Interlocutor identity affects language activation in bilinguals

https://doi.org/10.1016/j.jml.2015.01.002Get rights and content

Highlights

  • We familiarized early and late bilinguals with interlocutors speaking different languages.

  • In an audio–visual lexical decision task items were produced by the interlocutors.

  • The items were either congruent or incongruent with the interlocutors’ familiarization language.

  • During incongruent trials, the response latencies of early bilinguals increased.

  • Language activation of early bilinguals are modulated by interlocutor identity.

Abstract

In bilingual communities, individuals often communicate in one of their languages only, and they adjust to the linguistic background of different interlocutors with ease. What facilitates such efficiency? We investigated whether bilinguals’ language activation is supported by non-linguistic cues (e.g., interlocutor identity). First, in an audio–visual task, early (proficient) and late (less proficient) Basque–Spanish bilinguals were familiarized with six novel interlocutors who spoke either Spanish, Basque, or both languages. Then, participants completed an audio–visual lexical decision task, in which the interlocutors produced test items in Spanish or Basque. Early, but not late, bilinguals’ speed of processing decreased when the language that the interlocutors spoke during familiarization matched the language they spoke at test, relative to test trials when the interlocutors changed languages. Overall, results suggest that proficient and/or early bilinguals benefit from an association between language and interlocutor during (or even before) language comprehension, because they are able to predict the context-appropriate language based on non-linguistic cues, such as interlocutor context.

Introduction

Bilinguals from regions with two established languages are regularly exposed to both single and dual language interactional contexts. These interactional contexts (or language modes) are often defined by the interlocutor (e.g., Green and Abutalebi, 2013, Grosjean, 1998). For instance, a Basque–Spanish bilingual might communicate in Basque only with certain friends, but use both Basque and Spanish with their sibling. Here, we investigate whether the bilingual mind, during spoken language comprehension, actually adapts to interactional contexts based on prior knowledge about interlocutors. Specifically, we assess whether there is any bias in bilinguals’ language co-activation depending on interlocutor context. Then, we consider the consequences of such activation patterns for models of bilingual lexical access and language activation.

Proficient bilinguals’ language activation patterns are generally characterized as non-selective. That is, both languages of the bilinguals are active and processing is open to cross-language interaction during comprehension and production (for review, see Kroll, Bobb, & Wodniecka, 2006). Although the dual language system remains sensitive to language-specific cues, and bias toward one of the languages is possible at the conceptual, lexical, and phonological levels (e.g., Costa, 2005, Dijkstra, 2005, Elston-Güttler et al., 2005, Green, 1998, Libben and Titone, 2009, Schwartz and Areas Da Luz Fontes, 2008, Van Hell and De Groot, 2008, Weber and Cutler, 2004). Therefore, multiple loci for language selection are available (Kroll et al., 2006). Indeed, most models of bilingual language processing assume the operation of a language cue that alters co-activation at some point during comprehension or production. This has been primarily demonstrated using explicit linguistic cues (e.g., sentence context, lexical items, etc.; e.g., Libben and Titone, 2009, Schwartz and Kroll, 2006). However, ecologically-valid situations of language processing include a range of non-linguistic (contextual) information (e.g., identity of interlocutors, situations, world knowledge, etc.). Monolingual speech comprehension is greatly supported by such cues (e.g., Grosjean and Li, 2013, Strand, 1999), and it is reasonable to assume that bilingual comprehension too is facilitated by similar domain-general mechanisms. Thus far, no clear and systematic distinction of language cues exist in the literature of bilingual language processing (e.g., explicitly linguistic vs. non-linguistic language cues), particularly in bilingual spoken language comprehension.

The phenomenon that non-linguistic information can interact with activation patterns has been recently demonstrated with respect to emotions. Emotions conveyed by words in a second language bias the mechanisms of co-activation during written language processing: words with negative valence, but not words with positive or neutral valence, inhibit access to the native language (Wu & Thierry, 2012). In addition to emotions, culturally biased icons/images, recognized as non-linguistic cues, also alter co-activation during speech production (Jared et al., 2013, Li et al., 2013, Zhang et al., 2013). When Chinese–English bilinguals were asked to engage in a conversation via a computer program while viewing an image of either an Asian or a Caucasian face, they spoke English less fluently when a Chinese and not a Caucasian face was presented (Zhang et al., 2013). Similar results were obtained in a picture-naming paradigm. When the naming language (Chinese vs. English) and the cultural cue (images of Asian vs. Caucasian faces) presented alongside the object to be named were congruent, a facilitation effect was observed (Li et al., 2013). Therefore, still images of faces may function as cultural primes and affect bilingual speech production. In these studies, the participants’ only information about the faces was their socio-cultural identity, and participants had no direct experience with the language the person appearing in the images actually spoke. Considering real-world situations in the Unites States where the research took place, it is common that interlocutors of Asian heritage are native speakers of English (and do not speak Chinese); and vice versa, Caucasian interlocutors might speak little or no English. Hence, the faces presented in these studies could be considered less as real interlocutors and more as cultural icons or symbols associated with language communities.

No studies to date have investigated whether familiarity with the language background of interlocutors interact with bilingual language activation during comprehension. Thus far, only anecdotal evidence is available in support of this hypothesis. Bilinguals report that they develop an association between a language mode and specific interlocutors, and they often expect one specific language to be used with certain interlocutors (Grosjean, 1998). If bilinguals indeed develop associations between language modes and interlocutors or interactional contexts, it is a possibility that these associations, at some point during processing, shape language co-activation patterns (e.g., Green & Abutalebi, 2013). Such a finding would indicate that higher-level contextual information (interlocutor identity) interacts with the mechanisms of bilingual language activation and/or selection. Similar mechanisms that rely on higher-level contextual information have been proposed to support domain-general cognitive processing, including speech perception abilities in monolinguals (e.g., Elman and McClelland, 1988, Magnuson et al., 2008).

Yet, the role of non-linguistic factors (e.g., interlocutor identity) during the course of bilingual speech comprehension remains elusive. Most models of bilingual language control lack a clear account of how, when, or to what degree non-linguistic context might interact with co-activation patterns (for reviews, see Dijkstra, 2005, Dijkstra and van Heuven, 2002, Kroll et al., 2010). When considering comprehension, the models tend support the idea that activation is initially non-selective, because listeners have no control over which language is going to be used. Then, following the onset of the linguistic input, a bottom-up bias for the target language occurs, while cross-language representations still interact during (auditory) lexical access (e.g., Costa, 2005, Costa et al., 2006, Dijkstra, 2005, Green, 1998, Kroll and Dussias, 2004, Spivey and Marian, 1999; but see FitzPatrick and Indefrey, 2010, Rodriguez-Fornells et al., 2002). Whilst when considering spoken language comprehension, the bottom-up bias can occur rapidly, based only on a small amount of low-level auditory linguistic information. For instance, when word-initial stop consonants of second language (L2) words were pronounced with first language (L1) appropriate voice onset times, initial language activation of bilingual listeners were biased toward the L1 (Ju & Luce, 2004).

The more recent adaptive control hypothesis (Green & Abutalebi, 2013), however, considers bilingual language processing from broader perspective than the previous models. It proposes that bilinguals are sensitive to specific interactional contexts, such as single-language context (e.g., when only one language is used in the environment), dual-language context (e.g., when both languages are present in the environment but spoken by different interlocutors), and dense code-switching context (e.g., when interlocutors regularly interleave their languages within the same phrase or utterance). Bilinguals are able to adapt to these contexts via different cognitive control processes involved in language selection. Presumably, these cognitive control processes can exploit non-linguistic cues, such as voice, face, or gesture. For instance, in a dual-language context, different interlocutors can be associated with different languages or interactional contexts. When a bilingual addresses, or is addressed by, a new interlocutor, it might elicit a change in the bilingual’s language activation, depending on the interactional context linked to the given interlocutor. According to the adaptive control hypothesis, in this case, the cognitive process involved in language mode adjustment is salient cue detection, because an interlocutor might function as a salient (non-linguistic) cue aiding the bilinguals’ language control mechanisms (Green & Abutalebi, 2013). Whether bilingual listeners rely on interlocutors as salient cues during comprehension, and interlocutors in fact bias language co-activation is the focus of the current study.

The goal of the current study is to investigate the interaction between bilingual language co-activation and non-linguistic information during spoken language comprehension. Specifically, the aim is to test the effect of interlocutor identity on language activation during audio–visual lexical processing in bilinguals. In contrast to previous studies on bilingual auditory comprehension, the current research investigates this issue from a novel perspective because: it (1) evaluates whether a non-linguistic cue (such as an interlocutor) affects language regulation; (2) due to the comprehension nature of the task, it measures the state of language activation at the time (and not after) a linguistic stimulus is processed; (3) tests bilinguals of two languages (Spanish and Basque) that rely on almost identical phonetic inventories, meaning that bottom-up processes related to acoustic–phonetic information which might facilitate language activation are minimized.

To investigate these issues, we designed a task that included audio–visual presentation of interlocutors. Studying language processing in naturalistic settings, which include such real-life factors, is necessary to understand how the human mind functions in natural situations (Trueswell & Tanenhaus, 2005). In this task, first, we familiarized bilingual participants via video segments featuring six same-race (Caucasian) and same-gender (female) interlocutors who were associated with either Basque or Spanish (monolingual interlocutors), or with both languages (bilingual interlocutors). Next, participants completed an audio–visual lexical decision task. In this task, participants were presented with video segments in which the same interlocutors appeared producing words or pseudo-words. Occasionally, however, the monolingual interlocutors produced test items in the other language than they used in the familiarization phase (incongruent trials). Incongruent trials were produced with the interlocutors’ original voices, hence only the language of word changed across the congruent and incongruent trials (language-change condition). Therefore, comparisons of participants’ responses to trials in which the monolingual interlocutor used the predictable language (congruent trials) with the incongruent trials will indicate whether the identity of the interlocutor (as a contextual prime) modulates language co-activation. If language co-activation is constrained by interlocutor identity, we expect longer response latencies for incongruent trials as compared to congruent (or predictable) trials.

Besides measuring the congruency effect by directly contrasting responses to congruent and incongruent trials, we also considered the relationship between two types of consecutive trials. First, by comparing the responses to trials that were preceded by same-language trials with trials that were preceded by different-language trials (switch trials), we could measure whether the classic switch cost effect is present in the current comprehension paradigm: switch trials, compared to non-switch trials, are normally processed slower because of changes in language activation (e.g., in production: Meuter & Allport, 1999; in comprehension: Abutalebi et al., 2007). More relevant to the goal of the current study, is that it is also possible to measure whether there is a switch cost associated with changes in interlocutors (e.g., change from a Basque to a Spanish interlocutor) while the language of the trials remain the same (e.g., two consecutive Basque trials). If responses are slower to those same-language trials that are presented by a language inappropriate interlocutor (incongruent trials) as compared to same-language trials presented by a language appropriate interlocutor, then presumably interlocutor identity alters language co-activation (independently of the actual language of the trials), in a similar way to the classic switch cost effect.

Moreover, we measured responses to bilingual interlocutors (bilingual-interlocutor condition). As previously described, bilingual interlocutors were associated with both languages to the same extent during the familiarization and test phases, therefore they never produced incongruent trials. Whether they produced either Spanish or Basque words, it was congruent with their bilingual identity. However, the language of the upcoming congruent items produced by the bilingual interlocutors was unpredictable, unlike the predictable congruent trials produced by the monolingual interlocutors. For this reason, we propose that during responses to bilingual interlocutors, the highly proficient bilinguals’ two languages should be fully prepared or co-activated, because it is impossible to predict what language the interlocutor is going use. If this is the case, then response latencies to bilingual interlocutors, irrespective of the language of items, should be longer than to monolingual congruent trials.

Additionally, a voice-change condition was embedded in the experiment. In a subset of the congruent trials, the original voices of interlocutors were replaced with the voices of other interlocutors. If participants gained familiarity with the interlocutors’ identity in general (e.g., talker-specific voice information; Nygaard & Pisoni, 1998), then different responses were predicted for original-voice and voice-changed trials.

Two experiments were conducted. In Experiment 1, Basque–Spanish bilinguals, who reported relatively high proficiency levels across their languages, were assessed. In Experiment 2, bilinguals who acquired Spanish as their first language, learnt Basque in school-settings, and reported lower proficiency in Basque than participants in Experiment 1, were tested. Two different bilingual populations were recruited because it has been demonstrated that parallel language activation is constrained by language proficiency. High proficiency boosts parallel language activation, whereas lower proficiency in one of the languages generates asymmetrical language activation patterns (e.g., Blumenfeld and Marian, 2007, Brysbaert et al., 2000, Costa and Santesteban, 2004, Grosjean, 1998, Weber and Cutler, 2004). Consequently, if participants in Experiment 1 but not in Experiment 2 show interlocutor identity related congruency effects, then the observed effect is likely related to the bilinguals’ mental control over parallel language activation. If participants in both experiments show a significant congruency effect, then the observed pattern might be more due to association effects, and less related to language co-activation.

If participants in Experiment 1 and 2 represent two different groups of bilinguals in terms of their language co-activation abilities, it should be reflected in their responses in the bilingual-interlocutor condition as well. For instance, more proficient bilinguals are expected to display relatively equal levels of activation across Spanish and Basque words produced by bilingual interlocutors, as compared to the less proficient bilinguals, who are expected to exhibit higher activation in response to Spanish than Basque. In the voice-change condition, however, we predict that participants across experiments would respond similarly. Similar patterns (for instance slower responses to interlocutors with replaced voices) would indicate that participants were able to encode indexical information (not related to language type) related to interlocutors, hence both groups gained familiarity with the interlocutors to the same degree.

Section snippets

Participants

Thirty-one highly proficient bilingual participants were recruited from the province of Gipuzkoa in the Basque Country, a Basque–Spanish speaking region in Northern Spain. Data were excluded from 1 participant due to experimental error, and from 3 participants because of low level knowledge of Basque. The final sample of Experiment 1 consisted of 27 participants (mean age = 23.8 years).

Participants’ language proficiency was evaluated using a detailed bilingual language background questionnaire

Participants

Thirty low-proficient bilingual participants were recruited from the same region as the participants in Experiment 1. Data were excluded from 1 participant due to experimental error, and from 2 participants due to no knowledge of Basque. The final sample of Experiment 2 consisted of 27 participants (mean age = 23.08 years).

Language background was assessed using the same questionnaire as in Experiment 1. Participants in Experiment 2 included native speakers of Spanish who started to learn Basque in

Congruency effect

First, as in Experiment 1 & 2, average RT values were subjected to an ANOVA based on congruency status (congruent-predictable, incongruent, and congruent-unpredictable) and the language of item (Spanish vs. Basque) with groups as a within-subject factor (early/proficient vs. late/less proficient groups). This test produced a significant main effect of congruency (F1(2, 51) = 37.4, p = .001; F2(2, 297) = 164, p = .001) and language (F1(1, 53) = 37.4, p = .001; F2(1, 298) = 164, p = .001), in addition to a

Acknowledgments

This research has been partially supported by the CONSOLIDER-INGENIO2010 CSD2008-00048 Grant, and the ERC-2011-ADG-295362 grant to Manuel Carreiras. We thank Joana Acha, Leire Arietaleanizbeascoa, Vanessa Gallardo, Eider Gisasola, Joana Izurieta and Eider Juaristi for their help with stimuli production. We also thank Blair Armstrong and Peter Boddy for their help with manuscript preparation. Finally, we are grateful for the insightful comments provided by Albert Costa, Gerrit Jan Kootstra, and

References (53)

  • J.G. Van Hell et al.

    Sentence context modulates visual word recognition and translation in bilinguals

    Acta Psychologica

    (2008)
  • A. Weber et al.

    Lexical competition in non-native spoken-word recognition

    Journal of Memory and Language

    (2004)
  • J. Abutalebi et al.

    The neural cost of the auditory perception of language switches: An event-related functional magnetic resonance imaging study in bilinguals

    The Journal of Neuroscience

    (2007)
  • E.M. Altmann

    Advance preparation in task switching. What work is being done?

    Psychological Science

    (2004)
  • S. Bernolet et al.

    Shared syntactic representations in bilinguals: Evidence for the role of word-order repetition

    Journal of Experimental Psychology: Learning, Memory, and Cognition

    (2007)
  • H.K. Blumenfeld et al.

    Constraints on parallel activation in bilingual spoken language processing: Examining proficiency and lexical status using eye-tracking

    Language and Cognitive Processes

    (2007)
  • M. Brysbaert et al.

    The effects of age-of-acquisition and frequency-of-occurrence in visual word recognition: Further evidence from the Dutch language

    European Journal of Cognitive Psychology

    (2000)
  • C. Chandrasekaran et al.

    The natural statistics of audiovisual speech

    PLoS Computational Biology

    (2009)
  • A. Costa

    Lexical access in bilingual production

  • A. Costa et al.

    The dynamics of bilingual access

    Language and Cognition

    (2006)
  • C.J. Davis et al.

    BuscaPalabras: A program for deriving orthographic and phonological neighborhood statistics and other psycholinguistic indices in Spanish

    Behavior Research Methods

    (2005)
  • G.S. Dell et al.

    The P-chain: Relating sentence production and its disorders to comprehension and acquisition

    Philosophical Transactions of the Royal Society B: Biological Sciences

    (2014)
  • T. Dijkstra

    Bilingual word recognition and lexical access

  • T. Dijkstra et al.

    The architecture of the bilingual word recognition system: From identification to decision

    Bilingualism: Language and Cognition

    (2002)
  • G. Dreisbach et al.

    Preparatory processes in the task-switching paradigm: Evidence from the use of probability cues

    Journal of Experimental Psychology: Learning, Memory, and Cognition

    (2002)
  • I. FitzPatrick et al.

    Lexical competition in nonnative speech comprehension

    Journal of Cognitive Neuroscience

    (2010)
  • Cited by (43)

    • Perro or txakur? Bilingual language choice during production is influenced by personal preferences and external primes

      2022, Cognition
      Citation Excerpt :

      Similarly, other studies have found that bilinguals respond faster when the image of a face or cultural cue matched the target language (e.g., seeing an Asian face when using Chinese, Li, Yang, Scherf, & Li, 2013; Roychoudhuri, Prasad, & Mishra, 2016; Zhang, Morris, Cheng, & Yap, 2013). Furthermore, interlocutor identity does not only affect language production but can also affect how words are perceived in a language previously associated with a specific interlocutor (Martin, Molnar, & Carreiras, 2016; Molnar, Ibáñez-Molina, & Carreiras, 2015). In addition to these non-linguistic primes, the actual language used in a conversation affects language choice.

    • Neuro-dynamics of executive control in bilingual language switching: An MEG study

      2020, Cognition
      Citation Excerpt :

      To ensure that the switch effect we observe in this study more specifically reflects language switching, we used face cues, which are natural and direct language cues. In a real-world setting, the face of the interlocutor should naturally prompt the appropriate response language (Li, Yang, Scherf, & Li, 2013; Molnar, Ibáñez-Molina, & Carreiras, 2015; Woumans et al., 2015). Secondly, most previous language-switching studies involve a confound of cue-switching: if one cue is mapped to each language, whenever there is a language change, the cue also has to change.

    • Co-activation of the L2 during L1 auditory processing: An ERP cross-modal priming study

      2020, Brain and Language
      Citation Excerpt :

      It is by now a recognized phenomenon, however, that bilinguals display concurrent activation of both languages, regardless of whether they are reading, speaking, or listening to one language alone (e.g., Colomé, 2001; Costa, Miozzo, & Caramazza, 1999; Hermans, 2000; Macizo, Bajo, & Martín, 2010; Martín, Macizo, & Bajo, 2010; Spivey & Marian, 1999). Based on these results, recent accounts of bilingualism argue that the cognitive architecture of the bilingual is fundamentally nonselective, but under certain circumstances operates selectively (Kroll, Bobb, & Wodniecka, 2006; Martin, Molnar, & Carreiras, 2016; Molnar, Ibáñez-Molina, & Carreiras, 2015). The question now is no longer whether co-activation can occur, but under what (experimental and language) conditions it does or does not, and what levels of language processing are typically affected.

    View all citing articles on Scopus
    View full text