Effect of Age and Language Phonological Complexity on Children's Vocalizations

• In Event: Poster Session 11
  In Poster Session: PS 11 Topic - Language, Communication

Abstract

Although previous work suggests child vocalization changes on a universal maturational schedule, this has been studied in only a handful of languages (Cychosz et al., 2019). Researchers tend to concentrate on samples from Western, Educated, Industrialized, Rich, and Democratic (WEIRD) countries (Henrich et al., 2010), which represent communities with only a very limited range of language families. In fact, this lack of linguistic diversity in studies seriously affects our understanding of early language development and possibly leads us in the wrong direction. To contribute to filling this gap, this paper examines infant vocalizations in children learning languages varying in phonological complexity. Specifically, we looked at canonical proportion, a crucial index of phonological development (Oller et al., 1998).

We sampled data from daylong audio-recordings of 120 children (aged 1-40 months) primarily exposed to one of 16 different languages. Data annotation for this audio corpus was carried out in one of two ways: either manually by identifying segments of speech coming from the main child in short samples; or these segments were automatically extracted using the LENA proprietary software (Xu et al., 2009). Extracts were then cut into small 500ms clips to prevent leakage of the participants' personal information, and annotated by citizen scientists on the Zooniverse platform into canonical (containing speech-like consonant-vowel or vowel-consonant sequences) and non-canonical (including only vowels or weakly articulated consonants; Cychosz et al., 2019). We were able to derive the child's canonical proportion, defined as Canonical/(Canonical+Non-Canonical). Individual languages were classified into low, moderate, and high phonological complexity based on syllable complexity, consonant inventory, and vowel inventory (Maddieson, 2005).

A mixed model declaring age in interaction with the languages' syllable complexity revealed main effects of age (ß =0.029, SE = 0.008, p<0.001) and syllable complexity (ß = 0.290, SE = 0.103, p<0.01). Children learning low-syllable-complexity languages (where most syllables are consonant-vowel) had higher canonical proportion (M = 0.454, SD = 0.187) than children learning high-syllable-complexity languages (typically allowing complex onsets and codas; M = 0.353, SD = 0.208). In contrast, no significant effect was found of consonant or vowel inventory.

We found that the canonical proportion correlates with phonological properties of the language the child is learning, and specifically with syllable structure but not phonological inventory. This effect is compatible with an explanation whereby consistency in syllable structure, rather than simplicity in the phonological inventory, facilitates children's development of canonical syllables. Beyond the specific mechanistic explanation, this work highlights the importance of studying more diverse languages and cultures. Such research is necessary for the field to gain a more objective and universally valid image of child development.

Authors

• Natalia Kuzminykh, Università degli Studi di Siena
  Presenting Author

• Chiara Semenzin, Ecole Normale Superieure
  Non-Presenting Author

• Nadège Marin, Ecole Normale Superieure
  Non-Presenting Author

• Elika Bergelson, Duke Univeristy
  Non-Presenting Author

• Marisa Casillas, University of Chicago
  Non-Presenting Author

• Margaret E Cychosz, University of Maryland, College Park
  Non-Presenting Author

• Pauline Grosjean, University of New South Wales
  Non-Presenting Author

• Leslie Marie-Louise, Ecole Normale Superieure
  Non-Presenting Author

• Camila Scaff, University of Zurich
  Non-Presenting Author

• Sarah Walker, University of New South Wales, Sydney
  Non-Presenting Author

• Anne Warlaumont, University of California
  Non-Presenting Author

• Alejandrina Cristia, Centre National de la Recherche Scientifique (CNRS)
  Non-Presenting Author