Compensation for Phonological Assimilation in Bilingual Children

ABSTRACT

We investigate bilingual children’s perception of assimilations, i.e. phonological rules by which a consonant at a word edge adopts a phonological feature of a neighboring consonant. For instance, English has place assimilation (e.g., green is pronounced with a final [m] in green pen), while French has voicing assimilation (e.g., sac is pronounced with a final [g] in sac vert “green bag”). Previous research has shown that French and English monolingual toddlers compensate for the assimilation rule of their language, correctly recovering the intended words, but not for a rule that does not exist in their language. Using a word recognition videogame with French sentences, we show that French-English bilingual 6-year-olds perform exactly like French monolinguals of the same age: they compensate for voicing but not for place assimilation. Thus, despite their dual language input they have acquired French voicing assimilation and show no interference from English place assimilation.

Acknowledgments

We would like to thank all the parents and children that participated, as well as the kindergarten school that kindly allowed us to test monolingual children in their facilities, and Mr. Derek Ferguson from the Roaming School House in Paris for his help with the recruitment of bilingual children. We are also grateful to Alex Cristia for lending us the tablets, and to Anne Christophe for making the recordings. This work was supported by the Agence Nationale pour la Recherche under grants ANR-17-CE28-0007-01 and ANR-17-EURE-0017, and by the Ecole des Neurosciences de Paris Ile-de-France (ENP Graduate Program).

Disclosure statement

No potential conflict of interest was reported by the authors.

Data availability statement

The data described in this article are openly available in the Open Science Framework at http://doi.org/10.5281/zenodo.1205532 .

Notes

¹ They recognize it even less when it is pronounced with final [m] in an unviable context for place assimilation, such as lune rousse “red moon” (rousse starts with a uvular, not a labial, consonant), suggesting a small language-independent compensation effect.

² Although we are not aware of any research showing the mere absence of the application of L2 assimilation rules, we conjecture that bilingual children might also be exposed to L2 French in which neither voicing nor place assimilation applies.

³ These sentences had been recorded by the same speaker concurrently with the sentences used in the experiment.

⁴ In treatment coding, the estimate of the intercept corresponds to the mean of the baseline level, and the estimates for the independent variables correspond to simple effects (as opposed to main effects) of these variables with respect to the baseline (e.g., the estimate for context given voicing viable as baseline corresponds to the difference between voicing viable and voicing unviable trials).

⁵ In order to evaluate a potential effect of trial order, the model originally contained block as an additional fixed effect. This factor did not yield a significant effect (β = − 0.20, SE = 0.22, n.s.), and was therefore excluded from the model reported in the main text. We also checked that the overall model fit was not better when this factor was included (likelihood-ratio test: χ²(1) = 0.87, p = 0.35).

⁶ Numerically, the pattern actually went in the opposite direction, with 28% compensation for voiceless items, e.g., Montre la tasse maintenant ([zm]), and 53% for voiced ones, e.g., Montre la robe là-devant ([pl]).

⁷ Recall that even in our experiment with 6-year olds, the data from one bilingual child was rejected because according to parental report they did not know one of the training items, while for four other bilingual children one test item was excluded from analysis for the same reason.

⁸ The fact that neither we nor Skoruppa et al. (2013) and Skoruppa et al. (2013) found some small amount of compensation for non-native assimilation might be due to a lack of power. Indeed, experiments with toddlers and young children typically have a small number of trials (e.g., our experiment had 6 trials per condition, compared to 16 for adults in Darcy et al., 2009).

⁹ On the one hand, if both the familiar and the novel object were introduced by their names prior to the touching request, 5-year old children (as well as adults) performed the task at an acoustic level; that is, they showed a bias for the novel object. On the other hand, if neither of the objects was named prior to the request, they showed a bias for the familiar object, regardless of context.

Additional information

Funding

We would like to thank all the parents and children that participated, as well as the kindergarten school that kindly allowed us to test monolingual children in their facilities, and Mr. Derek Ferguson from the Roaming School House in Paris for his help with the recruitment of bilingual children. We are also grateful to Alex Cristia for lending us the tablets, and to Anne Christophe for making the recordings. This work was supported by the Agence Nationale pour la Recherche under grants ANR-17-CE28-0007-01 and ANR-17-EURE-0017, and by the Ecole des Neurosciences de Paris Ile-de-France (ENP Graduate Program).