This month's Scientific American has an article about preons, which are hypothetical particles that the standard-model particles would be built out of. They discuss a problem with confinement in these models. We know that, say, an electron has a size less than x. This requires an uncertainty in momentum of at least h/x. Say for simplicity that the preon is ultrarelativistic. Then its energy has to be at least h/x (in units with c=1), and this is equivalent to a rest mass of at least h/x. Putting in x<~10^-17 m for an electron gives m>~10^-25 kg, which is much too big for an electron. All of these arguments would seem to apply equally well to any theory in which standard-models have substructure. E.g., it would seem to apply to string theory. How is this not a showstopper for such models? The Sci Am article makes vague references to a resolution by some technical trick. The same problem occurs for a pi meson, and they say this was solved by Goldstone 1961 for bosons. Apparently 't Hooft extended the solution to fermions in 1979. Can anyone explain what's going on, using crayons?