The causes of the worldwide problem of encroachment of woody plants into grassy vegetation are elusive. The effects of soil nutrients on competition between herbaceous and woody plants in various landscapes are particularly poorly understood. A long-term experiment of 60 plots in a South African savanna, comprising annual applications of ammonium sulphate (146–1166 kg ha-1 yr-1) and superphosphate (233–466 kg ha-1 yr-1) over three decades, and subsequent passive protection over another three decades, during which indigenous trees encroached on different plots to extremely variable degrees, provided an opportunity to investigate relationships between soil properties and woody encroachment. All topsoils were analysed for pH, acidity, EC, water-dispersible clay, Na, Mg, K, Ca, P, S, C, N, NH4, NO3, B, Mn, Cu and Zn. Applications of ammonium sulphate (AS), but not superphosphate (SP), greatly constrained tree abundance relative to control plots. Differences between control plots and plots that had received maximal AS application were particularly marked (16.3 ± 5.7 versus 1.2 ± 0.8 trees per plot). Soil properties most affected by AS applications included pH (H2O) (control to maximal AS application: 6.4 ± 0.1 to 5.1 ± 0.2), pH (KCl) (5.5 ± 0.2 to 4.0 ± 0.1), acidity (0.7 ± 0.1 to 2.6 ± 0.3 cmol kg-1), acid saturation (8 ± 2 to 40 ± 5%), Mg (386 ± 25 to 143 ± 15 mg kg-1), Ca (1022 ± 180 to 322 ± 14 mg kg-1), Mn (314 ± 11 to 118 ± 9 mg kg-1), Cu (3.6 ± 0.3 to 2.3 ± 0.2 mg kg-1) and Zn (6.6 ± 0.4 to 3.7 ± 0.4 mg kg-1). Magnesium, B, Mn and Cu were identified using principal component analysis, boundary line analysis and Kruskal-Wallis rank sum tests as the nutrients most likely to be affecting tree abundance. The ratio Mn/Cu was most related to tree abundance across the experiment, supporting the hypothesis that competition between herbaceous and woody plants depends on the availability of anabolic relative to catabolic nutrients. These findings, based on more than six decades of experimentation, may have global significance for the theoretical understanding of changes in vegetation structure and thus the practical control of invasive woody plants.