Article
A compelling story can begin an investigation, but it cannot finish one. To test an extraordinary ability, researchers need to state what the person is claimed to do, what would count as success, and how the result differs from chance or ordinary information routes. The standards should be clear before the result is known.
Define the claim before testing it
An operational claim names the action, source or receiver if relevant, timing, outcome, and conditions. A telepathy test might specify which target a participant must identify and the chance level expected from guessing. A psychokinesis test might name the machine output and the direction or size of the predicted effect. If a claim shifts after a result, researchers cannot tell whether the original prediction succeeded.[1][5]
The measured outcome should be objective enough for another team to score the same way. Researchers should state what would count as a meaningful effect, not only whether a threshold was crossed. A p-value alone does not tell us the size, importance, precision, or truth of a claim; effect sizes and uncertainty belong beside it.[5]
Close ordinary routes for information
Blinding and randomization can reduce cues and selection effects when the design allows them. Teams should check for sensory leakage, target-handling clues, experimenter knowledge, scoring choices, and flexible analyses. These checks do not imply that a participant is deceptive. They ask whether ordinary information could have reached the outcome through a route the procedure did not intend.[1][2]
Preregistration records the hypothesis, exclusions, outcomes, and analysis before the data are known. Registered Reports move peer review before results, which can reduce pressure to favor striking findings. Neither practice makes a study immune to error, and exploratory results can still be valuable when identified as exploratory rather than presented as a prior prediction.[2][3]
Make the result repeatable and inspectable
Adequate power helps a study estimate a plausible effect with useful precision. Full reporting should include outcomes, effect sizes, uncertainty, deviations from plan, and null results. Sharing data and code can help others inspect an analysis when privacy, consent, and the study design permit it. None of these practices guarantees that an interpretation is correct, but each makes the evidence easier to examine.[1][3][5]
Independent replication matters because a finding can depend on one laboratory’s sample, materials, or procedure. The Open Science Collaboration tested 100 psychology studies, a coordinated effort that illustrates the value and difficulty of replication. Its results should not be treated as a universal rate for all fields or as a verdict on every contested topic.[4]
Let confidence grow with the evidence
A clear preregistered positive result is a reason to investigate further, not automatic proof of an ability. A null result limits the tested design and effect, not every imaginable version of a claim. Confidence becomes more reasonable when independent teams reproduce a defined result, the estimate is precise enough to interpret, and ordinary explanations have been seriously examined.[1][2][4]
If a surprising claim were tested again, which safeguard would most change your confidence: a blinded procedure, a predeclared analysis, open reporting, or replication by an independent team?