In a set of small Ontario lakes sampled from 1978 to 1981, the correlation between the mean ambient total phosphorus concentration, P, and phytoplankton biomass ranges from r2 = 0.29 to r2 = 0.46. The correlation is somewhat higher between P and nannoplankton corrected for turnover rate than between P and the total phytoplankton or straight nannoplankton biomass. The correlation is affected by the abundance of grazers and by lake morphology. Regressions based on log–log transformed values showed a decreasing return to scale (i.e. biomass increases at a slower rate relative to increases in P) in contrast with many earlier studies. We suggest that a decreasing return to scale is to be expected when primary production, rather than chlorophyll or phytoplankton biomass, is used as the dependent variable. Lakes were clustered on the basis of mean abundance in four functional plankton groups (nannoplankton, net phytoplankton, herbivorous zooplankton, and carnivorous zooplankton), as well as on the basis of water chemistry and watershed geology. Only clusterings based on the structure of the entire plankton community appear to have a real effect on the phosphorus–phytoplankton relationship, showing increasing simple correlations from r2 = 0.29 to r2 = 0.41 within clusters and multiple correlations to R2 = 0.97 overall. These relationships imply that while the typical phosphorus–phytoplankton relationship is generally valid, its specific form depends on the characteristics or "type" of lake ecosystem.