Connecting the Dots: Associative Transfer and Inference at 6 Months of Age

Infants continuously pick up incidental information about their environmental surroundings, expressing only a small fraction of what they learn. The present experiments examined whether 6-month-olds can transfer learning between stimuli they never see together through mnemonic networks. Using sensory preconditioning, deferred imitation, and associative inference paradigms, we characterized the properties and constraints of these emerging networks. Experiments 1A-1B revealed that 6-month-olds linked overlapping stimulus pairs encountered on successive days (AB, BC), enabling learning transfer between stimuli that never co-occurred (A-C). Experiments 2A-2B extended these findings to more complex four- and five-stimulus conditions (e.g., AB, BC, CD, DE). Infants' mnemonic networks exhibited bidirectional connections, response transfer regardless of element location (Experiments 1A-2B), and specificity, as they did not generalize to unrelated stimuli (Experiment 2B) or after exposure to unpaired stimuli (Experiment 1A). At the same time, we identified constraints in mnemonic networks; associations were pruned through extinction (Experiments 1A-1B) and persisted for 2 days but not 5 (Experiment 1C). We propose that 6-month-olds' ability to integrate multiple overlapping associative memories provides insights into the fundamental building blocks of episodic memory, including associative inference and relational generalization.