Fear conditioning research through the lens of meta-analyses: A systematic methodological review

Meta-analyses have become foundational tools for evidence synthesis and are receiving increasing attention across research disciplines, including fear conditioning - a central paradigm in affective, behavioral, and clinical neuroscience and psychology - where the number of meta-analyses has notably increased in recent years. However, substantial methodological heterogeneity among these meta-analyses complicates their interpretation and comparability. This review systematically summarizes existing meta-analyses in fear conditioning (n = 26), with particular emphasis on methodological choices and reporting standards. Our findings indicate substantial variability in reporting practices, particularly regarding review processes, applied methodological models, and the operationalization of certain variables, which may challenge reproducibility and interpretability. Most meta-analyses used aggregate data and random-effects models, while moderator analyses were frequently conducted but yielded few significant moderators. Outcome measures were frequently aggregated (e.g., skin conductance, startle responses, ratings) without a clear theoretical rationale. We discuss the implications of moderator analyses, the limited attention to temporal dynamics, and the aggregation of diverse outcome measures. We argue that the field would benefit from more precise reporting, theory-driven analytical decisions, and increased data sharing at both the primary and meta-analytic levels. Such improvements would facilitate individual participant data meta-analyses, which preserve within-study variability, enhance moderator analyses, and address challenges inherent to aggregate-level data synthesis. Establishing and expanding shared infrastructures represents a necessary step toward cumulative, reproducible, and clinically informative work in the future.