The Role of Consonantal Rhythm in Language Discrimination by German-learning 4-month-olds

• In Event: Poster Session 05
  In Poster Session: PS 05 Topic - Language, Communication

Abstract

Newborns can already tell languages apart (Mehler et al., 1988; Moon et al., 1993), if the languages belong to different rhythm classes: stress-, syllable- or mora-timed (Abercrombie, 1967). Meanwhile, it appears that babies begin distinguishing languages in the same rhythm class only at later ages (Christophe & Morton, 1998; Nazzi et al., 1998). Durational metrics characterizing systematic, segmental-level surface timing patterns have been investigated as an alternative to rhythm classes for operationalizing linguistic rhythm. Asking which surface cues would best predict language discrimination, an infant meta-analysis (Gasparini et al., 2020) and an adult study (White et al., 2012) found that language discrimination was more successful when languages differed in their consonantal interval variability, as indexed by the Pairwise Variability Index for consonants (rPVI-C, Grabe & Low, 2002). This contrasts with predictions that young babies would focus specifically on vowel intervals (Mehler et al., 1996; Nespor et al., 2003). The only known study to test durational metrics as a predictor of infants’ discrimination abilities was an accent discrimination study, which found that differences between English accents in utterance-final lengthening facilitated discrimination (White et al., 2014), but no known infant study has used this approach testing two distinct languages.
In this study we will test the extent to which rhythm class and rPVI-C predict infants’ language discrimination (Grabe & Low, 2002). For this, we will test typically-developing monolingual German-learning infants aged 3.5 to 4.5 months of age.
Based on power analysis, we will aim to test 99 babies in total (33 per variety of English, with preliminary results being prepared by April 2021). Using an infant-controlled visual habituation procedure (Molnar et al., 2013; Werker et al., 1998) and Habit2 software (Oakes et al., 2019), babies will be habituated to low-pass filtered recordings of either the native language German or one of the following varieties of English: stress-timed American English with rPVI-C similar to German; stress-timed British English with rPVI-C distinct from German, or syllable-timed Singaporean English with rPVI-C similar to German (see Table 1). During the test phase, babies will hear recordings of both languages in alternation. A discrimination score will be calculated as babies’ fixation times to the monitor when hearing the habituated minus fixation times when hearing the non-habituated language. Linear mixed models will establish the effects of rhythm class (syllable-timed, stress-timed) and difference in rPVI-C (similar, different) on these discrimination scores. Moreover, a continuous predictor of rPVI-C reflecting specific values of the test trials will be added as a regressor to assess infants’ immediate response to the sentences they hear.
We predict that babies are sensitive to differences in consonant duration variability (Gasparini et al., 2020; White et al., 2012), such that the differences in looking times will be greater when languages vary substantially in their rPVI-C. We expect differences in rPVI-C to be a better predictor of successful discrimination than rhythm class. This study will provide insight into the specific rhythmic cues used by babies’ for language discrimination, namely whether babies attend to consonantal information when listening to speech.

Authors

• Loretta Gasparini, University of Melbourne
  Presenting Author

• Paul Okyere Omane, University of Potsdam
  Non-Presenting Author

• Natalie Boll-Avetisyan, University of Potsdam
  Non-Presenting Author

• Alan Langus, University of Potsdam
  Non-Presenting Author