The classical Redfield ratio of carbon106 : nitrogen 16 : phosphorus1 is a cornerstone of biogeochemistry. With the use of >2,000 observations of the chemistry of particulate matter from small and large lakes, as well as nearand off-shore marine environments, we found that the best model to describe seston stoichiometry depended on the scale of analysis. We also found that there were better estimates for seston chemistry than the classical ratio for all habitats, whether freshwater or marine. Across the entire data set, a constant proportionality of C166 : N20 : P1 (±error) described the data, which implies higher C sequestration per unit of N and P in surface waters than given in the classical ratio. At a regional scale, however, C : P and C : N often declined with increasing seston abundance, rejecting a constant ratio model. Within both freshwater and marine habitats, higher seston abundance is often associated with lower C : P and C : N ratios (higher nutrient content). The difference in appropriateness of the constant ratio model with respect to the entire data compared with subsets of the data indicates a scale dependence in stoichiometric relationships in seston C:N: P ratios. Given these consistent shifts in seston chemistry with particle abundance, the narrower variation in seston chemistry associated with marine seston chemistry could occur because of a reduced range of particulate nutrient concentration. For all but the largest scales, the classical Redfield model of biogeochemical cycling should be replaced with a more general power function model.