ABSTRACT
Foreign accents can vary considerably in the degree to which they deviate from the listener’s native accent, but little is known about how the relationship between a speaker’s accent and a listener’s native language phonology mediates adaptation. Using an artificial accent methodology, we addressed this issue by constructing a set of three artificial accents (Near, Far, and Farther), varying in the number and magnitude of pronunciation deviations from standard Canadian English. These accents were presented to toddlers and adults in an eye-tracking task. Regardless of accent type, adults readily adapted to the exposed pronunciation change. Adults exposed to the Farther accent were also more willing to accept novel pronunciation changes. Young toddlers exposed to Far or Farther accents showed no evidence of acquiring the exposed pronunciation change and demonstrated worse word recognition for standard Canadian-accented words. These findings suggest that when a speaker’s accent deviates substantially from a young toddler’s native accent, this may lead to a significant decrement in their ability to recognize not only an unfamiliar accent but also native-accented speech. Overall, these findings provide a well-controlled test of competing models of accent adaptation and generate multiple hypotheses to be examined in the future using more ecologically valid stimuli.
Acknowledgments
We would like to thank Momina Raja, Yazad Bhathena, and Lisa Hotson, as well as the other members of the Child Language and Speech Studies Lab for their assistance in completing this study. We also would like to extend our thanks to the families who participated in this study.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 Although we acknowledge that mispronunciations of known words can be similarly detrimental to speech processing (see, Altvater-Mackensen & Mani, Citation2013; Mani & Plunkett, Citation2008; Paquette-Smith et al., Citation2016; Swingley & Aslin, Citation2002), here we distinguish the work on mispronunciations (which is concerned with online sensitivity to phonemic or sub phonemic deviations) from work on accent adaptation (which involves learning or adapting to the specific phonetic deviations made by a particular speaker).
2 Model structure: lmer(Difference Score ~ Accent Distance + Vowel Shift + (1 | Participant) + (1 | Item)).
3 Given that there was no main effect of Accent Distance in the model, here we report a single t-test to chance (0). If difference scores for exposed target items were computed separately for participants exposed to the Near, Far and Farther accents, participant’s scores would be greater than 0 in all three groups (all ts> 3.80, all ps< .001).
4 To reduce the likelihood of a type I error, the alpha level was Bonferroni adjusted to (α/3 = .017). Comparisons where p <.017 were considered statistically significant.
5 N = 68; The caregivers of four children did not submit their vocabulary forms and were excluded from this ANOVA.
6 Model structure: lmer(Difference Score ~ Accent Distance + Vowel Shift + (1 | Participant) + (1 | Item)).
7 To reduce the likelihood of a type I error, the alpha level was Bonferroni adjusted (α/3 = .017). T-tests where p < .017 were considered statistically significant.
8 Model structure: lmer(Difference Score ~ Accent Distance * Age + (1 | Participant) + (1 | Item)).