Comparison can be a powerful tool for learning for making inferences from one item to a seemingly distinct one - or even for abstracting new information about both items. For this to work, though, the learner must make the relevant alignment between the items. While introducing more items into the comparison could help isolate on the relevant similarities, it could alternately make these items harder to align. In this experiment, we're exploring the role of variability in making an initial alignment and abstraction, as well as in generalizing that abstraction. Previous work in our lab found that 3-month-olds failed to abstract a relation (e.g., XX or YZ) when they habituated to six toy pairs, despite succeeding with two repeating pairs (Anderson, Chang, Hespos & Gentner, 2018). Here we ask if habituating infants to a [six examples of a relation of that progress from more similar to each other to less similar ones][1] will allow 3-month-olds to align the shared relational structure between, despite the potential distraction of having new objects appear throughout habituation. If this is the case, we will test whether infants still make this alignment when the order is randomized. [Test trials will measure three levels of generalization][2]: (1) Do infants discriminate an exact pair from habituation from a new pair (Memory Check)?; Do infants generalize same or different to new similar-looking objects (Near Transfer)?; and do infants generalize same or different to dissimilar objects (Far Transfer)? [1]: https://osf.io/u2m8n/ [2]: https://osf.io/a84bf/