Materials, data, and analysis scripts for experiments investigating active sampling during word learning in children ("Act").
Zettersten, M., Choi, K., Kirkorian, H., & Saffran, J. (2025). Active sampling helps children learn new words by tuning their curriculum to past experiences. Preprint: https://osf.io/preprints/psyarxiv/ne43k_v1
Abstract
Young children are actively engaged in structuring their early language learning. But what sampling preferences do children have when actively seeking new information about unfamiliar words? And when and why is active control beneficial for learning words? Across 2 experiments (N = 303; M(age) = 4.6 years; range: 3.0-6.0; 166 female, 137 male; 85.8% White, 5.6% more than one race, 2.6% Asian, 0.7% Black; 6.2% Hispanic or Latino), we manipulated whether 3- to 5-year-old children were able to actively select their word learning input or were passively exposed to the active selections of another child. Experiment 1 used a yoked active-passive design. We found that having active control did not benefit children’s learning of unfamiliar novel words relative to receiving that same input passively. In Experiment 2, we first gave children experience with novel words, such that some object-label associations were experienced more frequently than others. Children who had active control systematically made more informative choices by selecting less frequently experienced object-label associations. By comparing their performance to children in two different passive conditions, we showed that children learn novel words better when experiencing input tuned to their past experience, regardless of whether they actively select this input themselves or passively experience the choices of a child sharing their same past word learning experience. Together, these studies demonstrate when and how active information-seeking benefits children’s word learning.
The repository contains the following documents:
- analysis: R scripts analyzing the data
- data: experimental data and codebook
- experiment_scripts: Experimental scripts used to present the experiment (JavaScript/html using jsPsych)
- figures: Figures created in analyses
- presentations: Files from past presentations
Here is a direct link to the html rendering of the R Markdown file in the analysis folder. This provides a walkthrough of the major analyses in the paper:
https://mzettersten.github.io/act/analysis/2_act_analysis.html
Click on the link below to view a demo of the experiment. Use the default option when prompted to enter information at the beginning of the demo:
- Version 1: https://mzettersten.github.io/act/experiment_scripts/version1/active_demo.html
- Version 2: https://mzettersten.github.io/act/experiment_scripts/version2
Feel free to contact me at [email protected] if you have any comments or questions about the data or the analyses.