Grassland degradation is one of the important environment and resource issues in North China. Different degrees of grassland degradation in alpine grassland are mainly caused by global climate change, overexploitation and other reasons, which threats to the sustainable development of grassland animal husbandry and improvement of living standards of farmers. Spatial structure of plant species is important in plant ecology as it affected the dynamics of plant populations, communities and ecosystem processes. Small-scale spatial association is the main sources of spatial heterogeneity, which can reveal the dynamic and stability of population and community structure. Stipa krylovii and Stellera chamaejasme are dominant species in alpine degraded grassland, have great significance for understanding sustainable utilization of grassland resources with studies on the spatial association of S. krylovii and S. chamaejasme. In this study for three main aims: First, to describe the scale transformation of spatial associations within the same gradient; Second, to study the spatial associations changes along with the degradation gradient; Third, to analyze the effect of bunch size structure and population patch on the spatial associations. Study sites locates in the Ma Changtan grassland in SuNan country of the north Qilian Mountain (38°47'22.1″-38°52'11″ N, 99°45'11.7″-99°57'41.2″ E). In July 2010, we recorded fine-scale (0-100 cm) spatial association of S. krylovii and S. chamaejasme from 2m×2m grids (2 cm×2 cm quadrats) in four plots under four kinds of degradation gradients: no degradation, light degradation, moderate degradation and heavy degradation. The point pattern analysis and Monte Carlo stochastic simulation method were used to quantitatively analyze spatial association. Based on bunch structure, patch characteristics, life history traits and strategies of ecological adaptation in the community, the scale transformation of spatial associations were discussed. The results showed that: (1) along with the natural pasture degeneration process extension, the population density and population territory area of S. krylovii gradually decreased, total area of hole gradually increased, and territory density increased firstly and then reduced. Meanwhile, the bunch size structure and population patch characteristics of S. chamaejasme showed the opposite change tendency; (2) under no degradation conditions, negative correlation was shown between S. krylovii and S. chamaejasme within 14 cm scale, whereas no significant was shown at 14 cm to 100 cm scales; negative correlation was shown under light degradation conditions within 51 cm scales, no significant correlation was shown at 51 cm to100 cm scales; under moderate degradation conditions, the spatial associations were no significant correlation within 86 cm scales, but positive correlation was shown at 86 cm to 100 cm scales; under heavy degradation conditions, there was no significant correlation as a whole and the scale ranges was from 0 cm to 100 cm. To sum up, the spatial associations between S. krylovii and S. chamaejasme were changed on the different scale and degradation gradient. The changes of population's bunch size structure, as well as the combining and annex of S. chamaejasme patches and the process of S. krylovii patch separated into many small "island" patches, are not only the antecedent conditions affecting scale transformation of spatial associations, but also the critical factors of population's status and function change.