Poster #74 - In the Right Ballpark: Systematic Acquisition of Partial Meaning for Hard Nouns From Their Observational Context

• In Event: Poster Session 14

Abstract

Introduction
How children learn words denoting abstract concepts (e.g., “tomorrow”) has long piqued the interest of cognitive and developmental scientists (Gleitman et al., 2005). This is because the machinery required to learn these words appears different from the machinery used to learn words for concrete objects (e.g., “ball”). In one Human Simulation Paradigm study (HSP), Kako (2005) revealed that although nouns denoting concrete objects could be acquired through their observational contexts, nouns denoting non-basic-level object categories (or “hard nouns”) could not be acquired through observational context alone (see also Gillette et al., 1999). This study asks whether the manner in which the HSP assesses learning (i.e., adult participants guess the precise words parents uttered in muted videos of child-directed speech) masks partial knowledge about hard nouns that can be acquired via their observational contexts.

Methods
Adult participants saw scenes from picture books (all text removed) that either did (target trials) or did not (distractor trials) contain a specific “hard noun” in its text. “Hard nouns” were defined as nouns denoting non-basic level object categories (see Kako, 2005; the nouns studied here included “hand”, “morning”, “school”, and “wind”). For each scene, participants were asked whether they thought the scene did or did not contain the target noun (Figure 1). Because participants did not know the noun’s identity in advance, participants had to guess at the start of the study. Participants received feedback however for each guess, allowing them to learn across trials. Of interest was whether participants succeeded at this categorization task without successfully identifying the target noun’s identity. Thus, at the end of forty categorization trials, we probed participants’ guesses for the target noun’s identity.

Results
We analyzed participants’ categorization as a function of their guesses in the word identity probe (Figure 1A-B). Participants who incorrectly guessed the target noun (70% of participants) performed above chance in the later categorization blocks, M = .58, SD = .12, t(13) = 2.37, p <.05, suggesting that, at least for some hard nouns, participants acquired systematic and useful partial knowledge from the observational contexts in which these nouns occurred. Additional evidence for this came from participants’ incorrect guesses in the word identity probe. Specifically, naïve coders judged incorrect guesses and their associated target nouns (M = 4.83, SD = 1.45) as more semantically related than incorrect guesses and the other target nouns (M = 3.17, SD = .52; t(11) = 4.60, p < .001; Figure 2C-D).

Discussion
How different sources of information shape word learning is of great interest as it sheds light on the mechanisms of learning and on strategies for intervention. The current work suggests that the value of different sources of information may depend on how word learning is defined and measured. In contrast to work showing that observational contexts are insufficient to yield learning of precise hard noun meaning (see Kako, 2005), this work reveals that observational contexts may yield systematic partial knowledge of hard noun meaning, knowledge that may lay the critical foundation for full word learning. 

Authors

• Kosta Boskovic, University of Connecticut
  Presenting Author

• Sumarga H. Suanda, University of Connecticut