The spatial pattern of seedlings, saplings and canopy trees was studied in two spruce (Picea abies (L.) Karst.) forests in central Sweden. Canopy and forest structure were determined in five 0.25 ha plots. Life stage classes were distinguished on the basis of age and size distributions. Ripley's K-function (1977) was used to analyze the spatial patterns within each class. A random distribution of seedlings gave way to a more aggregated pattern on a small scale during the establishment phase. Saplings and sub-canopy trees were strongly aggregated and canopy trees were again randomly distributed within the plots. The proportion of individuals growing in gaps was used as an index of association between the spatial pattern in saplings and sub-canopy trees and the occurrence of small (50-350 m 2) canopy gaps. Under the null hypothesis of independence the expected value of this statistic would equal the canopy gap ratio for the stand. Monte Carlo simulation of this statistic, using fixed sapling positions and randomly repositioned canopy gaps, confirmed the importance of canopy gaps for the final success of establishment of spruce. The association of understorey trees with gaps suggest that small gaps are typically closed by recruitment of new saplings from a sapling bank rather than by the release of larger suppressed trees.