The study was carried out along a well-documented soil-fertility gradient within the old-growth lowland tropical forest at La Selva Biological Station, in N.E. Costa Rica. This forest (50-250 m elevation) receives ca. 4 m of rain annually . More information about La Selva is available at http://www.ots.ac.cr. We sited our studies at seven forest plots of the CARBONO Project plot network [29, 30], chosen to represent the broadest range of ant density, ant nest sizes and soil nutrient concentrations in this landscape, based on previous studies [11, 21]. The plots were 0.5 ha and were located across ca. 500 ha of the old-growth forest landscape.
At each of the seven sampling sites, three colonies of the gypsy ant, Aphaenogaster araneoides, were located by feeding a forager and following her back to her nest. Each colony was selected to be at least 10 m from all other colonies used in the study, to prevent overlap or adjacency of home ranges to ensure independence of sampling, based on known measures of home range area . All colonies were collected in their entirety by excavation. We considered a nest to be fully excavated when we reached a discrete terminal chamber, typically containing early instar brood unattended by workers that had typically already been collected. Any nest that we judged to be not completely collected was excluded and another nest was collected in its stead. Freshly collected colonies were kept alive for at least eight hours inside re-sealable plastic bags before being frozen at -20°C for at least 24 h. We then counted all individuals in each colony to determine colony size (the number of adult workers) and an index of colony growth (the ratio of adult workers and worker pupae). After drying each colony at 60°C to constant mass, we took three replicate subsamples to characterize the colony with stable isotope analysis. Each subsample consisted of a thorax and six legs of a single individual ant selected to be ca. 1 mg to provide sufficient mass for accurate determination of isotope ratios. The gasters were intentionally excluded to prevent inclusion of residual food. For several other ant species, comparisons of stable isotope values between heads and thoraxes have indicated no significant differences between these tissue types .
The isotopic composition of the standing leaf litter at the time and site of each colony collection was used as the baseline of the food web for A. araneoides. At each ant sampling site, three forest-floor fine-litter samples (including leaf, twig, and reproductive materials), each ca. 10 L in volume, were collected from three haphazardly selected nearby locations. These samples were washed to remove soil and then dried at 60°C to constant mass. Each dried sample was ground to powder in a Wiley Mill and homogenized, three subsamples of ca. 2 mg were analyzed for isotopic composition, and the mean of these values was used to characterize each sampling site.
Stable isotope ratios are represented as δ15N and δ13C, representing per mil (‰) proportionality of heavy:light isotopes, relative to a universal standard for each. Samples were weighed into tin capsules in a microbalance to the nearest milligram. Ratios of stable isotopes in the ant tissue were measured with a PDZ Europa 20/-20 isotope ratio mass spectrometer at the UC Davis Stable Isotope Facility. For each sample, values of δ15N and δ13C were calibrated using values from established laboratory standards, run every 12 samples, calibrated against NIST Standard Reference Materials IAEA-N1, IAEA-N2, IAEA-N3, IAEA-CH7, and NBS-22.
We use (δ15NA.araneoides- δ15Nbaseline) as the δ15N discrimination factor of A. araneoides, in which the baseline is local (collection-site) standing leaf litter. The δ13C discrimination factor of A. araneoides was similarly evaluated, using (δ13CA.araneoides- δ13Cbaseline).
For each of the seven sampling sites, litter-dwelling ant density and richness were estimated with an "intensive sampling" protocol . Ten 1 m2 quadrats were established at 10 m intervals along a linear transect. Within each quadrat, all litter-ant nests were collected, sorted to species, and measured for size and growth estimates using the same method applied to A. araneoides. The species richness of the litter-ants was estimated using the mean number of species per quadrat; the density of litter ants was estimated using the mean number of adult litter ants nesting in quadrat.
We estimated the rate of leaf litter decomposition at each of the seven sampling sites with 2 successive 1-yr deployments of sets of litterbags containing a common litter, 10 g of freshly fallen and dried leaves of the pioneer tree Cecropia obtusifolia. Each litterbag was sized ca. 100 cm2 and was constructed with 55 μm vinyl mesh. Three sets of five litterbags each were placed at each site in June 2006 and June 2007. Bags were sampled destructively in the following time sequence: 2 wk, 4 wk, 8 wk, 26 wk, 52 wk, for a total of six time points including t = 0. Using the mean values for each time step per plot, decomposition rate, k, was calculated based on the exponential rate of decay, e-kt, and for each site we averaged the resulting k values from the two measurement series (2006-2007 and 2007-2008).
Total nutrients in the surface soil were determined for each of the seven sampling sites (plots) based on compositing six regularly-spaced soil cores (0-10 cm depth) from each 0.5 ha plot. After being air-dried, sieved (2 mm) and ground, the samples were analyzed at the Institute of Soil Science and Forest Nutrition, University of Göttingen, Germany by HNO3-pressure extraction and ICP-AES (see for details). While indices of available P have been found to be poorly related to ecosystem function at this site (e.g., , both stand structure and ecosystem processes significantly vary across the 3-fold within-forest range in total soil P at La Selva . Total soil P was also previously shown to be a robust predictor of the variation in density of litter arthropods across the La Selva landscape .
To test the hypotheses that resource base, competition and life history are associated with trophic level and C source, we created multiple regression models for all variables with δ15N discrimination factor (δ15NA.araneoides- δ15Nbaseline) and δ13C discrimination factor (δ13CA.araneoides- δ13Cbaseline) as dependent variables. These multiple regression models should be interpreted cautiously given the number of comparisons and degrees of freedom. We used AIC  to select the factor(s) most closely associated with the dependent variable and to evaluate competing models for within-forest variance in the δ15N and δ13C of A. araneoides.