Abstract
Introduction. Infants’ visual working memory (VWM) has a 3-item limit (Feigenson & Carey, 2003), but infants can "chunk" items using perceptually available properties, such as common features or spatial proximity, thereby overcoming that limit (Feigenson & Halberda, 2004; 2008). When such chunking cues are not available, adults can extract statistical regularities between visual arrays presented over time, and use these regularities to more efficiently group items in VWM (Brady, et al, 2009). Here we asked whether 13-month-old infants could also use statistical regularities between objects to increase VWM. Participants & Method. During 10 Familiarization trials, infants (N=21) saw 4 objects (red disk, blue cross, green pentagon, yellow squiggle) placed sequentially on a stage 2 at a time. In the No-Regularity condition, the objects were paired randomly on each Familiarization trial. In the Regularity condition, object identity was yoked such that, across Familiarization trials, the identity of one shape perfectly predicted the identity of the other. Next, infants saw 6 Test Trials in which all 4 objects were sequentially hidden behind an occluder, which was then removed to reveal either all 4 objects (Expected Outcome) or only 3 (Unexpected Outcome), on alternating trials. Results. As predicted by previous findings, infants in the No Regularity condition failed to look longer at the Unexpected Outcome when four objects were hidden and just three were revealed, F(1,10)=0.005, p=n.s.). This finding replicates earlier work demonstrating an upper limit on infants’ VWM. In contrast, infants in the Regularity condition increased their looking to the Unexpected Outcome, F(1,28)=6.37, p=0.024 (see Figure 1). This suggests that seeing statistical regularities between individual objects allowed infants to group or chunk the objects in VWM, and hence to remember more total information about the scene. Conclusions. Infants can chunk items in VWM based on the statistical regularities present as events unfold.
Meeting abstract presented at VSS 2013