Analysis of numbers of species (means in Fig. 2) to test effects of distance from the edge and of composition of experimental landscapes.

Site is a random factor with two levels; Composition is a fixed factor with four levels (Sparse, Dense, Sharp or Gradual). Distance is fixed with four levels (0–5, 5–10, 10–15 and 15–20 cm from edge of experimental landscapes). Site is a random factor with 2 levels; Patch is a random factor nested in Composition and Site and has three levels; n = 2 turfs sampled at each distance in each treatment. Non-significant terms at P >0.25 that were not pooled, did not change the outcome of any tests relevant for the hypotheses being tested.