Did physicists ever try to put empirical constraints on the properties of the aether? I recall reading a popularization by Isaac Asimov in which he remarked that the aether would have had to have a very low density, and then to explain the high speed of light it would have had to have a stiffness many orders of magnitude higher than that of steel. I wonder whether this is really the way 19th-century physicists would have approached the issue, or whether Asimov was just making up an imaginary analysis and retrospectively putting it into their minds. If the aether isn't entrained, then there is no obvious way to measure its inertia. But since experiments required partial entrainment, it seems like there should have been observable forces that were proportional to the density of the aether. E.g., if I accelerated a macroscopic object, it would have had to partially entrain some of the aether, and the aether's inertia should have added onto the physically observed inertia. But if the amount of aether entrained is proportional to the mass of the object (not its volume) then this is just a renormalization of the mass of all objects, and therefore not observable. From the WP article "Luminiferous aether," it sounds like this type of model also would have problems accounting for the wavelength-dependence of the index of refraction. Supposing that one can put an upper bound on the density of the aether, do we then get a lower bound on its shear modulus or something? Did anyone actually do this? I also seem to recall descriptions of failed attempts to make mechanical models of electromagnetic waves. Again, this was probably in some popularization. At the time I got the impression that these were actual mechanical models containing springs and gears, but I suppose that in reality they would have been models on paper -- if they existed at all ...? Actually it's not obvious to me that, even given modern knowledge, you can't make an actual mechanical model of a light wave, reproducing the transverse polarization and the superposition of waves. Is there some kind of no-go theorem that proves that this is impossible? It seems unlikely that such a model could correctly reproduce all the predictions of Maxwell's equations, since Maxwell's equations implicitly have SR built in, including things like relativistic addition of velocities.