A recently developed technique in soil ecology is the use of the bait lamina method, for assessing the feeding activity of soil animals in situ. Here the bait lamina method (slightly adapted to tropical conditions) was used for the comparison of four closely situated sites (a primary rain forest, a secondary rain forest and two mixed-species tree plantations) in the Brazilian Amazon (project SHIFT ENV-52). The total feeding activity (portion of empty apertures in a set of laminae) and the vertical distribution of the feeding activity were evaluated for differences between the sites. The feeding activities in the primary and secondary forest were at the same level and significantly lower than in the two plantations. Feeding activity in subplots where surface litter has been experimentally removed was significantly lower than in the controls with the litter layer intact, at all four sites. The abundance of soil macrofauna or Enchytraeidae was not correlated with the results of the bait lamina tests, while the abundance of mesofauna (probably mainly Oribatida) was consistent with the pattern of feeding activity. The feeding activities as determined with bait laminae did not correlate with the decomposition activities determined with litterbags. Due to the easy applicability in combination with its feasibility for statistical evaluation and its consistent responses to site differences in experimental treatments we consider the bait lamina method to be a promising approach for the biological assessment of tropical soils.