Feynman Was Right About the Paths

25 Aug, 2026

A new experimental result confirms Richard Feynman’s 1948 path integral formulation of quantum mechanics.

A single photon had 1,419,857 possible paths through an experiment at South China Normal University. Physicists reconstructed the quantum amplitude of every one of them, added those amplitudes together, and compared the result against where the photons actually landed. The two agreed.

Richard Feynman said in 1948 that quantum mechanics works exactly this way. What is new is not that he was right. It is that somebody finally measured the pieces.

Richard Feynman, circa 1960. From Engineering and Science, California Institute of Technology. Public domain.

The instruction nobody could check

Feynman published "Space-Time Approach to Non-Relativistic Quantum Mechanics" in Reviews of Modern Physics on April 1, 1948, opening on what he called a curious historical fact, that modern quantum mechanics began with two quite different mathematical formulations, Schrödinger's differential equation and Heisenberg's matrix algebra. His paper, he wrote, "will describe what is essentially a third formulation of non-relativistic quantum theory." It rests on two postulates. First, when an event can happen by alternative paths, you assign a complex contribution to each path and add them; the probability of the event is the absolute square of that sum. Second, the paths contribute equally in magnitude, and only the phase differs, set by the classical action along the path in units of Planck's reduced constant.

The second postulate carries the whole strangeness of the thing, because it explains where classical physics comes from. Classical mechanics says nature follows the path for which the action is stationary. Feynman's picture says nature does something much odder: every path contributes, but paths far from the classical trajectory acquire wildly varying phases and cancel each other out, and paths clustered near stationary action reinforce. The single familiar trajectory we call classical motion is what survives the interference of everything else.

That is a beautiful idea and a difficult one to test. Path integrals became one of the central calculational frameworks of modern physics, running through quantum field theory, particle physics, statistical mechanics, and condensed matter, and Feynman himself insisted the formulation was mathematically equivalent to the familiar one. Physicists trusted the sum over paths because it worked. What they could not do was reach into an experiment and pull out the individual path contributions the formula tells you to add.

What the experiment actually did

The new paper, "Direct experimental test of Feynman's path integral postulates with single photons," comes from Yong-Li Wen, Li-Man Tian, Yunfei Wang, Shanchao Zhang, Chang Li, Jianfeng Li, Enke Wang, Hui Yan, and Shi-Liang Zhu, and appeared in Science Advances on August 26, 2026. It builds directly on the same group's 2023 result in Nature Photonics, where they observed the propagators of single photons for the first time and used them to demonstrate the quantum principle of least action. A propagator is the amplitude for a quantum system to get from one position and time to another, and measuring it is the technical prerequisite for everything that follows.

For the new test, the team broke a photon's evolution into successive stages and measured 5 propagators along the chain. The apparatus resolves a finite ladder of possible positions at each stage, and combining those choices across the chain gives 17 to the fifth power, or exactly 1,419,857, discrete paths from start to finish. The number is not journalistic decoration; it falls out of the geometry of the instrument. From the measured propagators the researchers reconstructed the complex probability amplitude associated with each of those paths, coherently summed them the way Feynman's first postulate prescribes, and checked the resulting probability distribution against an independent measurement of where the photons appeared.

They then went after the second postulate separately, which is what makes this more than an elaborate consistency check. Feynman's rule requires the individual paths to contribute amplitudes of equal magnitude, with phase determined by classical action. Both relationships held, and the authors deposited the underlying figure data publicly, so the measurements behind that conclusion can be inspected rather than taken on trust.

Why nobody was shocked, and why that is the point

If this experiment had contradicted Feynman, modern physics would have a serious problem, and nobody expected one. Shi-Liang Zhu said as much: given the extraordinary success of the formulation over nearly 8 decades, the result itself was not unexpected, but seeing it work was still astonishing. His second remark is the one worth keeping. The familiar phrase "sum over all paths," he said, "was no longer just a symbolic instruction in a textbook, we could see its consequences emerge directly from experimental data." Physicists have not spent 78 years wondering whether the path integral was correct. They have spent 78 years using it, with results so consistently good that its authority came almost entirely from downstream success rather than from any direct look at its internal machinery. The achievement here is not a verdict. It is access.

If you want an intuition for what those amplitudes are doing, skip the road map and try a choir. Imagine 1.4 million singers, each at roughly the same volume, each starting at a slightly different point in the musical cycle. Some voices reinforce, others cancel, and what reaches your ear depends far less on how many singers there are than on where each one sits in its phase when the sound combines. Feynman's paths behave that way. Every alternative contributes; phase decides whether the contribution survives.

The analogy has limits, and so does the experiment. It is worth being precise about what the measurement does not show, because the temptation to overclaim is strong. It does not photograph a photon flying down 1,419,857 roads at once. Quantum mechanics permits several mathematically equivalent formulations, and an experiment can test predictions and structural relationships without settling what the mathematics ultimately says about reality. Nor did the researchers enumerate the infinite continuum of paths that appears in the mathematical path integral; they reconstructed every path in a finite, discretized version of it. And nothing here favors the many-worlds interpretation, the Copenhagen interpretation, or any other. What the experiment delivers is more modest and more useful than a metaphysical verdict: measurements of the amplitudes whose sum behaves exactly as Feynman said it should.

When instruments catch up to inference

Zhu says he hopes researchers from different backgrounds will adapt the experimental approach to their own systems, which is probably the most consequential thing about the result. A one-time confirmation is a headline. A reusable way to reconstruct path contributions experimentally is a tool, and tools are what turn an abstraction into a subject you can interrogate.

There is a pattern here that I keep running into from different directions. I wrote last week about chronostasis, the stopped-clock illusion, where careful measurement settled a question introspection could not: the distortion in perceived time is real but runs on the order of a tenth to a fifth of a second, nothing like the frozen full second of the folk description. That is one kind of boundary, between clock time and perceived time. The path integral is another, between an extraordinarily successful mathematical representation and the parts of that representation experiment can actually reach. Abstractions become interesting at exactly the moment instruments make them measurable, and the interesting part is rarely the confirmation. It is what becomes askable afterward.

Feynman gave physics a formulation organized around whole trajectories and the classical action rather than around wave functions or matrices, and it became enormously useful precisely because it let physicists reason in terms of alternative histories. Nearly 8 decades later, the experimental technology has caught up with the intuition, and researchers can reconstruct enough of those alternative histories to watch the mathematics assemble the answer.

For generations, physicists have learned one of quantum mechanics' strangest commands: sum over all paths. It worked so well that nobody seriously expected Feynman to fail this test. The surprise is that the instruction has finally become tangible. Feynman gave physics a way to calculate the impossible in 1948. In 2026, physicists began measuring the pieces.

My blog

Feynman Was Right About the Paths

The instruction nobody could check

What the experiment actually did

Why nobody was shocked, and why that is the point

When instruments catch up to inference

Further Reading