In field experiments with complex food webs, numerous species may respond to a simple perturbation through networks of direct and indirect effects. Uncovering major interactions in such a maze is a challenge. We utilized diverse statistical approaches (GLM, MANOVA, best-subsets inverse logistic regression, path analysis) coupled with laboratory feeding trials in order to uncover interaction pathways between ants (8 species), spiders (46 taxonomic categories) and shared prey (entomobryid and tomocerid Collembola).  In two field experiments, we baited open 1-m² plots in order to increase ant activities and densities in forest-floor leaf litter, with untreated plots serving as controls.  Only the 9 most-abundant spider taxa (> 1 ind/m²) were included in multivariate models testing responses of four spider functional groups, which we defined based upon laboratory feeding trials with ants as prey. The only functional group that responded to the experimental treatments consisted of active foragers that fed readily on ants –the spider genera Gnaphosa and Schizocosa.  In Exp. 1 Gnaphosa was 2X more abundant, and Schizocosa was 50% less abundant, in baited plots relative to controls; and in Exp. 2  Gnaphosa biomass-density was almost 2x higher, and Schizocosa biomass-density was 50% lower, in plots where the large ant Camponotus was more active. In Exp. 1 there were more tomocerid Collembola in baited plots, which suggested the hypothesis that Camponotus indirectly affected tomocerid Collembola through effects on spiders. Path analysis failed to support this hypothesis; however, path analysis suggested other indirect effects of Camponotus on tomocerid Collembola.