We analyzed nutrient data from the National Lakes Assessment (NLA), a probability survey of 1028 lakes >4 ha in lake area across the conterminous USA to quantify and contrast different methods of setting nutrient criteria. We calculated potential nutrient criteria for total P (TP), total N (TN), and chlorophyll a (Chl a) by 4 methods (25^th percentile of population, 75^th percentile of least-disturbed reference sites, diatom-based paleolimnological reconstruction, and stressor modeling) and compared them to existing draft US Environmental Protection Agency (EPA) criteria within national nutrient ecoregions. At the national scale, the ecoregional criteria derived from the different approaches were highly correlated. However, absolute values differed widely among approaches within ecoregions. Population 25^th percentiles were lower (often by a factor of 2–6) than values obtained with other approaches in almost all ecoregions, results indicating that population 25^th percentiles cannot be used as a surrogate for reference-site or paleolimnological approaches. Stressor regression models often did not explain much of the variance in nutrient concentration, especially in the less-disturbed ecoregions. For TP, diatom-inferred paleolimnological criteria were higher than reference-site-based criteria, which were higher than stressor-model criteria. For TN, these 3 approaches were very comparable. For Chl a, the reference-site and stressor-model approaches gave similar criterion values in low-nutrient ecoregions, but the reference-site 75^th-percentile approach had much higher criterion values than the stressor-modeling approach in the high-nutrient ecoregions. Use of NLA reference-site 75^th percentiles as nutrient criteria showed that 42% of the assessed lakes exceeded TP criteria, 47% exceeded TN criteria, and 32% exceeded Chl a criteria. Survey results also suggest that most lakes are P limited. Ninety-three percent of the lakes in the population had molar TN∶TP ratios > 16, and 52% had TN∶TP > 50.