The elusive link between language control and executive control: A case of limited transfer

Abstract

We investigated the relationship between language control and executive control by testing three groups of bilinguals (104 participants) and 54 monolinguals in a training and transfer paradigm. Participants practised either a language or a non-linguistic colour/shape switching task and were tested one week later on both tasks. The colour/shape task produced significant immediate improvement with training, which was maintained one week later, but exhibited no cross-task transfer effects. In the dominant language, training effects did not persist after one week, and there were no transfer effects. In the non-dominant language there were significant training effects that lasted one week, and there was also transfer facilitation from prior practice with the colour/shape task, which was limited to a reduction in mixing costs. Despite limited transfer, there were significant correlations between tasks in mixing costs for bilinguals, in switching costs for monolinguals, and in intrusion errors for all participants. Finally, the pattern of costs observed for the two tasks exhibited both similarities and differences across participants. These results imply a limited but significant role for executive control in bilingual language control, possibly playing a stronger role in facilitating non-dominant-language production and in supporting the ability to monitor response outcomes to avoid errors.

Keywords:

• Bilingualism
• Cognitive control
• Task switching
• Language switching

This research was supported by EU-FP7 Grant IRG-249163 to Anat Prior, and by R01s from NICHD (HD050287 to Tamar Gollan and HD051030 to Vic Ferreira) and NIDCD (R01 DC011492 to Tamar Gollan).

The authors thank Adva Sharabi, Samer Andrea, Mayra Murillo and Tiffany Ho for diligent data collection and coding, and Andrei Markus for programming assistance. The authors also thank Iring Koch, Jared Linck and an anonymous reviewer for comments on an earlier draft.

Notes

¹ Note that exclusion of outliers did not change these patterns substantially. When we eliminated the highest and lowest observation for each cost type for each participant group, the correlation in mixing costs for the entire sample was reduced somewhat but remained significant (r(152)=.24, p<.01). A comparison of the two language groups again yielded a significant correlation for bilinguals (r(101)=.31, p<.01) but not for monolinguals (r(51)=.03, p=.81). The general correlation in switching costs was also somewhat reduced (r(150)=.12, p=.13) and was no longer significant in the entire sample. However, after separating the two language groups, the correlation remained significant for the monolinguals (r(50)=.36, p<.01) but not for the bilinguals (r(100)=−.03, p=.75. Eliminating the outliers in the analysis of intrusion errors again somewhat reduced the magnitude of the correlation (r(150)=.24, p<.01), but it remained significant.

² The present study allowed us to probe whether the previously reported switching-cost advantage in non-linguistic switching for bilinguals (Prior & MacWhinney, 2010; Prior & Gollan, 2011) was replicated in the current data set, because it included the same participant groups as tested by Prior and Gollan (2011).In our previous study (Prior & Gollan, 2011), Spanish-English, but not Mandarin-English, bilinguals exhibited significantly smaller switching costs than monolinguals in the colour/shape task, and the advantage was present only after controlling for between-group differences in socio-economic status (SES) (parent education level). In addition, Spanish-English bilinguals exhibited significantly smaller switching costs than Mandarin-English bilinguals in both linguistic and non-linguistic tasks, a finding we linked to the fact that Spanish-English bilinguals reported switching languages more often in daily life. In the current data set, the Spanish-English bilinguals reported switching languages about as often as Spanish-English bilinguals in the 2011 paper (3.4, see Table 1 in this paper, versus 3.2 on the same five-point scale in Prior & Gollan, 2011).To replicate the analysis reported in Prior & Gollan (2011), we looked at RTs from the Training1 and Transfer sequences and calculated the relative switching cost for each individual in the colour/shape task, and also corrected for group differences in SES by including parental education as a covariate in the analysis. To that end we divided the switching cost (the difference between switching and repeat times) by the mean RT on repeat trials. This relative switching cost was calculated to correct for the slower response times of the bilinguals. In an ANCOVA on the relative switching cost, with primary caregiver education entered as a covariate, we did not find an effect of language group (F<1).A key consideration in interpreting this failure to replicate is that in the current study bilinguals produced vocal responses in the colour/shape task, whereas previous studies relied on manual responses. Well-documented bilingual disadvantages in verbal responses might have masked switching-cost advantages in the current data, and indeed the Spanish-English and Mandarin-English bilinguals were significantly slower than monolinguals in overall response speed in the colour/shape task (p<.001).