Falsification
What would change our minds?
A claim that cannot fail a test cannot be supported by one. Before each study we state, in advance, what pattern of data would count against the continuity hypothesis and what pattern would count against a conventional explanation.
Observations that would count against continuity
- Judges cannot match blinded readings to the intended target above the chance baseline defined in advance.
- Material introduced in an anchor session does not reappear in later independent sessions more often than in unrelated control sessions.
- Consistency across sessions disappears once transcripts are stripped of generic, high-base-rate statements.
- Apparent effects track how much information was available to the medium rather than which identity was tested.
- Effects are present in one team's sessions and absent under independent replication.
Observations that would count against a conventional explanation
- Blinded judges match readings to the intended target reliably, at effect sizes that survive preregistered analysis and replication.
- Specific, low-base-rate material recurs across mediums who never had access to each other or to the target record.
- Sealed answers, recorded before any session, are matched at rates that sampling error and cueing cannot explain.
- Effects persist when protocol audits, leakage checks and exposure logs show no route by which the information could have been obtained.
Why state this in advance
If a test can be reinterpreted after the fact, it is not a test. Writing the failure conditions down before data collection removes the freedom to move the goalposts, in either direction, once results are visible.
The commitments that make this enforceable
- Hypotheses, thresholds and analysis plans are filed before data collection.
- Exclusions are defined in advance; any later exclusion is reported with its reason.
- Analysts work on coded data and do not choose between analyses after seeing outcomes.
- Every deviation from the protocol is published with the study report.