Thus, although 7-month-old infants in Experiment 1 and in Experiment 3 showed a preference for a “familiar” stimulus that embodied the second-order correlations and 11-month-olds in Experiment 1 and 4 showed a preference for a “novel” stimulus that violated the second-order correlations, both of these patterns of preference indicate that infants at each of these ages learned the correlation implied by the other two correlations that they experienced during familiarization. That infants as young as 7 months of age are capable of learning second-order correlations is particularly impressive because this is the earliest age at which they have been shown to associate the correlated features of objects that are presented together (Younger & Cohen, 1986). This implies that second-order correlation learning may be a process that develops in parallel with infants’ ability to associate the different features of the objects around them. There is also evidence that neonates are capable of learning correlations among features (Slater et al., 1991), and therefore an important unanswered question is whether infants younger than 7 months of age are also capable of second-order correlation learning.