The reduction of the surface oxide on Rh(111) by H(2) was observed in situ by scanning tunneling microscopy (STM) and high-resolution core level spectroscopy (HRCLS). At room temperature, H(2) does not adsorb on the oxide, only in reduced areas. Reduction starts in very few sites, almost exclusively in stepped areas. One can also initiate the reduction process by deliberately creating defects with the STM tip allowing us to examine the reduction kinetics in detail. Depending on the size of the reduced area and the hydrogen pressure, two growth regimes were found. At low H(2) pressures or small reduced areas, the reduction rate is limited by hydrogen adsorption on the reduced area. For large reduced areas, the reduction rate is limited by the processes at the border of the reduced area. Since a near-random distribution of the reduction nuclei was found and the reduction process at defects starts at a random time, one can use Johnson-Mehl-Avrami-Kolmogoroff (JMAK) theory to describe the process of reduction. The microscopic data from STM agree well with spatially averaged data from HRCLS measurements.