Edited by: Esther Ruigendijk, University of Oldenburg, Germany
Reviewed by: Francesca Pesciarelli, University of Modena and Reggio Emilia, Italy; Carolina Gattei, Universidad Torcuato Di Tella, Argentina
This article was submitted to Language Sciences, a section of the journal Frontiers in Psychology
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
Prior studies using the event-related potential (ERP) technique show that integrating sentential code-switches during online processing leads to a broadly distributed late positivity component (LPC), while processing semantically unexpected continuations instead leads to the emergence of an N400 effect. While the N400 is generally assumed to index lexico-semantic processing, the LPC has two different interpretations. One account suggests that it reflects the processing of an improbable or unexpected event, while an alternative account proposes sentence-level reanalysis. To investigate the relative costs of semantic to language-based unexpectancies (i.e., code-switches), the current study tests 24 Spanish-English bilinguals in an ERP reading study. Semantically constrained Spanish frames either varied in their semantic expectancy (high vs. low expectancy) and/or their language continuation (same-language vs. code-switch) while participants’ electrophysiological responses were recorded. The Spanish-to-English switch direction provides a more naturalistic test for integration costs to code-switching as it better approximates the code-switching practices of the target population. Analyses across three time windows show a main effect for semantic expectancy in the N400 time window and a main effect for code-switching in the LPC time window. Additional analyses based on the self-reported code-switching experience of the participants suggest an early positivity linked to less experience with code-switching. The results highlight that not all code-switches lead to similar integration costs and that prior experience with code-switching is an important additional factor that modulates online processing.
Over the last decade, interest in the psycholinguistic processes underlying the integration of code-switched speech, defined as the fluid alternation of both languages within the same conversation or in text (
Critically, linguists have demonstrated that code-switching is not a random or chaotic process, and that instead it is systematic and constrained (
Due to the heterogeneity of bilingual acquisition, proficiency in the component languages is one such individual-level factor. Higher proficiency bilinguals are more likely to engage in intra-sentential code-switches, whereas lower proficiency bilinguals are more restricted to inter-sentential and single-word code-switches (
The current psycholinguistic studies of code-switching highlight three broad themes of study: (1) Its relationship to other switching phenomena such as cued-language switching (e.g.,
At the neurocognitive level,
Building from this seminal work, Van Hell and colleagues have continued to explore the individual-level factors that may contribute to the emergence of the N400 and the LPC as they relate to the processing of code-switches.
Following these important lines of research, the study reported here extends the paradigm first reported in
If the LPC is linked to the processing of code-switches more generally, we should replicate
If the LPC, however, is linked to improbability, we should not find the LPC for the semantically-expected code-switched targets because the direction of the code-switch in our materials respects linguistic ecology (i.e., switches are from Spanish into English).
Additionally, if the added semantic unexpectancy adds difficulty to semantic integration for the bilingual participants, we should find an N400 associated with the integration of the semantically-unexpected code-switched target.
Finally, we will use self-reports on code-switching experience to investigate whether we find modulation of an early frontal positivity in our bilingual sample.
Twenty-four Spanish-English highly proficient bilinguals (17 female; mean age = 23.83;
One hundred and sixty sentences constituted the materials in the reading task. All sentence contexts were semantically constrained and represented four conditions: (1) a sentence with a semantically expected same-language target word (same-language, expected continuation); (2) a sentence with a semantically unexpected but plausible same-language target word (same-language, unexpected continuation); (3) a sentence with the English translation of the semantically expected target word (code-switched, expected continuation); and (4) a sentence with the English translation of the semantically unexpected target word (code-switched, unexpected continuation). Sample stimuli are provided in
Example stimuli.
Same-language (Spanish) continuation | Code-switched continuation | |
---|---|---|
Highly expected target | Los jóvenes se reunieron para ver el partido y apoyar al equipo. |
Los jóvenes se reunieron para ver el partido y apoyar al team. |
Low expected target | Los jóvenes se reunieron para ver el partido y apoyar al entrenador. |
Los jóvenes se reunieron para ver el partido y apoyar al coach. |
Frequency (log frequency from NIM database,
Frequency and length values for critical nouns.
Frequency | Length | ||||
---|---|---|---|---|---|
High cloze | Spanish | 1.58 | 0.60 | 6.70 | 2.10 |
English | 1.44 | 0.69 | 5.80 | 2.31 | |
Low cloze | Spanish | 1.31 | 0.57 | 6.45 | 1.92 |
English | 1.49 | 0.59 | 6.48 | 2.42 |
Sentence completion norms were collected using a cloze procedure in order to verify the semantic constraint of the experimental materials. To this end, the 160 experimental sentences were truncated immediately before the target word. Twenty-six Spanish-English bilinguals who did not take part in the main study (15 females; mean age = 36.64,
Each of the four versions of a sentence was randomly assigned to one of four experimental lists. The lists were comprised of 40 sentences per condition for a total of 160 experimental trials. Participants read only one of the lists in the procedure. Twenty “catch trial” sentences (10 unilingual Spanish, 10 with a single-word insertion from English) describing horse-related content were included to each list to ensure that participants were paying attention to the task (see “Procedure”). Samples sentences are provided below:
Catch trial-unilingual:
Catch trial-code-switched:
Participants were seated in the recording chamber at a distance of 3.5 ft away from a 19-in LCD monitor (60 Hz refresh rate) enclosed in a Faraday cage to minimize electrical noise (
The electrophysiological activity of the brain (EEG) was recorded during the experimental task from 32 electrodes mounted on an elastic cap. The location of the electrodes was based on the 10/20 International System (
EEG data were processed with Edit 4.3 software (Compumedics, El Paso, TX). The processing of the data consisted of the following steps: (1) visual inspection of the continuous signal and rejection of sections with artifacts, (2) eve-movement corrections, employing the spatial filter transform implemented in Edit 4.3 (
After visual inspection of the epochs and following prior studies on the processing of code-switches (
Participants responded correctly to 95.51% of catch trials (
Neither the main effects of Language or Expectancy nor the interactions reached significance in this early time window (all
The ANOVA on the mean amplitudes corresponding to this time window showed a main effect of Expectancy [
The Expectancy effect in Fz, Cz, and Pz electrodes. Boxes indicate time windows included in the analysis and dark outlines the regions where the effect is statistically significant.
A late positivity arose in response to language continuation but not Expectancy as revealed by a main effect of Language [
Switched vs. non-switched conditions in Fz, Cz, and Pz electrodes. Boxes indicate time windows included in the analysis and dark outlines the regions where the effect is statistically significant.
Despite the lack of a significant Expectancy × Language interaction in the ANOVAs above, we directly compared the combination of semantic unexpectancy and code-switches to highly expected same-language continuations. We compared the mean amplitudes of the baseline condition (same-language, expected continuation) to those belonging to our new critical condition (code-switched, unexpected continuation) in the N400 and LPC time windows. Because the results show evidence for an N400 for the Expectancy manipulation and an LPC for the Language manipulation, we expected to observe a combination of an N400 and an LPC to the critical condition as a result of the combination of the two forms of unexpectancy relative to the baseline condition. However, the planned comparisons of the two conditions only unveiled significant brain response differences in the time window corresponding to the LPC, between 500 and 700 ms post noun (main effect
The baseline (non-switched, highly expected continuation) vs. the double unexpectancy (code-switched and semantically unexpected continuation) in Fz, Cz, and Pz electrodes. Boxes indicate time windows included in the analysis and dark outlines the regions where the effect is statistically significant.
To test whether prior code-switching experience affected the emergence of an early positivity between 250 and 350 ms, we split the sample into two subgroups based on their self-reported code-switching experience. Although the number of participants is limited, the subgrouping led to a group who reported to code-switch “often” (
In this study, we sought to replicate and extend one of the first ERP studies on the online processing of code-switching by
Our findings broadly replicate the original
Moreover, we report a broadly distributed and extended LPC for the code-switched conditions. The LPC has now been documented in several studies on code-switching (e.g.,
Although our results revealed a strong N400 effect for unexpected continuations, the direct planned comparison between our baseline condition (same-language, expected continuation) and the code-switched, unexpected continuation, only evinced an LPC effect and not an N400 effect. The lack of an N400 effect may be due to statistical power, especially since the bilingual sample did show some variation in their own code-switching experience or may be due to the conflation of a possible N400 effect, as visually suggested in
Following recent results suggesting that an early positivity component may be tied to prior experience with code-switching (
The current study used the ERP technique to directly compare two forms of unexpectancy: semantic unexpectancy with language-based unexpectancy. The results complement the now emerging picture from the nascent literature on the neurocognitive processes involved in the online processing of code-switching. Code-switches broadly elicit an LPC even when they match the code-switching patterns found in the targeted community of speakers. This late positivity suggests that the successful integration of code-switches requires sentence-level reanalysis. At the same time, additional factors, such as semantic expectancy and individual differences in exposure to code-switching, may affect the presence of additional neurocognitive processes. These findings suggest that not all sentential code-switches are processed with similar integration costs. Likewise, not all bilinguals experience similar integration costs. While these initial results require further replication, they point toward the increasing need to incorporate bilingual experience into experimental work on code-switching.
The datasets generated for this study are available at the Open Science Framework repository located at https://osf.io/py78j/.
The studies involving human participants were reviewed and approved by The Pennsylvania State University Institutional Review Board. The participants provided their written informed consent to participate in this study.
JV, PR, and PD contributed to the conception and design of the study. JV, PR, and PD were involved in the implementation of the study. PR was involved in data collection and data preprocessing and statistical analysis. JV and PR wrote the first draft of the manuscript. All authors contributed to manuscript revision, and read and approved the submitted version.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
We thank Javier Solivan and Maya Waide for their help during data acquisition and the audience of the
1
2
3