Assessing the effect of object locations on word learning during naturalistic adult–child interactions.

The aim of this study was to assess previously observed relationships between object locations and early word learning within an ecologically valid design. Adults labeled objects during openly structured interactions with their child. Fine-grained distributions of object locations were determined by a convolutional neural network. However, an analysis of the distributions of these locations during the interactions revealed that none of the spatial variables-the areas, overlap, or distances-were able to significantly predict which corresponding labels were learned by the child. A consequent power analysis enabled specific predictions for future work. An interaction between the child's age and the area or the overlap of these locations would predict word learning in a sample of around 50 dyads or more. (PsycInfo Database Record (c) 2026 APA, all rights reserved).