The predictions of classical theory for the establishment of a network by free-radical polymerization have been tested under conditions for which those predictions should be simple, i.e., short pregel regimes in which the composition and chain length distribution of the primary chains should be nearly invariant and for a nearly equally reactive system (methyl methacrylate/ethylene glycol dimethacrylate). At the moderate to high concentrations of cross-linker used, the classical theory is inapplicable, each of its predictions being disobeyed. Gelation occurs 1-2 orders of magnitude later than predicted, the gel point is relatively insensitive to the amount of cross-linker, the dependence on the primary chain length is weaker than the predicted inverse dependence, and the divergence of the weight-average molecular weight is more consistent with the percolation picture. All of these point to a violation of the mean-field picture, probably through cyclization and (size-dependent) reduction in the reactivity of the pendants.