Mapping and tracking the development of visual and semantic information use in object perception

What information do people use for object perception, and how does this change across development? In this large-scale study (total n = 333), we investigated the development of object processing in a large sample of children (ages 4: n = 69; 6: n = 90; 8: n = 104) and a comparison group of adults (n = 70). Participants completed a fun and engaging object foraging task, requiring discrimination of target objects from distractors. We used human ratings as well as activation patterns from a neural network to quantify the visual and semantic information available for object discrimination. In line with our preregistered hypotheses (https://osf.io/84wxp), older children relied more on semantic and high-level visual information than younger children. Also as predicted, they tuned into the visual cues most diagnostic for the task—shape for basic-level distinctions (e.g., brush vs. pipe) and features for subordinate-level distinctions (e.g., two different bowties). Critically, an exploratory representational similarity analysis (RSA) nonetheless revealed remarkably similar underlying object representations across all age groups, suggesting that developmental changes in object perception are primarily quantitative rather than qualitative. These findings suggest that children's improved object discrimination abilities reflect increased access to relevant visual and semantic information. Overall, our results support a model of continuous development of high-level vision across the lifespan.