Early Verb Learning: How Do Children Learn How to Compare Events?

Abstract

An important problem verb learners must solve is how to extend verbs. Children could use cross-situational information to guide their extensions; however, comparing events is difficult. In 2 studies, researchers tested whether children benefit from initially seeing a pair of similar events (“progressive alignment”) while learning new verbs and whether this influence changes with age. In Study 1, 2.5- and 3.5-year-old children participated in an interactive task. Children who saw a pair of similar events and then varied events were able to extend verbs at test and differed from a control group; children who saw 2 pairs of varied events did not differ from the control group. In Study 2, events were presented on a monitor. Following the initial pair of events that varied by condition, a Tobii x120 eye tracker recorded 2.5-, 3.5-, and 4.5-year-olds’ fixations to specific elements of events (areas of interest) during the 2nd pair of events, which were the same across conditions. After seeing the pair of events that were highly similar, 2.5-year-olds showed significantly longer fixation durations to agents and to affected objects as compared with the all-varied condition. At test, 3.5-year-olds were able to extend the verb, but only in the progressive alignment condition. These results are important because they show children’s visual attention to relevant elements in dynamic events is influenced by their prior comparison experience, and they show that young children benefit from seeing similar events as they learn to compare events to each other.

Notes

¹ Because proportional data may be subject to instability of error term variances, arc sine transformations were applied to all proportional data (Netter, Wasserman, & Kutner, 1985). These analyses with transformed data showed the same patterns as reported here.

² One reason children may have failed to point on one or more trials was that the test trials as presented could not be paused using the Tobii Studio software. Most (but not all) of the children who were excluded because they refused point at test were in the youngest 2.5-year-old age group, with approximately equal representation from both conditions.

³ Study 2 did not include a control group because the task of recognizing an action at test (even with new objects) seemed so easy to perform that we did not predict that the advantage of seeing multiple events versus a single event would emerge. Given this study design, it will not be possible to conclude from the results that the comparison of multiple events is more useful than is seeing a single event, but it will be possible to show whether the types of events that are compared influence performance.

⁴ Additional eye-tracking details:Size of stimuli: The scenes fit into a 24 cm × 14 cm area (visual angle: 32.9º × 19.6º). The agent was approximately 18 cm × 7 cm (visual angle: 25º × 9.85º), and the objects ranged in size and included objects that were 3 cm × 2 cm (4.33º × 2.8º), 9 cm × 1 cm (12.6º × 1.4º), 4 cm × 3 cm (5.6º × 4.2º), and 5 cm × 2.5 cm (7º × 3.5º).Fixation filter: We used the standard I-VT fixation filter in the Tobii Studio software, which has been set to yield accurate fixation data for the most common eye-tracking uses. The noise reduction setting was disabled. The minimum fixation duration was set at 60 ms, which is a conservative setting to allow for complex visual behavior.Processing of eye-movement data: As reported, we averaged individuals’ fixation duration for a specific AOI across scenes to be sure a particular scene or object did not have a major effect on the results. We also excluded participants as reported in the Participant section. However, beyond these considerations and our use of the standard I-VT fixation filter, we did not further process the eye-tracking data.

Additional information

Funding

Funding for this research was provided by a grant from the National Institutes of Health (NIH; 2R15 HD044447), the Howard Hughes Medical Institute summer research program, the McNair Program, and the support of Trinity University. The project was supported by the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD), and the content is solely the responsibility of the authors and does not necessarily represent the views of the NICHD or NIH.