The replication scorecard

Popular psychology is full of famous findings. Some survived rigorous re-testing. Some didn't. This page is the honest ledger behind everything written here, updated as the evidence moves.

Held up

• Anchoring. The first number drags every number after it. One of the most reliably reproduced effects across labs (relevant anchors, not the random-number party trick).
• Framing. Save a third versus lose two thirds flips choices with the math unchanged. Replicated at scale.
• Social proof. We read the crowd to decide what is normal, across countries and decades.
• Default effects. Whatever is pre-chosen tends to stay chosen, from organ donor forms to pension plans.
• The compromise effect. Adding a worse or extreme option shifts what the middle looks like.
• Loss aversion. Losses weigh roughly twice as much as same-size gains in most tested settings.

Did not hold up (don't build on these)

• Social and money priming. Words on a desk making you walk slower, images of money making you selfish. Failed under automated measurement and preregistration.
• Ego depletion. Willpower as a draining tank. Twenty-three labs, about two thousand people, pooled effect near zero.
• Power posing. No hormone or risk-taking effects in the large re-test; the original lead author publicly stopped believing it.
• The deterministic marshmallow story. The famous life-outcome claims shrink drastically with controls.
• The random-number anchor. Wheel-of-fortune style incidental anchors are far shakier than the textbook version.
• The God helmet. The sensed presence tracked suggestibility, not the magnets, under double-blind testing.

Why trust this site: because it keeps this list. Believing a famous finding was never the mistake. Refusing to check is the mistake.