Development in Preschoolers’ Learning from Naturalistic Overheard Speech

• In Event: 1-099 - Factors that promote third-party learning in young children

Integrative Statement

Research on overhearing has demonstrated that children as young as 18 months can learn a novel word from a brief overhearing exposure. Such experiments have found learning of single nouns, verbs, and facts (e.g., Floor & Akhtar, 2006). This work suggests that young children require neither joint attention nor directed speech to acquire new word-object mappings. However, previous studies have primarily tested overhearing in highly constrained paradigms, which include pedagogical cues (e.g., child-directed speech, referential gaze) likely absent in daily overhearing. Here, we test learning from naturalistic overheard speech to shed light on the potential utility of overheard input “in the wild.”

In Experiment 1, 68 4.5-6-year-olds were assigned to either an Overhearing or Didactic condition, where they learned a set of four nouns and six facts corresponding to four novel and two familiar toys. In the Overhearing condition, a confederate played with the toys with the child. After a minute, the experimenter received a phone call and delivered a script describing the toys, as if to a friend. She neither looked to the objects, nor made eye contact with the child, and tested the child’s learning upon “hanging up." In the Didactic condition, children heard the same script, but in enthusiastic child-directed speech, with pedagogical cues. 4.5-6-year-olds learned four words and six facts above chance. While there was no difference in word-learning between conditions, children performed better on fact-learning overall (X2(1) = 81.88, p < .001), and in the Didactic condition in particular (X2(1) = 7.14, p < .01).

In Experiment 2, 64 3-4.5-year-olds were taught three novel words and five facts. They learned facts better (X2(1) = 31.84, p < .001) and above chance in both conditions, but performed better overall in the Didactic condition (X2(1) = 20.21, p < .001), only learning words above chance when taught them directly. In both experiments, facts associated with familiar objects (e.g., “the dog I’ve had for two years”) were learned better than those associated with novel objects (“the zav I found in the garden”) (ps < .001), likely because those facts could be mapped directly to the noun without attending to the scene to first disambiguate the referent.

As a behavioral proxy of children’s online tracking of the speech, we coded videos of the overhearing exposure. Younger and older children touched individual novel objects more often when the experimenter was talking about them than when she wasn’t, suggesting they were aware the speech was relevant to their context, tuning in to it, and doing so more with age (p < .05).

Taken together, our results point to children’s ability to marshal their own attentional resources as a driver of development in overhearing ability. The learning advantage of child-directed speech appears to lie in part in its decreased referential ambiguity, particularly in contexts with co-present referents. Overheard may be more equivalent to child-directed speech as a source of learning for abstract words whose meaning can be inferred from intralinguistic context, and/or precursors to word-meaning mappings like recognition and semantic domain.

Authors

• Ruthe Foushee, UC Berkeley
  Presenting Author

• Mahesh Srinivasan, UC Berkeley
  Non-Presenting Author

• Fei Xu, University of California-Berkeley
  Non-Presenting Author