Poster #123 - Why Don’t our 32-month-olds Show Word Learning? A Cautionary Tale in Object Salience and Selection

• In Event: 2-055 - Poster Session 06
  In Poster Session: PS 06 Section - Language, Communication

Integrative Statement

Word learning in the lab is typically assessed using infant-suited looking-based methods (e.g., Werker et al., 1998; Golinkoff et al., 1987). In infants, object salience affects looking (Libertus et al., 2013). Word learning theories posit that object salience can influence infants’ linking of words and objects (Hollich et al., 2000), which is crucial for word learning (Horst & Samuelson, 2008). However, how salience affects infants’ demonstration of word learning in preference-looking procedures has been less explored (Pruden et al., 2006). In a study designed to test unrelated manipulations – language background and fluency – on word learning, we demonstrate how object salience disproportionately affects gaze during word learning tasks.

Methods

In a preferential-looking eyetracking paradigm, 33 monolingual and bilingual children aged 32 months saw trials in their native (monolinguals) or dominant (bilinguals) language. The study comprised two consecutive blocks, each with a Learning Phase (8 trials) and a Test Phase (4 trials). Children were exposed to two novel labels, moba and voopi. Side of presentation was counterbalanced, although due to unrelated manipulations, each novel label was always paired with the same object. The two novel objects were judged by two adults to have similar salience.

On Learning Phase trials, children saw familiar-novel object pairs, labeled in a fluent (e.g., “Look at the key/moba!”) or disfluent (e.g., “Look at the uh… banana/voopi!”) sentence, and had to infer the novel referent via mutual exclusivity (Markman & Wachtel, 1988). On Test Phase trials, participants saw the two novel objects side-by-side, and one object was labeled twice (e.g., “Moba! Moba!”). Word learning was assessed by the proportion of looking to the target object.

Results & Discussion

Learning Phase. An inspection of looking curves suggested that infants used mutual exclusivity, increasing looking to the novel over familiar target when hearing the novel word (Figure 1). However, comparisons with chance (.50) masked this effect due to greater interest in the moba than the voopi (moba above chance: M=.60, SD=.09; voopi below chance, M=.44, SD=.09; ps <.05).
Test Phase. Data were collapsed across block, language background, and fluency condition after preliminary analyses showed no effects. Averaging across objects, infants looked at chance to the target (M=.51, SD=.13, t=.51, p=.61). However, when we analyzed data by object (Figure 2), children looked more at the moba (M=.70, SD=.17, p <.001 above chance) than the voopi (M=.34, SD=.21, p <.001 below chance), whether they heard the label “moba” or “voopi”, t= 8.18, p <.001.

Our results highlight issues with differing object salience: In a learning phase, children possibly demonstrated mutual exclusivity, but the effect was more apparent for the more salient object. At test, pairing more-and-less-salient novel objects together revealed object preference, and masked any word learning. These results are surprising, since our 32-month-olds were much older than theories of when object salience might override other factors in word learning (Pruden et al., 2006). These findings call for caution: Novel object pairs destined for word learning studies should be piloted for salience, by children, and be considered in analyses.

Authors

• Elizabeth Morin-Lessard, Concordia University
  Presenting Author

• Krista Byers-Heinlein, Concordia University
  Non-Presenting Author