I'm not sure that claim is true. Take this experiment, where A sends B a message following two paths: /-->--B (/ and \ are mirrors) | | | | \--

I'm not sure that claim is true. Take this experiment, where A sends B a message following two paths: /-->--B (/ and \ are mirrors) | | | | \--<--A with speeds: (assume lengths are all 1m) /--cr--B | | cu cu | | \--cl--A The time it takes for path 1 (left,up,right) is cl+cu+cr. The time it takes for path 2 is cu. B can measure the difference cl+cu+cr-cu = cl+cr. A can compare cl+cr to cu: if cl+cr != 2cu the velocity is not isotropic. To see that's always possible, A and B can simply bounce back path 2, so B receives pulses every 2cu (and hence can measure and compare to other time intervals). Directional speed of light would be very weird, I think it'd show up everywhere in experiments if it weren't true. --- Also, I think there's a notable distinction in physics: they are usually called postulates to distinguish from axioms. A postulate is an assumption about a physical theory (usually something simple and "beautiful" -- mathematically neat and satisfying Occam's razor); if a theory doesn't match reality, one of its postulates is incorrect. An axiom in mathematics of course can't be proven wrong. Because axioms are the basis for your mathematical theory describing reality, they can't be incorrect (as long as they form a mathematically consistent theory); the most could happen is they're insufficient to describe reality (you need other axioms and another mathematical theory), but they're not (somewhat) falsifiable in the sense of physical postulates.