Abstract: The knowledge on the habitual preference of stream fish is the key to understand the species-environment relationship and is important for the protection and management of the diversity in fish. However, to date it is still unclear how a fish population varies in different habitat patches in the shallow streams. In this study, we investigated the fish species diversity and the population structures in different patches and months in local habitual conditions, based on the data collected monthly in 10 habitat patches (involving five pools and five riffles) in the headwaters of the Qingyi River from September 2012 to August 2013. The principal component analysis showed that pools and riffles varied substantially in local habitual conditions, such as the substrate size, the current velocity, the water depth and the concentration of dissolved oxygen. According to the results of paired t-test, the coefficients of the variability of the water depth, the current velocity and the concentration of dissolved oxygen in the riffles were significantly higher than those in the pools. This suggested that the habitual conditions in the pools were more stable than those in the riffles. Fifteen species were collected including 8 species of Cyprinidae fish that accounted for more than 50% of the entire samples in this study. We applied two-way crossed ANOSIM based on discontinuous variables of the occurrence of fish and found that the habitat patches and time (months) did not significantly affect the species composition. However, analysis based on continuous variables of the abundance of each species showed that the assemblage structure varied significantly in different patches and months. We performed SIMPER analysis to identify the key species that contributed to the inter-patch dissimilarity of the assemblage structure, and found that Vanmanenia stenosoma was more abundant in the riffles, and that other seven key species (e.g., Zacco platypus, Acrossocheilus fasciatus and Phoxinus oxycephalus) were more abundant in the pools. According to the results of ANOVA test there were more species in the pools than in the riffles, but there was no difference in fish abundance between the two patches. We observed that there were marked monthly changes in the fish abundance in the pools, and this may be associated with the fish cohort recruitment and the overwintering death which caused an increase in the abundance in October and a decrease in December. We also found that both the diversity and the abundance of fish in the riffles varied significantly over months, which may be caused by the cohort recruitment and death, as well as the overwintering shift in the habitat-patch. These suggested that in this area the distribution of fish species in different habitat patches might not conform to the conventional habitat-guild model because most fish species were habitat-generalists. However, the spatial and temporal changes in fish abundance may lead to a dynamic assemblage structure in different patches and months.