Most significant psychology findings are likely real?

Using nearly 36k psychology papers (1975–2017) and almost 488k test statistics, the author quantifies trends in statistical power, publication selection bias, and p‑hacking and estimates that, under plausible assumptions, roughly 80%+ of reported significant results reflect genuine effects rather than pure artifacts. The paper shows that power has been low and p-hacking/selection bias substantial, but that even so the share of false discoveries may be well under the majority in many scenarios. — This reframes the reproducibility debate away from 'everything is false' toward a quantified, conditional assessment that affects public trust, funding priorities, and how policymakers should treat single-study claims.