The effect of background variability on referent selection in fast mapping tasks: evidence from eye movements

Abstract

Introduction In word learning studies children are presented with a novel object alongside two or more familiar competitor objects and are asked to point to one of the objects in response to a label (e.g. Which one’s the blicket?). Research shows that even young infants can reliably select a novel object in response to a novel label (Houston-Price, Plunkett & Harris, 2005). Previous research on referent selection has mainly investigated how properties of objects themselves, such as the number of known competitors or variability of novel targets, affect children’s ability to learn word-object mappings (e.g., Horst, Scott & Pollard, 2010; Twomey, Ranson & Horst, 2014). However, while related research shows that background context affects infants’ categorization and noun generalization (Vlach & Sandhofer, 2011; Goldenberg & Johnson, 2015; Goldenberg & Sandhofer, 2013), the effect of background context on referent selection is unknown. The current study therefore examined infants’ referent selection performance when objects were presented on single or variable-colored backgrounds. Method Two-year-old infants (N = 31) were presented with three combinations of novel label-object associations presented on a computer screen. Half of the infants were shown the objects on a white background (constant condition) and half on five different colored backgrounds (variable condition). Each trial consisted of one novel and two familiar objects side-by-side. Infants saw each combination five times (once with each background color in the variable condition) and were asked for each novel object three times (using a recording played through speakers) and every familiar stimulus once. An eye tracker measured infants’ looking toward the objects and the background. Results In trials where the novel target was named, infants in the constant condition looked at the novel target longer (M = 0.56, SD = 0.22) than infants in the variable condition (M = 0.48, SD = 0.25), t (283) = 2.85, p = .005, d = 0.17. In contrast, infants in the variable condition spent more time looking at the background (M = 0.29, SD = 0.6) than those in the constant condition (M = 0.19, SD = 0.46), t (455) = -1.90, p = .05, d = 0.09. Time course analysis of the proportion of target looking between the two groups revealed that looking preferences of the infants in the variable condition reverted back to chance levels (0.33) faster than in the constant condition. Discussion These findings highlight the differences between the processes involved in referent selection and generalization tasks. In category generalization tasks, infants encounter a single stimulus on each training trial, whereas the in referent selection tasks infants encounter multiple objects simultaneously. Referent selection tasks therefore require infants to process more information. While variability in generalization tasks help infants decontextualize, the changing backgrounds in the current study interfered with infants’ attention during the more complex referent selection task.