Object representation without object files at 10 months of age

• In Event: 2-031 - New advances in the study of object representations in infancy: Formation, structure, and content

Integrative Statement

Infants at 10 months of age can individuate objects around them based on spatiotemporal (e.g., Xu and Carey, 1996) and conceptual/linguistic (Xu et al., 2004) properties. Most current theories claim that both of these individuation processes rely on the same representational format (object files), which are organized around a spatiotemporal pointer. This implies that spatiotemporal information should always be prioritized over conceptual information in object individuation (e.g., Kibbe and Leslie, 2011). In contrast, we argue that infants can also represent objects in a different, exclusively conceptual format, one that is not location-bound. Given previous results that suggest that infants can remember objects while forgetting their locations (Yoon et al., 2008) and that spatiotemporal individuation may be content-sensitive (Kuhlmeier et al., 2004), we hypothesized that object individuation by conceptual vs. spatiotemporal properties rely on different representational formats, and predicted that young infants would be successful with either one in isolation, but would fail when the two kinds of properties are in conflict.

We report three studies with 10-month-olds (N = 24 each), using the manual search object individuation paradigm (Van de Walle et al., 2000). In the presentation phase of the two-object conditions, two unfamiliar objects were placed on the top of an opaque box so that both were simultaneously visible. Then the objects were hidden in the box. Infants were allowed to retrieve only one of the objects, and were then allowed to search for the missing second object (hidden in a secret compartment). The dependent variables were search duration and number of reaches into the box after the first object was retrieved. These measures were compared to the one-object condition, in which a second object would not have been expected in the box. Study 1 was a replication of the original study, where the objects were placed on the top of the box with no accompanying verbalization. We found significantly longer search duration in the two-object (4.32 s) than in the one-object (2.71 s) condition (t(23) = 2.914, p = .008. In Study 2, the object pairs were labeled with the same novel label to induce conceptual object representation. We found no difference in search duration between the conditions (two-object: 3.69 s, one-object: 3.40 s). This is consistent with the prediction that the conceptual information, which indicated the presence of one kind of object, would conflict with the spatiotemporal information about two entities. In Study 3, the two objects were labeled with different novel labels. Here, we expected that the infants would search longer for the second object in the two-object condition. This prediction was confirmed (two-object: 2.42 s, one-object: 1.72 s; t(23) = 2.116, p = .045), showing that the infants' failure in Study 2 was not due to the linguistic input per se. We also report a follow-up study aimed to test how infants encode object location when the conceptual system individuates the objects, where our findings were inconclusive. Taken together, these studies support the proposal that spatiotemporal and conceptual individuation might rely on different representational systems.

Authors

• Gabor Brody, Central European University
  Presenting Author

• Gergely Csibra, Central European University
  Non-Presenting Author