From representation to inference: Rethinking generalization in open-access developmental science

Abstract Widely reused open-access data sets shape theory, training, and syntheses of evidence in child and adolescent development, making them infrastructure for cumulative science. In this article, I suggest that representation is an inferential constraint: Generalization depends on who is sampled and what contexts and constructs are measured and documented. Synthesizing cross-data set patterns across developmental data sources, I identify recurring limits, including overrepresentation of higher-resourced and majority-culture families, inconsistent socioeconomic measures, uneven capture of contextual variables, and partial visibility of disability and neurodevelopmental variation. These features bound what secondary analyses can support and make limits on generalization easy to overlook. Transparent metadata and privacy-sensitive governance help make those limits visible while protecting participants. I conclude with practical steps for data set teams, funders, journal editors and reviewers, and researchers.