Birds were censused from the top of a 2- 5-m ladder positioned at the centre of circles of radius 20 m. Every 1 min the observer visually scanned one half of a circle, recording the identity of and the distance to every bird seen. This procedure was carried out for about eight scans and then repeated for the complementary semicircle. Birds tended to be repulsed from the observer and/or ladder. Consequently the census method cannot be used to estimate absolute densities. However, the frequency distribution of bird-observer distances was found to be independent of time of day, time of year, and bird species. The method can, therefore, be used at a particular location to compare different bird species and different times. Vegetation height affected the frequency distribution of bird-observer distances and so the method cannot provide accurate comparisons of bird densities in different areas. Fewer birds were seen per scan when the census was carried out 4-6 h after sunrise. The two observers in the study saw the same average number of birds per scan.