The study used the mobile reinforcement task with 6-month-olds in a design analogous to that used by Dwyer et al. (1998). Infants were taught three sets of correlations: (1) two hand puppets (A and B) that went together; (2) a mobile that was movable by the infant’s kicking that went with a particular crib context; (3) one of the hand puppets (A) that went with the crib context. Cuevas et al. found that infants began to associate the other hand puppet (B) with the mobile—despite never being exposed to the two simultaneously—through the activated memories of puppet A and the crib context. In a related vein, there is also evidence that infants can learn nonadjacent dependencies in language that occur over one or more intervening units and require infants to track discontinuous sequential relationships. For example, by 6 months of age infants are able to track nonadjacent dependencies among vowels in natural language, and by 10 months of age they can track nonadjacent relationships among consonants (Gonzalez-Gomez & Nazzi, 2012; for a review see Sandoval & Gomez, 2013). Finally, Mou, Province, and Luo (2014)found evidence that infants at 16 months of age are capable of transitive inference. Thus infants who saw an agent prefer a red object over a yellow object (A > B) and a yellow object over a green object (B > C), inferred that the agent should prefer the red object ahead of the green one (A > C). Although the processes involved in transitive inference are likely different from the associative processes that we posit support second-order correlation learning, this study demonstrates that by 16 months, at least, infants are able to generalize from their experience about an unobserved relationship (in this case, that A > C).