Forests often show clear vertical structure, which has significant influences on a series of structural and functional properties of forest ecosystems, such as seed dispersal ability, canopy radiation condition, photosynthetic capacity, and seedling growth. However, how forest vertical structure changes with horizontal scale has been less studied. The schrenk spruce (Picea schrenkiana) forest is the most wide-spreaded forest type in Xingjiang Province, northwest China. It has profound significance in soil and water conservation, timber harvesting, and climate regulation to the region. Analyzing the horizontal spatial patterns of different tree height classes can provide important insights into understanding the responsible processes that have generated the pattern. We conducted an across-scale analysis of forest structure to characterize the spatial characteristics of two contrasting Schrenk Spruce experimental forests in the campus of Xinjiang Agriculture University. All trees were mapped within a 2.4hm² rectangular plot for the establishing and the mature experimental forests. Species within each plot were identified, and the number of species, tree height, tree age, and DBH (diameter at breast height) were recorded for each individual tree. Based on these measurements, we categorized all the tree individuals into four height classes: the upper, intermediate, lower and sapling classes. The spatial distribution patterns and spatial associations of trees among different height classes were first analyzed using a point pattern analysis method, the Ripley's K statistic. Then, the Moran spatial correlograms were constructed to quantify the spatial pattern of tree height. Our results showed that: i) Schrenk Spruce tree individuals in different height classes distributed in an aggregative fashion. All tree-classes were clumped but at different distances. Bigger trees showed clumped distributions at greater distances. The sapling height class was the most clumped strata and had a positive association with other tree species; ii) Intra- and inter-species spatial associations of Schenk Spruce varied with the height of other tree species and the scale of analysis. Betula tianschanica Rupr. and Salix tianschanica Rgl. inhibited the regeneration of Schrenk spruce at all scales in the establishing population plot, whereas positive associations occurred between the scales of 10 to 40 m in the mature population plot; The intermediate height class and upper height class were spatially independent of Betula tianschanica Rupr. and Salix tianschanica Rgl; iii) Different spatial interactions existed between different height classes and also varied with the scale. Dead Schrenk spruce trees were spatially dependent of other trees at larger scales but independent at smaller scales. Negative repulsion interactions were found between saplings and lower living trees,whereas positive attraction interactions existed between saplings and dead trees.. Saplings were negatively associated with adult trees at almost all scales, iv) The correlogram analysis revealed the spatial gradients for different height classes. The distances of high positive autocorrelation values were 50 meters in both plots. Three distinct peaks of significant positive spatial autocorrelation were found for the sapling height class in the mature population plot, but no particular spatial autocorrelation in the establishing population plot. Negative spatial autocorrelation occurred above the 60m scale for both the intermediate height class and the upper height class, reflecting trees at upper strata tend to occure in larger patches.