The Multisensory Nature of Verbal Discourse in Parent–Toddler Interactions

ABSTRACT

Toddlers learn object names in sensory rich contexts. Many argue that this multisensory experience facilitates learning. Here, we examine how toddlers’ multisensory experience is linked to another aspect of their experience associated with better learning: the temporally extended nature of verbal discourse. We observed parent–toddler dyads as they played with, and as parents talked about, a set of objects. Analyses revealed links between the multisensory and extended nature of speech, highlighting inter-connections and redundancies in the environment. We discuss the implications of these results for our understanding of early discourse, multisensory communication, and how the learning environment shapes language development.

Acknowledgments

We thank many members of the Computational Cognition and Learning Laboratory, especially Maddie Bruce, Danielle Rosenstein, and Jessica Steinhiser, for their assistance in this research.

Funding

This research was supported in part by the National Science Foundation (SBE-BCS0924248, SBE-SMA1203420) and the National Institutes of Health (R01-HD074601, R21-EY017843, K99-HD082358).

Notes

1. Mean proportion of time objects were visually dominant for Cameras 1 (M = .18, SD = .06) and 2 (M = .19, SD = .09) were not significantly different from each other, p = .63.

2. We treated visual dominance as a binary variable to match the other nonverbal variables (parent touch and toddler touch). All current results hold when different threshold of visual dominance (3% or 7% of Field of View) were employed.

3. Other reasonable methods for distinguishing extended and brief discourse revealed similar trends (see Suanda, Smith, & Yu, 2016).

4. Simulations were performed in MATLAB.

5. These Odds Ratios are to be interpreted as the increase in likelihood that a nonverbal event occurred during an utterance within extended discourse over what would be expected by chance. We present the mean Odds Ratios across subjects along with the 95% confidence intervals around those means.

6. In deriving chance-level overlap, we constructed randomly sampled segments of the entire interaction that matched each extended discourse in duration and target object, computed the proportion of time within those samples that overlapped with the nonverbal events, and then repeated that process across the 1000 iterations. Chance level overlap was the mean proportion of overlap across all iterations.

7. The fact that the results for visual dominance mirror the results for toddlers’ manual actions is likely due to the fact that visual dominance is more tightly coupled to toddlers’ manual actions (see Yu & Smith, 2012).

Additional information

Funding

This research was supported in part by the National Science Foundation (SBE-BCS0924248, SBE-SMA1203420) and the National Institutes of Health (R01-HD074601, R21-EY017843, K99-HD082358).