The role of spatial variables, soil properties and overstorey structure in spatial variation of the herb-layer community in a riparian mixed forest was shown. The research were conducted in the "Dnipro-Orils’kiy" Nature Reserve (Ukraine). The research polygon was laid in a forest in the floodplain of the River Protich, which is a left inflow of the River Dnipro. Plant abundance was quantified by measuring cover within experimental polygon. The experimental polygon consisted of 7 transects and each transect was made up of 15 test points. The distance between rows in the site was 3 m. At the site we established a plot of 45×21 m, with 105 subplots of 3×3 m organized in a regular grid. Vascular plant species lists were recorded for each 3×3 m subplot along with visual estimates of species cover projection. The plant community was represented by 43 species, of which 18.6% were phanerophytes, 39.5% were hemikriptophytes, 9.3% were therophytes, 7.0% were geophytes. An overall test of random labelling revealed the total nonrandom distribution of the tree stems within the site. Constrained correspondence analysis (CCA) was applied as ordination approach. The forward selection procedure allowed us to select 6 soil variables, which explain 28.3% of the herb-layer community variability. The list of the important soil variables includes soil mechanical impedance (at the depth 0–5, 30–35, 75–80, and 95–10 cm), soil moisture, and soil bulk density. The variation explained by pure spatial variables is equal to 11.0 %. The majority of the tree-distance structured variation in plant community composition was broad-scaled. The significant relationship was found between the pure spatial component of the community variation and a lot of phytoindicator estimations of which the variability of damping and humidity were of the greatest importance. Trees stand was demonstrated to be a considerable factor structuring both the herb-layer community and spatial variation of the physical properties of soil.