Abstract: The spatial pattern of soil seed bank of Fraxinus mandshurica were examined by geostatistics in two stands with different densities, tree species composition and ground plants, in the Northeast of China. Asampling grid of 559 points, at each point a sample was located, was established according to regular grid design at three spatial scales (cell size were 0.5m×0.5m, 1m×1m, and 2m×2m) in each plot. Spatial pattern were quantified by using semivariance analysis, fractal dimension and Kriging maps. The object of this paper is to quantify the spatial pattern of seed bank, to compare its variations between two stands, and to explore what sort of mechanism would control its variations. The results indicated seed distribution showed appreciable heterogeneity within stand and clearly difference between the two stands.In plot 1 (mixed stand), seed bank, with a mean of 158.1±96.7 seeds·m^-2, higher sill (C₀+C=0.549) and 4.25m scales, has a stronger spatial pattern. There were a high autocorrelated variation (89.4%) and lowly random variation (10.6%) at this scale. However, at scale from 5m to 31.3m, its autocorrelated variation ranged 51.4%～69.8% and random variation ranged 30.2%～47.6%, also has distinctly mosaic patterns. Fractal dimension (D) analysis showed that the Dvalues was relatively lower(1.728～1.865), their coefficient of variation only about 26.7%～68.35%, indicated that there was a high self-similarity of patch shapes of seed bank.In plot 2 (pure stand), with a mean of 57.4±44.7 seed·m^-2, and lower sill (C₀+C=0.125), spatial patterns were not so strong as plot 1, which has a smaller scale (2.59m). Its spatially autocorrelated variation was 56.8% and randomly variation was 43.2%. Over the scale there were no clearly autocorrelated spatial patterns. Fractal dimension analysis showed higher Dvalue (1.906～1.960), and their coefficient of variation varying wildly (69.51%～134.14%), implying highly random variation and more complex patch shapes. The differences of spatial pattern of seed bank between the two stands were likely to be due to spatial heterogeneity at the level of stands.