What ahrkon states is quite correct. Since all the information we might extract is encoded in the wavefunction we cannot ascribe a radius to the electron. The wavefunction is always spread out over some volume. In a metal for instance the wavefunction of an electron is a plane wave that spreads out over the entire volume of the metal. Strictly speaking, the electron is just as large as the metal you're looking at.
The lengthscales in such a problem are defined by the wavelenghts of the particles involved. In the metal the wavenumbers of the electron states are more or less given by k=n*pi*a/L, n=0,1,2,.. a is the lattice constant and L the total size of the system. The smallest lengthscale is thus the lattice constant. In a superconductor the electrons form Cooper-pairs. These cooper-pairs are far more spread out (even though they are composite objects they are regarded as a single (quasi) particle) with a "coherence length" of up to a couple of microns.
In the condensed state it is often not even useful to speak of "a" electron. In solids there are no longer electrons, but "dressed" objects called quasi-particles. An example of such an object is the Cooper-pair, or the small polaron.
In any case, it is not really useful to speak of the size of an elementary particle. This is a little different in string theory (not a subject I'm really familiar with) where particles are represented by one dimensional objects (loops). If i remember correctly these can have a size, but given the fact that these things live at the Planck scale they are probably a lot smaller then 10^-15 m