Human infants are born with their eyes open and an otherwise limited motor repertoire; thus, studies measuring infant looking are commonly used to investigate the developmental origins of perception and cognition. However, scholars have long expressed concerns about the reliability and interpretation of looking behaviours. We evaluated these concerns using a pre-registered (https://osf.io/jghc3), systematic meta-analysis of 76 published and unpublished studies of infants’ early physical and psychological reasoning (total n = 1,899; 3- to 12-month-old infants; database search and call for unpublished studies conducted July to August 2022). We studied two effects in the same datasets: looking towards expected versus unexpected events (violation of expectation (VOE)) and looking towards visually familiar versus visually novel events (perceptual novelty (PN)). Most studies implemented methods to minimize the risk of bias (for example, ensuring that experimenters were naive to the conditions and reporting inter-rater reliability). There was mixed evidence about publication bias for the VOE effect. Most centrally to our research aims, we found that these two effects varied systematically—with roughly equal effect sizes (VOE, standardized mean difference 0.290 and 95% confidence interval (0.208, 0.372); PN, standardized mean difference 0.239 and 95% confidence interval (0.109, 0.369))—but independently, based on different predictors. Age predicted infants’ looking responses to unexpected events, but not visually novel events. Habituation predicted infants’ looking responses to visually novel events, but not unexpected events. From these findings, we suggest that conceptual and perceptual novelty independently influence infants’ looking behaviour.

人类婴儿出生时眼睛是睁着的，否则运动能力有限;因此，测量婴儿外表的研究通常用于调查感知和认知的发育起源。然而，学者们长期以来一直对观看行为的可靠性和解释表示担忧。我们使用预先注册 （https://osf.io/jghc3） 的系统荟萃分析来评估这些问题，该分析涉及 76 项已发表和未发表的婴儿早期生理和心理推理研究 （总计 n = 1,899;3 至 12 个月大的婴儿;数据库搜索和征集未发表的研究2022 年 7 月至 8 月进行）。我们在同一数据集中研究了两种效应：关注预期与意外事件（违反期望 （VOE））和关注视觉上熟悉的事件与视觉上新颖的事件（感知新颖性 （PN））。大多数研究实施了将偏倚风险降至最低的方法（例如，确保实验者对条件不了解并报告评分者间的可靠性）。关于 VOE 效应的发表偏倚的证据不一。我们研究目标的最核心是，我们发现这两种效应系统性地变化——效应量大致相等（VOE，标准化平均差 0.290 和 95% 置信区间 （0.208， 0.372）;PN，标准化均数差 0.239 和 95% 置信区间 （0.109， 0.369））——但独立地基于不同的预测变量。年龄预测了婴儿对意外事件的视觉反应，但不能预测视觉上新奇的事件。习惯化预测了婴儿对视觉新奇事件的看向反应，但不能预测意外事件。从这些发现中，我们认为概念和知觉新奇性独立影响婴儿的观看行为。