Abstract In wood pastures spatial associations of tree saplings with nurse structures such as unpalatable plants are generally explained as associational resistance; i.e. reduced herbivory by association with a defended neighbour. However, these associations may result from other underlying processes that occur at the seed stage. Here we tested whether the observed associations between Picea abies tree saplings and nurse structures could also be explained by higher soil seed densities under versus outside nurse shrubs, or by reduced seed removal inside versus outside nurse structures. Further we tested for differential effects of the main seed predator groups on seed removal and establishment, both in gaps and in dense vegetation. We sampled in total 640 soil cores from inside and outside canopies of eight Crataegus monogyna shrubs and compared densities of P. abies seeds. Seed removal was studied inside and outside the canopies of four different nurse structures: C. monogyna, Gentiana lutea, rocks and tree stumps. The effects of cattle, birds, rodents and insects on seed removal and seedling establishment were studied using selective exclosures both in gaps and in dense vegetation. Soil seed densities under and outside shrubs did not differ significantly. Seed removal was considerable (85%), indicating the importance of seed removal as limiting factor for tree regeneration in wood pastures. In contrast to our expectations, seed removal was significantly higher under all four nurse structures then outside. The subsequent exclusion of cattle, birds and rodents reduced seed removal and increased establishment, but the individual contribution of each of these groups was relatively small, while the contribution of insects was greater. Seed removal was higher and, consequently, seedling establishment lower in dense vegetation than in the gaps. Our findings show that the earlier observed spatial associations between tree saplings and unpalatable plants, rocks and tree stumps are unlikely to be caused by an initially higher soil seed density or by reduced seed removal under these structures. Those structures were reported earlier to enhance tree sapling survival by offering protection against cattle grazing, but they appear not to protect tree seeds. This study shows the complexity of tree establishment in wood pasture ecosystems: apparently, safe sites for tree saplings are not safe for seeds.