Reaumuria soongorica, a drought-tolerant shrub typical of Tamaricaceae, is one of the dominant species in desert areas in northwest China. It functions as a key biotic barrier against desertification due to its strong effects of sand fixation, therefore is critical for maintaining the stability of desert ecosystems. Gansu Corridor (in west Gansu province), which is the main range of distribution for R. soongorica in China, is characterized by environmental harshness associated with long-term aridity, rare precipitation, high evaporation, extremely high temperature in summer and freezing conditions in winter. As a consequence, variations in plant morphological and physiological traits in this area are primarily accounted for by abiotic factors. In October, 2007, we located 11 natural populations of R. soongorica in Gansu Corridor along a geographic gradient from west to east, south to north. In each located population, we sampled seeds from 15 healthy, representative individuals which were at least 5 m apart from each other. Variations in four characteristics of seed activity including germination rate, germination potential, water-absorbing capacity and thousand-kernel weight (TKW) of sampled seeds were assessed using a combination of Analysis of Variance (ANOVA), correlation analysis and cluster analysis to reveal the pattern of variability in seed viability. The results showed that the germination rate, germination potential and TKW of R. soongorica seeds differed significantly among populations at the 0.01 level of significance, and the water-absorbing capacity at 42 h varied significantly at the 0.05 level of significance. In contrast, all of the four traits examined showed no significant difference within populations. The average Coefficient of Variance (CV) among populations across the four traits was 0.03, with the maximum value being 0.07 for TKW and the minimum being 0.01 for germination rate. TKW, germination rate and germination potential were positively correlated with each other (P < 0.01). All traits except water-absorbing capacity of seeds were positively correlated with longitude, and showed a weak (non-significant) positive correlation with mean annual precipitation and mean temperature in July; they also were negatively correlated with latitude and altitude, and showed a weak (non-significant) negative correlation with mean annual temperature and mean temperature in January. These results indicate that the four tested traits of R. soongorica seeds have a clear pattern of varying primarily along geographic gradients related to longitude, latitude and altitude. Cluster analysis (using Euclidean distances) including all four seed traits revealed that the 11 sampled populations can be grouped into five clusters: (1) Jiuquan 2 and Jiuquan 3 populations; (2) Jiuquan 1, Zhangye 1 and Zhangye 3 populations; (3) Wuwei 3 and Zhangye 2 populations; (4) Wuwei 2, Wuwei 5 and Wuwei 1 populations; (5) Wuwei 4 population. Information on traits related to the viability of R. soongorica seeds in different populations, and on how these traits vary over broad geographic regions in Gansu Corridor can be useful in doing future research projects for the conservation and utilization of its germplasm and can also help design provenances experiments.