When One Exemplar Behaves Like Many

• In Event: 1-198 - Acquisition of Same/Different Relations by Human and Nonhuman Groups

Integrative Statement

For infants to abstract relations in both visual and linguistic domains, it may be necessary for them to compare multiple exemplars (Ferry, Hespos & Gentner, 2015; Gerken & Bollt, 2008; Marcus, Vijayan, Rao, & Vishton, 1999). In a compelling example to the contrary, 9-month-old infants generalized an ABA or AAB phoneme structure to new words after hearing a single word with the target structure (Gerken, Dawson, Chatila, & Tenenbaum, 2015). Gerken, et al. (2015) argued that this unusual finding was due to the linguistic nature of the stimuli. We argue instead for a domain-general explanation: that early in infancy, encoding processes are somewhat variable; thus, infants unfamiliar with a stimulus might encode repetitions as distinct exemplars of the target structure. In this scenario, even repeating a single exemplar could support relational abstraction, because the infant would experience a series of highly similar, but unique pairs. If this proposal is correct, it makes a counter-intuitive prediction that young infants, who have less efficient visual processes, may be more able to generalize from repetitions of a single exemplar, because they perceive variability in the input. In contrast, older infants, with more consistent encoding strategies, would experience the series as repetition of a single item, and would not show generalization.

To test whether a younger age group (3 months, n = 48) and an older age group (7- and 9-months, n = 48) would show different outcomes when given a repeated exemplar, we presented infants with a single exemplar of a same or different relation repeated for 6 to 9 habituation trials, then tested infants on a mix of novel and familiar relations (see Figure 1).

Our findings are consistent with the prediction that 3- but not 7- and – 9-month-olds would abstract the relation from a single repeated exemplar (see Figure 2). Seven- and 9-month-olds recognized a pair they had seen in habituation, t(46) = 4.653, p < .001, but did not discriminate between novel and familiar relations when they were instantiated with new objects, t(46) .905, p = .370. In contrast, 3-month-olds both recognized the habituation pair, t(45) = 2.792, p = .008, and discriminated between novel and familiar relations in new object pairs, t(34) = 2.337, p = .024. This suggests that repeating a single exemplar allowed younger infants to form a relational abstraction. However, a repeated measures ANOVA across the two age groups revealed a significant main effect of relation, F(1,77) = 4.844, p = .031, but no interaction between age and relation, F(1,77) = 1.074, p = .303

Taken together, these findings support the theory that alignment at the learner’s level is critical for abstracting relations. For younger infants, this may mean that even one exemplar is enough. Further, we show how the development of attention and encoding processes can interact with this learning mechanism.

Authors

• Erin M Anderson, Northwestern University
  Presenting Author

• Susan Hespos, Northwestern University
  Non-Presenting Author

• Dedre Gentner, Northwestern University
  Non-Presenting Author