Stimulus and expectation-driven novelty independently guide infant looking behaviour: a systematic review and meta-analysis

Abstract

Infants look longer at events that are unexpected (e.g. a ball floating in midair) and at events that are different from what they have previously seen (e.g. rightward motion, after repeated exposure to leftward motion). This has led to debate about whether shared or distinct mechanisms lead infants to look longer toward unexpected events (the violation-of-expectation response, or VOE) versus visually novel events (the perceptual novelty response, or PN). Here, we studied both looking behaviors using meta-analysis of condition-level data (from 34 papers, 78 conditions, and 1915 infants) and mega-analysis of infant-level data (from 25 papers, 60 conditions, and 1482 infants) from prior studies of infants’ understanding of agents and objects. First, we estimated VOE and PN in the same data and found that showing infants an unexpected event drove their looking behavior as reliably as showing infants a visually novel event. Second, we tested whether VOE and PN are supported by similar or distinct mechanisms. Under the hypothesis that both responses are driven by a common mechanism, the same factors should predict the size of both effects. However, these responses were moderated by different predictors. The PN effect was larger after infants were habituated instead of familiarized, and did not differ depending on infant age. The VOE effect, in contrast, was not moderated by habituation vs familiarization and was larger for younger infants. These findings suggest that these two responses reflect distinct drivers of infant visual attention

An asynchronous, hands-off workflow for looking time experiments with infants.

“…Particularly in developmental studies, where behavioral readouts are often noisy and indirect and power is therefore low, small sample sizes inflate both false positive and false negative results (Button et al, 2013). Power analysis of the existing literature suggests that the sample size required for 80% power in VoE studies ( N = 80–100) is substantially larger than current standard study samples ( N = 20–30; Kunin et al, 2023). This meta-analytic estimate converges with our estimate for the Zoom and hands-off VoE study.…”

“…In the automated paradigm, all infants saw six trials, whereas in-lab infants watched up to nine trials until habituated (looking time reduced by half). Some meta-analytic evidence suggests that the familiarization procedure does not influence the size of observed VoE effects (Kunin et al, 2023). However, future studies could test whether the effect size in the hands-off workflow is more similar to the in-lab study when infants are allowed to habituate.…”

Naive psychology depends on naive physics

“…They recognize agents' intentions and goals, looking longer when agents change their goals or act inefficiently Luo & Baillargeon, 2005;Woodward, 1998). For reviews and meta-analyses of this literature, see: Carey (2009); Kunin et al (2023); Liu and Almeida (2023); Margoni et al (2023);. These expectations persist throughout our lives, guiding the judgments, predictions, and explanations of adults (Gao et al, 2009;Scholl, 2001;Scholl & Tremoulet, 2000;Shu et al, 2021;Smith et al, 2020).…”