In a 2025 Proceedings of the National Academy of Sciences article, Jeremy Freese and colleagues revisit a long-standing concern in the social sciences: the file drawer problem. This term refers to the selective non-publication of studies with null or non-significant results, which biases the evidentiary record.

Focusing on social science survey experiments—a domain that has rapidly expanded in recent decades—the authors provide systematic evidence that such publication bias remains both prevalent and consequential. They show that the apparent robustness of experimental findings is often inflated, not because of methodological flaws in individual studies, but because negative or null results are disproportionately excluded from the published literature.

Freese and co-authors argue that addressing this problem requires both cultural and institutional change: greater uptake of pre-registration, incentives for journals to publish null results, and norms of transparency that recognize the scientific value of well-executed studies regardless of outcome. By demonstrating the extent of bias and its implications for cumulative knowledge, the article makes a forceful case for reform.