Many studies have shown that silicon oxidizes by interstitial transport of oxygen, that there is an oxygen exchange between the gas (O-2) and the SiO2-O-2 interface, and that an oxygen exchange between the oxygen molecules in O-2 is catalyzed by the SiO2 surface. A question remains on whether there is exchange between O-2 and the Si-SiO2 interface. We have investigated this by oxidizing silicon in three steps: in 210 mbar O-2 at 1100 degrees C; (1) in O-16(2) + (2) in O-18(2) + (3) in O-16(2) and measuring the amount of O-18 at the Si-SiO2 interface with nuclear reaction analysis after the secund and third steps. There was a 20% loss of the O-18 at the Si-SiO2 interface in the third step. The lost O-18 was transported interstitially through the oxide to the gas phase, without reacting with the lattice. This cannot be explained by the predominant Deal-Grove oxidation model. We propose a modification to the Deal-Grove model which allows us to account for the phenomena. The modification can also be used to explain why the oxidation of Si is very sensitive to changes of the surface reactions (contaminants).