Prophetic dreams, coincidence and the arithmetic of a very large number of nights

The precognitive dream is among the most persistent claims in the folklore of sleep. It is also one where the arithmetic is worth doing carefully, because the intuitive sense of how unlikely a coincidence is turns out to be badly calibrated.

Start with the numbers

People have several dreams a night — four to six REM periods, plus non-REM mentation. Even counting only remembered dreams conservatively, a person might recall one or two a week.

Across a population of, say, a hundred million adults, that is on the order of hundreds of millions of remembered dreams a week. Billions a year.

Against that denominator, dreams that appear to anticipate events are not merely likely — they are statistically inevitable. Given enough trials, extremely improbable individual events occur routinely. The right question is never "what are the odds this dream matched?" but "given how many dreams are generated, how many matches would we expect by chance?" — and the answer is a great many.

The psychological mechanisms

Probability alone does not explain the strength of the experience. Several well-documented biases do.

Confirmation bias and selective recall. Dreams that match subsequent events are memorable and get retold. The overwhelming majority that match nothing are forgotten within minutes. The sample being reasoned from is filtered to remove all the failures.

Retrospective matching. Dreams are ambiguous, and interpretation is generous after the fact. A dream about "something bad involving water" can be matched to a flood, a drowning, a burst pipe, a shipwreck on the news, or a difficult conversation about a holiday. Before the event, no specific prediction was made.

Reconstructive memory. Human memory is not a recording. Recalling a dream after learning what happened introduces the new information into the memory itself — a robust finding in eyewitness research. People sincerely remember the dream as having been more specific than it was, and there is no dishonesty in it.

Anxiety-driven content. The most commonly reported precognitive dreams involve accidents, illness and death — which are also the most common contents of ordinary anxiety dreams. Someone worrying about an elderly parent will dream about their death repeatedly over years. When the parent eventually dies, one of those dreams will have been recent.

Everyday inference. Considerable knowledge is processed without conscious awareness. A dream that a colleague will leave, or that a relationship will end, or that a relative is unwell, may reflect accurate reading of evidence already available.

The experimental record

Precognition has been tested formally. J.B. Rhine's Duke laboratory from the 1930s, and later the Maimonides dream telepathy studies of the 1960s and 1970s run by Montague Ullman and Stanley Krippner, produced positive results that generated considerable interest.

Subsequent assessment has been unkind. The Maimonides work has been criticised on methodological grounds — judging procedures, potential sensory leakage, and multiple analyses — and independent replication attempts have generally failed. Meta-analyses in this area are contested and the field has not produced a protocol that yields consistent results in the hands of sceptical investigators.

Daryl Bem's 2011 precognition papers, published in a mainstream psychology journal, prompted large-scale replication attempts that were negative, and are now cited more often as a case study in the replication crisis than as evidence for precognition.

Why the experience matters anyway

Explaining a phenomenon statistically does not make it unimportant. Two things are worth saying.

The subjective conviction is genuinely powerful. People who have had a striking dream coincidence frequently describe it as the most compelling experience of their lives, and the feeling of significance is not reduced by knowing the base rates.

And such dreams tend to occur around things that matter — illness, death, separation. They are, on the continuity hypothesis, an accurate reflection of what the dreamer was already preoccupied with. That is meaningful information about the dreamer, even where it carries none about the future.

A test you can run

If you want to evaluate your own, the method is straightforward and rarely attempted.

Write dreams down before anything happens, in specific terms, with dates. Record every dream, not only the striking ones. Then, after some months, count: how many were followed by matching events, and how many were not?

Almost nobody does this, and the reason is instructive. Without a written record made in advance, the phenomenon depends entirely on memory reconstructed after the fact — which is precisely the condition under which the biases above operate most freely.