Appendix S1. (a) Summary data and (b) correlation matrix for measures of tree cover within cultivated sites used in multivariate modelling (n = 19 sites). Correlations are Pearson's r with significance from pairwise tests (* P < 0.05; ** P < 0.01; *** P < 0.001; otherwise non-significant).

Appendix S2. Methods for calculation of individual species densities [reproduced from the methods outlined in the complementary study of Hulme et al., (2013)].

Appendix S3. Abundance of trees recorded in 19 × 1 km2 study sites used in analyses. Summary values of minimum, mean and maximum are calculated only across those sites where the respective tree was recorded.

Appendix S4. Forest-dependent bird species recorded in cultivated sites. Shown are habitat preference (F = forest generalist, FF = forest specialist), number of sites in which each species was recorded, and primary and secondary diet preferences determined from literature searches (Fr = fruits, S = seeds, I = insects and other invertebrates, N = nectar), current IUCN global status (LC = least concern, VU = vulnerable), current estimated population trend (De = decreasing, St = stable, Un = unknown) and whether species is restricted to the Guinea-Congo Forest Biome.

Appendix S5. Univariate tests of three measures of bird species richness in relation to seven measures of tree cover. Each explanatory variable was fitted as a linear term. Significant terms at P < 0.05 were taken forward into multivariate modelling, accounting for correlations between variables (see Methods).

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.