ABSTRACT
Prior research suggests that prediction supports language processing and learning. However, the ecological validity of such findings is unclear because experiments usually include constrained stimuli. While theoretically suggestive, previous conclusions will be largely irrelevant if listeners cannot generate predictions in response to complex and variable perceptual input. Taking a step toward addressing this limitation, three eye-tracking experiments evaluated how adults (N = 72) and 4- and 5-year-old children (N = 72) generated predictions in contexts with complex visual stimuli (Experiment 1), variable speech stimuli (Experiment 2), and both concurrently (Experiment 3). Results indicated that listeners generated predictions in contexts with complex visual stimuli or variable speech stimuli. When both were more naturalistic, listeners used informative verbs to generate predictions, but not adjectives or number markings. This investigation provides a test for theories claiming that prediction is a central learning mechanism, and calls for further evaluations of prediction in naturalistic settings.
Acknowledgements
We thank all participants, as well as Dominick Reuter, Claire Robertson, and Cynthia Lukyanenko for assistance with stimuli, Mia Sullivan for assistance with data collection, and other members of the Princeton Baby Lab for assistance with participant recruitment. We are also grateful to Adele Goldberg for comments on a previous version of this paper. This research was supported by grants from the National Institute of Child Health and Human Development to Casey Lew-Williams (R01HD095912, R03HD079779) and from the National Science Foundation to Tracy Reuter (DGE-1656466).
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 However, it is important to distinguish behavioral measures of prediction from prediction itself: Although prediction is typically operationalized via anticipatory eye movements, prediction may occur before or in the absence of overt behaviours. For example, a listener might accurately predict a speaker’s referent before they visually locate it within the surrounding scene. Similarly, listeners could presumably anticipate an abstract or absent referent.