Infants Detect Patterns of Choices Despite Counter Evidence, but Timing of Inconsistency Matters

Abstract

ABSTRACTUnderstanding others’ preference for a relational category of objects (e.g., prefer darker colored shirts) can be challenging for young children, as it involves comparison of choice options within and across exemplars. Adding to the challenge is occasional inconsistency in choices made by others. Here the authors examined whether 14-month-olds could detect an experimenter’s preference for taller objects when they observed choices that were somewhat inconsistent. Infants watched four familiarization events involving different object pairs: The experimenter chose the taller of two objects thrice and the shorter object once—the inconsistent choice was presented at different time points of familiarization. The infants detected the experimenter’s preference for taller objects only when they had observed three consistent choices consecutively from the beginning. This finding is in line with relational learning, specifically the significant role of initial data in the extraction of relational commonality. It also connects to the hierarchical Bayesian models of rational learning: Inconsistency can be discounted when the initial data allow learners to distinguish a highly probable hypothesis.