Precognition

← Encyclopedia

Precognition is the claim that the future can be known before it happens — not inferred from present signs, but perceived directly, through a channel no one has named. The experience that keeps the claim alive is ordinary and unsettling: a dream vivid enough to be told at breakfast, and an event, days later, that seems to have been its original. Antiquity built oracles and arts of divination on it; the dream has remained the claim’s natural home ever since.

The scrutiny is nearly as old. Aristotle allowed in On Divination in Sleep that a few dreams might foretoken what followed, but classed most as “mere coincidences.” The word itself surfaced in the seventeenth century, then waited three hundred years for an instrument. What any instrument would face was clear by 1932, when psychologists collected the public’s dreams about the kidnapped Lindbergh baby: of some 1,300, about five percent saw the child dead, and just four placed the grave among trees — the detail that proved true. Enough dreamers generate previews of everything. The question was never whether the future is sometimes dreamed; it is whether it is dreamed more often than chance requires.

The first modern theory came from an engineer. J. W. Dunne — British soldier and aeronautical designer — was troubled for years by dreams that seemed to anticipate his waking life, and in March 1927 published An Experiment with Time, arguing that everyone dreams the future and almost no one notices. His sharpest observation cut against the prophetic tradition itself: a dream of volcanic catastrophe matched not the eruption but the inaccurate newspaper account he later misread — what a dream foresees, Dunne concluded, is never the event, only the dreamer’s own future experience of it. Over this he raised “Serialism,” a many-levelled theory of time; a 1932 test with the Society for Psychical Research ended in disagreement. The philosophers were kinder than the experiments — C. D. Broad judged Dunne’s the only theory of the matter “worth consideration” — and the book marked Wells, Priestley, Stapledon and, later, Nabokov.

The laboratory era began in the 1930s at Duke University, where J. B. and Louisa Rhine ran the first sustained experimental program: forced-choice guessing, in advance, of the order of a twenty-five-card, five-symbol deck, chance five hits. Rhine’s positive results won some academic notice; his methods did not survive it — flawed mathematics, unblinded procedures, cards thin enough to read from behind — and more rigorous successors could not reproduce the scores. Yet when Charles Honorton and Diane Ferrari pooled the whole forced-choice tradition in 1989 — 309 experiments by 62 investigators between 1935 and 1987, more than 50,000 participants — a minute overall effect of 0.02 carried a Stouffer Z of 6.02, p = 1.1 × 10⁻⁹, needing, by their calculation, 46 unreported null studies for every published one to fall to chance. The field’s lasting shape was set: single experiments that collapse on inspection, and aggregates that refuse to go away.

Daryl Bem is the reason precognition now appears in histories of psychology that never mention card decks. A Cornell social psychologist of long standing, who had reviewed the telepathy experiments sympathetically in the early 1990s, Bem sent the Journal of Personality and Social Psychology, a flagship that rejects some 85 percent of submissions, a paper titled “Feeling the Future.” Its design was simple: run well-established psychological effects in reverse, recording the response before the stimulus that should cause it. Nine experiments and more than 1,000 participants time-reversed familiar effects — approach and avoidance, priming, habituation — along with a “retroactive facilitation of recall,” in which memory-test scores appeared to improve with practice taken after the test. Eight of nine experiments reached significance; the mean effect size was d = 0.22, the combined Stouffer Z 6.66, p = 2.68 × 10⁻¹¹. Bem framed it all as “the anomalous retroactive influence of some future event on an individual’s current responses.” The paper passed ordinary peer review — “The level of proof here was ordinary,” the social psychologist Lee Ross explained. That was the alarm. Either standard methods had just detected the future leaking backward into the mind, or standard methods could not be trusted.

The same journal issue carried the counterattack. Eric-Jan Wagenmakers — who later said reading the paper made him physically unwell — and three colleagues argued that Bem’s analyses were partly exploratory, that one-sided p-values overstated the case, and that a default Bayesian t-test rated the evidence for psi “weak to nonexistent”; their conclusion aimed not at precognition but at psychology, which must change how it analyses data. Jeffrey Rouder and Richard Morey, pooling the Bayesian accounting across studies, found somewhat more — a Bayes factor near 40 for emotionally charged images — and judged even that orders of magnitude below what appropriate skepticism demands for ESP. Others noted that eight significances in nine attempts is, at these effect sizes, itself improbable — the signature of a hidden file drawer. None of it faulted Bem’s arithmetic, only what arithmetic of that kind could show.

Then came the replications. Stuart Ritchie, Richard Wiseman and Christopher French pre-registered three independent replications of Bem’s ninth experiment — the recall study, his largest effect at d = 0.42 and, he had said, among the easiest to repeat. All three came up empty: combined n 150, combined one-tailed p .83. Publishing the null proved harder than producing it. JPSP declined the paper without review, saying it “does not publish replications”; Science and Psychological Science declined likewise; the British Journal of Psychology sent it to two referees, one favourable and one not — the unfavourable referee, both parties later confirmed, was Bem — and rejected it. The study finally appeared in PLoS ONE in 2012, the refusals by then a story in their own right: the literature’s front door had opened to a demonstration of precognition and closed to its disconfirmation. Jeff Galak and colleagues followed at scale — seven experiments, 3,289 participants — and found nothing either; a meta-analysis of all replication attempts put the recall effect at d = 0.04, indistinguishable from zero.

Bem’s answer was aggregation. With three co-authors he meta-analysed 90 feeling-the-future experiments from 33 laboratories in 14 countries — 12,406 participants — and reported a small, stubborn signal: Hedges’ g = 0.09, z = 6.33, a Bayes factor of 5.1 × 10⁹; excluding his own experiments left 69 studies at g = 0.06, still formally decisive by his critics’ own Bayesian conventions. The reply was structural, not arithmetical: a meta-analysis of heterogeneous, partly unpublished studies inherits the biases of its inputs — in the flat verdict of the methodologists Leif Nelson, Joseph Simmons and Uri Simonsohn, “meta-analytical thinking increases the prevalence of false positives.” There the record rests, each camp regarding the other’s instrument as the artifact: pre-registered replications that find nothing, meta-analyses that keep finding the same small thing.

The affair left precognition where it found it; its lasting effects fell on how psychology tests anything at all. In the history psychology tells about itself, Bem’s paper opens the replication crisis: the journalist Daniel Engber wrote in 2017 that if one had to choose a single moment that set the crisis off, this might be it — though it was one of several roughly coincident shocks of 2010–2012. The reforms that followed read as a list of what Bem’s critics had found wanting: confirmatory and Bayesian analysis; pre-registration, which Ritchie, Wiseman and French had already turned on Bem; Registered Reports, in which a journal accepts a study’s design before any data exist, decoupling publication from result. When the Open Science Collaboration replicated 100 published studies in 2015, 97 percent of the originals had been significant; 36 percent of the replications were, at about half the original effect sizes. Bem’s results, it turned out, were not an aberration of parapsychology. They were ordinary — which had been Lee Ross’s defense, read the other way.

One wing of the field had meanwhile gone physiological. In “presentiment” experiments — pioneered by Dean Radin in 1997, later relabelled predictive anticipatory activity — the body is the instrument: skin conductance, heart rate, pupil size in the seconds before a randomly selected image, arousing or calm, appears on a screen. The claim under study: that physiology shifts before the random draw, by a margin the subject never feels. A 2012 meta-analysis by Julia Mossbridge, Patrizio Tressoldi and Jessica Utts gathered 26 reports from seven laboratories, 1978–2010, and found a combined effect of 0.21, z = 6.9 — the higher-quality studies showing slightly larger effects, not smaller. The authors kept the framing modest: the cause, they wrote, “undoubtedly lies within the realm of natural physical processes,” and remains to be determined. The neuroscientist Samuel Schwarzkopf answered that claims this seismic require commensurate evidence and that artifacts in how each trial’s baseline is computed could manufacture the anticipation — while conceding that the most popular counter-explanation, drifting expectations, was a “straw man.” The exchange is the literature in miniature: a small effect, an unsettled artifact question, no mechanism on either side.

Behind every experiment stands the oldest objection: an event influencing a mind before the event happens reverses the antecedence that runs through physics and nearly all of neuroscience. The objection is strong; it is not, philosophers note, a proof of impossibility. Backward causation has a small but serious literature — the “bilking” argument, tachyons, the Wheeler–Feynman absorber theory, retrocausal readings of delayed-choice quantum experiments — and Tim Maudlin has priced the admission honestly: accept it, and “the metaphysical picture of the past generating the future” must be abandoned. None of this work endorses precognition; it establishes only that the claim is no contradiction in terms, leaving the burden on evidence — where it has not yet been carried.

The end of the matter divides in two. That small anomalous deviations recur in meta-analyses of these experiments is a live empirical question, under study and unexplained in any direction. That precognition has been demonstrated is a different claim altogether — contested at every joint and, by the mainstream’s standing verdict, without accepted scientific evidence. Dunne’s old rule, on one editorial reading, held one more time. A precognitive dream, he insisted, never shows the event itself, only the dreamer’s own future experience of it. A decade of staring at “Feeling the Future” changed the discipline more than it changed the claim: pre-registration, replication, reform.

→ Related: Telepathy · Remote Viewing · J B Rhine · Divination · Synchronicity