A German retroactive priming task shows mixed effects_

The task

In a retroactive priming task, a person judges a picture as pleasant or unpleasant as fast as they can, and a word appears only after they have pressed the button. If a word shown after the response still speeds up a matching judgement, the sequence appears to run backwards in time. This study rebuilt the task in German and ran it in two laboratory studies, with Bem as a co-author.

The confirmatory result

Under the test that was fixed in advance, the predicted effect did not appear. Study 1 was not significant. Study 2, which added a purpose-built sham control, produced a weak significant effect in the opposite direction to the one predicted. On the confirmatory criteria the psi hypothesis was rejected in both studies.

The after-the-fact signal

An exploratory split by gender found a significant effect for men only in Study 1, with nothing in women. The authors label this clearly as post hoc, not as a confirmed finding.

Why it matters to us

This is what an honest anomalies result usually looks like. The planned test comes back null, a subgroup lights up afterwards, and a sceptic and a believer can read the same table and each feel confirmed. The authors say as much.

It is exactly why we fix the analysis before the data exists and never promote a subgroup found after the fact to a headline. A Wittmann paper, co-authored by Bem, reporting its own null is the standard we hold ourselves to.