On the Delayed-Choice Experiment, the Death of the Moving Present, and What It Means That the Past Was Never Quite Where We Thought It Was
After the Black Hole, Another Rabbit Hole
A few weeks ago I wrote an essay called The Prison and the Map, in which a two-in-the-morning question about GPS coordinates carried me, by what felt like reasonable steps at the time, to the proposition that we are almost certainly living inside the interior geometry of a black hole. I thought I was finished with cosmological time for a while. The universe, as it tends to, had other ideas.
The trouble began with a video. One of those algorithm-suggested physics explanations slipped into my queue while I was actually trying to look up something else, and forty minutes later I was staring at footage of a French laboratory in Orsay with the uneasy expression of a man who has just realized the floor is not where he left it. The experiment they were running should not, by any classical understanding, have produced the result it kept producing. It had been producing that result for nearly twenty years. Nobody seemed especially troubled by this except me.
The standard caveat applies, as it does whenever I write about physics. I am not a physicist. I am a painter, a writer, a student of the Arts who has always found the hard sciences more interesting from the side door than through the front. I will get some things wrong. The argument I want to make does not, fortunately, depend on technical mastery. It depends on what the experimental result implies about time, which is something I have been thinking about, in one form or another, since the 2004 paper. Bear with me.
What Wheeler Was Actually Asking
The experiment I am circling is called the delayed-choice experiment, devised in 1978 by John Archibald Wheeler. Wheeler was one of the great physicists of the twentieth century, a student of Bohr’s, the man who coined the term black hole, and someone temperamentally inclined to ask questions that other physicists had quietly agreed not to ask. He was eighty-six when the experiment was finally realized in a laboratory. He lived just long enough to see it work.
The setup goes like this. You take the classic double-slit experiment, the textbook demonstration of wave-particle duality. A photon is fired at a barrier with two slits in it, and behind the barrier is a detection screen. If you do not measure which slit the photon went through, you get an interference pattern, the unmistakable signature of wave behavior. If you do measure which slit, you get two simple bands, particle behavior. Acquiring the information appears to change what the photon was doing.
The conventional answer to the obvious follow-up question, when does the photon decide whether it is a wave or a particle, is that it decides at the slits. Wheeler asked something sharper. What if you waited until after the photon was past the slits, after it was nearly upon the detector, and only then chose whether to measure which path it took? The classical intuition is that the photon could not possibly know what you were going to do later, and therefore must have committed to one description or the other when it passed the slits. Wheeler’s prediction, following standard quantum mechanics rigorously, was that the photon would behave according to whatever choice the experimenter made, no matter how late the choice was made.
In its cleaner version, Wheeler proposed using a Mach-Zehnder interferometer. A photon enters, hits a beam-splitter, takes two possible paths, and arrives at a recombination point. If a second beam-splitter is in place at the recombination, you get interference, wave behavior. If you remove the second beam-splitter, two detectors at the path ends record which path the photon took, particle behavior. The trick is to insert or remove that second beam-splitter after the photon is already in the apparatus. The trick, that is, is to make the choice late.
For thirty years this lived as a thought experiment, with the technical objections circling. The principal worry was that the choice, however quick, might still in some sense reach the photon at its entry. A genuinely random and genuinely delayed choice was needed, and the technology was slow in coming.
The Photons Did the Thing
In 2007, the team at the Institut d’Optique in Orsay, led by Vincent Jacques, performed the experiment in its strict form. The choice of configuration was made by a quantum random-number generator, which is about as random as anything physics will let you have, and the choice was triggered after the photon had entered the interferometer, with sufficient spatial separation that no light-speed signal could connect the choice to the entry. They ran it thousands of times. The photons behaved as waves when the wave configuration was selected and as particles when the particle configuration was selected, and they did this with a precision that would satisfy anyone with a stopwatch and a strong opinion about causality.
I want to underline this, because it took me a few minutes to absorb. The photon enters the apparatus. It commits to whatever it is going to commit to. Then, after that commitment, a quantum random number generator decides what configuration the apparatus will be in when the photon arrives at the detector. The photon’s behavior conforms to the configuration. Every. Single. Time. As if the question had been asked before the photon could have heard it, except the question demonstrably was not asked before the photon could have heard it, because the entire point of the experiment is to ask it later.
There is a still more elaborate version, published in 2000 by Yoon-Ho Kim and his collaborators under Marlan Scully’s supervision, called the delayed-choice quantum eraser. The version of this that broke my brain involves entangled photon pairs. One photon of each pair, the signal photon, passes through a double-slit apparatus and gets registered. Its entangled partner, the idler, is sent to a separate detector array configured to either preserve which-path information or to erase it. The choice of preservation or erasure is made after the signal photon has already been registered. After. Past tense. Done.
If you look at the signal photon’s record alone, you see no interference. If you sort that record according to what later happened to the idler, the signal photons whose partners had their which-path information erased show interference fringes, and those whose partners did not show none. This is not, technically, retroactive causation. The signal photons are not changing what they did. The correlation that was always there in the data becomes visible, or invisible, depending on how the later choice was made. But to my eye, sitting at a kitchen table at one in the morning, that distinction looks like the kind of distinction physicists make when they have run out of other distinctions to make.
A 2017 satellite-based experiment by Francesco Vedovato and his colleagues at Padua extended this to thousands of kilometers, with photons reflected from an orbiting reflector and the choice made on Earth while the photon was already on its return journey. It still worked. It works at every scale anyone has tested. It is one of the more thoroughly replicated counter-intuitive results in modern physics, and most people have never heard of it. Including me, until a few weeks ago. Welcome to the club.
Three Ways to Read This, None of Them Comforting
What does it mean? There are three principal interpretations on offer, and each charges a different ontological price. I will lay them out as honestly as I can. I have my preference, but I want to walk through each before showing my hand.
The first reading is Wheeler’s own. It holds that no quantum phenomenon possesses concrete reality until it has been registered by a measurement, and that the act of measurement participates constitutively in fixing what is registered. Wheeler called this participatory realism, and at its more austere edges it shades into a kind of operationalism in which physical reality consists in the relations between measurement outcomes rather than in any underlying substance whose properties those outcomes reveal. The past, on this view, is genuinely indeterminate prior to its observational closure. The advantage of this reading is its fidelity to the data without imposed metaphysical surplus. The cost is that the observer acquires a status uncomfortably close to ontological partner with the universe being observed, and the past loses its standing as a settled domain. Wheeler did not flinch from these conclusions. I find myself wanting to.
The second reading is the eternalist or block-universe construal, which I described at length in The Prison and the Map and which I am increasingly inclined to think is the correct one. On this view, all events in spacetime exist as a single four-dimensional structure. What we call the flow of time is a feature of our embedded perspective rather than of the manifold itself. The delayed-choice paradox dissolves under this reading because there was never a moment at which the photon’s behavior was undetermined and a later moment at which it became determined. The photon’s traversal and the experimenter’s later choice are both coordinates within a single consistent four-dimensional structure, mutually fixed from the outset. Nothing causes anything backward, because the categories of before and after are local features of an observer’s worldline rather than fundamental relations in the manifold. The cost of this reading is the surrender of the moving present as a real feature of reality. The advantage is that the strangeness becomes geometrical rather than metaphysical. The past is fully real, fully determinate, and fully inaccessible. We have been here before.
The third reading is genuine retrocausality, formalised in Yakir Aharonov’s two-state vector formalism and earlier in John Cramer’s transactional interpretation. Here the state of the universe at any moment is fixed by boundary conditions imposed from both temporal directions, and microphysical causation runs both ways at the level beneath ordinary experience. Our perception of unidirectional flow is held to be an emergent thermodynamic feature rather than a primitive of physical law. I find this reading atmospheric and stylish, but I distrust it for the same reason I distrust most attempts to save the appearances by multiplying mechanisms. It does the work. So does eternalism, with fewer moving parts.
A common thread runs through all three. The classical conviction that the past is fixed and the future open finds no clear support in the underlying physics. The delayed-choice result joins relativistic simultaneity, the time-reversibility of the Schrödinger equation, and the CPT theorem as quiet erosions of the folk picture of time. What survives across every interpretation is the strict prohibition on signalling into the past. Whatever the apparatus is doing, you cannot use it to send a message backward. The strangeness lives wholly in the interpretation. The world remains, for ordinary purposes, well-behaved. Cold comfort, but I will take it.
What This Does to Time
Here is the part that should bother any thoughtful person, and which I have been turning over since the night the algorithm sent me down this hole.
The moving present, the ordinary common-sense conviction that now is ontologically privileged and that the past is settled while the future is open, finds no natural home in any of the three readings above. Wheeler dispenses with the determinacy of the past. The eternalist dispenses with the privileged present. The retrocausalist dispenses with the asymmetry of causation. All three recover the appearance of ordinary temporal experience as something local, perspectival, or thermodynamic. None of them grants the moving now the standing it claims for itself in everyday consciousness.
This is not a new conclusion in the history of philosophy. Parmenides anticipated it, Boethius refined it, McTaggart formalised it in his 1908 argument for the unreality of time. Hermann Weyl pressed it. Gödel pressed it harder. A long line of physicists whose mathematical commitments outran their psychological ones have arrived at versions of the same position. The delayed-choice experiment is a recent addition to a long catalogue of reasons for doubting that time, as it appears to us, is what time, in fact, is. The novelty is that the doubt now has a laboratory address.
For practical purposes, the question of which interpretation is correct may admit of no resolution. The three frameworks are empirically equivalent. They cannot be distinguished by experiment, since any experiment they could disagree about turns out to be one for which they predict the same outcome. The choice among them is a metaphysical preference, conditioned by which classical intuition the chooser is least willing to surrender. What no defensible reading permits is the retention of all of them at once.
I want to be clear about something. I am not arguing that science has dissolved time, or that the experience of duration is somehow unreal, or that we should adopt some flatland resignation in the face of the manifold. The experience of duration is, whatever else it is, the medium in which I write this and you read it. What I am saying is that the structure underwriting that experience is not what the experience advertises. The moving present is a local convenience. The deeper architecture is something else.
The Codex Was There First
I want to address something that readers familiar with my Codex of the Outer Realms will recognize as a recurring theme, because it bears directly on what the experiment seems to be saying.
The Codex was written, across its five volumes, on the premise that human temporal categories do not carry through to the deeper structure of reality, and that the contemplative posture appropriate to that condition is something other than the busy linear striving of ordinary life. The Heretical Shape of the Universe makes the case in terms drawn from cosmology and philosophy. Chaos Unveiled makes it in terms drawn from a particular reading of Azathoth as the anti-origin, the cycle without inception, the condition prior to causal structure itself. The argument was philosophical, and made no appeal to laboratory data because none was required. The contemplative traditions had been making versions of it for two thousand years before quantum mechanics arrived to embarrass us.
What the delayed-choice experiment supplies is not vindication, exactly. The Codex does not need vindication from physics, and I would distrust it if it did. What the experiment supplies is an empirical anomaly, conducted in well-lit rooms by sober technicians, whose implications can be stated with mathematical precision and whose results can be checked by anyone with the requisite apparatus. It is a piece of laboratory evidence consistent with every interpretation under which the everyday picture of time is mistaken, and inconsistent with no defensible interpretation under which that picture is preserved. That is, by the standards of empirical inquiry, a fairly substantial result.
A point of doctrinal precision worth noting. Wheeler’s participatory reading of his own experiment risks a slide into idealism, in which the observer’s act of registration confers reality upon the past. This is not the position the Codex takes. The Codex insists that the universe is indifferent rather than mind-dependent, and that the strangeness lies precisely in the irrelevance of consciousness to what obtains. The eternalist reading is the natural ally of this stance. The four-dimensional manifold contains us. We do not move through it. Its alien temporal structure exists whether or not anyone is watching. The horror, if horror is the appropriate word, is not that the past is malleable to our intentions. It is that the past was never the kind of thing the moving present takes it to be, and our experience of agency within time is a parochial feature of being embedded in something that does not, structurally, work the way we think it works.
The Codex was making this argument in the register of contemplative philosophy. The laboratory has joined it from a different direction, with no awareness that the conversation was already in progress. I find this more satisfying than any deliberate convergence I could have engineered. It is the same satisfaction the Prison and the Map essay arrived at, from a different angle. The physics keeps catching up.
Conclusions, or What I Worked Out Instead of Doing Anything Useful
The apparatus sits in a darkened laboratory. A photon enters. Somewhere downstream, a beam-splitter is inserted or withdrawn by the click of a quantum random-number generator, after the photon’s entry, after its commitment, after the moment by which any classical account would insist its fate had been sealed. The detector registers. The result conforms to the configuration that obtained at the moment of registration, with no regard for the order in which that configuration and the photon’s traversal occurred.
It is a small fact, in the scheme of things. It concerns a single photon, in a single apparatus, on a single afternoon. It implies, however, that the universe does not honor the temporal asymmetry on which our sense of agency depends, and that the past, properly understood, is something other than the receptacle of fixed events we have always taken it to be.
My 2004 paper concluded that time travel should not be attempted on ethical grounds. The Prison and the Map concluded that the geometry forbids it whether or not we have the manners. The delayed-choice experiment, as I read it, suggests something more disquieting: that the moving present is not the place where reality is being written. The writing has already happened. We are reading.
Whatever else the experiment teaches, it teaches that the deeper architecture is something other than what the experience advertises. The contemplative traditions have been saying this for centuries. The Codex says it in the register of philosophy. The laboratory now agrees, in the register of photon detectors and quantum random-number generators clicking away in a French basement at four in the morning. I find this conclusion both unsettling and rather marvelous, in the same proportion the Prison and the Map essay arrived at. The universe is, on every reading I can find, considerably stranger than the moving present permits us to see. I will take that.
I think I closed the laptop somewhere near three.
Sources
Wheeler, J. A. (1978). The “Past” and the “Delayed-Choice” Double-Slit Experiment. In A. R. Marlow (ed.), Mathematical Foundations of Quantum Theory. Academic Press.
Jacques, V., et al. (2007). Experimental Realization of Wheeler’s Delayed-Choice Gedanken Experiment. Science, 315(5814), 966–968.
Kim, Y.-H., Yu, R., Kulik, S. P., Shih, Y., and Scully, M. O. (2000). Delayed “Choice” Quantum Eraser. Physical Review Letters, 84(1), 1–5.
Vedovato, F., et al. (2017). Extending Wheeler’s Delayed-Choice Experiment to Space. Science Advances, 3(10), e1701180.
Aharonov, Y., Bergmann, P. G., and Lebowitz, J. L. (1964). Time Symmetry in the Quantum Process of Measurement. Physical Review, 134(6B), B1410–B1416.
Cramer, J. G. (1986). The Transactional Interpretation of Quantum Mechanics. Reviews of Modern Physics, 58(3), 647–687.
McTaggart, J. M. E. (1908). The Unreality of Time. Mind, 17(68), 457–474.
Medley, R. (2004). Ethics in Time Travel. OC-615.
Medley, R. (2026). The Prison and the Map. robmedley.com.