Vis sammendrag
Vocal learning is a key foundation of language development, involving the gradual acquisition of language-specific speech sounds through exposure to surrounding speech. Although children’s language development is strongly shaped by their social environments, research has focused mainly on individual caregivers rather than broader social networks (Goldstein, King & West, 2003; Athari et al., 2021). Features of a child’s social network, including the number of close relationships and the diversity of speakers, may influence speech perception and production, suggesting a complex link between social interaction and vocal learning (Okocha et al., 2024; Casillas et al., 2020). Current evidence is inconclusive and most of the research is on indirect parental measures focusing on receptive skills (word comprehension, vocabulary), with little research assessing infants' language skills directly and expressive language. While some studies suggest benefits from speaker variability, others show no correlation with language outcomes, highlighting the necessity for further investigations to clarify how diverse social networks contribute to infants' language acquisition (Bulgarelli & Bergelson, 2023).
We hypothesize that individual differences in the social diversity of interactions of 11-month-old infants relate to the maturity of their early vocalizations, i.e. proportion of canonical babbling, as measured during naturalistic parent-child play interactions. We thus predict that a larger speaker variability (different unique speakers regularly speaking to the child and in the presence of the child, weighted by their frequency of interactions) should be associated with more mature vocalizations at the same age.
We recruited thirty families with monolingual Norwegian infants, i.e., exposed to 90 percent of Norwegian. Parents provided, in an online interview, information about the child's social network that was used to compute the speaker diversity measure, through Shannon's entropy formula (Shannon, 1948). Parent-child play interactions were recorded at home (with an average duration of 15 min). Infants’ vocalizations were manually classified into canonical, non-canonical, and other.
To calculate the speaker diversity variable, we developed three quantitative variables from Shannon’s entropy applied to a child’s social network. These measures are obtained from a questionnaire asking both parents to jointly reconstruct the social network of the child, namely who speaks to the child and/or in the presence of the child, and in what amounts of time.
We compute three entropy measures:
1. Child network entropy is a holistic measure of network structure that includes both child-directed speech, as well as any substantial adult-to-adult and adult-to-sibling amount of interactions occurring in the presence of the child.
2. Generation entropy to examine potential differential effects of input provided by children versus adults.
3. Input type entropy: child-directed versus overheard speech.
We analyze the data using mixed-effects regression models, with infants’ age and social network diversity measures (child network, generation, and input type entropy) as predictors of vocalization maturity. Model comparisons are conducted, allowing us to evaluate the contribution of social diversity to infant vocal learning. Data analysis is ongoing and will be presented at the conference.
This study contributes to a better understanding of the role of the child social network in the emergence of human speech, and the trajectories of language development.
