As can be seen in the many creative questions here, it is hotly disputed whether you can measure the speed of light in only one direction to the point where a infamous postulate was made that the speed of light might be instantaneous in one direction; you can never know.

So here is yet another idea that seems to be capable of falsifying above. The question is, what seems to be wrong here / how might the universe conspire?

This is a variation of the original measurement, close to what is described here: Measuring the one-way speed of light?

Instead of a cogwheel we just rotate a path and shine a collimated beam of photons through it. At the end is a flux integrator. The predictions are then:

  1. there should be some rotational speed where no light can make it through the path, if the speed of light is finite.

  2. if the speed of light is finite, the integrator will show a particular shape (amplitude over time) for a particular angular frequency.

expanding on 2, at any given speed you should be able to measure the deviation between an instantaneous assumed speed of light (that will always have the same shape) and the experimentally measured value (which will have start and ends increasingly cut off), thus be able to derive the speed of light.

Of course, if you are able to rotate the device at close to relativistic speeds, you can also derive the speed from 1). But again, the idea is any rotational speed should show a deviation.

Goes without saying the apperatus can be rotated on any axis.

Here's a simulation where the speed of light is slowly increased.


Here are two pictures that sort of tell the story from the simulation.

the burger rotates, thus cutting off light

light enters the apparatus

  • 1
    $\begingroup$ Does this answer your question? Can One-Way Speed of Light be Instantaneous? $\endgroup$
    – Max
    Nov 1, 2021 at 23:44
  • 3
    $\begingroup$ I didn’t understand your proposal until I watched part of the linked video. Please consider editing the post so that it’s clearer to people who can’t or won’t watch a video (by e.g. adding an image here). $\endgroup$
    – rob
    Nov 1, 2021 at 23:47
  • 1
    $\begingroup$ What exactly is your question? Your predictions seem to be correct as well as your simulation (bar modelling light as a wave / relativistic effects) - without a clock or a mirror, the "one way speed of light" idea doesn't work in this case. $\endgroup$
    – Max
    Nov 1, 2021 at 23:51
  • 1
    $\begingroup$ rob, great idea I did add those. I did read all the other one way speed of light questions. I don't take issue with physics not caring about c being isotropic, but generally it seems acknowledged that you cannot form a test of the speed of light in one direction, only by measuring back and forth. This seems to do it? $\endgroup$
    – Shaggi
    Nov 1, 2021 at 23:53
  • 1
    $\begingroup$ maybe one can use your idea of interrupted light propagation in an experiment with a non rotating setup, timing a parapet interrupting the beam of light and timing when the detector sees the interruption. A thought experiment, as I have not put down numbers. Maybe at a LIGO leg. $\endgroup$
    – anna v
    Nov 2, 2021 at 9:11

4 Answers 4


Unfortunately, there is simply no possible way to measure the speed of light independently of your synchronization convention. In this case, if you use the standard isotropic synchronization convention then length contraction is also isotropic and so is the one way speed of light. If you use an anisotropic synchronization convention then length contraction is anisotropic and so is the one way speed of light.

In this case the anisotropic length contraction leads to the wall of the device becoming curved. Because of this curvature the light will pass through at the same rotational velocity regardless of the one way speed of light.

  • 2
    $\begingroup$ This seems like the correct answer, but it is admittedly a pretty subtle effect that leads to the conspiracy. $\endgroup$
    – Jagerber48
    Nov 2, 2021 at 3:11
  • 1
    $\begingroup$ Which one is the reference system that sees the wall curved? The disc rotation can be non-relativistic $\endgroup$
    – user65081
    Nov 2, 2021 at 4:29
  • 1
    $\begingroup$ I would love to see this plausible answer fleshed out. It’s not immediately clear to me that the curved spokes on a relativistic wheel are relevant, because the derivations of that shape are done in the frame where the wheel is rolling. My intuition for anisotropic length contraction is not very good. $\endgroup$
    – rob
    Nov 2, 2021 at 4:33
  • 2
    $\begingroup$ @cmaster-reinstatemonica your expectation is wrong. There is fundamentally no possible way to measure the one-way speed of light. More specifically, you cannot even define a one way speed without defining a synchronization convention. Then you only get out what you assumed. This is not a question of clever experimental design. It is simply a question of definition. This was proven decades ago by Reichenbach. $\endgroup$
    – Dale
    Nov 2, 2021 at 15:40
  • 2
    $\begingroup$ Here is Reichenbach’s original 1924 book: books.google.com/books/about/… but I don’t actually use it. This paper is a much broader overview: sciencedirect.com/science/article/abs/pii/… It describes many different approaches that are equivalent but I just use Reichenbach’s name since he has priority. Note, this is paywalled but there are non paywalled versions on the web that I won’t link to directly. $\endgroup$
    – Dale
    Nov 2, 2021 at 16:21

Let's simplify the experiment slightly by replacing the rotation of the shutter with linear movement perpendicular to the light path. Say the width of the gap, in the rest frame measured perpendicular to the light path, is $y$.

Since the shutter is moving at relativistic speed $v$, the contracted width $y'$ becomes


where $c_y$ is the speed of light along the axis perpendicular to the light path. This value is needed in order to calculate the original value $c_x$ which is the one-way speed of light along the horizontal x-axis:

$$c_x=L/t = L v / y'$$

where $L$ is the horizontal length of the shutter passage the light beam moves through and $t$ is the time it takes for the light to barely pass from the opening edge to the opposite closing edge of the moving passage as measured in the rest frame (we can experimentally select a speed $v$ for this to be the case.)

But now you have shifted the problem to the perpendicular axis - how do you measure $c_y$?

  • $\begingroup$ Thank you, I can see it's essentially the same for a moving shutter, and that relativistic contraction plays a role in deriving a value for c, defeating the purpose (because it is circular as you say). I guess the relativistic geometry for something spinning is really weird as it would contract more the further out the disc you go. $\endgroup$
    – Shaggi
    Nov 2, 2021 at 23:25
  • $\begingroup$ Why do you assert that c_x and c_y are different from c, which is the same in any direction? If you do this experiment in a vacuum, isn't there only one c? $\endgroup$
    – ttonon
    Nov 18, 2021 at 20:23
  • $\begingroup$ @ttonon The average back-and-forth speed is c, but in any one direction it is still unknown. See youtube.com/watch?v=pTn6Ewhb27k or en.wikipedia.org/wiki/One-way_speed_of_light $\endgroup$
    – Matt
    Nov 18, 2021 at 20:55

The way we determine something's shape ultimately boils down to interacting with it by bouncing a c-propagating signal off of it. If the apparatus looks like a circle with a straight-line aperture through it to us, interrogating it with c-propagating signals, then the apparatus looks like a circle with a straight-line aperture through the middle of it to the c-propagating signal you're trying to shoot through the apparatus.

If the apparatus looks like a circle with a wavy line (like a Pepsi logo or a yin-yang) through the middle of it (because it's spinning fast enough to deform), then it will look like a circle with a wavy line through the middle of it to the laser you're trying to shoot through the apparatus.

You get Dale's answer if we let the apparatus be perfectly rigid, that is, the speed at which forces propagate through the medium is the speed of light, so the apparatus deforms as the speed of light vs the rotational velocity.

You can run an argument that the speed of sound in any given medium scales linearly with the speed of light, but it's unnecessary. If the photons bouncing off of the apparatus and into our retinas tell us there's a straight path from one end to another, then there's a path from one end to the other that's straight for photons.

  • 5
    $\begingroup$ My answer has nothing to do with rigidity. You could make the apparatus out of jello and my answer would hold. $\endgroup$
    – Dale
    Nov 2, 2021 at 4:00

I assume this question was inspired by the same Veritasium video as all the others.

As explained in answers to earlier questions, all that he does in that video is define "velocity" differently, so that the same physical phenomena are assigned different velocity numbers, some of which are $\infty$ instead of $c$. It can't be falsified because it's just a word game.

See, e.g., this answer. To quote from it:

What we mean when we say that the speed of light is constant is that there exist coordinates with respect to which it's constant. In a Newtonian corpuscular world, no such coordinates would exist, so the fact that they do exist in the real world is physically meaningful. [...]

There also exist coordinates with respect to which the speed of light isn't constant. This is not physically meaningful, because no theory could ever avoid them; you can always do a formal substitution of variables [...]

If $(x,t)$ are standard inertial coordinates, then with respect to coordinates $(x,t')$ where $t'=t-x$, the speed of light $|dx/dt'|$ ranges from $c/2$ to $\infty$ depending on direction. Why don't we see this as an anisotropy in the sky? Because the universe in different directions has aged by different amounts, and their ages differ by just the right amount to compensate for the different light travel times.

So, let's apply that here. You wrote:

there should be some rotational speed where no light can make it through the path, if the speed of light is finite.

If the light is allowed to change direction, then it can always make it through, because the speed of the apparatus is everywhere less than the speed of light, so light can make radial progress while also moving transversely to avoid the sides.

If the light is not allowed to change direction, then it can't make it through at high enough rotational speeds even if its "velocity" is infinite, because with respect to the $t'$ used to define the "infinite" velocity, the path is curved enough that there is no straight-line way through. Why is it curved? Because (supposing the $(x,t)$ coordinates we started with were the center-of-momentum frame of the path), $t' = t-x$ cuts through different isotropic times $t$ at different $x$ positions, catching, in effect, different times in the path's rotation.

The path isn't consistently curved in these coordinates. It's most curved when it's roughly parallel to the $x$ axis, and straight when it's in the $yz$ plane. But you can't get the light through when it's in the $yz$ plane because the speed of light in that direction isn't $\infty$, but $c$.

This directional anisotropy isn't real. The problem is intrinsically symmetric. It appears anisotropic in these coordinates because the coordinates themselves are anisotropic for no reason.

  • $\begingroup$ Thank you for your perspective. I've come to terms with me misunderstanding exactly what was meant (like so many before me), I guess it's easy for us laymen to take it at face value. I'll study the second half of your post and get back. $\endgroup$
    – Shaggi
    Nov 2, 2021 at 23:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.