Poster #193 - How children learn from others: An analysis of selective word learning

• In Event: 2-110 - Poster Session 07
  In Poster Session: PS 07 Section - Social Cognition

Integrative Statement

Many studies show that children can engage in selective word learning (e.g., Koenig, Clement, & Harris 2005, Koenig & Harris 2005), and there are now several reviews that offer different theoretical positions on how children engage in such learning (e.g., Harris, Koenig, Corriveau, & Jaswal, 2018; Heyes, 2017; Mills, 2013; Sobel & Kushnir, 2013). These studies vary in terms of its methodological and theoretical approaches. Our goal is to synthesize the findings of this literature to elucidate the mechanisms behind children’s selective word learning.

We conducted a metaanalysis of 63 published articles that totaled 105 separate studies, 666 groups/conditions within these studies, and 6525 individual participants. Participants were roughly 2-5-year-olds and primarily the research was conducted in the United States and Canada. Following Wellman et al. (2001), each age group/condition was coded for the proportion of appropriate responses to the test question asked. We also examined the role of a number of different methodological and theoretical factors (described below) on performance.

We found that the three types of questions used to assess selective trust (Ask: Whom should we ask for information? Endorse: Who generated correct information? Explicit Judgment (EJ): Who was better at generating information?) each had different developmental trajectories. Moreover, different theoretical and methodological factors influenced performance on each question. Some factors had little relation to performance, such as year of publication, informant type (person, video, etc.), how many informants were presented to the child, whether an explicit judgment question was asked, and when during the procedure it was asked.

To look at theoretical factors, we coded each study as to whether it manipulated information about the epistemic competence of an informant – the accuracy with which the informant generated information – and/or nonepistemic factors – factors that could indicate competence only based on prior knowledge possessed by the child. For example, speaking in a non-native accent could indicate less linguistic knowledge; child informants generally possess less knowledge than adult informants; attractive informants tend to be rated more trustworthy. All of these manipulations require that children have knowledge about the relation between what is manipulated and its potential for epistemic competence, and should be overridden by accuracy information. This factor played an important role in performance, with studies manipulating epistemic competence showing better performance on Ask and Endorse, but not EJ questions.

Finally, we considered overall performance on each test question in terms of three contrasting theories of selective word learning: A more nativist approach based on default biases (Harris et al., 2018; Jaswal et al., 2010), a more empiricist approach based on reinforcement learning (Heyes, 2017), and an approach based on rational constructivism (Sobel & Kushnir, 2013). Each theory had explanatory power for a different type of question. Ask questions were best explained by a more nativist account; Endorse questions were best explained by rational constructivism and EJ questions were best explained by reinforcement learning. This suggests that children’s selective learning is potentially governed by multiple mechanisms and that different types of questions potentially prompt responses from those different mechanisms.

Authors

• Zoe Finiasz, Brown University
  Presenting Author

• David Sobel, Brown University
  Non-Presenting Author