Abies faxoniana forests create a zone of vegetation in the northwestern Sichuan alpine valley area and are important ecologically and as a center of biodiversity. This zone is sensitive to global climate change and is located in the eastern portion of the Qinghai-Tibet Plateau and the upper reaches of Yangtze River. After long-term extensive logging on a broad scale, the extent of these forests has been greatly reduced and the forests themselves degraded. More recently, forest restoration and regeneration are being emphasized in the region including old growth natural forests, plantation forests and natural secondary forests. Wolong Nature Reserve and Miyaluo Forest District were selected as an area to study soil faunal dynamics during the restoration of A. faxoniana forests. The study was conducted in old-growth A. faxoniana forest (OAFF) and 50-year-old A. faxoniana forest (50AFF) in April 2008. This study was conducted in northwestern Sichuan Province (27°57'-34°21'N, 97°26'-104°27'E), China. A total of 287 soil macrofauna individuals and 4681 soil meso-microfauna individuals, belonging to 33 and 57 families, respectively, were collected during the study. Three dominant groups, the Doratodesmidae, Paradoxosomatidae and Curculionidae, accounted for 62.02% of the soil macrofauna individuals. Nematoda dominated the soil meso-microfauna, accounting for 81.01% of the individuals. The number of taxonomic groups and the individual density of soil macrofauna in the 50AFF were significantly higher than those of the OAFF (P<0.05). The number of taxonomic groups of soil meso-microfauna in the 50AFF were significantly lower and the individual density was significantly higher than those of the OAFF (P<0.05). PCA ordination suggests the community structures of the two soil macrofauna communities were more different than those of the soil meso-microfauna communities. The similarities between the two soil macrofauna communities were lower than those of soil meso-microfauna, and the values of the Sørenson Index were lower than those of the Morisita-Horn Index. This implies the recovery rate is slower for the soil macrofauna than the soil meso-microfauna, and the restorative process has a relatively stronger influence on the taxonomic compositions of the soil macrofauna community than on the individual densities of the dominant and common taxonomic groups. Several ecological indices, such as species richness, abundance, and a diversity index, were analyzed. The Shannon (H), Margalef (D) and Simpson (C) indices of soil meso-microfauna communities in 50AFF were higher and the Pielou Index (E) was lower than that of the OAFF. But a significant difference (P<0.05) between the two soil macrofauna communities was only recorded with the Margalef Index (D). Correlation analysis suggests the group number and mean density of the soil macrofauna were remarkably positively correlated with organic matter content (P<0.05); the group number was remarkably positively correlated with litter depth content, but was also markedly negatively correlation with total N content (P<0.05). The group number and mean density of the soil meso-microfauna were only positively correlated with soil bulk density content (P<0.05). All results above showed the restoration of A. faxoniana forests can significantly improve soil macrofauna diversity; the taxonomic composition of the soil fauna community recovered slowly, but the individuals of the dominant and common taxonomic groups recovered rapidly.