In the Right Neighborhood: Semantic Coherence in Parents’ Ambiguous Naming Events

• In Event: Poster Session 09
  In Poster Session: PS 09 Topic - Language, Communication

Abstract

Word learning theories differ in their depiction of the extralinguistic environment and its role in learning. While some theories stress the invaluable role of the social context for learning (Tamis-LeMonda et al., 2014; Tomasello, 1988), others stress its insufficiency (Gillette et al., 1999; Medina et al., 2011). Consistent with the latter view is Human Simulation Paradigm research (HSP; Cartmill et al., 2013; Gillette et al., 1999; Medina et al., 2011), where naïve observers watch muted videos of parent-toddler interactions and guess “mystery words” parents uttered in the videos. This research shows that despite access to the social and extralinguistic contexts, even adults struggle to identify those mystery words (e.g., Medina et al., 2011). Here, we asked whether the errors provided by HSP participants help reconcile the different perspectives of the extralinguistic context and its role in learning.
From the recordings of ten two-year-olds and their parents during free-play, we created vignettes of sixty nouns uttered by parents (e.g., Medina et al., 2011; see Figure 1B). Forty adults watched 30 of those vignettes and were asked to identify the mystery word in each. Similar to previous HSP research (e.g., Medina et al., 2011), participants struggled (M = 0.35, SD = 0.32; see Figure 2A), suggesting substantial referential uncertainty in the input.
Participants’ HSP errors, however, paint a more nuanced picture. We followed-up the HSP study with a semantic-ratings task, where a separate group of participants (N = 28) rated the semantic relatedness of mystery words and their associated HSP errors. Participants saw a “probe” word (one of the words in a mystery word-error pair) and had to guess which of two words was more similar in meaning (one of which was the other word in the mystery word-error pair; see Figure 1C). Each of 734 HSP error pairings were rated eight times.
Mystery words and their HSP errors were rated as more similar in meaning in both subject-level (Msubject= 0.76, SD= 0.05) and word-level (Mword= 0.78, SD= 0.17) analyses (see Figure 2B). To investigate whether this effect was driven by the few referentially clear HSP mystery words, we correlated HSP performance for a mystery word with performance in the semantic ratings task involving that mystery word (see Figure 2C). We also compared the semantic ratings involving High Informative vignettes (i.e., vignettes that were referentially clear; see Medina et al., 2011) to those involving Low Informative vignettes. Both the continuous (Figure 2C) and categorical analyses (Figure 2D) revealed no link between vignette referential clarity (HSP task) and semantic-relatedness of the words conjured up in error by those vignettes (Semantic Ratings task).
Understanding the nature of the input is central to theories of early word learning. In the current study, we suggest that although the noisy non-verbal context may be insufficient in providing precise word meaning, it may provide learners with candidate meanings that are semantically adjacent. Ongoing research in our lab is exploring the role these meaning-adjacent instances play in word learning.

Authors

• Sara Johnson, University of Connecticut
  Presenting Author

• Tiffany Schalla, University of Connecticut
  Non-Presenting Author

• Sumarga H. Suanda, University of Connecticut
  Non-Presenting Author