The permanganate/oxalate reaction has been known for more than a century; however, its mechanism is still subject to debate. The latest general publication by a French group established a model that involved 14 steps including 8 equilibria. The model was found to be able to simulate experimentally observed phenomena and to account for the bistability in a continually stirred tank reactor (CSTR). However, some earlier reported observations that we found in the literature seem to be inconsistent with this model. We performed electron paramagnetic resonance (EPR) measurements and stopped-flow studies with spectrophotometrical detection to shed light on these contradictions. We found that one of the key steps of the model describing the decomposition of Mn(VII) is not acceptable at least with the indicated rate constant. The only other step in which permanganate is involved is not capable of accounting for the autocatalytic nature. Our striking observation that the reaction still can be autocatalytic when applying a large stoichiometric excess of manganous ions points out that autocatalysis cannot be purely explained by a positive Mn2+ feedback loop. Thus, we propose that the surface-catalyzed formation of colloidal manganese dioxide from Mn(II) and Mn(VII) provides a second positive feedback loop in the reaction.