Litter invertebrates are notoriously patchy at small scales. Here we show that the abundance of 10 litter taxa also varies 100-fold at landscape and regional scales across 26 forest stands in Peru and Panama. We contrast three hypotheses that link gradients of abundance to ecosystem biogeochemistry. Of 14 factors considered (12 chemical elements, plus fiber and litter depth), four best predicted the abundance of litter invertebrates. In the Secondary Productivity Hypothesis, phosphorus limits abundance via the conversion of detritus to microbial biomass. Two of four microbivore taxa, collembola and isopods, increased with the percentage of P (%P) of decomposing litter. However, percentage of S (correlated with %P) best predicted the abundance of collembola, oribatids, and diplopods (r2 = 0.38, 0.33, 0.21, respectively). In the Structural Elements Hypotheses, N and Ca limit the abundance of silk-spinning and calcareous taxa, respectively. Mesostigmatids, pseudoscorpions, and spiders, all known to make silk, each increased with percentage of N of litter (r2 = 0.22, 0.31, 0.26, respectively). Calcareous isopods, but not diplopods, increased with percentage of Ca of litter (r2 = 0.59). In the Ecosystem Size Hypothesis, top predators are limited by available space. The abundance of the three remaining predators, chilopods, staphylinids, and ants, increased with litter depth (r2 = 0.31, 0.74, 0.69, respectively), and food webs from forests with deeper litter supported a higher ratio of predators to microbivores. These results suggest that biogeochemical gradients can provide a mechanism, through stoichiometry and trophic theory, shaping the geography of community structure.