When one is not Enough: Children’s Ability to Combine Cues in Novel Word Learning

• In Event: Lexical Development: environment and mechanisms

Abstract

Children can successfully identify the referent of a novel word by using individual cues like a speaker’s gaze (e.g., Baldwin, 1993) or their knowledge of familiar objects’ names (e.g., Markman & Wachtel, 1988) to rule out other potential referents. Recent work has put multiple cues in conflict to examine whether children preferentially weight cues (e.g., Grassmann & Tomasello, 2010; Jaswal & Hansen, 2006). In naturalistic environments, however, there are usually multiple cues that cooperate in identifying the correct referent of novel words. The current research examined whether the presence of multiple, cooperating cues improves children’s accuracy in word learning.

Three-year-old children watched videos of an adult labeling one of several objects placed before her on a table. In Study 1 (Figure 1a-1c), children (n=51) could identify the referents of some novel words by using the speaker’s gaze (both objects were novel, but the speaker looked at the referent). For other words, children could identify the referent using mutual exclusivity (ME; the speaker did not look at the referent, but the distractor was familiar with a known name). For other words, children could identify the referent using both cues. In Study 2 (Figure 1d), children (n=40) could identify the referent only by combining both cues – each cue eliminated some, but not all potential referents. Children in each study then completed test trials. On test trials, children were shown images of two novel objects and heard a sentence labeling one. Children’s eye movements were tracked throughout the experiment to measure their accuracy in identifying the referents of novel words (teaching videos) and their success in learning and retaining the word-referent mappings (test trials).

In Study 1, children’s accuracy in fixating the target object during teaching videos significantly differed between conditions, F(2,48.4)=10.3, p<.001 (see Figure 2). Children’s accuracy when both cues were present (M=49%) was significantly higher than when they could use just gaze (M=34.7%), F(1,49.4)=16.5, p<.001 and marginally higher than when they could use just ME (M=43%), F(1,49.7)=3.6, p=.06. Children’s accuracy in fixating the target object during test trials, however, did not significantly differ between conditions, F(2,47.8)=1.3, p=.27.

In Study 2, children’s accuracy in fixating the target object during teaching videos could not be reliably tracked, because of the close proximity of objects in the video. Children’s accuracy in fixating the target object during test trials (M=57.9%) was significantly higher than chance, F(1,37.1)=14.0, p<.001.

When a single cue will suffice (Study 1), the presence of multiple cues improves children’s ability to identify the referents of novel words, but not their ability to encode/retain the word-referent mapping. When a single cue is insufficient (Study 2), the presence of multiple cues enables word learning that would otherwise not be possible. When children first hear a word, there is a wealth of information in their environment. The current work suggests that multiple cues may be crucially important for learning words in naturalistic environments and represents an important step in exploring when and how an abundance of information affects word learning.

Author

• Ron Pomper, Boys Town National Research Hospital
  Presenting Author