A complete summary of the data sets, including original citations, can be obtained at http://www.nceas.ucsb.edu/ecostoichiometry. These datasets include: summary of insect nitrogen content; phylogenetic distributions of body nitrogen composition (percent N) among insects; atomic ratios of lake C, N and P; and references for each. Autotrophs: We supplemented the literature with unpublished data to develop databases documenting the C:N:P stoichiometry of terrestrial plants and suspended particulate matter ('seston') in lakes. We restricted our terrestrial autotroph database to elemental analyses of foliage collected under field conditions, excluding agricultural and greenhouse studies. Multiple data fora single species were averaged before analysis. Terrestrial data were most frequently reported in percentage dry weight terms (percent N, percent P); when percent C values were not reported we converted percent N and percent P data to C:N and C:P ratios using the mean percentage C of reported values (46.4 percent C). To evaluate whether this procedure introduced any bias to observed patterns in foliage C:P and C:N, we calculated the mean and variability (coefficient of variation, c.v.) of the C:N and C:P ratios for that subset of species for which percent C, percent N, and percent P were all reported (n = 44). The mean and c.v. of C:N for this limited data set were 35.9 and 0.57, respectively, and 805 and 0.78 for C:P. These values are reasonably close to those for the remaining entries (for C:N, mean was 36.5, c.v. was 0.64; for C:P, mean was 990, c.v. was 0.75); thus, using a fixed percentage C value to estimate C:N and C:P probably did not influence the major patterns observed. A total of 501 plant species from 358 genera, 107 families, 62 orders, 20 subclasses, 8 classes and 5 divisions were included. We assessed C:N:P stoichiometry at the base of freshwater pelagic food webs by compiling a database of seston elemental composition in 226 lakes from published and unpublished reports. Only data for surface waters during the summer growing season were included; multiple observations during a year were averaged, and thus a 'lake-year' was the primary observation unit. Data were generally for lakes of small to moderate size but information for several of the world's great lakes was also included. Lakes were primarily located in North America but seston data for lakes in Europe, Africa and Asia were also obtained. Seston contains a mixture of living algae but also bacteria, protozoa and detritus and forms the food base for relatively indiscriminate planktonic filter-feeders. Although the contribution of these different components probably differs among lakes, various data indicate that, in general, seston particles in stratified lakes are dominated by phytoplankton biomass. For example, even in some lakes where seston C:P was high (and thus the contribution of low-nutrient detritus might be thought relatively important), algae contributed about 70 percent of total seston biomass (bacteria and protozoa contributed, 20 percent and, 5 percent, respectively, implying little influence of detritus). Thus, the freshwater and terrestrial data sets for 'autotrophs' differ in that the terrestrial data involve observations for particular plant species while the lake data correspond to a mixture of particles, living and non-living. Finally, if different seston particles have substantially different C:N:P ratios, bulk seston C:N:P measurements may not accurately quantify actual stoichiometric food quality for particular herbivores that can discriminate among particles, such as some calanoid copepods. Herbivores: Data for the C:N:P stoichiometry of terrestrial herbivorous insects and lake zooplankton were compiled from published and unpublished sources. Multiple data for a single species were averaged before analysis. As for terrestrial plants, when values of percentage C were not given, data reported as percent N and percent P were converted to C:N and C:P ratios using the mean percentage C value for the remainder of the herbivore database (48 percent C). We followed the same procedure used in analyses of the foliage data to evaluate possible bias introduced by assuming this fixed percentage C value. However, data for few species included all three parameters (percent C, percent N, percent P); we thus confined our assessment of possible biases to data on herbivore C:N. The mean and c.v. values of C:N for the data subset with direct measurements of percent C and percent N (n = 67) were 5.9 and 0.21 whereas values for entries for which the fixed percentage C value was used (n = 97) were 6.7 and 0.28. Here again, using a fixed value of percentage C to estimate C:N and C:P from percent N and percent P probably did not unduly influence theobserved patterns. A total of 130 species of insects from 93 genera, 40 families and 7 orders were included. By far, most insects included were leaf-eating, though a minority were phloem-feeding herbivores (such as aphids). Leaf-eaters and phloem-feeders did not differ in C:N:P ratios and therefore all taxa were analysed together. Predatory zooplankton were excluded from the compilation but several omnivorous taxa were retained. A total of 43 species of zooplankton from 23 genera, 12 families, 8 orders, 4 classes and 2 phyla were included. The majority of the taxa were crustaceans (mainly branchiopods ('cladocera'), malacostracans and copepods) but data for several rotifers were also compiled. All stoichiometric ratios were calculated on an atomic basis.