The derivation of this law can be found in a variety of references. To begin with, it is given in the original work by Mott: Journal of Non-Crystalline Solids 1, 1-17 (1968). The law has been re-derived by Ambegaokar, Halperin and Langer (Journal of Non-Crystalline Solids 8-10, 492-496 (1972)); from this re-derivation one can almost immediately see how the T-1/4 comes about. See also the following references: Metal-Insulator Transitions, by N.F. Mott, 2nd edition (Taylor & Francis, London, 1990), Sec. 1.15, and  Electronic Properties of Doped Semiconductors, by B.I. Shklovskii and A.L. Efros (Springer, Berlin, 1984), Ch. 9.

A lot of papers have indeed been published in the 80's and 90's. It is very easy, experimentally (graphically), to find  different exponent, such as -1/3 or more complex fractions conjuctured by a fractal structure of the percolating  network. I believe that this really is the main problem. When trying to fit the results with slightly different exponents, the correlation coefficient R2 is always excellent ... and -1/4 is not always the one leading to the highest value of R2.

Alain