Spatial vegetation patterns are associated with ecosystem stability and multifunctionality in drylands. Changes in patch size distributions (PSDs) are generally driven by both environmental and biological factors. However, the relationships between these factors in driving PSDs are not fully understood. We investigated 80 vegetation plots along an aridity gradient in the Alxa plateau, Northwest China. The sizes of vegetation patches were obtained from aerial images, and the heights of patch‐forming species were measured in the field. Soil samples were collected on the bare ground between patches for determination of physiochemical properties. Point pattern analysis was used to infer plant–plant interactions. A model selection procedure was employed to select the best predictors for the shape of PSDs and biological factors (vegetation total cover, community plant height, and plant–plant interactions). We then used structural equation modeling to evaluate the direct and indirect effects of environmental and biological factors on the shape of PSDs. In our study area, two types of PSDs coexisted, namely those that best fit to power law distributions and those that best fit to lognormal distributions. Aridity was the main environmental factor, while community mean height and competition between plants were the main biological factors for the shape of PSDs. As aridity and community mean height increased, power law‐like PSDs were exhibited, whereas competition led to deviations of PSDs from power laws. Aridity affected the shape of PSDs indirectly through changes in community mean height. Community mean height was correlated with competition, thereby indirectly affecting the shape of PSDs. Our results suggest the use of community functional traits as a link between the environment and plant–plant interactions, which may improve the understanding of the underlying mechanisms of PSD dynamics.