Poster #195 - Children's Attention to Errors in Native and Foreign-Accented Speech

• In Event: 1-115 - Poster Session 03
  In Poster Session: PS 03 Section - Social Cognition

Integrative Statement

Listeners are constantly confronted with speakers varying in age, accent, and fluency. In some cases, listeners might improve their comprehension by learning about the specific patterns of variation they hear. However, when speech differs in numerous ways from expected patterns, or if the variation is not consistent (as is true of some foreign-accented speech), an alternative strategy might be to place less weight on the speech signal itself and more weight on contextual information. To date, there is little evidence regarding whether children and adults adjust their weighting of the speech signal in this way. Recently, adult listeners seemed to demonstrate this type of flexible processing for foreign-accented speakers (Gibson et al., 2017), reinterpreting unlikely sentences produced by foreign-accented speakers (but not native-accented speakers) in a form that made logical sense. This suggests that adults expect foreign-accented speakers to make errors and, therefore, weigh top-down information more heavily. It is unknown whether children develop and use expectations about groups of speakers, such as those with foreign accents, during language processing. One situation in which children’s expectations of foreign-accented speakers may become apparent is when choosing who to learn from. If children expect foreign-accented speakers to produce more speech errors than native-accented speakers, they may assess speech errors differently for these two types of speakers when trying to select a reliable informant.
This study investigates whether 5- and 6-year-old children are more tolerant of grammatical errors in foreign-accented speech than in native speech. In Experiment 1A (N = 32, Mage = 5.97 years), 5- and 6-year-olds listened to pairs of grammatical (e.g., “that is a dog”) and ungrammatical (e.g., “that am a shoe”) native-accented speakers label familiar objects. Following this exposure, children were asked to choose which of the two speakers had correctly labeled a novel object. Children chose labels produced by the grammatical speaker significantly more often than chance (p = .009), suggesting that they use syntactic proficiency as a cue to reliability for native-accented speakers (Sobel & Macris, 2013). Experiment 1B examines children’s expectations and evaluation of speech errors in the context of foreign-accented speech. Five and 6-year-olds (N = 16, Mage = 5.63 years, data collection ongoing), hear the same grammatical and ungrammatical utterances as in Experiment 1A. However, both informants speak with foreign accents (matched for type and strength). Preliminary results suggest that children do not use syntactic proficiency when assessing the reliability of foreign-accented speakers, selecting between informants at chance (p = .432). If this trend continues, it suggests that children might expect foreign-accented speakers to produce grammatical errors and therefore place less weight on these errors when assessing speaker reliability. If so, future work will assess whether children also place less weight on semantic errors produced by foreign-accented speakers in comparison to native-accented speakers. If they do not, it would suggest that children do not simply ignore all types of speech errors produced by accented speakers, but instead are more willing to “forgive” errors that have less effect on the overall meaning of an utterance.

Authors

• Alexandra Rett, University of California San Diego
  Presenting Author

• Katherine S. White, University of Waterloo
  Non-Presenting Author