Retrocausality

← Encyclopedia

An effect arriving before its cause sounds less like a thesis than a mistake. Philosophers have a name for the cleanest objection to it. Suppose some later action C is supposed to bring about an earlier event E; once E has been seen to happen, an agent could simply refuse to perform C — and the alleged backward link breaks, every time, on demand. Max Black laid this trap in 1956 and called it bilking, and for a generation it stood as the reason no serious person need entertain a cause that works in reverse. The interesting thing about retrocausality is not that the objection is wrong but how narrowly it holds, and where the loopholes lead: into the equations of physics, into one of the most misdescribed experiments in the popular literature, and into a notorious failure of the psychology lab.

Michael Dummett answered Black first, in 1964, by noticing that the bilking contradiction needs an incoherent recipe to get going — one that never says in what order C and E actually occur — so the trouble is not special to backward causation. Huw Price sharpened the diagnosis in 1996. The objection assumes the agent can learn that E has happened without that knowledge itself having altered E, and for ordinary past events that independence is exactly what human beings enjoy and what makes the past feel fixed. With respect to the future, no such independent access exists; one cannot peek at tomorrow the way one reads a record of yesterday. The asymmetry the bilking argument leans on, then, is a fact about epistemic access, not a law of metaphysics — and in the quantum case it fails outright. Any correlation between a future measurement setting and the hidden state of a particle measured now cannot be bilked, because the agent can never gain the advance knowledge of that state the machine would require. The condition that would let one cheat is, as Price put it, perennially violated — and that single fact governs everything below: wherever retrocausality has a foothold in real physics, it carries no usable signal into the past.

The foothold is older than quantum theory’s modern form. In 1945 John Wheeler and Richard Feynman rebuilt classical electrodynamics with no independent field — only charges acting on charges, half through the ordinary retarded wave that runs forward in time and half through an advanced wave that runs backward. The advanced half was not an embarrassment to be discarded; it did work, deriving the drag an accelerating charge feels as the response of future absorbers reaching back to the present. The mathematics was time-symmetric by construction, and it left behind a vocabulary — the advanced wave, the future absorber — that every later model would borrow. John Cramer borrowed it in 1986 for his transactional interpretation of quantum mechanics. A quantum event, on his account, is a handshake: the source emits an offer wave forward in time, an absorber answers with a confirmation wave backward in time, and the standing pattern of the two completes the transaction, the Born probabilities falling out of the product. Yet Cramer was emphatic about how to read this. The back-and-forth describes an event itself outside time, not a literal sequence of telegrams sent up and down the years, and nothing in it permits a message to travel backward.

The same restraint marks the most rigorous version of the idea. In 1964 Yakir Aharonov, with Peter Bergmann and Joel Lebowitz, described a quantum system between two measurements using two state vectors at once — the familiar one evolving forward from how the system was prepared, and a second evolving backward from how it is finally found. A full account, on this two-state picture, needs both boundaries, the future condition as much as the past one. What must be said plainly is that this formalism predicts nothing the standard theory does not; it is empirically equivalent to ordinary quantum mechanics, a re-description and the natural language for weak measurements, compatible with a retrocausal world without compelling one. That is the honest status of the whole serious strand: the equations are real and respected; the claim that the universe literally contains a backward influence is an interpretive choice, and a minority one.

Why make the choice at all? Because of Bell’s theorem, which forces any theory matching the quantum correlations to surrender either locality — the principle that nothing here instantly disturbs the facts there — or the independence of a particle’s hidden state from a measurement setting not yet selected. Most physicists give up the second only by accident, through superdeterminism, which few find palatable. The retrocausal program gives it up on purpose and for a reason: if a future setting can causally shape the earlier hidden state, the needed correlation appears for a plain mechanical cause, and locality survives. Olivier Costa de Beauregard proposed the earliest such picture, entangled partners staying correlated along a zigzag through spacetime — forward to the source, backward to the other wing — rather than by a spacelike leap. Price built the philosophical case that this is what relativity’s time-symmetry was asking for all along, and Ken Wharton developed all-at-once models in which a measurement acts as a future boundary condition on a field, the quantum statistics emerging from ignorance of the whole trajectory. No working physicist is obliged to accept any of it; but it is a coherent research program, not a confusion, earning its place as the one move that saves locality after Bell.

None of which is what the famous experiments show, though they are forever enlisted to show it. Wheeler proposed his delayed-choice setup in 1978: send a single photon into an interferometer and decide only after it has entered whether to measure wave-like interference or which path it took. The photon obliges the later choice every time, as if its earlier nature waited on a decision not yet made — and in the cosmic version the light has traveled for billions of years before anyone on Earth picks the question. The pull toward backward causation is obvious, and Wheeler himself refused it. One should not credit the photon with a definite history at all. “It is wrong,” he wrote, “to attribute a tangibility to the photon in all its travel from the point of entry to its last instant of flight.” There is no recorded wave-or-particle past sitting there to be rewritten. The later choice does not reach back and edit a trajectory; it determines what, if anything, a trajectory was ever a fact about.

The delayed-choice quantum eraser, proposed by Marlan Scully and Kai Drühl in 1982 and realized by Kim and colleagues in 2000, tightens the illusion. Here the which-path information for a detected photon is carried by an entangled partner, and a measurement on that partner — performed after the first photon has already struck its screen — can seem to erase the information and restore interference. The crucial fact is where the interference lives. It appears only in coincidence counts, in subsets of the detections sorted according to the partner’s later result. The full, unsorted pattern on the screen shows no fringes whatever, no matter what is chosen afterward, so the marginal record of the earlier detector never changes. The later choice alters not the past but how an existing dataset may be partitioned; the correlations were present all along, waiting to be revealed by the sorting. The scheme sends no signal backward, and standard quantum mechanics reproduces every count without a whisper of time-travel. As the physicist Jonte Hance has put it, the eraser shows that the correlations entanglement provides “have to be able to fit every possible way you could measure a system.” It rewrites the bookkeeping of observers, and leaves the past exactly as it found it.

These two animals must not be confused: the interpretations of the previous strand postulate a backward influence as a deliberate theoretical price, while the delayed-choice experiments postulate nothing of the kind and are merely misread as if they did. And both must be kept clear of the third strand, where retrocausality entered the laboratory as an empirical claim and did not survive. In 2011 the social psychologist Daryl Bem published “Feeling the Future” in a leading journal — nine experiments, more than a thousand participants, each one a well-known psychological effect run in reverse, so that a stimulus selected at random in the future appeared to influence a response already given. Eight of the nine reached significance. A claim of precognition from a respected researcher in a top-tier journal, it landed like a detonation, and it did not hold. Statisticians, Eric-Jan Wagenmakers among them, found one-sided tests where two were due and exploratory analyses dressed as confirmations; the psychologist James Alcock judged that “just about everything that could be done wrong in an experiment occurred here.” Ritchie, Wiseman, and French ran three pre-registered replications in 2012 and found nothing, the combined result indistinguishable from chance. Galak and colleagues followed with seven more failures and a meta-analysis of every known attempt, more than four thousand participants, returning an average effect of essentially zero. The episode now serves less as evidence for anomalous cognition than as one of the founding cases of psychology’s replication crisis — the fringe face of an idea whose serious versions never claimed anything so usable.

That contrast is where the strands sort themselves, and where the older cosmologies the archive keeps may seem, for a moment, to find an echo in the physics. The hermetic and magical traditions took for granted that an act could reach across time as freely as across space — a rite securing a harvest not yet sown, the end present in the beginning and drawing it on. Retrocausality is the one place modern physics has held the door for backward influence at all, and it is instructive how it holds it: only in time-symmetric mathematics that makes no new prediction, only as a minority reading chosen to spare locality, and always under the iron condition that nothing usable passes back. The advanced wave of the transaction carries no command to the past; the eraser sorts data it cannot alter; the precognition study, the one promise of a future felt before it arrived, dissolved on replication. The traditions wanted a past one could still work upon. What the formalism allows is symmetric and silent — a backward arrow that, wherever it is real, can be written into the equations and never sent.

→ Related: Quantum Measurement Problem · Block Universe · Precognition · Quantum Entanglement · Many Worlds Interpretation