Dreams and prediction claims: how to test a match

Define the measure before calling it a prediction

If both a dream and a later event can be interpreted flexibly, many matches can be constructed. Ask whether the dream was recorded before the event and whether that record was preserved unchanged. An interpretation made after an event is a different type of data.

Count the non-matches too

Selecting only examples that matched leaves the comparison incomplete. How many records existed, how many events did not match and what was the matching criterion? Without these, you cannot calculate a success rate. Selecting one interesting example from many accounts by many people over a long period differs from testing a prediction specified in advance.

A proposed investigation design

1. Write the claim and time interval before the investigation begins.
2. Preserve the records without changing them afterwards.
3. Set the matching rules before looking at the examples.
4. Where possible, keep the person assessing the records unaware of the outcome.
5. Report all records and ambiguous examples.

This outline is an exercise proposed for the ELYADAL web guide. It is not a completed experiment or laboratory result. Preregistration helps distinguish the planned work from analyses developed later. If there are no supporting data, use “not demonstrated” and explain why.