Self‐generated variability in object images predicts vocabulary growth

We take a novel, objective approach to investigate the link between object manipulation and vocabulary development. Object names are a major component of early vocabularies and learning object names depends on being able to visually recognize objects in the world. However, the fundamental visual challenge of the moment-to-moment variations in object appearances that learners must resolve has received little attention in word learning research. We provide the first evidence that image-level object variability matters and may be the link that connects infant object manipulation to vocabulary development. Using head-mounted eye tracking, the present study objectively measured individual differences in the moment-to-moment variability of visual instances of the same object, from the first-person views of 22 15-month-old infants as they engaged in free-flowing toy play with six objects with their parent. We measured the frame-by-frame visual variability of objects on which infants fixated their gaze using a raw image measure: "mask orientation" (MO; orientation of the most elongated axis of the object pixels in the scene camera image). This measure is unencumbered by assumptions about the nature of varying object views or the specific forms of visual information that may be critical to object recognition. We also collected vocabulary measures via MacArthur-Bates Communicative Inventories (MCDIs) completed by parents when their infant was 15 and 21 months of age. Infants who generated more variable visual object images through manual object manipulation at 15 months of age experienced greater vocabulary growth over the next six months; other potential sources of object image variability, such as parent manipulation, and other measures of infants behavior, such as amount of object manipulation, did not predict vocabulary growth. These results provide a direct link from object manipulation to the variability of the visual information at the level of 2-dimensional images to later object name learning. The data contained in this volume include both raw frame-by-frame object manipulation and MO data, as well as the aggregated data and MCDIs for each subject that went into linear mixed effects models.