1. Results are analysed from 11 experiments in which effects of fish addition and nutrient loading on shallow lakes were studied in mesocosms. The experiments, five in 1998, six in 1999, were carried out in six lakes, distributed from Finland to southern Spain, according to a standard protocol.
2. Effects of the treatments on 29 standard chemical, phytoplankton and zooplankton variables are examined to assess the relative importance of bottom-up (nutrient enrichment) and top-down (fish predation) effects. For each year, the experiments in different locations are treated as replicates in a meta-analysis. Results of individual experiments are then compared in terms of the patterns of significant influences of nutrient addition and fish predation with these overall results (the baseline), and between years in the same location.
3. The overall meta-analysis gave consistent results across the 2 years, with nutrient loading influencing all of the chemical variables, and on average 31% of primary producer and 39% of zooplankton variables. In contrast, fish influenced none of the chemical variables, 11% of the primary producer and 44% of the zooplankton variables. Nutrient effects on the system were thus about three times greater than fish effects, although fish effects were not inconsiderable.
4. The relative importance of nutrients and fish in individual experiments often differed between years at the same location and effects deviated to varying degrees from the baseline. These deviations were treated as measures of consistency (predictability) of conclusions in repeat experiments. Consistency increased southwards and this is interpreted as a consequence of more variable annual weather northwards.
5. The influence of nutrient loading was greater southwards and this was probably manifested through naturally greater annual macrophyte abundance in warmer locations in consequence of the longer plant growing-season. There was no trend in the relative importance of fish effects with latitude but this may partly be an artefact of the simple fish community used. These findings suggest that nutrient control should be a greater priority than biomanipulation in the restoration of eutrophicated shallow lakes in warm temperate regions.
6. Starting conditions affected the outcome of experiments. High initial concentrations of total phosphorus and planktonic chlorophyll a concentration (created by local conditions prior to the experiment) led to de-emphasis of the importance of nutrient loading in the experiment.