We differentially increased the numbers of ants of different genera in open, replicated 1x1-m patches of forest floor by placing, in the center of each patch, baits surrounded by exclusion screens of differing mesh size.  Small ants (Crematogaster sp., Aphaenogaster sp. and Prenolepis sp.) were more abundant on the baits with the smaller mesh, than in the leaf litter associated with the baits enclosed with large-mesh screening that permitted entry of the larger ants (Camponotus sp. and Formica sp.).  After 30 days of baiting, we started sifting all the leaf litter in each plot and counted the number of individuals in 40 spider genera and 2 groups of centipedes. We then used logistic regression to reveal the impact (positive, negative or none) of ants of different genera on the local occurrence and relative abundance of different spiders and centipedes, which are major generalist predators in the forest-floor food web.