Background and aims The changing soils is a neverending process moderated by numerous biotic and abiotic factors. Among these factors, trees may play a critical role in forested landscapes by having a large imprint on soil texture and chemical properties. During their evolution, soils can follow convergent or divergent development pathways, leading to a decrease or an increase in soil spatial complexity. We hypothesized that trees can be a strong local factor intensifying, blocking or modifying pedogenetic processes, leading to local changes in soil complexity (convergence, divergence, or polygenesis). These changes are hypothetically controlled by regionally predominating soil formation processes. Methods To test the main hypothesis, we described the pedomorphological features of soils under tree stumps of fir, beech and hemlock in three soil regions: Haplic Cambisols (Turbacz Reserve, Poland), Entic Podzols (Žofínský Prales Reserve, Czech Republic) and Albic Podzols (Upper Peninsula, Michigan, USA). Soil profiles under the stumps, as well as control profiles on sites currently not occupied by trees, were analyzed in the laboratory for 20 physical and chemical properties. In total, we analyzed 116 soil samples. The age of trees and time of tree death were determined using the radiometry (14 C), dendrochronology and repeated tree censuses. To process the data, we used multivariate statistics, namely, redundancy analyses (RDAs) and principal component analyses (PCAs). The statistical significance of variables was tested using Kruskal-Wallis, Dunn, and permutation tests. To reach the main aims of the present study, we examined the dataset at three levels of data complexity: 1) soil regions, 2) microsite (i.e., tree stump versus control site), and 3) soil horizon. Results Living tree roots and empty or infilled root channels were the most important pedogenic factors that affected the dimensions of soil horizons and the moisture in the root zone under tree stumps. Microsites explained almost 6% of the soil variability (p < 0.001,