We applied various multivariate statistical methods to analyze data collected from fisheries-independent bottom trawl surveys during 1997 to 2000 on the continental shelf (30-200m) of the East China Sea (ECS). The multivariate statistical analyses include cluster analysis, non-metric multidimensional scaling (NMDS)), SIMPER and BIO-ENV analyses. Based on the analyses, we identified spatial structures of fish communities and their relationships with environmental variables. Fish communities were classified into the three groups, offshore community in ECS, mixture community in ECS, and outer edge community of continental shelf in ECS. Fish community structures changed along depth gradients. Depth tended to be a limiting factor for the distribution of each fish species considered in this study. Currents had different characteristics of temperature and salinity and influenced the distribution and migration of fishes, and subsequently spatial and temporal structuring of fish species compositions. The spatial structure and compositions of fish species in the outer edge community remained relatively stable over all seasons. However, for the offshore and mixture communities, their boundaries disappeared in the summer, but maintained in other seasons. There were only a few species that were solely found in one community. The depth ranges of different fish species overlapped, but biomass and presence/absence of the species varied significantly along depth gradients. Different fish species tended to have different optimal habitats. Moreover, for a given fish species they might require different habitats in different stages of their life history. Thus, they might migrate within the continental shelf area to search optimal habitats. As a result, classification of the three communities should be considered relative, and the boundary between fish communities could change with seasons. We concluded that the determination of permanent geographic distribution area of fish community might not be possible.