[Objectives] The small-island effect (SIE), which challenges the understanding of the traditional species-area relationship (SAR), has become one of the hot topics in the field of island biogeography and habitat fragmentation. However, the existing researches have neglected to explore the SIE in fragmented urban habitats with high intensity of human disturbance. Our study examined SIE in bird assemblages on 30 fragmented woodlots (0.25﹣290.40 hm2) in Huaxi University Areas, Guizhou, Southwestern China (Fig. 1). [Methods] We downloaded Google imagery of the study area and used ArcGIS 10.7 to outline the boundaries of the 30 woodlots and calculate the area of each woodlot. We used the line-transect method to survey breeding birds on the 30 fragmented woodlots during the breeding seasons from 2017 to 2021, with the length of line-transects established (82.5﹣7 073.7 m) roughly proportional to the woodlots’ area. We employed package “sars” in R to construct four key regression models (i.e., the left-horizontal with one-threshold model, the continuous with one-threshold model, the linear model and the null model) to predict species-area relationship, and used the Akaike’s information criterion corrected for small sample size (AICc) to determine the most plausible model. [Results] Totally, we surveyed each line transect 11 times and recorded 98 bird species belonging to 11 orders and 41 families. Excluding high-flying birds, non-forest dwelling birds, and species that were recorded only once, the number of bird species recorded in different woodlots ranged from 12 to 49, with an average of 24 species per woodlot. Among the four key regression models we constructed, the left-horizontal with one threshold regression model, which had the lowest AICc value, was the best model for predicting species-area relationship, and the difference in AICc value between it and the next best model was more than 2 (Table 1). The results indicated the left-horizontal with one threshold regression model was the most plausible model for predicting species-area relationship. The model showed that above the area threshold of 1.16 hm2, the species richness increased with the woodlot’s area (Fig. 2), consistent with the area effect proposed by the traditional island biogeography. Whereas, below the area threshold of 1.16 hm2, species richness did not change significantly with the area (Fig. 2), suggesting a SIE. [Conclusion] In conclusion, our study revealed the existence of SIE in bird assemblages in Huaxi University Areas, Guizhou. We speculate that the emergence of SIE in bird assemblages may be related to the unique karst environment, food resources, and ecological function of “transfer station” and “stepping stone”. Although further studies that target at examining the underlying mechanisms for the SIE are needed, we suggest that habitats protection and fragmentation greenways as strategies in urban planning and construction and that protecting small woodlots with important ecological value should not be neglected while giving priority to the protection of large woodlots.