Models describing the dependence oflake trophic state on external phosphorus loading are of two general types: empirical, steady state models derived from statistical treatment of data from large numbers of lakes, and theoretical, dynamic models based on more or less detailed mathematical descriptions of kinetics of nutrient and population dynamics.In the former type of models, inlake P concentrations have been found to depend mainly on inflow P concentrations and on water residence time. Efforts to explain the remaining variance by other factors such as iron, calcium, alkalinity, etc. have generally not been successful. Relations between lake P concentrations and chlorophyll may be useful for the prediction of maximum chlorophyll levels with some confidence but large variations in the actual chlorophyll concentrations are caused by, e.g. zooplankton grazing, varying chlorophyll content of the algae and other limiting nutrients.Dynamic, theoretical models are more suited for the prediction of rates of response to perturbations than the empirical steady state models. Due to the much larger demand for input data and labour investment, these models are generally more suited for the scientist as tools for formalizing and testing hypotheses, than for the water manager. Simple models with flexible structure generally have advantages over complex models.Both empirical and theoretical models are potentially powerful tools in eutrophication research and management. It is important for both scientists and water managers to select the optimum complexity with regard to the problem to be solved and to the resources available in each case.