Do children extract more from their environment than adults?

• In Event: Poster Session 05
  In Poster Session: PS 05 Topic - Attention, Learning, Memory

Abstract

Statistical learning is a fundamental learning system which allows learners of all ages to extract structure from our complex worlds (Saffran, Aslin, & Newport, 1996); it has been credited with feats ranging from learning linguistic patterns to action sequences. Despite this widespread involvement in learning, there are situations where adult learners fail. For example, when presented with two patterns back-to-back, adults only learn the first, unless presented with an overt cue to the existence of the second (e.g. voice change or long pause) (Gebhart, Aslin, & Newport, 2009). Adults also fail to learn more than one structure when multiple are presented simultaneously (Creel, Newport, & Aslin, 2004; Forest & Finn, under review; Turk-Browne, Jungé, & Scholl, 2005; Vuong, Meyer, & Christiansen, 2016). It is currently unknown whether children display the same learning failures, but there are reasons to think that they may not.
Specifically, ongoing development in children’s memory (Barr & Brito, 2014) and attentional abilities (Plude, Enns, & Brodeur, 1994) could lead them to learn all structures they encounter, even when they are not presented with an overt cue. For example, if children’s knowledge of the first structure is weaker because their attention fluctuates more during learning, they may learn a second structure when unknowingly presented with one. Similarly, children’s worse selective attention might allow them to learn patterned information adults ignore; thereby learning multiple patterns simultaneously where adults learn one. The limited work on statistical learning in school-aged children suggests their learning may be worse than adults’ (Raviv & Arnon, 2018). This worse learning, however, may benefit children who have much to learn. At the moment we do not know whether children learn multiple patterns in situations where adults cannot.
Here, we tested the hypothesis that children may learn more from complex learning environments than adults. In two experiments, 5-9 year-old children and adults watched two statistical patterns either simultaneously (Experiment 1, n = 128) or sequentially (Experiment 2, n = 103), and were then tested on their knowledge of each. Experiment 1 results indicate that most participants (both adults and children) only learned one of two patterns. However, the pattern adults learned was equally likely to be the first or second one tested; children, instead, only demonstrated learning of the pattern they were tested on first (Figure 1). While this does not show that children learn more, it does suggest that children have a representation of both patterns at the start of the test phase. In Experiment 2—when structures were presented sequentially, not simultaneously—adults only showed learning of the first structure they saw (replicating past results in the adult literature), but children demonstrated above chance learning of each pattern. Interestingly, in Experiment 2 children also showed significantly better learning of the first structure they were tested on (Figure 2).
In tandem, these results suggest that children may show broader but more fragile knowledge of their environments than adult learners, and highlight that children may, counterintuitively, be better suited for complex learning environments than adults are.

Authors

• Tess Allegra Forest, University of Toronto
  Presenting Author

• Chuyun Shen, University of Toronto
  Non-Presenting Author

• Amy S Finn, University of Toronto
  Non-Presenting Author