Long-term continuous cropping of cotton has caused dramatic soil-borne diseases in many places, leading to substantial agricultural losses. However, in some areas of Xingjang Province, the obstacles caused by continuous cotton cropping can spontaneously restore and maintain high yields of cotton for many years. To analyze the variable spectrum of soil bacterial communities and changes in the community structure in these spontaneously restored fields during cropping, soils at depths from 1 to 30 cm were sampled from cotton fields with a history of 0, 1, 3, 5, 10, 15 or 20 years of cotton cropping in the Akesu region of Xinjiang Province. The bacterial communities in these samples were studied using 16S rRNA-based polymerase chain reaction-density gradient gel electrophoresis (PCR-DGGE) with samples from uncultivated land as a control. Bacterial community diversity indices including the Shannon-Wiener diversity (H), Abundance index (S) and Evenness index (E_H) were compared among these samples. Samples from uncultivated land had relatively high levels of the richness indices but both the Diversity and Evenness Indices were at lower levels. With increasing years of cotton cropping, both the bacterial Diversity and Evenness Indices increased, whereas the richness indices showed a general decrease. However, after 10 years of continuous cropping all these indices were restored to their original values or reached a relative stable level. Cluster analysis of DGGE fragments indicated that the seven samples were clustered into three branches: fragments from samples under successional cropping for 0, 10 and 20 years formed one small branch with a similarity of approximately 50%; fragments from successional cropping for 1, 5 and 15 years formed another branch with a similarity of 53%; and the last branch comprised fragments from successional cropping for 3 years with a similarity of 44%. Principal component analysis (PCA) showed that all of the samples were statistically correlated with the major component and fluctuated on the right of the major component between the positive and negative axes of the second principal component. Both cluster analysis and PCA results suggested that, compared to those from original uncropped fields, the bacterial community structure showed the most variation in samples from the field of 3-year cropping, whereas similar patterns of bacterial community structure were found between samples from fields of 10 years of cotton cropping and those from uncropped fields. Nineteen clones were sequenced from each band and among them one sequence was selected and submitted to GenBank (accessory no. JN572545-JN572563). By aligning with the GenBank database, all sequences from DGGE were classified into four groups: Microbacterium, Uncultured Chloroflexi bacterium, TM7 Phylum sp. canine, and Flavobacteria. Further analysis demonstrated that the isolated V3 sequences showed a homology of 88%-100% to known sequences in GenBank and 47% of the sequences belonged to bacteria which were not cultured. No microbial data were correlated with soil-borne plant diseases of cotton. The study demonstrated that the age of cotton fields had significant effects on soil bacterial diversity. Continuous cotton cropping exerted significant influences on the community structure of soil bacteria in Xinjiang Province, with an initial suppression effect on bacterial diversity. However, the bacterial community reached a stabilized or even increased level compared with its original state after 5 years of continuous cropping. In addition, correlations between variations in the bacterial community structure at a depth of 1-30 cm and the yield of cotton and pest disease attacks were also found in this study.