Progressively Aligning Multiple Exemplars in Infancy

• In Event: Poster Session 13
  In Poster Session: PS 13 Topic - Cognitive Processes

Abstract

A growing number of studies demonstrate relational learning early in infancy (Anderson et al., 2018; Ferry et al., 2015). We interpret these findings as evidence of analogical comparison because the signatures of the structure-mapping process that are evident in children and adults are also apparent in infants. One such signature is that comparison of alignable exemplars facilitates relational learning. In this study, we strengthen this claim by testing whether progressive alignment can benefit early relational learning among infants. In progressive alignment, a challenging comparison is facilitated by initially giving learners highly similar (and readily alignable) exemplars of the relation before presenting them with more disparate pairs (Gentner et al., 2007; Kotovsky & Gentner, 1996). In this case, the learner is likely to compare and spontaneously align the highly similar initial pairs. This alignment will boost the salience of the common relational structure (Markman & Gentner, 1993; Gick & Holyoak, 1983), allowing the learner to perceive the relation even in a very different pair. If progressive alignment holds for infants as well as children, then giving infants a sequence that begins with highly similar (and alignable) pairs will highlight the common relation, allowing the infants to generalize the relation to less overtly similar pairs. If so, then when tested after habituation, these young infants should be able to discriminate between familiar and novel relations of objects seen during the habituation phase (memory check), and demonstrate a novelty preference when generalizing the relation onto other new items from the same object category (near transfer) and different object category (far transfer).

To test whether progressive alignment facilitates relational transfer, we presented 3-month-olds (n = 30; target n = 48) with 6 highly similar pairs of objects in either a same or different relation across 6 to 9 habituation trials. In the test phase, all infants saw novel and familiar relations across three test trial types (Figure 1). Preliminary findings show that infants look longer at the novel relations in every test trial – memory check: MNovel = 27.26s, MFamiliar = 16.42s, near transfer: MNovel = 28.25s, MFamiliar = 25.69s, and far transfer: MNovel = 26.12s, MFamiliar = 19.76s (Figure 2). This pattern not only suggests that 3-month-olds might be able to discriminate familiar pairs at test when presented with 6 highly similar exemplars, but that they may also generalize to new objects out of the habituation category.

Should these trends continue, they will support our prediction that progressively aligning multiple exemplars facilitates 3-month-olds' abstraction of the critical relation, compared to the same age group's failure to learn it from 6 distinct exemplars as demonstrated in Anderson et al. (2018). Therefore, if our predictions bear out, the phenomenon of progressive alignment will constitute further evidence that very young infants can abstract a common relation from multiple alignable exemplars via structure-mapping, and that this might guide relational learning early in infancy.

Authors

• Apoorva Shivaram, Northwestern University
  Presenting Author

• Erin Morgan Anderson, Indiana University - Bloomington
  Non-Presenting Author

• Susan J Hespos, Northwestern University
  Non-Presenting Author

• Dedre Gentner, Northwestern University
  Non-Presenting Author