

Participants

Thirty two native German-speaking students (16 men) participated in the experiment. All participants were right-handed, had normal or corrected-to-normal vision, were paid €20 for their participation, and signed a written informed consent. A description of participant data (sex, age, and language ability based on an extensive language questionnaire and vocabulary and comprehension tests) can be found in Table 2. As seen in Table 2, each group of participants was split into two proficiency groups (high and low) on the basis of a median split based on months of English exposure abroad. This enabled us to introduce the factor “language proficiency” in the statistical analyses to rule out whether effects were affected by general language ability. Despite this split, all participants can be considered advanced, late learners of English.

Table 2. 

Participant Data Provided by the Language Questionnaire and Posttesting


Experiment 1: English Film, Phonology (n = 16)
Experiment 1: English Film, Noise (n = 16)
Experiment 2, German Film, Noise (n = 16)
Experiment 3, English Film, German Words (n = 16)
Age (years) 25.4 (25.7, 25.3) 24.8 (24.4 25.3) 25.6 (26.5, 24.6) 24.3 (24.5, 24.0) 
Sex 9 females (56.3) 11 females (68.8) 9 females (56.3) 9 females (56.3) 
(75.0, 37.5) (87.5, 50.0) (50.0, 62.5) (37.5, 75.0) 
Age of acquisition (English, years) 11.3 (11.4, 11.1) 10.6 (11.0, 10.3) 11.6 (12.0, 11.3) 10.5 (10.1, 10.9) 
Months abroad*** 7.1 (13.5, 0.6) 7.6 (12.6, 2.5) 6.5 (12.3, 0.8) 4.2 (8.3, 0.1) 
Listeninga,8.0 (8.0, 8.0) 8.6 (8.9, 8.3) 7.9 (8.5, 7.4) 8.3 (8.9, 7.6) 
Readinga,** 7.4 (7.8, 7.0) 7.9 (8.5, 7.3) 7.3 (7.8, 6.9) 7.3 (7.9, 6.6) 
Speakinga,** 8.1 (8.3, 7.3) 8.6 (8.9, 8.3) 8.3 (8.6, 8.0) 7.9 (9.0, 6.8) 
Writinga,** 7.8 (8.3, 7.3) 8.2 (8.9, 7.5) 7.4 (8.0, 6.8) 7.4 (8.3, 6.5) 
Independencea,*** 7.9 (8.5, 7.3) 8.4 (9.3, 7.6) 7.5 (8.3, 6.8) 7.4 (8.4, 6.4) 
Vocabulary test (%)*** 92.9 (93.8, 92.0) 94.8 (96.0, 93.5) 90.3 (93.8, 86.8) 93.0 (95.9, 90.1) 
Memory task (%) 78.1 (78.1, 78.1) 77.8 (81.4, 74.1) 78.3 (79.5, 77.1) 79.0 (77.5, 80.5) 
Film comprehension test (%) 95.3 (90.6, 100) 94.3 (96.9, 91.8) 97.0 (97.0, 97.0) 94.7 (98.7, 91.8) 
test in English test in English test in German test in English 

Experiment 1: English Film, Phonology (n = 16)
Experiment 1: English Film, Noise (n = 16)
Experiment 2, German Film, Noise (n = 16)
Experiment 3, English Film, German Words (n = 16)
Age (years) 25.4 (25.7, 25.3) 24.8 (24.4 25.3) 25.6 (26.5, 24.6) 24.3 (24.5, 24.0) 
Sex 9 females (56.3) 11 females (68.8) 9 females (56.3) 9 females (56.3) 
(75.0, 37.5) (87.5, 50.0) (50.0, 62.5) (37.5, 75.0) 
Age of acquisition (English, years) 11.3 (11.4, 11.1) 10.6 (11.0, 10.3) 11.6 (12.0, 11.3) 10.5 (10.1, 10.9) 
Months abroad*** 7.1 (13.5, 0.6) 7.6 (12.6, 2.5) 6.5 (12.3, 0.8) 4.2 (8.3, 0.1) 
Listeninga,8.0 (8.0, 8.0) 8.6 (8.9, 8.3) 7.9 (8.5, 7.4) 8.3 (8.9, 7.6) 
Readinga,** 7.4 (7.8, 7.0) 7.9 (8.5, 7.3) 7.3 (7.8, 6.9) 7.3 (7.9, 6.6) 
Speakinga,** 8.1 (8.3, 7.3) 8.6 (8.9, 8.3) 8.3 (8.6, 8.0) 7.9 (9.0, 6.8) 
Writinga,** 7.8 (8.3, 7.3) 8.2 (8.9, 7.5) 7.4 (8.0, 6.8) 7.4 (8.3, 6.5) 
Independencea,*** 7.9 (8.5, 7.3) 8.4 (9.3, 7.6) 7.5 (8.3, 6.8) 7.4 (8.4, 6.4) 
Vocabulary test (%)*** 92.9 (93.8, 92.0) 94.8 (96.0, 93.5) 90.3 (93.8, 86.8) 93.0 (95.9, 90.1) 
Memory task (%) 78.1 (78.1, 78.1) 77.8 (81.4, 74.1) 78.3 (79.5, 77.1) 79.0 (77.5, 80.5) 
Film comprehension test (%) 95.3 (90.6, 100) 94.3 (96.9, 91.8) 97.0 (97.0, 97.0) 94.7 (98.7, 91.8) 
test in English test in English test in German test in English 

Figures listed in parentheses represent the mean of the high-proficiency participants and the low-proficiency participants, respectively, for each film version (n = 8 in each group).

Significant difference between the total high (n = 32) and total low (n = 32) proficiency group means in a two-tailed t test: *p < .05, **p < .01, ***p < .001.

a

Self-rating of ability in English on a scale of 1 to 10.

Materials

The Film

In order to induce a language context before the experiment, we created a 20-min silent film that was supplied with a German and an English narrative describing an obvious crime plot. Both the German and the English versions were spoken by native speakers (for more details on film preparation and characteristics, see Elston-Güttler, Gunter, et al., 2005).

Acoustically Presented Nonderivational Pseudowords

During the visual experiment, the participants heard either pseudowords or pseudowords played backward. This manipulation was done in order to explore the extent to which the phonology of the L1 has an influence on zooming in to the L2. In order to be sure that only phonology played a role, a special kind of pseudoword was used, the so-called nonderivational pseudowords (for a recent study, see Deacon, Dynowska, Ritter, & Grose-Fifer, 2004). The typical pseudoword used in experimental studies is derived from real words by changing two or more characters. Data by Deacon et al. (2004) suggest that such derivational pseudowords are capable of activating the real word root at a lexical and semantic level. From that perspective, such derivational pseudowords are less optimal to be used as a pure manipulator of phonological information. Nonderivational pseudowords, however, are not derived from actual words and therefore will not activate the root of a real word. The nonderivational pseudowords in our study are “German,” in that they did not resemble English words, are pronounceable in German, and are orthographically legal in German. All had two syllables and the morphological characteristics of nouns. Examples of the 50 words created are kirzo, sampor, and kefflis. The words were spoken in by six different German native speakers (students, 3 women) and were electronically edited in order to equalize for loudness (digitalization with 44 kHz). The acoustic material was presented at a comfortable sound level of approximately 60 dB.

During the forward condition, participants were confronted with a continuous stream of acoustically presented pseudowords (with an interval of approximately 100 msec between the items), which were randomly taken out of the complete set of 300 items (6 × 50). During the presentation of the critical part in the visual modality (i.e., the interlingual homograph and the following target word), the acoustic stream stopped until the lexical decision was made. In the reversed pseudoword condition, the procedure was the same, except that acoustic items were manipulated such that no phonological information was present. As merely playing the words backward is not sufficient for this, we used the LINUX software SOX (SOX phaser 1 1 5 0 2 −S) to manipulate the acoustic information further.

The Language Material

During the experiment, participants were presented with the same sentence material as in Elston-Güttler, Gunter, et al. (2005). After each sentence, the subjects had to perform an LDT. A total of 54 interlingual homographs (cf. gift = “poison” in German) were used during the experiment. For each item, an English target word was created that was a direct translation of the German meaning of the interlingual homograph (i.e., poison). In addition, two sentences per item were produced, having a relatively open context without a reference toward the target word meaning (i.e., 108 sentences in total). These sentences could either end with the interlingual homograph prime word or with an unrelated prime that was matched in length and word frequency to the interlingual homograph prime (see Table 3). This resulted in a total of 216 experimental sentences (see Table 1).

Table 3. 

Sentence Length and Letter Length and Frequency of Primes and Targets

ConditionSentence Word CountPrime FrequencyaPrime Letter LengthTarget FrequencyaTarget Letter Length
Related (interlingual homographs) 6.4 (1.0) 50.7 (87.1) 5.0 (1.6) 174.3 (137.1) 5.8 (2.2) 
Unrelated (control) 6.4 (1.0) 52.1 (87.4) 5.3 (1.7) 174.3 (137.1) 5.8 (2.2) 
Mean 6.4 (1.0) 51.4 (86.6) 5.2 (1.7) 174.3 (137.1) 5.8 (2.2) 
ConditionSentence Word CountPrime FrequencyaPrime Letter LengthTarget FrequencyaTarget Letter Length
Related (interlingual homographs) 6.4 (1.0) 50.7 (87.1) 5.0 (1.6) 174.3 (137.1) 5.8 (2.2) 
Unrelated (control) 6.4 (1.0) 52.1 (87.4) 5.3 (1.7) 174.3 (137.1) 5.8 (2.2) 
Mean 6.4 (1.0) 51.4 (86.6) 5.2 (1.7) 174.3 (137.1) 5.8 (2.2) 

Standard deviations are given in parentheses.

a

Mean frequency per million of test and corresponding control primes, along with the targets used in both test and control conditions, using the English lemma frequency dictionary in the CELEX lexical database (Baayen, Piepenbrock, & Van Rijn, 1993).

To control for order effects in the block analysis, each presentation list was presented in two orders, with the first and second halves switched. Each of the resulting four presentation lists consisted of 432 trials comprising 108 critical trials, 132 filler trials, and 192 pseudoword trials, divided into 12 short blocks. In each presentation list, targets were repeated once, whereas interlingual homographs only appeared once. Pseudoword targets were created by changing existing words by one or two letters (i.e., blude) and abiding by English orthographic rules (all pseudowords were checked to make sure they were not German words). To ensure that participants fully read sentences, a memory recall task was included after every block (36 trials), in which participants indicated whether they had read certain sentences in the experiment.

A trial started off by presenting the sentence excluding its final word. When the participant pressed the “yes” button, the sentence was removed and a fixation cross appeared at the position of the sentence-final word. After 200 msec, the fixation cross was replaced by the sentence-final word, either the interlingual homograph or the control prime. This sentence-final word was presented for 200 msec. Then the target word was presented immediately at the same position of the sentence-final word. Now the subject was required to do the LDT. After the lexical decision (maximum time was 3000 msec), the target word disappeared, and after a 2500-msec intertrial interval, the next trial started. Thus, the actual position of the sentence-final and target word depended on the length of the sentence. During the presentation of a trial, a continuous stream of acoustic items was presented, which was interrupted during the embedded LDT. Thus, the acoustic stream stopped around the time of the fixation cross and continued after the lexical decision response was given.

Procedure

Participants were seated in a dimly lit, soundproof electrically shielded cabin, facing a color video screen at a distance of 110 cm. Half of the participants received the forward pseudoword (phonology) condition, the other half the backward pseudoword (noise) condition. The experiment started off with a short training session in which the subjects were required to do the LDT up to the level of at least 85% correct responses. During the training, the interfering acoustic stimuli (in either of the two conditions) were also presented. Participants were instructed that they would see a crime movie and would have to complete a comprehension test about it, but as they were to be tested half an hour after viewing, they would perform a totally unrelated lexical decision experiment in the meantime. After the practice session, they viewed either the German or the English version of the film on the computer screen. Immediately after the film's end, the LDT started without any interfering instructions of the researchers. The LDT typically lasted 30 min and was presented in 12 blocks, ending with a memory task as described above. After all 12 blocks were carried out, a posttest was given in which participants had to indicate if they knew the English meanings of all the interlingual homographs that appeared in the experiment. Unknown items were discarded from further analysis on a subject-by-subject basis (see Table 2 for results). Last, subjects had to fill out 10 multiple choice questions related to the film they viewed.

ERP Recording

The electroencephalogram (EEG) was recorded with 52 Ag–AgCl electrodes (Electro-Cap) from Fp1, Fpz, Fp2, AF7, AF3, AFz, AF4, AF8, F7, F5, F3, Fz, F4, F6, F8, FT7, FC5, FC3, FCz, FC4, FC6, FT8, T7, C5, C3, Cz, C4, C6, T8, TP7, CP5, CP3, CPz, CP4, CP6, TP8, P9, P7, P5, P3, Pz, P4, P6, P8, PO7, PO3, POz, PO4, PO8, O1, Oz, O2, and the right mastoid, each referred to the left mastoid (nomenclature as proposed by the American Electroencephalographic Society, 1991). Bipolar horizontal and vertical electrooculograms were recorded for artifact rejection purposes. Electrode resistance was kept under 5 kΩ. The signals were recorded continuously with a band pass between DC and 125 Hz and were digitized at a rate of 500 Hz. ERPs were filtered off-line with a 7-Hz low pass for graphical display. All statistics were computed on nonfiltered data.

Data Analysis

Average ERPs starting 200 msec before and lasting 1000 msec after the presentation of the target words were computed for each electrode position and each of the two conditions (related and unrelated). The first and second halves of the experiment were averaged separately in order to investigate the development of the zooming-in effect. Incorrectly discriminated targets and trials containing an unknown prime or target (5%) were excluded from analysis. Due to ocular artifacts, an additional 15% of the trials were excluded. Averages were aligned to a 200-msec prestimulus baseline. Statistical analyses were performed using seven regions of interest (ROIs): left frontal: F7, F5, F3, FT7, FC5, FC3; right frontal: F8, F6, F4, FT8, FC6, FC4; left central: T7, C5, C3, TP7, CP5, CP3; right central: T8, C6, C4, TP8, CP6, CP4; left parietal: P7, P5, P3, PO7, PO3, O1; right parietal: P8, P6, P4, PO8, PO4, O2; midline: Fz, FCz, Cz, CPz, Pz, POz.

For each experiment, statistical analyses were carried out in three latency windows determined by visual inspection: 150–250 msec for N200; 250–550 msec for N400, and 550–850 msec for P600 or extended N400 effects. For Experiments 1 to 3, the within-subject factors for the repeated measure analyses of variance were block (first half vs. second half of the experiment), relatedness (related vs. unrelated), and ROI (7), whereas the between-subjects factor was English proficiency (high vs. lower proficiency in English as determined by the language questionnaire and posttests). For Experiment 1, an additional between-subjects factor phonology (forward vs. backward pseudowords) and was introduced to directly compare the two critical conditions. The Geisser–Greenhouse correction (Vasey & Thayer, 1987; Greenhouse & Geisser, 1959) was always applied when evaluating effects with more than one degree of freedom in the numerator. Note that in the omnibus analyses of both the ERP data and behavioral data, the factor of proficiency was not significant as a main effect, nor did it interact significantly with other factors (all Fs < 1.0) in Experiment 1, so it is not reported below.

Behavioral Data

Overall, subjects carried out the LDT quite well (95.0% correct). The analysis of correct RTs revealed a significant main effect of block [F(1, 14) = 8.24, p < .008], only indicating that RTs were 33 msec faster in the second block (672 vs. 639 msec). No other effects were significant [F(1, 14) = 0.2–2.18, ns]. Therefore, there were no behavioral priming effects.

Electrophysiological Data

N200 Time Window (150–250 msec)

The omnibus analysis of the 150–250 msec time window showed a main effect of block [F(1, 28) = 14.3, p < .0007] as well as an interaction between relatedness and ROI [F(6, 174) = 2.46, p < .026]. On the basis of this interaction, separate analyses were carried out for each of the seven ROIs, and these analyses revealed no significant main effects for any of the ROIs [F(1, 28) = 0.12–2.89, p < .1]. All other main effects and interactions involving relatedness were not significant (F < 1.0).

N400 Time Window (250–550 msec)

The omnibus analysis of the 250–550 msec time window showed main effects of relatedness [F(1, 28) = 4.95, p < .034] and block [F(1, 28) = 17.12, p < .0003]. On the basis of the three-way interaction between relatedness, block, and phonology [F(1, 28) = 5.5, p < .026] separate analyses were carried out for both phonology conditions (all other main effects and interactions involving relatedness were not significant, F < 1.0). The analysis of the condition in which pseudowords were presented backward (noise) showed main effects of block [F(1, 14) = 17.9, p < .0008] and ROI [F(6, 84) = 8.31, p < .0001] only. The main effect of relatedness was not significant [F(1, 14) = 0.38, ns]. The analysis of the condition in which German pseudowords were played forward (phonology) showed main effects of relatedness [F(1, 14) = 5.81, p < .03] and ROI [F(6, 84) = 5.89, p < .0001], but also an interaction between relatedness and block [F(1, 14) = 5.1, p < .04]. On the basis of this interaction, separate analyses were performed for the first block and the second block for the phonology condition. These analyses revealed a significant main effect of relatedness for the first block [F(1, 14) = 11.2, p < .005], in which L1-related targets were less negative in amplitude than unrelated ones. In the second block, however, no such relatedness effect was significant [F(1, 14) = 0.12, ns].

P600 Time Window (550–850 msec)

The omnibus analysis of the 550–850 msec time window showed a marginally significant interaction between relatedness and phonology [F(1, 28) = 3.42, p < .074] (all other main effects and interactions involving relatedness were not significant, F < 1.0). On the basis of this interaction, separate analyses were carried out for both phonology conditions. Both conditions, in which German pseudowords were played forward (phonology) and in which they were played backward (noise), showed no significant effects of relatedness [phonology: F(1, 14) = 0.46, ns; noise: F(1, 14) = 1.97, ns].

In summary, these analysis showed that the N400 time window showed effects of relatedness, but only in the first block of the German phonology condition in which German pseudowords were played during the experiment.

Discussion and Rationale for Experiment 1

As can be seen in Figure 1, the ERPs of the related and unrelated target word did not show influence of L1 on the L2, as a modulation of the N400 (priming) was not obtained in either block of the experiment in the reversed pseudoword condition. We can therefore confirm that, as found in the original Elston-Güttler, Gunter, et al. (2005) study, the English film viewed before the experiment proper zoomed the participants into the L2, and the addition of phonologically neutral stimuli did not impact that. The reversed pseudoword condition in Experiment 1 therefore replicates Elston-Güttler, Gunter, et al., in which a null result was obtained when an English film preceded the experiment, but Experiment 1 extends this finding to apply to cases in which background noise is presented during the visual experiment (for the replication of the experiment when preceded by the German film, see Experiment 2).

Figure 1. 

ERPs elicited by critical targets at selected electrode sites. Waveforms show the average for related (solid line) and unrelated (dotted line) targets from 200 msec prior to stimulus onset up to 800 msec poststimulus onset. This figure shows the ERPs elicited after viewing the English (L1) film version and while participants heard noise during the experiment.

Figure 1. 

ERPs elicited by critical targets at selected electrode sites. Waveforms show the average for related (solid line) and unrelated (dotted line) targets from 200 msec prior to stimulus onset up to 800 msec poststimulus onset. This figure shows the ERPs elicited after viewing the English (L1) film version and while participants heard noise during the experiment.

From Figure 2, it becomes clear that when the acoustic stream does contain L1 phonology, the process of zooming in is affected. That is, despite an English film before the experiment that helps in adjusting to the L2, the presence of L1 phonology during the experiment itself caused a modulation of the N400 component 250–550 msec poststimulus onset, reflecting priming of L1 meanings during the all-L2 visual task. As in the German film version reported in Elston-Güttler, Gunter, et al. (2005), however, this effect is limited to the first half of the experiment. This indicates that L1 phonology affects the zooming process, but L1 phonology does not prevent effective zooming in: In the second half of the experiment, all measurable activation of L1 meanings during L2 processing disappeared. The block effect stresses the importance of looking at language control and zooming in as dynamic processes. In the BIA+ model (Dijkstra & Van Heuven, 2002), it is assumed that the system continually recalibrates based on incoming input. In the present experiment, we start with an L2 language setting based on the film, but as the visual experiment begins, L2 phonology is present, so the system does not immediately calibrate threshold levels to rule out L1 relevance as it does when the background is phonologically neutral noise. Instead, it recalibrates threshold levels such that L1 relevance is still allowed, and only as more and more trials in the visual experiment, along with the background phonology itself, confirm that the background phonology is completely irrelevant to the task does the system recalibrate to allow full zooming in to the L2.

Figure 2. 

ERPs elicited by critical targets at selected electrode sites. Waveforms show the average for related (solid line) and unrelated (dotted line) targets from 200 msec prior to stimulus onset up to 800 msec poststimulus onset. This figure shows the ERPs elicited after viewing the English (L1) film version and while participants heard German (L2) phonology during the experiment. Only while hearing German phonology did participants show priming of the L1 interlingual homograph meaning as shown by the modulation of the N400 component, and this effect was qualified by experimental block.

Figure 2. 

ERPs elicited by critical targets at selected electrode sites. Waveforms show the average for related (solid line) and unrelated (dotted line) targets from 200 msec prior to stimulus onset up to 800 msec poststimulus onset. This figure shows the ERPs elicited after viewing the English (L1) film version and while participants heard German (L2) phonology during the experiment. Only while hearing German phonology did participants show priming of the L1 interlingual homograph meaning as shown by the modulation of the N400 component, and this effect was qualified by experimental block.

It should be noted that the critical N400 block effect from Elston-Güttler, Gunter, et al. (2005) was replicated, but Experiment 1 differs in some aspects. First, the RT data do not show clear semantic priming of L1 meanings in either the noise or L1 phonology condition. Here, we assume that RTs may not be sensitive enough to capture the more subtle zooming-in process, especially when RTs are particularly slow as they were in Experiment 1, possibly due to the additional processing effort caused by the background stimuli. As the mean RTs in Experiment 1 were rather long (655 msec), and the primes were presented for 200 msec, the RT was made approximately 450 msec after prime onset. Several studies on homonyms (see Elston-Güttler & Friederici, 2005 for a review; Elston-Güttler & Friederici, 2007) suggest that the longer the time lapse between the prime and the target, the more likely context plays a role in processing. If the initial activation of the L1 interlingual homograph meanings is short-lived, then it is detectable during the N400 (ERP), but may disappear by the time the lexical decision (RT) is made. A similar RT–ERP dissociation was reported in Elston-Güttler and Friederici (2005), in which learners showed use of context cues in homonym processing in the RTs measured later in time, but not in the ERPs.

In Experiment 1, we did not obtain an additional modulation of the N200 from 150 to 250 msec as we did in Elston-Güttler, Gunter, et al. (2005). In our previous study, we assumed that the N200 effects were caused by enhanced form-level or orthographic processing of interlingual homographs (see Elston-Güttler, Gunter, et al., 2005; Elston-Güttler, Paulmann, & Kotz, 2005; Bentin, Mouchetant-Rostaing, Giard, Echallier, & Pernier, 1999; Compton, Grossbacher, Posner, & Tucker, 1991; for studies on the N200 as a reflection of phonological mismatch, see Van den Brink, Brown, & Hagoort, 2004; Connolly & Phillips, 1994; Connolly, Phillips, Stewart, & Brake, 1992), whereas the N400 reflected the semantic priming effects. In Experiment 1, we can only assume that the additional N200 effects were not as pronounced due to the more complex nature of the input including the auditory stimuli, and because there was no clear N200 modulation, the N400 itself may have started slightly earlier. The relatively broad distribution of the N400 (ROI did not significantly interact with condition effects), along with the extended time window, is consistent with N400 effects previously reported in L2 semantic priming (Kotz & Elston-Güttler, 2004). Although the N200 effects were not obtained, the more crucial measure of L1, the N400 modulation in Block 1 of the experiment, is clearly observed in Experiment 1 when L1 phonological information was presented.

However, we cannot be sure that the lack of RT priming and the absence of N200 effects are really driven by the presence of background stimuli. In order to test this assumption, we need to test the condition in Elston-Güttler, Gunter, et al. (2005), in which block effects were obtained (i.e., the experiment preceded by the German film), but this time with background noise during the task. This is the aim of Experiment 2: Here, we will present participants with the German film before the experiment, and they will hear the reversed German pseudowords during the English visual experiment. This way, we can explore whether the more crucial N400 effects from Elston-Güttler, Gunter, et al. are replicable with background noise. If we obtain the N400 in the absence of the N200 and RT results, then we also have further evidence that the differences between Experiment 1 and Elston-Güttler, Gunter, et al. are indeed caused by the additional processing load of background auditory stimuli. A replication of the N400 in the first block of the experiment would also provide more reliable evidence that preceding language context (the German film) makes zooming in to the L2 more difficult. Furthermore, a replication of the N400 effects with background noise is important to establish whether the reversed German pseudowords were really an appropriate control condition (i.e., causing no additional L1 activation) in Experiment 1.

EXPERIMENT 2: REPLICATION OF THE ORIGINAL ZOOM EFFECT

Methods, Materials, and Procedure

In Experiment 2, the same materials and procedure were used as in Experiment 1. The participants were 16 right-handed native German-speaking students (8 men) who had normal or corrected-to-normal vision. A description of participant data can be found in Table 2.

As in Experiment 1, incorrectly discriminated targets and trials containing an unknown prime or target (10%) were excluded from analysis. Due to ocular artifacts, an additional 15% of the trials were excluded. Statistical analysis used the factors block, relatedness, language proficiency, and ROI (for ERPs). However, as proficiency was not significant as a main effect, nor did it interact significantly with other factors (all Fs < 1.0) in Experiment 2, this factor was not reported in the statistical analyses. Before Experiment 2, participants were presented with the film narrated in German, then during the visual experiment, they were presented with the backward pseudoword (noise) auditory manipulation.

Behavioral Data

Overall, participants carried out the LDT well (92.8% correct). The statistical analysis of correctly discriminated targets revealed a main effect of block [F(1, 14) = 6.43, p < .023] only, indicating that the lexical decision mean was 39 msec faster for the second block of the experiment (760 vs. 721 msec). All other effects were not significant [F(1, 14) = 0.04–3.34]. Therefore, as in Experiment 2, we observed no behavioral priming effects.

Electrophysiological Data

N200 Time Window (150–250 msec)

The omnibus analysis of the 150–250 msec time window only showed a significant main effect of block [F(1, 14) = 4.76, p < .047]. Neither the main effect of relatedness [F(1, 14) = 0.96, ns] nor the interaction between relatedness and block [F(1, 14) = .16, ns] was significant.

N400 Time Window (250–550 msec)

The omnibus analysis of the 250–550 msec time window showed main effects of relatedness [F(1, 14) = 11.16, p = .005] and block [F(1, 14) = 12.14, p = .004]. On the basis of the interaction between relatedness and block [F(1, 14) = 9.73, p = .0075], separate analysis were done for the first and second experimental halves. These analyses revealed a significant main effect of relatedness for the first block [F(1, 14) = 23.8, p < .0002], in which L1-related targets were less negative in amplitude than unrelated targets. In the second block, this relatedness effect was not present [F(1, 14) = .33, ns]. All other main effects and interactions involving relatedness were not significant in the omnibus analyses (Fs < 1.0).

P600 Time Window (550–850 msec)

The omnibus analysis of the 550–850 msec time window showed a marginally significant main effect of block [F(1, 14) = 3.57, p < .08]. All other main effects and interactions with relatedness were not significant (Fs < 1.8).

In summary, only the N400 time window showed effects of relatedness, and only in the first block of the experiment.

Discussion and Rationale for Experiment 2

As can be seen from Figure 3, the modulation of the N400 component in the latency window 250–500 msec was obtained in Experiment 2 when participants viewed a German film before, and heard phonologically neutral background noise during, the visual English experiment. Just as in Elston-Güttler, Gunter, et al. (2005), the N400 priming of L1 meanings is only significant in the first half of the experiment. This replication provides strong evidence that the German film indeed creates a language context that affects the zooming-in process, and this is the case even when phonologically neutral noise is played in the background.

Figure 3. 

ERPs elicited by critical targets at selected electrode sites. Waveforms show the average for related (solid line) and unrelated (dotted line) targets from 200 msec prior to stimulus onset up to 800 msec poststimulus onset. This figure shows ERPs after viewing the German (L2) film and hearing noise during the experiment. Modulations in the ERP in the N400 time window were significant in Block 1.

Figure 3. 

ERPs elicited by critical targets at selected electrode sites. Waveforms show the average for related (solid line) and unrelated (dotted line) targets from 200 msec prior to stimulus onset up to 800 msec poststimulus onset. This figure shows ERPs after viewing the German (L2) film and hearing noise during the experiment. Modulations in the ERP in the N400 time window were significant in Block 1.

The latency and distribution is similar to the N400 effects obtained in Experiment 1, and there were no significant N200 modulations or RT priming effects obtained. This supports our assumption that the differences between Elston-Güttler, Gunter, et al. (2005) and the present study may be due to the presence of background stimuli. Note also that the reversed pseudowords did not affect the level of L1 activation as reflected by the N400, so we can assume that the phonologically neutral noise serves as an adequate control condition.

It seems clear that both the mere presence of L1 phonology (Experiment 1) and the global language context (Experiment 2) affect zooming in and support the idea that bilingual lexical access may be fundamentally nonselective, but can be manipulated. However, in both experiments, no N400 effects reflecting L1 influence on processing were found in the second half of the experiment. As the English trials built up and helped the system recalibrate, it appears as though the degree of L1 activation caused by the preceding film or L2 background phonology can indeed be greatly reduced; the system apparently recognizes over time that the L1 stimuli are irrelevant to the task. But can we always zoom in to the L2, no matter how strong the context? What happens if we add semantics to the L1 phonology by playing real German words in the background during the L2 experiment?

In Experiment 3, we present the English film before the experiment, then play real German words in the background during the visual experiment (of course, no interlingual homographs or words that are formally similar to any English words were included; see Methods). If phonology itself, with or without meaning, affects zooming in, then we would predict N400 effects in the first half of the experiment only, as in Experiments 1 and 2. However, if semantics adds an additional layer to L1 phonology and creates an even stronger context, then we would expect the system to have more difficulty zooming in by the second half of the experiment, and we may observe L1 influence throughout the entire experiment. This result would be in line with several studies that argue for nonselective access in mixed-stimuli lists (cf. Von Studnitz & Green, 2002; De Moor, 1998; Dijkstra et al., 1998). Presumably, even if the background L1 stimuli are irrelevant to the task, the mere presence of L1 words in the experiment may keep activation thresholds high for both languages.

EXPERIMENT 3: SEMANTIC INTERFERENCE OF THE L1

Methods, Materials, and Procedure

In Experiment 3, the same materials and procedure were used as in Experiment 1. Before the experiment, participants were presented with the film narrated in English, but instead of receiving pseudoword or noise stimuli as interference, subjects were presented with real German words. A total of 48 interfering words were used, which had two syllables and a mean word frequency of 110 (SD = 58). They were semantically unrelated to any of the other test material and did not resemble English words in their orthography and phonology. As with the pseudowords, the real German words were produced by (the same) six speakers and edited for loudness. The acoustic presentation format was similar to that used in Experiment 1.

The participants were 16 right-handed native German-speaking students (8 men) who had normal or corrected-to-normal vision (see Table 2). Incorrectly discriminated targets and trials containing an unknown prime or target (7%) were excluded from analysis. Due to ocular artifacts, an additional 15% of the trials were excluded. Statistical analysis used the factors block, relatedness, language proficiency, and ROI (for ERPs).

Behavioral Data

Overall, subjects carried out the LDT quite well (93.3% correct). As in the previous experiments, there was a main effect of block [F(1, 14) = 4.89, p < .044], indicating that the lexical decisions in the second block were 42 msec faster (788 vs. 746 msec) than in the first block of the experiment. No other effects were significant (Fs < 1.0). Therefore, as in both Experiments 1 and 2, there was no behavioral priming effect observed.

Electrophysiological Data

N200 Time Window (150–250 msec)

The omnibus analysis of the 150–250 msec time window revealed a significant main effect of block [F(1, 14) = 8.34, p = .012], whereas the main effect of relatedness was only marginally significant [F(1, 14) = 3.76, p = .073]. All other main effects and interactions with relatedness were not significant (all Fs < 1.6).

N400 Time Window (250–550 msec)

The omnibus analysis of the 250–550 msec time window showed main effects of relatedness [F(1, 14) = 7.94, p = .014] and block [F(1, 14) = 7.42, p = .016]. The interaction between relatedness and block was not significant [F(1, 14) = 0.09, ns]. However, there were significant interactions between relatedness and proficiency [F(1, 14) = 5.62, p = .033], between relatedness and ROI [F(6, 84) = 2.44, p = .032], and between relatedness, proficiency, and ROI [F(6, 84) = 2.87, p = .014]. On the basis of these interactions, separate analysis were carried out for both high and low English proficiency groups, and effects of both relatedness and ROI were considered in each group. For high-proficiency participants, the effect of relatedness was not significant [F(1, 7) = 0.24, ns], nor was the interaction between relatedness and ROI [F(6, 42) = 0.42, ns]. For the low-proficiency participants, however, the main effect of relatedness was significant [F(1, 7) = 17.89, p = .0039] as was the interaction between relatedness and ROI [F(6, 42) = 4.38, p = .0016]. Analyses by ROI for the low group revealed significant effects of relatedness for left frontal, left central, midline, right frontal, and right central ROIs [F(1, 7) = 26.23–10.0, p < .02], whereas no effect of relatedness was found for the left and right parietal ROIs [F(1, 7) = 1.68–0.47, ns].

P600 Time Window (550–850 msec)

The omnibus analysis of the 550–850 msec time window showed a significant interaction between relatedness and ROI [F(6, 84) = 2.57, p = .025] as well as a significant interaction between relatedness and proficiency [F(1, 14) = 13.10, p = .0028]. All other main effects and interactions with relatedness were not significant [Fs < 1.7, ns, except for relatedness: F(1, 14) = 3.30, p = .09]. On the basis of the first interaction, separate analyses for each of the seven ROIs were performed. A significant main effect of relatedness was found for the right frontal ROI [F(1, 14) = 5.22, p = .0384]. The relatedness effect in all other ROIs was not significant [F(1, 14) = 3.1–0.01, ns]. On the basis of interaction between relatedness and proficiency, separate analyses were performed for the high and low groups. For the low group, the main effect of relatedness reached significance [F(1, 7) = 12.43, p = .0097], whereas for the high group, it did not [F(1, 7) = 2.0, ns].

In summary, only low-proficiency participants showed a significant modulation in the N400 time window, and this was most prominent at frontal and central electrode sites. Furthermore, the low group showed a sustained modulation into the late time window. The priming effects shown by the low group were observed over both blocks of the experiment.

GENERAL DISCUSSION

As can be seen in Figures 4and 5, there was a modulation of the N400 in the latency window 250–550 msec over both halves of the experiment when participants heard real German words in the background in Experiment 3, but for low-proficiency participants only. This indicates that, in the course of the English experiment, the less proficient participants were not able to zoom in to the L2 as L1 meanings were active during the entire experiment. If given more time, intermediate learners may indeed be able to zoom in, but it clearly takes longer than the 20-min experiment for intermediate learners to recalibrate the system in the presence of L1 semantics combined with phonology.

Figure 4. 

ERPs elicited by critical targets at selected electrode sites. Waveforms show the average for related (solid line) and unrelated (dotted line) targets from 200 msec prior to stimulus onset up to 800 msec poststimulus onset. Both Figures 4 and 5 show the ERPs after viewing the English (L1) film version and while hearing real German (L2) words during the experiment; Figure 4 shows high-proficiency participants.

Figure 4. 

ERPs elicited by critical targets at selected electrode sites. Waveforms show the average for related (solid line) and unrelated (dotted line) targets from 200 msec prior to stimulus onset up to 800 msec poststimulus onset. Both Figures 4 and 5 show the ERPs after viewing the English (L1) film version and while hearing real German (L2) words during the experiment; Figure 4 shows high-proficiency participants.

Figure 5. 

ERPs elicited by critical targets at selected electrode sites. Waveforms show the average for related (solid line) and unrelated (dotted line) targets from 200 msec prior to stimulus onset up to 800 msec poststimulus onset. Both Figures 4 and 5 show the ERPs after viewing the English (L1) film version and while hearing real German (L2) words during the experiment; Figure 5 shows low-proficiency participants. Only the low-proficiency participants showed N400 modulations of L1-related targets, and this effect was significant over both halves of the experiment.

Figure 5. 

ERPs elicited by critical targets at selected electrode sites. Waveforms show the average for related (solid line) and unrelated (dotted line) targets from 200 msec prior to stimulus onset up to 800 msec poststimulus onset. Both Figures 4 and 5 show the ERPs after viewing the English (L1) film version and while hearing real German (L2) words during the experiment; Figure 5 shows low-proficiency participants. Only the low-proficiency participants showed N400 modulations of L1-related targets, and this effect was significant over both halves of the experiment.

Interestingly, however, the high-proficiency learners did not show any significant L1 influence in Experiment 3. Recall that under similar conditions, with pure L1 phonology from Experiment 1, there were no proficiency effects; high-proficiency learners showed L1 influence in the first half of the experiment just as the low group did. Why, with real words, would seemingly stronger L1 influence on a semantic level appear to actually help higher-proficiency learners? We tentatively suggest that purely phonological versus semantic interference may operate differently for the two proficiency levels. For the low-proficiency participants, for whom a completely predicted pattern of results was found, we saw phonological interference in Block 1 of Experiment 1 and over the entire experiment in Experiment 3, in which real German words were auditorily presented in the background. In this case, it appears as if semantics adds to the interference, as one would predict. However, with the high-proficiency participants, the addition of semantics to the background stimuli may have changed the response of participants entirely. With added semantics, the distraction became much more obvious: In response to this, the word recognition system may have greatly reduced L1 activation levels, causing no interference effects. Evidence for a very similar type of increased control of L1 activation in conjunction with semantic contexts (in sentences), on the part of high-proficiency participants only, has already been reported in Elston-Güttler, Gunter, et al. (2005) and Elston-Güttler, Paulmann, et al. (2005). Note also that the high-proficiency participants indeed showed interference in the phonological condition, which was more difficult to detect as L1 interference. The more obvious the distraction, the more higher-proficiency learners may actually be aided in achieving language control. This explanation remains speculative, however, as the varied manner in which high- and lower-proficiency learners respond to semantic and form-related interference still needs to be systematically outlined in future studies. Even with this open question, though, we can still confidently claim that both high- and low-proficiency participants are subject to phonological zoom-in effects (Experiment 1), and that at least low-proficiency participants show even more profound interference once semantics are included in the background interference stimuli (Experiment 3).

Because we found an effect of L1 activity (N400 priming effect) in the first block of Experiment 1, we can argue that irrelevant phonetic information of the L1 has been taken into account. This suggests that the L1 lexicon was reactivated by this phonological information. Note that in the second block, the word recognition system can resist the irrelevant information and deactivate the L1 lexicon. All three experiments together show that the L1 lexicon will only be activated when phonological and phonological/semantic information from the L1 is presented in the irrelevant channel, but not with a manipulated reversed speech signal. After a while, the system is able to focus again on the main task (possibly by inhibiting the L1 influence) and is zoomed-in again.

Although the issue concerning the automaticity of the zoom-in effect is interesting, it is clear that the present selective attention paradigm cannot help us to shed light on the underlying nature of the zooming-in effect. During the experiments, irrelevant (acoustic) information needs to be discarded (or inhibited) by the cognitive system in order to process the relevant (visual) information efficiently, that is, without any interference. Typically, this type of paradigm is used to explore the automaticity of irrelevant information processing. Thus, if one wants to know whether semantic information is processed automatically, one looks for interference in the attended channel of irrelevant semantic information. One major criticism on using this type of paradigm in this way is that attention switching cannot be ruled out as a potential explanation. It is possible that participants switched their attention to the irrelevant channel, thereby putting it into focus of their processing (Holender, 1986; for a review on automatic semantic processing, see Deacon & Shelley-Tremblay, 2000). That is, the irrelevant information received a certain amount of attention, causing the interference. In the present study, an attentional switch of this kind cannot be ruled out either. It is very well possible that during the second block, participants switched their attention back to the main task, thereby getting rid of the interference. A possible test for this hypothesis would be to explore the processing of the irrelevant items. We know that the early components N1 and P2 are enhanced when attention is given to a particular stimulus (Heinze, Münte, & Mangun, 1994). If an attentional explanation was true, one would expect that the early components of the irrelevant information in Block 1 will show larger endogenous components compared to Block 2. Unfortunately, our experimental design did not make such explorations possible because the points in time when the irrelevant stimuli were presented were not explicitly recorded in the running EEG. Note, however, that we do not want to argue that the processing of L1 stimuli in the irrelevant channel is necessarily automatic. Our question relates to whether phonological information from the L1 influences the access of L1 meanings of interlingual homographs. Whether this process is affected by attentional aspects or operates automatically in the bilingual word recognition system is an interesting question, but it is not at stake here. In the present experiment, we aimed to explore whether the bilingual word recognition system encounters problems due to the processing of irrelevant information, and which information leads to these (distraction) problems. By using the present design, we have found clear evidence that the zooming-in process is affected by L1 phonology.

Our results outline, for the first time in the literature, that the presence of L1 purely phonological stimuli affects language control in the same manner as preceding global context. This is an important result, as it indicates that context types other than semantic or global language context affect the bilingual word recognition system. It is also quite striking that after an English film, in an all-English experiment including English sentences, the addition of semantics to L1 phonology (both completely irrelevant to the task) made zooming-in impossible within the experiment itself for less proficient learners. Further studies might indicate whether zooming-in is possible at all under these conditions, or if the time course is merely extended.

In terms of the BIA+ model, language context would be considered a task manipulation, whereas the sentences included in the visual experiment clearly create a semantic context. It is not clear, however, whether L1 phonology or real L1 words are purely task or linguistic parameters. On the one hand, they operate like manipulations of list composition (task changes) by creating a mixed-stimuli environment. On the other hand, they do so by activating the phonological and/or semantic information in the L1 (thus affecting the recognition system itself). However they are defined, it is clear that these types of context work together to create subtle differences in how well we can zoom in to and out of the L2, and as users of more than one language, we can feel these effects in our everyday experience.

Acknowledgments

We thank Sven Gutekunst and Sylvia Stasch for technical support and data acquisition as well as Suse Wagner, Stefan Girgsdies, and Christian Obermeier for help with experimental preparation and data processing. The vampire-film was compiled and edited by Daniel Hofmann, who also narrated the German commentary. Adam Ussishkin narrated the English version.

Reprint requests should be sent to Kerrie E. Elston-Güttler, Max Planck Institute for Human Cognitive and Brain Sciences, PO Box 500 355, 04303 Leipzig, Germany, or via e-mail: guettler@cbs.mpg.de.

Note

1. 

Note that several authors discuss this effect as the irrelevant sound effect because reduction in serial recall can also be elicited by nonspeech sounds (for a review, see Jones, 1995). For present purposes, we will, however, focus on the impact of irrelevant speech/speech-like sounds.

REFERENCES

REFERENCES
American Electroencephalographic Society
.
(
1991
).
Guidelines for standard electrode position nomenclature.
Journal of Clinical Neurophysiology
,
8
,
200
202
.
Baayen
,
R. H.
,
Piepenbrock
,
R.
, &
Van Rijn
,
H.
(
1993
).
The CELEX lexical database [CD-ROM].
Philadelphia, PA
:
Linguistic Data Consortium, University of Pennsylvania
.
Beauvillain
,
C.
, &
Grainger
,
J.
(
1987
).
Accessing inter-lexical homographs: Some limitations of language-selective access.
Journal of Memory and Language
,
26
,
658
672
.
Bentin
,
S.
,
Mouchetant-Rostaing
,
Y.
,
Giard
,
M. H.
,
Echallier
,
J. F.
, &
Pernier
,
J.
(
1999
).
ERP manifestations of processing printed words at different psycholinguistic levels: Time course and scalp distribution.
Journal of Cognitive Neuroscience
,
11
,
235
260
.
Broadbent
,
F. E.
(
1982
).
Task combination and selective intake of information.
Acta Psychologica
,
50
,
253
290
.
Buchner
,
A.
, &
Erdfelder
,
E.
(
2005
).
Word frequency of irrelevant speech distractors affects serial recall.
Memory & Cognition
,
33
,
86
97
.
Chen
,
H. C.
, &
Ho
,
C.
(
1986
).
Development of Stroop interference in Chinese–English bilinguals.
Journal of Experimental Psychology: Learning, Memory, and Cognition
,
12
,
397
401
.
Colle
,
H. A.
, &
Welsh
,
A.
(
1976
).
Acoustic masking in primary memory.
Journal of Verbal Learning and Verbal Behavior
,
15
,
17
32
.
Compton
,
P. E.
,
Grossbacher
,
P.
,
Posner
,
M. I.
, &
Tucker
,
D. M.
(
1991
).
A cognitive–anatomical approach to attention in lexical access.
Journal of Cognitive Neuroscience
,
3
,
304
312
.
Connolly
,
J. F.
, &
Phillips
,
N. A.
(
1994
).
Event-related potential components reflect phonological and semantic processing of the terminal word of spoken sentences.
Journal of Cognitive Neuroscience
,
6
,
256
266
.
Connolly
,
J. F.
,
Phillips
,
N. A.
,
Stewart
,
S. H.
, &
Brake
,
W. G.
(
1992
).
Event-related potential sensitivity to acoustic and semantic properties of terminal words in sentences.
Brain and Language
,
43
,
1
18
.
De Bruijn
,
E.
,
Dijkstra
,
T.
,
Chwilla
,
D.
, &
Schreifers
,
H.
(
2001
).
Language context effects on interlingual homograph recognition: Evidence from event-related potentials and response times in semantic priming.
Bilingualism: Language and Cognition
,
4
,
155
168
.
De Groot
,
A. M. B.
,
Delmaar
,
P.
, &
Lupker
,
S. J.
(
2000
).
The processing of interlexical homographs in translation recognition and lexical decision: Support for non-selective access to bilingual memory.
Quarterly Journal of Experimental Psychology, Section A: Human Experimental Psychology
,
53A
,
397
428
.
De Moor
,
W.
(
1998
).
Visual word recognition in bilinguals.
Unpublished Master's Thesis, University of Ghent.
Deacon
,
D.
,
Dynowska
,
A.
,
Ritter
,
W.
, &
Grose-Fifer
,
J.
(
2004
).
Repetition and semantic priming of nonwords: Implications for theories of N400 and word recognition.
Psychophysiology
,
41
,
60
74
.
Deacon
,
D.
, &
Shelley-Tremblay
,
J.
(
2000
).
How automatically is meaning accessed: A review of the effects of attention on semantic processing.
Frontiers in Bioscience
,
5
,
82
94
.
Dijkstra
,
T.
,
Grainger
,
J.
, &
Van Heuven
,
W. J. B.
(
1999
).
Recognition of cognates and interlingual homographs: The neglected role of phonology.
Journal of Memory and Language
,
41
,
496
518
.
Dijkstra
,
T.
,
Timmermans
,
M.
, &
Schreifers
,
H.
(
2000
).
On being blinded by your other language: Effects of task demands on interlingual homograph recognition.
Journal of Memory and Language
,
42
,
445
464
.
Dijkstra
,
T.
, &
Van Heuven
,
W. J. B.
(
2002
).
The architecture of the bilingual word recognition system: From identification to decision.
Bilingualism: Language and Cognition
,
5
,
175
197
.
Dijkstra
,
T.
,
Van Jaarsveld
,
H.
, &
Brinke
,
S. T.
(
1998
).
Interlingual homograph recognition: Effect of task demands and language intermixing.
Bilingualism: Language and Cognition
,
1
,
51
66
.
Elston-Güttler
,
K. E.
(
2000
).
An enquiry into cross-language lexical–conceptual relationships and their effect on L2 lexical processing.
PhD Dissertation, University of Cambridge.
Elston-Güttler
,
K. E.
, &
Friederici
,
A. D.
(
2005
).
Native and L2 processing of homonyms in sentential context.
Journal of Memory and Language
,
52
,
256
283
.
Elston-Güttler
,
K. E.
, &
Friederici
,
A. D.
(
2007
).
Ambiguous words in sentences: Brain indices for native and non-native disambiguation.
Neuroscience Letters
,
414
,
85
89
.
Elston-Güttler
,
K. E.
,
Gunter
,
T. C.
, &
Kotz
,
S. A.
(
2005
).
zooming into L2: Global language context and adjustment affect processing of interlingual homographs in sentences.
Cognitive Brain Research
,
25
,
57
70
.
Elston-Güttler
,
K. E.
,
Paulmann
,
S.
, &
Kotz
,
S. A.
(
2005
).
Who's in control?: Proficiency and L1 influence on L2 processing.
Journal of Cognitive Neuroscience
,
17
,
1593
1610
.
Farley
,
L. A.
,
Neath
,
I.
, &
Allbritton
,
D. W.
(
2007
).
Irrelevant speech effects and sequence learning.
Memory & Cognition
,
35
,
156
165
.
Gerard
,
L. D.
, &
Scarborough
,
D. L.
(
1989
).
Language-specific access of homographs by bilinguals.
Journal of Experimental Psychology: Learning, Memory, and Cognition
,
15
,
305
315
.
Grainger
,
J.
, &
Dijsktra
,
T.
(
1992
).
On the representation and use of language information in bilinguals.
In R. J. Harris (Ed.),
Cognitive processing in bilinguals
(pp.
207
220
).
Amsterdam
:
North-Holland
.
Green
,
D. W.
(
1998
).
Mental control of the bilingual lexico-semantic system.
Bilingualism: Language and Cognition
,
1
,
67
81
.
Greenhouse
,
S.
, &
Geisser
,
S.
(
1959
).
On methods in the analysis of profile data.
Psychometrika
,
24
,
95
112
.
Grosjean
,
F.
(
2001
).
The bilingual's language modes.
In J. Nicol (Ed.),
One mind, two languages: Bilingual language processing
(pp.
1
22
).
Malden, MA
:
Blackwell Publishers
.
Heinze
,
H.-J.
,
Münte
,
T. F.
, &
Mangun
,
G. R.
(
1994
).
Cognitive electrophysiology.
Basel
:
Birkhäuser
.
Hernandez
,
A.
,
Dapretto
,
M.
,
Mazziotta
,
J.
, &
Bookheimer
,
S.
(
2001
).
Language switching and language representation in Spanish–English bilinguals: An fMRI study.
Neuroimage
,
14
,
510
520
.
Holender
,
D.
(
1986
).
Semantic activation without conscious identification in dichotic listening, parafoveal vision, and visual making: A survey and appraisal.
Behavioral and Brain Sciences
,
9
,
1
66
.
Hughes
,
R. W.
,
Tremblay
,
S.
, &
Jones
,
D. M.
(
2005
).
Disruption by speech of serial short-term memory: The role of changing-state vowels.
Psychonomic Bulletin & Review
,
12
,
886
890
.
Jones
,
D.
(
1995
).
The fate of the unattended stimulus: Irrelevant speech and cognition.
Applied Cognitive Psychology
,
9
,
S23
S38
.
Jones
,
D. M.
, &
Macken
,
W. J.
(
1995
).
Phonological similarity in the irrelevant speech effect: Within-or between-stream similarity?
Journal of Experimental Psychology: Learning, Memory, and Cognition
,
21
,
103
115
.
Jones
,
D. M.
,
Miles
,
C.
, &
Page
,
J.
(
1990
).
Disruption of proofreading by irrelevant speech: Effects of attention, arousal or memory?
Applied Cognitive Psychology
,
4
,
89
108
.
Kotz
,
S. A.
, &
Elston-Güttler
,
K.
(
2004
).
The role of proficiency on processing categorical and associative information in the L2 as revealed by reaction times and event-related brain potentials.
Journal of Neurolinguistics
,
17
,
215
235
.
LeCompte
,
D. C.
,
Neely
,
C. B.
, &
Wilson
,
J. R.
(
1997
).
Irrelevant speech and irrelevant tones: The relative importance of speech to the irrelevant speech effect.
Journal of Experimental Psychology: Learning, Memory, and Cognition
,
23
,
472
483
.
Martin
,
R. C.
,
Wogalter
,
M. S.
, &
Forlando
,
J. G.
(
1988
).
Reading comprehension in the presence of unattended speech and music.
Journal of Memory and Language
,
27
,
382
398
.
Neely
,
C. B.
, &
LeCompte
,
D. C.
(
1999
).
The importance of semantic similarity to the irrelevant speech effect.
Memory & Cognition
,
27
,
37
44
.
Oswald
,
C. J. P.
,
Tremblay
,
S.
, &
Jones
,
D. M.
(
2000
).
Disruption of comprehension by the meaning of irrelevant sound.
Memory
,
8
,
345
350
.
Paulmann
,
S.
,
Elston-Güttler
,
K. E.
,
Gunter
,
T. C.
, &
Kotz
,
S. A.
(
2006
).
Is bilingual lexical access influenced by language context?
NeuroReport
,
17
,
727
731
.
Price
,
C. J.
,
Green
,
D. W.
, &
Von Studnitz
,
R.
(
1999
).
A functional imaging study of translation and language switching.
Brain
,
122
,
2221
2235
.
Proverbio
,
A. M.
,
Leoni
,
G.
, &
Zani
,
A.
(
2004
).
Language switching mechanisms in simultaneous interpreters: An ERP study.
Neuropsychologia
,
42
,
1636
1656
.
Rodriguez-Fornells
,
A.
,
Schmitt
,
B. M.
,
Kutas
,
M.
, &
Münte
,
T.
(
2002
).
Electrophysiological estimates of the time course of semantic and phonological encoding during listening and naming.
Neuropsychologia
,
40
,
778
787
.
Salamé
,
P.
, &
Baddeley
,
A. D.
(
1982
).
Disruption of short-term memory by unattended speech: Implications for the structure of working memory.
Journal of Verbal Learning and Verbal Behavior
,
21
,
150
164
.
Salamé
,
P.
, &
Baddeley
,
A. D.
(
1989
).
Effects of background music on phonological short term memory.
Quarterly Journal of Experimental Psychology: Human Experimental Psychology
,
41
,
107
122
.
Scarborough
,
D. L.
,
Gerard
,
L.
, &
Cortese
,
C.
(
1984
).
Independence of lexical access in bilingual word recognition.
Journal of Verbal Learning and Verbal Behavior
,
23
,
84
99
.
Smith
,
M. C.
, &
Kirsner
,
K.
(
1982
).
Language and orthography as irrelevant features in colour–word and picture–word Stroop interference.
Quarterly Journal of Experimental Psychology, Section A: Human Experimental Psychology
,
34A
,
153
170
.
Van den Brink
,
D.
,
Brown
,
C. M.
, &
Hagoort
,
P.
(
2001
).
Electrophysiological evidence for early contextual influences during spoken-word recognition: N200 versus N400.
Journal of Cognitive Neuroscience
,
13
,
967
985
.
Van Hell
,
J. G.
, &
Dijkstra
,
T.
(
2002
).
Foreign language knowledge can influence native language performance in exclusively native contexts.
Psychonomic Bulletin & Review
,
9
,
780
789
.
Van Heste
,
T.
(
1999
).
Visual word recognition in bilinguals: Selective and nonselective activation processes as a function of the experimental task.
Unpublished Master's Thesis, University of Leuven.
Van Heuven
,
W. J. B.
,
Dijkstra
,
T.
, &
Grainger
,
J.
(
1998
).
Orthographic neighborhood effects in bilingual word recognition.
Journal of Memory and Language
,
39
,
458
483
.
Vasey
,
M. W.
, &
Thayer
,
J. F.
(
1987
).
The continuing problem of false positives in repeated measures ANOVA in psychophysiology: A multivariate solution.
Psychophysiology
,
4
,
479
486
.
Venetjoki
,
N.
,
Kaarlela-Tuomaala
,
A.
,
Keskinen
,
E.
, &
Hongisto
,
V.
(
2006
).
The effect of speech and speech intelligibility on task performance.
Ergonomics
,
49
,
1068
1091
.
Von Studnitz
,
R. E.
, &
Green
,
D. W.
(
2002
).
Interlingual homograph interference in German-English bilinguals: Its modulation and locus of control.
Bilingualism: Language and Cognition
,
5
,
1
23
.
Williams
,
J. N.
(
1994
).
The relationship between word meanings in the first and second language: Evidence for a common, but restricted, semantic code.
European Journal of Cognitive Psychology
,
6
,
195
220
.