Although dissociative adsorption of molecular hydrogen on Si(100) is thermodynamically favored by an adsorption energy of almost 2 eV, the sticking probability is immeasurably small (less than 10@super -11@) at room temperature, indicating the presence of a large energy barrier to adsorption. An adsorption barrier is expected to manifest itself in desorption as well by imparting hyperthermal amounts of kinetic energy to the desorbing molecules. Surprisingly, however, H@sub 2@ molecules associatively desorbing from Si(001) show no signs in their translational or rotational kinetic energy distributions of having traversed such a barrier, in apparent contradiction with microscopic reversibility. We have obtained experimental and theoretical results resolving this long-standing puzzle. Using surface second harmonic generation as a sensitive coverage probe, we observed that the dissociative sticking probability increases markedly with hydrogen coverage, and decreases with exposure pressure. Both dependencies are very unusual and impose severe constraints on the adsorption mechanism. By combining detailed measurements of the adsorption and desorption kinetics with statistical mechanical modeling and ab initio calculations, we arrived at a quantitative, mechanistic description of adsorption/desorption consistent with all observations and providing a natural explanation of the barrier puzzle. The model involves two distinct reaction pathways. At intermediate to high hydrogen coverages, thermal adsorption and desorption are dominated by an adsorption-barrier free, autocatalytic pathway, while a non-autocatalytic, bare-dimer pathway with a ~0.7 eV adsorption barrier dominates at very low coverages. Fitted model parameters are in quantitative agreement with density functional theory calculations.