Background and aims: Soil stability is a key ecosystem function provided by agricultural landscapes. A multitude of influential factors such as soil texture and plant community structure have been suggested, but few studies compare the relative importance of these factors for soil stability in the field. In addition, studies on effects of plant traits on soil stability have ignored intraspecific trait variability despite growing evidence of its importance for ecosystem functioning. Methods: Using path model analysis, we quantified the effect of plant functional traits (PFTs), abiotic soil characteristics and vegetation characteristics on three soil stability measures in 30 field margins of an agriculture landscape of Korea. We compare models with and without intraspecific trait variability. Results: Variance in soil stability was relatively well explained by our conceptual path model (81 % explained variance for soil aggregate stability, 50 % for penetration resistance and 35 % for soil shear vane strength). The overall most influential variable was root density while vegetation cover and species richness was much less important. Accounting for intraspecific trait variability improved the goodness-of-fit of all path models but not the overall explained variance. However, intraspecific trait variability allowed identifying important direct and indirect effects of PFTs on soil stability that would have remained hidden otherwise. Conclusion: We have demonstrated that the consideration of intraspecific trait variability – even though measuring it could strongly limit achievable sample sizes – is essential for uncovering the substantial effect of plant functional community composition on a key ecosystem function, soil stability.