The replication crisis: where are we in 2024?

Five years after the 'crisis' became mainstream discourse, it's worth asking what actually changed. What improved: pre-registration has become standard in several fields (clinical trials were already required; psychology, economics, and political science have followed). Registered Reports — where journals accept a paper based on the design before results are known — have eliminated a meaningful chunk of p-hacking. What hasn't improved: publication bias is still substantial. The incentive to produce positive novel findings remains. Lab culture at most institutions hasn't changed. The careers of people who ran influential non-replicating studies have mostly been fine. The most interesting development is field-specific heterogeneity. Psychology has had significant reforms. Preclinical biomedical research (animal models for drug targets) remains deeply problematic — Begley and Ellis (2012) estimated ~89% of landmark cancer findings failed to replicate and the situation has probably not improved dramatically. What I'm watching: multi-site replication projects as a normal part of any high-stakes finding, and stronger incentives for replication studies as citable research.