Bilingual language switching costs in auditory comprehension

ABSTRACT

Previous research on bilingual language switching and lexical access has demonstrated a consistent reaction time cost associated with producing a switched token. While some studies have shown these costs to be asymmetrical, with bilinguals evidencing a greater delay when producing switches into their dominant language relative to the non-dominant language, others have shown symmetrical costs, depending on individual (e.g. proficiency) and contextual (e.g. language mode) factors. The current study, employing an eye-tracking paradigm, extends this line of research by examining the potential for switch costs during auditory comprehension. Paralleling previous production-oriented research, results of the current study demonstrate flexible switch costs during auditory comprehension. Switch costs were asymmetrical in monolingual mode, with greater costs incurred when switching into the dominant language, and uniformly absent in bilingual mode. Results are discussed with respect to bilingual language selection mechanisms in both production and comprehension.

KEYWORDS:

• Bilingualism
• language switching
• comprehension
• lexical access

Acknowledgements

I am indebted to Zenzi Griffin for feedback on an early version of this project, as well as technical assistance and lab access. In addition, I am grateful to Steven Alcorn and Ariel Sibille for their dedicated efforts on this project. All errors are my own.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1. Language mode has been shown to impact the type and frequency of language switches (e.g. Treffers-Daller, 1998) and phonetic production (e.g. Khattab, 2003, 2009; Olson, 2013, 2015; Simonet, 2014).

2. Worth noting, the potential role of inhibition in language selection is the subject of some debate (for review, see Bobb & Wodniecka, 2013). Others have proposed a more direct access mechanism for balanced bilingual populations (e.g. Costa, 2005; Costa et al., 2006; Finkbeiner, Almeida, Janssen, & Caramazza, 2006; Finkbeiner, Gollan, & Caramazza, 2006; Paradis, 1980, 2004) or a system in which asymmetrical switch costs are motivated by the relative strength of activation perseverance for the L1 and L2 (Verhoef et al., 2009). While an inhibitory framework is adopted for the current discussion, in part due to the focus on L1-dominant bilinguals, it is possible that the results of the current study would find explanation within alternative approaches.

3. It is readily acknowledged that the blocks termed “monolingual mode” are likely not truly monolingual. Many researchers have noted the limitations on creating a monolingual mode in an experimental setting (e.g. Marian & Spivey, 2003). However, paralleling previous research (Olson, 2016; Simonet, 2014), it is assumed that the monolingual modes are relatively more monolingual than the bilingual mode.

4. As noted by a reviewer, in the current design there is a conflation between recency and language mode, with targets in the bilingual mode more recently preceded by a language switch relative to those in the monolingual mode (for evidence for the role of recency, see Blanco-Elorrieta & Pylkkänen, 2016). While the current study was not expressly designed to assess the relative contributions of language mode and recency, the variability in target token placement allowed for a post-hoc analysis. To examine the potential role of recency, in the bilingual mode, tokens were coded with respect to the number of intervening lexical items between the preceding language switch and target token, and grouped as either high recency (2–3 words post-switch) or low recency (4–6 words post-switch). A LME model was conducted on reaction times (log transformed), with response type, target language, and recency as main effects and subject as a random effect with random slopes and intercepts for each of the main effects. Results demonstrated no significant impact of recency on reaction times (β = .009, t = −0.80), and no significant two-way or three-way interactions involving recency (|t| < 2.00). In monolingual mode, distance from the utterance onset was employed as a proxy for distance from the previous switch (high recency: 3–5 word; low recency: 6–11 word). Similar analysis revealed that while there was a significant effect of recency (β = .037, t = 3.32), there were no two- or three-way interactions between recency and any of the other main effects (|t| < 2.00). These results suggest that while recency may play a role in switch cost magnitude, language mode is the key factor in determining the relationship between switch costs in the L1 and L2 (i.e. symmetry vs. asymmetry).

5. Again, here L1 and L2 in the current study are representative of the dominant and non-dominant languages, respectively. Drawing on previous findings (e.g. Schwieter & Sunderman, 2008), proficiency may be a more significant factor than order or age of acquisition.

6. It should be noted that applying inhibition to a given token does not completely eliminate cross-linguistic activation (e.g. Marian & Spivey, 2003).

7. While the BIA+ Model (Dijkstra & van Heuven, 2002), like the original BIA model (Dijkstra & van Heuven, 1998), is rooted in orthographic word comprehension, it represents a valuable point of comparison with the fundamentally production-driven ICM (Green, 1998).

8. It would be plausible to posit that recency may be considered a local effect, while language mode a global effect. Further research should seek to confirm the potential relationship between recency and language mode.

Additional information

Funding

This work was supported by Purdue University under a Research Initiative Grant.