Soil microbial communities are important indicators for evaluating the effectiveness of vegetational restoration. The ability of soil microbial communities to recover under six types of restoration was examined using Biolog, phospholipid fatty acid (PLFA), and denaturing gradient gel electrophoresis (DGGE). Soil samples were collected from eroded loessial soils that had been restored for 30 years on the Loess Plateau with natural grass, black locust, korshinsk peashrub, Chinese pine, and mixed forest of Chinese pine-indigo and black locust-indigo. Part of sloped farmland that had not been restored represented pre-restoration conditions. An 80-year-old forest of Chinese arborvitae represented the end point of restoration. Soils that were originally farmed but have been under restoration treatments for the past 30 years had significantly higher average well-color development, total PLFAs, bacterial PLFAs, fungal PLFAs, and Shannon indices of catabolic diversity, structural diversity, and bacterial phylogenetic diversity. These indicators, though, were lower than those for the native forest. Principal component analysis significantly separated the sloped farmland, Chinese arborvitae, and the six types of restoration. Redundancy analysis indicated that pH, total organic carbon, total nitrogen, available nitrogen, and available phosphorus were the most important environment factors in affecting soil functional diversity, and C/N ratio and available nutrients were the main properties in determining microbial community structure. Vegetational restoration substantially increased microbial biomass and microbial physiological activity, and shifted microbial community structure, but the community structures and compositions did not recover to the status of the natural forest during the early stage (<30 years) of vegetational restoration.