Silvopastoral agroforestry can satisfy some objectives required of European land-use systems: reduced agricultural production, increased timber production, increased product diversity and environmental enhancement. A national network experiment was set up on four sites, each representative of a UK grassland farming area, with three replicates of common treatments: sycamore (Acer pseudoplatanus L.) at two silvopastoral planting densities [100 (S100) and 400 (S400) stems ha–1] protected by tree shelters and an agricultural control (ACONT) all with grazing sheep; a woodland control (WCONT, 2500 trees ha–1) without grazing. Common management protocols were applied and common measurements recorded. Results are provided for the first six years. There were no significant differences between S100, S400 and ACONT in agricultural productivity, though there were significant differences between sites (P <0.001). There were no significant differences in tree survival between the silvopastoral treatments and WCONT (mean 92.5% ± 0.74) but there was a difference between S100 and S400 (90.8 vs 94.7%: P <0.001). There were significant differences between the sites (range 86.5–96.2%: P <0.001) and between the first three years, when replacement of dead trees took place, (82.5, 95.1 and 96.9% for years 1, 2 and 3 respectively: P <0.001). There were significant differences in the total height of the trees in years two to four between WCONT, S100 and S400 (113.5, 154.1 and 194.5 cm respectively in year four: P <0.001). However, by year six WCONT and S100 were similar (180.7 ± 17.31 cm) while S400 were taller (219.0 ± 22.80 cm: P <0.05). It is concluded that tree shelters maintained silvopastoral tree survival at the level of conventional woodland. Tree height extension was compromised on S100 where a higher animal:tree ratio resulted in greater animal activity and soil compaction around trees compared to S400. The site with poorly-drained soil proved to be unsuitable for sycamore-based silvopastoral agroforestry.