There are various roads from SR to GR. In a couple of years I may get a chance to teach a semester-length class on relativity for liberal arts students. Any comments on what story line works best? Some possibilities: (1) Thought experiments with elevators suggest that the Newtonian distinction between inertial and noninertial frames rests on a shaky foundation. This leads to the equivalence principle, and more thought experiments with elevators show that there must be gravitational time dilation, as verified in Pound-Rebka and Hafele-Keating. Clocks and rulers change their behavior from point to point (i.e., the metric is not constant). (2) Instantaneous propagation of signals is inconsistent with SR. Therefore Newtonian gravity can't be right, and we must have phenomena such as gravitational waves. Thought experiments such as Feynman's sticky bead and the "Atlas" argument in Taylor and Wheeler's spacetime physics show that gravitational waves carry energy. SR tells us that mass and energy are equivalent, so gravitational fields must themselves act as sources of gravitational fields. This is pretty much the Einstein field equations put into words. (3) Maxwell's equations aren't invariant under Galilean boosts, so we're forced to use Lorentz transformations instead. But Newtonian gravity isn't invariant under Lorentz transformations, so again we need to make a new theory. This is summarized from the introduction to General Relativity from A to B by Geroch. (4) The following is summarized from the introduction to Einstein's paper "The foundation of the general theory of relativity" (annotated translation at the end of the pdf version of this book http://www.lightandmatter.com/genrel/ ). Thought experiments such as the parable of the two planets lead us to Mach's principle. Applying Mach's principle to a rotating frame gives noneuclidean spatial geometry. Of course more than one of these could be presented in a semester-length course -- maybe all of them could. Geroch doesn't really spell out #3 very explicitly, and I'm not clear on exactly what he has in mind. When we apply the Lorentz transformation to electrical interactions, we're forced to invent magnetism. (I think Purcell was the first to do this at the undergrad level, and it can be done at a gen ed level, e.g., http://www.lightandmatter.com/html_books/7cp/ch06/ch06.html#Section6.2 [Broken] .) So naively I guess this would suggest simply making Newtonian gravity into a twin of electromagnetism, and this is actually qualitatively a pretty decent picture, since it gives gravitational waves, although obviously it's wrong in detail (wrong polarization properties, ...) What is the crucial difference between the gravitational case and the EM case? Opposite sign of the coupling constant? The equivalence principle? The fact that mass, unlike charge, has special logical status in SR?